CodeMirror/waifu2x
1
0
Fork 0
mirror of synced 2024-10-04 20:23:43 +13:00
Code Issues Releases Wiki Activity
waifu2x/models/vgg_7/photo/scale2.0x_model.t7

1651 lines
4.2 MiB
Text
Raw Normal View History

Add 2x model for photo scaling
2015-11-30 21:29:34 +13:00
4
1
3
V 1
13
nn.Sequential
3
2
9
2
Convert model files; Add new pretrained model - Add new pretrained model to ./models/upconv_7 - Move old models to ./models/vgg_7 - Use nn.LeakyReLU instead of w2nn.LeakyReLU - Add useful attribute to .json New JSON attribute: The first layer has `model_config` attribute. It contains: model_arch: architecture name of model. see `lib/srcnn.lua` scale_factor: if scale_factor > 1, model:forward() changes image resolution with scale_factor. channels: input/output channels. if channels == 3, model is RGB model. offset: pixel size that is to be removed from output. for example: (scale_factor=1, offset=7, input=100x100) => output=(100-7)x(100-7) (scale_factor=2, offset=12, input=100x100) => output=(100*2-12)x(100*2-12) And each layer has `class_name` attribute.
2016-05-15 06:04:08 +12:00
13
w2nn_channels
Add 2x model for photo scaling
2015-11-30 21:29:34 +13:00
1
3
2
Convert model files; Add new pretrained model - Add new pretrained model to ./models/upconv_7 - Move old models to ./models/vgg_7 - Use nn.LeakyReLU instead of w2nn.LeakyReLU - Add useful attribute to .json New JSON attribute: The first layer has `model_config` attribute. It contains: model_arch: architecture name of model. see `lib/srcnn.lua` scale_factor: if scale_factor > 1, model:forward() changes image resolution with scale_factor. channels: input/output channels. if channels == 3, model is RGB model. offset: pixel size that is to be removed from output. for example: (scale_factor=1, offset=7, input=100x100) => output=(100-7)x(100-7) (scale_factor=2, offset=12, input=100x100) => output=(100*2-12)x(100*2-12) And each layer has `class_name` attribute.
2016-05-15 06:04:08 +12:00
17
w2nn_scale_factor
Add 2x model for photo scaling
2015-11-30 21:29:34 +13:00
1
1
2
6
output
4
Convert model files; Add new pretrained model - Add new pretrained model to ./models/upconv_7 - Move old models to ./models/vgg_7 - Use nn.LeakyReLU instead of w2nn.LeakyReLU - Add useful attribute to .json New JSON attribute: The first layer has `model_config` attribute. It contains: model_arch: architecture name of model. see `lib/srcnn.lua` scale_factor: if scale_factor > 1, model:forward() changes image resolution with scale_factor. channels: input/output channels. if channels == 3, model is RGB model. offset: pixel size that is to be removed from output. for example: (scale_factor=1, offset=7, input=100x100) => output=(100-7)x(100-7) (scale_factor=2, offset=12, input=100x100) => output=(100*2-12)x(100*2-12) And each layer has `class_name` attribute.
2016-05-15 06:04:08 +12:00
3
Add 2x model for photo scaling
2015-11-30 21:29:34 +13:00
3
V 1
16
torch.CudaTensor
0
1
0
2
9
gradInput
4
Convert model files; Add new pretrained model - Add new pretrained model to ./models/upconv_7 - Move old models to ./models/vgg_7 - Use nn.LeakyReLU instead of w2nn.LeakyReLU - Add useful attribute to .json New JSON attribute: The first layer has `model_config` attribute. It contains: model_arch: architecture name of model. see `lib/srcnn.lua` scale_factor: if scale_factor > 1, model:forward() changes image resolution with scale_factor. channels: input/output channels. if channels == 3, model is RGB model. offset: pixel size that is to be removed from output. for example: (scale_factor=1, offset=7, input=100x100) => output=(100-7)x(100-7) (scale_factor=2, offset=12, input=100x100) => output=(100*2-12)x(100*2-12) And each layer has `class_name` attribute.
2016-05-15 06:04:08 +12:00
4
Add 2x model for photo scaling
2015-11-30 21:29:34 +13:00
3
V 1
16
torch.CudaTensor
0
1
0
2
Convert model files; Add new pretrained model - Add new pretrained model to ./models/upconv_7 - Move old models to ./models/vgg_7 - Use nn.LeakyReLU instead of w2nn.LeakyReLU - Add useful attribute to .json New JSON attribute: The first layer has `model_config` attribute. It contains: model_arch: architecture name of model. see `lib/srcnn.lua` scale_factor: if scale_factor > 1, model:forward() changes image resolution with scale_factor. channels: input/output channels. if channels == 3, model is RGB model. offset: pixel size that is to be removed from output. for example: (scale_factor=1, offset=7, input=100x100) => output=(100-7)x(100-7) (scale_factor=2, offset=12, input=100x100) => output=(100*2-12)x(100*2-12) And each layer has `class_name` attribute.
2016-05-15 06:04:08 +12:00
7
modules
Add 2x model for photo scaling
2015-11-30 21:29:34 +13:00
3
Convert model files; Add new pretrained model - Add new pretrained model to ./models/upconv_7 - Move old models to ./models/vgg_7 - Use nn.LeakyReLU instead of w2nn.LeakyReLU - Add useful attribute to .json New JSON attribute: The first layer has `model_config` attribute. It contains: model_arch: architecture name of model. see `lib/srcnn.lua` scale_factor: if scale_factor > 1, model:forward() changes image resolution with scale_factor. channels: input/output channels. if channels == 3, model is RGB model. offset: pixel size that is to be removed from output. for example: (scale_factor=1, offset=7, input=100x100) => output=(100-7)x(100-7) (scale_factor=2, offset=12, input=100x100) => output=(100*2-12)x(100*2-12) And each layer has `class_name` attribute.
2016-05-15 06:04:08 +12:00
5
14
Add 2x model for photo scaling
2015-11-30 21:29:34 +13:00
1
1
4
Convert model files; Add new pretrained model - Add new pretrained model to ./models/upconv_7 - Move old models to ./models/vgg_7 - Use nn.LeakyReLU instead of w2nn.LeakyReLU - Add useful attribute to .json New JSON attribute: The first layer has `model_config` attribute. It contains: model_arch: architecture name of model. see `lib/srcnn.lua` scale_factor: if scale_factor > 1, model:forward() changes image resolution with scale_factor. channels: input/output channels. if channels == 3, model is RGB model. offset: pixel size that is to be removed from output. for example: (scale_factor=1, offset=7, input=100x100) => output=(100-7)x(100-7) (scale_factor=2, offset=12, input=100x100) => output=(100*2-12)x(100*2-12) And each layer has `class_name` attribute.
2016-05-15 06:04:08 +12:00
6
Add 2x model for photo scaling
2015-11-30 21:29:34 +13:00
3
V 1
Convert model files; Add new pretrained model - Add new pretrained model to ./models/upconv_7 - Move old models to ./models/vgg_7 - Use nn.LeakyReLU instead of w2nn.LeakyReLU - Add useful attribute to .json New JSON attribute: The first layer has `model_config` attribute. It contains: model_arch: architecture name of model. see `lib/srcnn.lua` scale_factor: if scale_factor > 1, model:forward() changes image resolution with scale_factor. channels: input/output channels. if channels == 3, model is RGB model. offset: pixel size that is to be removed from output. for example: (scale_factor=1, offset=7, input=100x100) => output=(100-7)x(100-7) (scale_factor=2, offset=12, input=100x100) => output=(100*2-12)x(100*2-12) And each layer has `class_name` attribute.
2016-05-15 06:04:08 +12:00
23
nn.SpatialConvolutionMM
3
7
16
2
2
dH
1
1
Add 2x model for photo scaling
2015-11-30 21:29:34 +13:00
2
2
dW
1
1
2
Convert model files; Add new pretrained model - Add new pretrained model to ./models/upconv_7 - Move old models to ./models/vgg_7 - Use nn.LeakyReLU instead of w2nn.LeakyReLU - Add useful attribute to .json New JSON attribute: The first layer has `model_config` attribute. It contains: model_arch: architecture name of model. see `lib/srcnn.lua` scale_factor: if scale_factor > 1, model:forward() changes image resolution with scale_factor. channels: input/output channels. if channels == 3, model is RGB model. offset: pixel size that is to be removed from output. for example: (scale_factor=1, offset=7, input=100x100) => output=(100-7)x(100-7) (scale_factor=2, offset=12, input=100x100) => output=(100*2-12)x(100*2-12) And each layer has `class_name` attribute.
2016-05-15 06:04:08 +12:00
11
nInputPlane
Add 2x model for photo scaling
2015-11-30 21:29:34 +13:00
1
Convert model files; Add new pretrained model - Add new pretrained model to ./models/upconv_7 - Move old models to ./models/vgg_7 - Use nn.LeakyReLU instead of w2nn.LeakyReLU - Add useful attribute to .json New JSON attribute: The first layer has `model_config` attribute. It contains: model_arch: architecture name of model. see `lib/srcnn.lua` scale_factor: if scale_factor > 1, model:forward() changes image resolution with scale_factor. channels: input/output channels. if channels == 3, model is RGB model. offset: pixel size that is to be removed from output. for example: (scale_factor=1, offset=7, input=100x100) => output=(100-7)x(100-7) (scale_factor=2, offset=12, input=100x100) => output=(100*2-12)x(100*2-12) And each layer has `class_name` attribute.
2016-05-15 06:04:08 +12:00
3
Add 2x model for photo scaling
2015-11-30 21:29:34 +13:00
2
Convert model files; Add new pretrained model - Add new pretrained model to ./models/upconv_7 - Move old models to ./models/vgg_7 - Use nn.LeakyReLU instead of w2nn.LeakyReLU - Add useful attribute to .json New JSON attribute: The first layer has `model_config` attribute. It contains: model_arch: architecture name of model. see `lib/srcnn.lua` scale_factor: if scale_factor > 1, model:forward() changes image resolution with scale_factor. channels: input/output channels. if channels == 3, model is RGB model. offset: pixel size that is to be removed from output. for example: (scale_factor=1, offset=7, input=100x100) => output=(100-7)x(100-7) (scale_factor=2, offset=12, input=100x100) => output=(100*2-12)x(100*2-12) And each layer has `class_name` attribute.
2016-05-15 06:04:08 +12:00
6
output
Add 2x model for photo scaling
2015-11-30 21:29:34 +13:00
4
Convert model files; Add new pretrained model - Add new pretrained model to ./models/upconv_7 - Move old models to ./models/vgg_7 - Use nn.LeakyReLU instead of w2nn.LeakyReLU - Add useful attribute to .json New JSON attribute: The first layer has `model_config` attribute. It contains: model_arch: architecture name of model. see `lib/srcnn.lua` scale_factor: if scale_factor > 1, model:forward() changes image resolution with scale_factor. channels: input/output channels. if channels == 3, model is RGB model. offset: pixel size that is to be removed from output. for example: (scale_factor=1, offset=7, input=100x100) => output=(100-7)x(100-7) (scale_factor=2, offset=12, input=100x100) => output=(100*2-12)x(100*2-12) And each layer has `class_name` attribute.
2016-05-15 06:04:08 +12:00
8
Add 2x model for photo scaling
2015-11-30 21:29:34 +13:00
3
V 1
16
torch.CudaTensor
Convert model files; Add new pretrained model - Add new pretrained model to ./models/upconv_7 - Move old models to ./models/vgg_7 - Use nn.LeakyReLU instead of w2nn.LeakyReLU - Add useful attribute to .json New JSON attribute: The first layer has `model_config` attribute. It contains: model_arch: architecture name of model. see `lib/srcnn.lua` scale_factor: if scale_factor > 1, model:forward() changes image resolution with scale_factor. channels: input/output channels. if channels == 3, model is RGB model. offset: pixel size that is to be removed from output. for example: (scale_factor=1, offset=7, input=100x100) => output=(100-7)x(100-7) (scale_factor=2, offset=12, input=100x100) => output=(100*2-12)x(100*2-12) And each layer has `class_name` attribute.
2016-05-15 06:04:08 +12:00
0
Add 2x model for photo scaling
2015-11-30 21:29:34 +13:00
1
Convert model files; Add new pretrained model - Add new pretrained model to ./models/upconv_7 - Move old models to ./models/vgg_7 - Use nn.LeakyReLU instead of w2nn.LeakyReLU - Add useful attribute to .json New JSON attribute: The first layer has `model_config` attribute. It contains: model_arch: architecture name of model. see `lib/srcnn.lua` scale_factor: if scale_factor > 1, model:forward() changes image resolution with scale_factor. channels: input/output channels. if channels == 3, model is RGB model. offset: pixel size that is to be removed from output. for example: (scale_factor=1, offset=7, input=100x100) => output=(100-7)x(100-7) (scale_factor=2, offset=12, input=100x100) => output=(100*2-12)x(100*2-12) And each layer has `class_name` attribute.
2016-05-15 06:04:08 +12:00
0
Add 2x model for photo scaling
2015-11-30 21:29:34 +13:00
2
2
kH
1
3
2
Convert model files; Add new pretrained model - Add new pretrained model to ./models/upconv_7 - Move old models to ./models/vgg_7 - Use nn.LeakyReLU instead of w2nn.LeakyReLU - Add useful attribute to .json New JSON attribute: The first layer has `model_config` attribute. It contains: model_arch: architecture name of model. see `lib/srcnn.lua` scale_factor: if scale_factor > 1, model:forward() changes image resolution with scale_factor. channels: input/output channels. if channels == 3, model is RGB model. offset: pixel size that is to be removed from output. for example: (scale_factor=1, offset=7, input=100x100) => output=(100-7)x(100-7) (scale_factor=2, offset=12, input=100x100) => output=(100*2-12)x(100*2-12) And each layer has `class_name` attribute.
2016-05-15 06:04:08 +12:00
9
gradInput
Add 2x model for photo scaling
2015-11-30 21:29:34 +13:00
4
Convert model files; Add new pretrained model - Add new pretrained model to ./models/upconv_7 - Move old models to ./models/vgg_7 - Use nn.LeakyReLU instead of w2nn.LeakyReLU - Add useful attribute to .json New JSON attribute: The first layer has `model_config` attribute. It contains: model_arch: architecture name of model. see `lib/srcnn.lua` scale_factor: if scale_factor > 1, model:forward() changes image resolution with scale_factor. channels: input/output channels. if channels == 3, model is RGB model. offset: pixel size that is to be removed from output. for example: (scale_factor=1, offset=7, input=100x100) => output=(100-7)x(100-7) (scale_factor=2, offset=12, input=100x100) => output=(100*2-12)x(100*2-12) And each layer has `class_name` attribute.
2016-05-15 06:04:08 +12:00
4
2
2
kW
1
Add 2x model for photo scaling
2015-11-30 21:29:34 +13:00
3
Convert model files; Add new pretrained model - Add new pretrained model to ./models/upconv_7 - Move old models to ./models/vgg_7 - Use nn.LeakyReLU instead of w2nn.LeakyReLU - Add useful attribute to .json New JSON attribute: The first layer has `model_config` attribute. It contains: model_arch: architecture name of model. see `lib/srcnn.lua` scale_factor: if scale_factor > 1, model:forward() changes image resolution with scale_factor. channels: input/output channels. if channels == 3, model is RGB model. offset: pixel size that is to be removed from output. for example: (scale_factor=1, offset=7, input=100x100) => output=(100-7)x(100-7) (scale_factor=2, offset=12, input=100x100) => output=(100*2-12)x(100*2-12) And each layer has `class_name` attribute.
2016-05-15 06:04:08 +12:00
2
12
nOutputPlane
1
32
2
4
padW
Add 2x model for photo scaling
2015-11-30 21:29:34 +13:00
1
0
2
6
weight
4
Convert model files; Add new pretrained model - Add new pretrained model to ./models/upconv_7 - Move old models to ./models/vgg_7 - Use nn.LeakyReLU instead of w2nn.LeakyReLU - Add useful attribute to .json New JSON attribute: The first layer has `model_config` attribute. It contains: model_arch: architecture name of model. see `lib/srcnn.lua` scale_factor: if scale_factor > 1, model:forward() changes image resolution with scale_factor. channels: input/output channels. if channels == 3, model is RGB model. offset: pixel size that is to be removed from output. for example: (scale_factor=1, offset=7, input=100x100) => output=(100-7)x(100-7) (scale_factor=2, offset=12, input=100x100) => output=(100*2-12)x(100*2-12) And each layer has `class_name` attribute.
2016-05-15 06:04:08 +12:00
9
Add 2x model for photo scaling
2015-11-30 21:29:34 +13:00
3
V 1
16
torch.CudaTensor
2
32 27
27 1
1
4
Convert model files; Add new pretrained model - Add new pretrained model to ./models/upconv_7 - Move old models to ./models/vgg_7 - Use nn.LeakyReLU instead of w2nn.LeakyReLU - Add useful attribute to .json New JSON attribute: The first layer has `model_config` attribute. It contains: model_arch: architecture name of model. see `lib/srcnn.lua` scale_factor: if scale_factor > 1, model:forward() changes image resolution with scale_factor. channels: input/output channels. if channels == 3, model is RGB model. offset: pixel size that is to be removed from output. for example: (scale_factor=1, offset=7, input=100x100) => output=(100-7)x(100-7) (scale_factor=2, offset=12, input=100x100) => output=(100*2-12)x(100*2-12) And each layer has `class_name` attribute.
2016-05-15 06:04:08 +12:00
10
Add 2x model for photo scaling
2015-11-30 21:29:34 +13:00
3
V 1
17
torch.CudaStorage
864
0.00319613935 0.385936052 0.154560208 -0.226256311 -0.111868329 0.210532621 0.142217129 -0.433677047 -0.0434815362 0.16460593 -0.0805575848 -0.243565097 0.359835029 0.0825311169 -0.424066395 -0.139719605 0.0418730751 0.0732983723 -0.0533087291 -0.331745595 0.1710345 -0.465663075 0.09218546 0.282856435 0.0704289973 0.190882131 -0.00597122777 -0.0280030575 0.0250577126 -7.88146353e-05 -0.0629380643 -0.0537395179 0.254756391 -0.0138474936 0.178419068 0.182388455 -0.0400946289 0.0765707493 -0.0283307694 0.0238939766 -0.0389560983 -0.221013486 0.0202924628 -0.1302239 -0.0947881117 0.0177350752 -0.039759405 0.0196579546 0.0507140458 0.051788155 0.000732009241 -0.00157982763 -0.0292589422 -0.0940610915 -0.0984597951 -0.147135854 0.081451118 0.0552935041 -0.0853654742 -0.042059876 -0.0456411578 0.0546464473 0.00647929125 -0.136268094 -0.178324372 -0.249754444 -0.0596325807 -0.110588178 -0.133965358 -0.0125132538 0.0820946544 -0.0230593178 0.0687591508 -0.0734574795 -0.089750126 0.0440542176 -0.0300320443 0.06794741 0.00908509456 -0.0808430985 0.0228329301 -0.00868495088 -0.0966249034 0.00904992409 0.0700424835 -0.00375666935 -0.0198186673 -0.0129138995 -0.00743022282 0.0345512368 -0.0583307929 0.11782302 0.0202076193 -0.101700276 0.180238247 -0.0821271762 0.0297952741 -0.0804032907 -0.0596643686 -0.0126981344 -0.01082219 0.00603096792 -0.129403651 0.0123206396 0.00280569913 0.0567342713 -0.248886406 0.0996902734 -0.0107798921 -0.0699014291 0.145679921 0.0234801471 0.071899429 -0.0553116724 0.0187216774 -0.0651869923 -0.0502634905 0.0280625001 -0.129783303 -0.136776522 -0.0203822274 -0.105611846 0.188128814 0.0351774842 -0.0456914939 0.0868926942 -0.0199443493 0.14976275 0.0500422902 -0.00775712868 0.101769045 0.102029152 -0.032691855 -0.167025402 -0.0804062113 -0.0521236658 -0.0920556635 -0.0762363747 0.027291717 -0.146486312 -0.14705874 0.0522372648 -0.0655959994 -0.189267322 0.0434191003 -0.0477990285 -0.0333269723 -0.0207669772 -0.267122447 -0.231416792 -0.0228656959 -0.0134868929 -0.103909612 0.0181147102 0.0428573042 0.0173908714 0.0449054725 -0.016360322 -0.0214377847 -0.0596830361 -0.0577857941 0.0427619442 0.0660886243 -0.0508728623 0.0180474091 -0.105272308 -0.0464422815 0.0480238274 -0.061216224 0.0449287184 0.0911002085 0.194061324 0.0266480036 -0.147735298 0.0451427847 -0.0275215618 -0.0745429546 -0.0778698176 0.0705130249 0.0456512012 -0.049664177 -0.0404105149 0.0380478948 -0.0900534391 0.0186118409 0.0124187311 0.0649260059 -0.0915073082 0.082780458 0.0429866426 -0.0569231249 0.0279757511 -0.0494377427 0.0691875443 -0.0376473516 -0.00141032843 -0.0287753828 0.0292691234 -0.0420424715 0.0207478963 -0.0208278168 0.0358044356 0.0129559962 0.0718667358 0.0677964389 -0.294619322 -0.1008498 0.0453634411 -0.114476129 0.0353640132 -0.0385543741 0.0753991902 -0.0438310578 0.0167727154 0.0594782233 0.0338023454 -0.0393698998 0.0208939202 0.00624796608 -0.00243287045 0.0484042838 -0.0318230204 -0.013782776 -0.0815613046 0.00820495002 0.0491944179 0.0172826666 -0.00112441508 -0.111486457 -0.0907194912 0.0210788418 -0.0676244348 -0.117431827 0.00134240999 0.0230772737 -0.0179099552 -0.00892999768 0.060750097 -0.0180201177 0.0146458596 -0.0508028604 0.0135505283 -0.0105846021 0.0533247367 -0.0246313736 -0.0274317786 -0.0364539996 0.0106004737 0.00311781629 0.0157746095 -0.0166240633 0.0194131378 -0.457280755 -0.0802552849 -0.00864348467 -0.0447495356 0.104437478 0.0215352066 0.0166262463 -0.0228977986 -0.0258988906 -0.120233335 0.103652738 0.0433478169 0.0339284725 -0.00238839746 -0.0695558935 0.00649555773 -0.00839226507 0.0275769979 0.00726352958 -0.0105558122 0.0449342951 -0.0146041876 -0.0494973622 0.0142917596 -0.0878995657 -0.0103174904 -0.0170356072 0.0203096252 -0.236406907 -0.164451346 0.0777543634 -0.143420622 -0.287133843 0.0679773763 0.0507716201 0.0533377193 0.0256505217 -0.0650997981 0.0261502266 -0.000391386711 -0.0297729298 0.0843436569 -0.0540001616 -0.0291272365 0.0815765411 -0.00303735072 0.0376446545 -0.0324782357 -0.00655954005 -0.057436008 0.0701829121 0.0056912927 0.0351031721 -0.0480551645 0.0108280387
2
5
train
5
0
2
10
gradWeight
4
Convert model files; Add new pretrained model - Add new pretrained model to ./models/upconv_7 - Move old models to ./models/vgg_7 - Use nn.LeakyReLU instead of w2nn.LeakyReLU - Add useful attribute to .json New JSON attribute: The first layer has `model_config` attribute. It contains: model_arch: architecture name of model. see `lib/srcnn.lua` scale_factor: if scale_factor > 1, model:forward() changes image resolution with scale_factor. channels: input/output channels. if channels == 3, model is RGB model. offset: pixel size that is to be removed from output. for example: (scale_factor=1, offset=7, input=100x100) => output=(100-7)x(100-7) (scale_factor=2, offset=12, input=100x100) => output=(100*2-12)x(100*2-12) And each layer has `class_name` attribute.
2016-05-15 06:04:08 +12:00
11
Add 2x model for photo scaling
2015-11-30 21:29:34 +13:00
3
V 1
16
torch.CudaTensor
2
32 27
27 1
1
4
Convert model files; Add new pretrained model - Add new pretrained model to ./models/upconv_7 - Move old models to ./models/vgg_7 - Use nn.LeakyReLU instead of w2nn.LeakyReLU - Add useful attribute to .json New JSON attribute: The first layer has `model_config` attribute. It contains: model_arch: architecture name of model. see `lib/srcnn.lua` scale_factor: if scale_factor > 1, model:forward() changes image resolution with scale_factor. channels: input/output channels. if channels == 3, model is RGB model. offset: pixel size that is to be removed from output. for example: (scale_factor=1, offset=7, input=100x100) => output=(100-7)x(100-7) (scale_factor=2, offset=12, input=100x100) => output=(100*2-12)x(100*2-12) And each layer has `class_name` attribute.
2016-05-15 06:04:08 +12:00
12
Add 2x model for photo scaling
2015-11-30 21:29:34 +13:00
3
V 1
17
torch.CudaStorage
864
Add upconv_7/photo models. Update vgg_7/photo/noise* models
2016-07-04 07:51:22 +12:00
0 0 0 0 2.00878077e-07 8.83790868e-17 -1.0032074e-05 0.000134123227 6.78975364e-16 4.02195781e-07 -5.64576163e-18 -6.3107537e-09 -4.82995657e-11 7.5557304e-14 -8.27901463e-12 4.86363447e-08 4.89801816e-14 1.97747198e-07 -6.65547873e-07 8.15943204e-05 -5.12787235e-09 1.76546194e-10 8.38362908e-13 9.84428337e-12 5.60982585e-07 -3.45519729e-05 9.01648605e-07 9.46736145e-06 -9.16021148e-10 -5.66364802e-11 -3.94458984e-05 -4.25848176e-11 -9.30360372e-11 -8.62138192e-16 1.2183034e-06 -2.86939084e-10 -5.55941693e-09 1.48311537e-11 -1.04848487e-12 -1.15325729e-05 4.72590708e-20 -5.92280447e-09 -8.32336688e-09 2.8694811e-09 -3.3588976e-09 4.9002411e-09 1.00775115e-13 -1.79409843e-09 1.7262905e-10 2.56119847e-06 2.03077292e-07 -2.86144602e-14 -1.46325237e-05 -2.41008324e-09 -1.89015248e-09 -2.67405949e-06 6.65873722e-06 3.98253912e-12 3.18276239e-10 -6.66453985e-08 6.91706553e-11 -7.94525544e-15 1.42738127e-10 9.2912969e-11 2.76935425e-05 2.3841875e-07 -1.82506511e-07 1.47487054e-08 7.46112016e-17 -4.19074237e-12 1.02524855e-09 -3.80317218e-08 1.14980622e-10 4.54878034e-07 3.24746381e-08 2.39931467e-07 -7.34286143e-09 -2.68278256e-07 5.18027954e-10 2.1885147e-09 3.26459348e-09 -6.30195842e-08 3.57949957e-14 -1.12689567e-12 2.48390597e-09 4.10219615e-14 -2.62980693e-10 1.75987848e-12 5.36376273e-19 -7.22445104e-10 6.41793174e-10 -1.79847769e-11 -3.14350546e-09 -9.79498947e-12 1.32657622e-12 -4.4536326e-11 -1.20552926e-07 -1.88840978e-08 -4.03996905e-11 2.72183831e-09 -1.90190003e-10 -6.14233404e-12 9.63411425e-13 -4.21379112e-11 1.31928934e-09 -4.21232563e-11 5.65605635e-12 -2.426302e-08 3.63747877e-12 -2.08398291e-07 -1.08964248e-12 7.56439135e-07 -9.78881872e-05 1.87098648e-09 0.000236777618 1.33264422e-07 -1.20415474e-18 2.97980129e-08 3.55106806e-07 -2.13071311e-10 1.79408721e-06 3.48306592e-11 4.33246827e-10 -1.01381992e-09 1.11982061e-07 -5.58092211e-07 -8.26567081e-09 1.131435e-26 -2.33026782e-07 -3.09976179e-14 4.43423701e-11 -9.05624038e-05 1.80269089e-05 2.22605337e-07 8.41654191e-10 1.16128418e-09 1.04609115e-17 -8.47844128e-10 1.10773834e-14 -4.27492962e-17 -2.11263932e-06 -8.64334299e-11 1.25049814e-06 4.59903569e-15 -4.22878089e-17 5.12684445e-14 4.13753416e-19 8.05521982e-10 -5.68944529e-07 -8.86580698e-10 -1.71964909e-09 -2.23491314e-09 6.13233988e-06 7.09707987e-11 6.18011894e-33 1.52034241e-17 -4.89568368e-12 1.81278231e-14 -1.12607602e-25 -2.11509314e-06 -1.74144731e-07 1.32968148e-06 -6.98839475e-09 -4.49221128e-15 -1.44142002e-14 3.25842748e-08 1.46342623e-15 2.2343626e-13 3.84374133e-08 -3.58442095e-11 -2.2646887e-07 4.83962107e-11 -9.43366416e-08 -1.65074954e-09 -1.88129379e-09 1.03174802e-08 -6.4384123e-11 -4.27680835e-10 3.34994673e-13 -5.92918798e-20 5.48874937e-07 1.89378362e-08 1.74479292e-10 -3.19999539e-12 6.0174743e-10 1.09602183e-06 -1.48975499e-08 7.74733593e-08 7.27862548e-10 2.95342659e-07 7.56339769e-10 3.86028898e-09 -9.56476242e-07 -6.99146145e-13 -1.44621897e-14 -1.34226987e-11 1.02711919e-08 4.09910208e-05 -5.87477889e-09 -6.90327569e-12 -1.11258652e-20 1.03133391e-09 1.41728799e-11 -7.50275859e-08 7.04061365e-10 -3.75217795e-08 6.3716028e-09 1.26203686e-10 4.86400381e-07 4.02472288e-05 -1.23222276e-06 4.98700956e-06 -8.64520788e-10 -1.69126878e-07 8.01571809e-10 3.06212377e-22 -3.02392036e-05 -1.81186556e-06 9.47863919e-06 -9.1683286e-11 6.74389527e-16 1.33834593e-11 -6.17259716e-11 -3.52064111e-09 -2.96637758e-12 7.83816745e-07 -7.32537501e-06 1.78617057e-07 -1.71564984e-08 -3.16418962e-08 3.17426139e-14 -5.53331993e-05 9.57663957e-08 1.24794736e-14 1.21436301e-06 1.36937817e-09 -6.7057e-09 -2.97508684e-08 -3.13674966e-20 2.18006455e-20 6.56492224e-17 -3.31340388e-18 -7.32214446e-14 -6.74734679e-07 -7.32004196e-07 1.0633552e-07 4.765743e-08 -2.42565246e-10 3.63369452e-07 -9.38715945e-12 5.22472413e-20 -7.48762091e-07 2.84092616e-09 -1.61390096e-10 -2.13804365e-12 2.79890351e-08 6.8541095e-10 9.2245088e-08 6.03117695e-11 3.22379845e-10 -2.81691901e-12 4.47711326e-14 -1.8047038e-07 1.66747882e-13 4.59713021e-07 6.47230464e-11 -8.24375464e-12 6.73460094e-13 -4.5659112e-09 1.24764279e-06 -1.13
Convert model files; Add new pretrained model - Add new pretrained model to ./models/upconv_7 - Move old models to ./models/vgg_7 - Use nn.LeakyReLU instead of w2nn.LeakyReLU - Add useful attribute to .json New JSON attribute: The first layer has `model_config` attribute. It contains: model_arch: architecture name of model. see `lib/srcnn.lua` scale_factor: if scale_factor > 1, model:forward() changes image resolution with scale_factor. channels: input/output channels. if channels == 3, model is RGB model. offset: pixel size that is to be removed from output. for example: (scale_factor=1, offset=7, input=100x100) => output=(100-7)x(100-7) (scale_factor=2, offset=12, input=100x100) => output=(100*2-12)x(100*2-12) And each layer has `class_name` attribute.
2016-05-15 06:04:08 +12:00
2
5
_type
2
16
torch.CudaTensor
Add 2x model for photo scaling
2015-11-30 21:29:34 +13:00
2
4
Convert model files; Add new pretrained model - Add new pretrained model to ./models/upconv_7 - Move old models to ./models/vgg_7 - Use nn.LeakyReLU instead of w2nn.LeakyReLU - Add useful attribute to .json New JSON attribute: The first layer has `model_config` attribute. It contains: model_arch: architecture name of model. see `lib/srcnn.lua` scale_factor: if scale_factor > 1, model:forward() changes image resolution with scale_factor. channels: input/output channels. if channels == 3, model is RGB model. offset: pixel size that is to be removed from output. for example: (scale_factor=1, offset=7, input=100x100) => output=(100-7)x(100-7) (scale_factor=2, offset=12, input=100x100) => output=(100*2-12)x(100*2-12) And each layer has `class_name` attribute.
2016-05-15 06:04:08 +12:00
bias
4
13
Add 2x model for photo scaling
2015-11-30 21:29:34 +13:00
3
V 1
16
torch.CudaTensor
1
Convert model files; Add new pretrained model - Add new pretrained model to ./models/upconv_7 - Move old models to ./models/vgg_7 - Use nn.LeakyReLU instead of w2nn.LeakyReLU - Add useful attribute to .json New JSON attribute: The first layer has `model_config` attribute. It contains: model_arch: architecture name of model. see `lib/srcnn.lua` scale_factor: if scale_factor > 1, model:forward() changes image resolution with scale_factor. channels: input/output channels. if channels == 3, model is RGB model. offset: pixel size that is to be removed from output. for example: (scale_factor=1, offset=7, input=100x100) => output=(100-7)x(100-7) (scale_factor=2, offset=12, input=100x100) => output=(100*2-12)x(100*2-12) And each layer has `class_name` attribute.
2016-05-15 06:04:08 +12:00
32
1
1
4
14
3
V 1
17
torch.CudaStorage
32
-0.464619279 0.00201993715 0.000628183421 -0.0447127037 0.00761218136 0.00716312602 -0.16316776 0.026375398 0.089718923 0.0235591289 -0.0583189465 0.00781400129 -0.00234992872 -0.0250623245 0.00974455103 -0.0568613298 0.00461343024 3.57015379e-05 0.0308940504 -0.0055342936 -0.00583036756 -0.147389829 0.00526255602 -0.0788191259 -0.149562985 0.00441113766 -0.0376145728 0.0201702155 0.00117593398 0.000996135408 -0.185849369 -0.179666892
Add 2x model for photo scaling
2015-11-30 21:29:34 +13:00
2
Convert model files; Add new pretrained model - Add new pretrained model to ./models/upconv_7 - Move old models to ./models/vgg_7 - Use nn.LeakyReLU instead of w2nn.LeakyReLU - Add useful attribute to .json New JSON attribute: The first layer has `model_config` attribute. It contains: model_arch: architecture name of model. see `lib/srcnn.lua` scale_factor: if scale_factor > 1, model:forward() changes image resolution with scale_factor. channels: input/output channels. if channels == 3, model is RGB model. offset: pixel size that is to be removed from output. for example: (scale_factor=1, offset=7, input=100x100) => output=(100-7)x(100-7) (scale_factor=2, offset=12, input=100x100) => output=(100*2-12)x(100*2-12) And each layer has `class_name` attribute.
2016-05-15 06:04:08 +12:00
8
gradBias
Add 2x model for photo scaling
2015-11-30 21:29:34 +13:00
4
Convert model files; Add new pretrained model - Add new pretrained model to ./models/upconv_7 - Move old models to ./models/vgg_7 - Use nn.LeakyReLU instead of w2nn.LeakyReLU - Add useful attribute to .json New JSON attribute: The first layer has `model_config` attribute. It contains: model_arch: architecture name of model. see `lib/srcnn.lua` scale_factor: if scale_factor > 1, model:forward() changes image resolution with scale_factor. channels: input/output channels. if channels == 3, model is RGB model. offset: pixel size that is to be removed from output. for example: (scale_factor=1, offset=7, input=100x100) => output=(100-7)x(100-7) (scale_factor=2, offset=12, input=100x100) => output=(100*2-12)x(100*2-12) And each layer has `class_name` attribute.
2016-05-15 06:04:08 +12:00
15
3
V 1
16
torch.CudaTensor
Add 2x model for photo scaling
2015-11-30 21:29:34 +13:00
1
Convert model files; Add new pretrained model - Add new pretrained model to ./models/upconv_7 - Move old models to ./models/vgg_7 - Use nn.LeakyReLU instead of w2nn.LeakyReLU - Add useful attribute to .json New JSON attribute: The first layer has `model_config` attribute. It contains: model_arch: architecture name of model. see `lib/srcnn.lua` scale_factor: if scale_factor > 1, model:forward() changes image resolution with scale_factor. channels: input/output channels. if channels == 3, model is RGB model. offset: pixel size that is to be removed from output. for example: (scale_factor=1, offset=7, input=100x100) => output=(100-7)x(100-7) (scale_factor=2, offset=12, input=100x100) => output=(100*2-12)x(100*2-12) And each layer has `class_name` attribute.
2016-05-15 06:04:08 +12:00
32
Add 2x model for photo scaling
2015-11-30 21:29:34 +13:00
1
1
Convert model files; Add new pretrained model - Add new pretrained model to ./models/upconv_7 - Move old models to ./models/vgg_7 - Use nn.LeakyReLU instead of w2nn.LeakyReLU - Add useful attribute to .json New JSON attribute: The first layer has `model_config` attribute. It contains: model_arch: architecture name of model. see `lib/srcnn.lua` scale_factor: if scale_factor > 1, model:forward() changes image resolution with scale_factor. channels: input/output channels. if channels == 3, model is RGB model. offset: pixel size that is to be removed from output. for example: (scale_factor=1, offset=7, input=100x100) => output=(100-7)x(100-7) (scale_factor=2, offset=12, input=100x100) => output=(100*2-12)x(100*2-12) And each layer has `class_name` attribute.
2016-05-15 06:04:08 +12:00
4
16
3
V 1
17
torch.CudaStorage
32
Add upconv_7/photo models. Update vgg_7/photo/noise* models
2016-07-04 07:51:22 +12:00
0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0
Add 2x model for photo scaling
2015-11-30 21:29:34 +13:00
2
Convert model files; Add new pretrained model - Add new pretrained model to ./models/upconv_7 - Move old models to ./models/vgg_7 - Use nn.LeakyReLU instead of w2nn.LeakyReLU - Add useful attribute to .json New JSON attribute: The first layer has `model_config` attribute. It contains: model_arch: architecture name of model. see `lib/srcnn.lua` scale_factor: if scale_factor > 1, model:forward() changes image resolution with scale_factor. channels: input/output channels. if channels == 3, model is RGB model. offset: pixel size that is to be removed from output. for example: (scale_factor=1, offset=7, input=100x100) => output=(100-7)x(100-7) (scale_factor=2, offset=12, input=100x100) => output=(100*2-12)x(100*2-12) And each layer has `class_name` attribute.
2016-05-15 06:04:08 +12:00
4
padH
Add 2x model for photo scaling
2015-11-30 21:29:34 +13:00
1
Convert model files; Add new pretrained model - Add new pretrained model to ./models/upconv_7 - Move old models to ./models/vgg_7 - Use nn.LeakyReLU instead of w2nn.LeakyReLU - Add useful attribute to .json New JSON attribute: The first layer has `model_config` attribute. It contains: model_arch: architecture name of model. see `lib/srcnn.lua` scale_factor: if scale_factor > 1, model:forward() changes image resolution with scale_factor. channels: input/output channels. if channels == 3, model is RGB model. offset: pixel size that is to be removed from output. for example: (scale_factor=1, offset=7, input=100x100) => output=(100-7)x(100-7) (scale_factor=2, offset=12, input=100x100) => output=(100*2-12)x(100*2-12) And each layer has `class_name` attribute.
2016-05-15 06:04:08 +12:00
0
Add 2x model for photo scaling
2015-11-30 21:29:34 +13:00
1
2
4
Convert model files; Add new pretrained model - Add new pretrained model to ./models/upconv_7 - Move old models to ./models/vgg_7 - Use nn.LeakyReLU instead of w2nn.LeakyReLU - Add useful attribute to .json New JSON attribute: The first layer has `model_config` attribute. It contains: model_arch: architecture name of model. see `lib/srcnn.lua` scale_factor: if scale_factor > 1, model:forward() changes image resolution with scale_factor. channels: input/output channels. if channels == 3, model is RGB model. offset: pixel size that is to be removed from output. for example: (scale_factor=1, offset=7, input=100x100) => output=(100-7)x(100-7) (scale_factor=2, offset=12, input=100x100) => output=(100*2-12)x(100*2-12) And each layer has `class_name` attribute.
2016-05-15 06:04:08 +12:00
17
Add 2x model for photo scaling
2015-11-30 21:29:34 +13:00
3
V 1
Convert model files; Add new pretrained model - Add new pretrained model to ./models/upconv_7 - Move old models to ./models/vgg_7 - Use nn.LeakyReLU instead of w2nn.LeakyReLU - Add useful attribute to .json New JSON attribute: The first layer has `model_config` attribute. It contains: model_arch: architecture name of model. see `lib/srcnn.lua` scale_factor: if scale_factor > 1, model:forward() changes image resolution with scale_factor. channels: input/output channels. if channels == 3, model is RGB model. offset: pixel size that is to be removed from output. for example: (scale_factor=1, offset=7, input=100x100) => output=(100-7)x(100-7) (scale_factor=2, offset=12, input=100x100) => output=(100*2-12)x(100*2-12) And each layer has `class_name` attribute.
2016-05-15 06:04:08 +12:00
12
nn.LeakyReLU
Add 2x model for photo scaling
2015-11-30 21:29:34 +13:00
3
Convert model files; Add new pretrained model - Add new pretrained model to ./models/upconv_7 - Move old models to ./models/vgg_7 - Use nn.LeakyReLU instead of w2nn.LeakyReLU - Add useful attribute to .json New JSON attribute: The first layer has `model_config` attribute. It contains: model_arch: architecture name of model. see `lib/srcnn.lua` scale_factor: if scale_factor > 1, model:forward() changes image resolution with scale_factor. channels: input/output channels. if channels == 3, model is RGB model. offset: pixel size that is to be removed from output. for example: (scale_factor=1, offset=7, input=100x100) => output=(100-7)x(100-7) (scale_factor=2, offset=12, input=100x100) => output=(100*2-12)x(100*2-12) And each layer has `class_name` attribute.
2016-05-15 06:04:08 +12:00
18
6
Add 2x model for photo scaling
2015-11-30 21:29:34 +13:00
2
Convert model files; Add new pretrained model - Add new pretrained model to ./models/upconv_7 - Move old models to ./models/vgg_7 - Use nn.LeakyReLU instead of w2nn.LeakyReLU - Add useful attribute to .json New JSON attribute: The first layer has `model_config` attribute. It contains: model_arch: architecture name of model. see `lib/srcnn.lua` scale_factor: if scale_factor > 1, model:forward() changes image resolution with scale_factor. channels: input/output channels. if channels == 3, model is RGB model. offset: pixel size that is to be removed from output. for example: (scale_factor=1, offset=7, input=100x100) => output=(100-7)x(100-7) (scale_factor=2, offset=12, input=100x100) => output=(100*2-12)x(100*2-12) And each layer has `class_name` attribute.
2016-05-15 06:04:08 +12:00
7
inplace
Add 2x model for photo scaling
2015-11-30 21:29:34 +13:00
5
Convert model files; Add new pretrained model - Add new pretrained model to ./models/upconv_7 - Move old models to ./models/vgg_7 - Use nn.LeakyReLU instead of w2nn.LeakyReLU - Add useful attribute to .json New JSON attribute: The first layer has `model_config` attribute. It contains: model_arch: architecture name of model. see `lib/srcnn.lua` scale_factor: if scale_factor > 1, model:forward() changes image resolution with scale_factor. channels: input/output channels. if channels == 3, model is RGB model. offset: pixel size that is to be removed from output. for example: (scale_factor=1, offset=7, input=100x100) => output=(100-7)x(100-7) (scale_factor=2, offset=12, input=100x100) => output=(100*2-12)x(100*2-12) And each layer has `class_name` attribute.
2016-05-15 06:04:08 +12:00
1
2
Add 2x model for photo scaling
2015-11-30 21:29:34 +13:00
5
Convert model files; Add new pretrained model - Add new pretrained model to ./models/upconv_7 - Move old models to ./models/vgg_7 - Use nn.LeakyReLU instead of w2nn.LeakyReLU - Add useful attribute to .json New JSON attribute: The first layer has `model_config` attribute. It contains: model_arch: architecture name of model. see `lib/srcnn.lua` scale_factor: if scale_factor > 1, model:forward() changes image resolution with scale_factor. channels: input/output channels. if channels == 3, model is RGB model. offset: pixel size that is to be removed from output. for example: (scale_factor=1, offset=7, input=100x100) => output=(100-7)x(100-7) (scale_factor=2, offset=12, input=100x100) => output=(100*2-12)x(100*2-12) And each layer has `class_name` attribute.
2016-05-15 06:04:08 +12:00
_type
Add 2x model for photo scaling
2015-11-30 21:29:34 +13:00
2
16
torch.CudaTensor
Convert model files; Add new pretrained model - Add new pretrained model to ./models/upconv_7 - Move old models to ./models/vgg_7 - Use nn.LeakyReLU instead of w2nn.LeakyReLU - Add useful attribute to .json New JSON attribute: The first layer has `model_config` attribute. It contains: model_arch: architecture name of model. see `lib/srcnn.lua` scale_factor: if scale_factor > 1, model:forward() changes image resolution with scale_factor. channels: input/output channels. if channels == 3, model is RGB model. offset: pixel size that is to be removed from output. for example: (scale_factor=1, offset=7, input=100x100) => output=(100-7)x(100-7) (scale_factor=2, offset=12, input=100x100) => output=(100*2-12)x(100*2-12) And each layer has `class_name` attribute.
2016-05-15 06:04:08 +12:00
2
6
negval
Add 2x model for photo scaling
2015-11-30 21:29:34 +13:00
1
Convert model files; Add new pretrained model - Add new pretrained model to ./models/upconv_7 - Move old models to ./models/vgg_7 - Use nn.LeakyReLU instead of w2nn.LeakyReLU - Add useful attribute to .json New JSON attribute: The first layer has `model_config` attribute. It contains: model_arch: architecture name of model. see `lib/srcnn.lua` scale_factor: if scale_factor > 1, model:forward() changes image resolution with scale_factor. channels: input/output channels. if channels == 3, model is RGB model. offset: pixel size that is to be removed from output. for example: (scale_factor=1, offset=7, input=100x100) => output=(100-7)x(100-7) (scale_factor=2, offset=12, input=100x100) => output=(100*2-12)x(100*2-12) And each layer has `class_name` attribute.
2016-05-15 06:04:08 +12:00
0.10000000000000001
Add 2x model for photo scaling
2015-11-30 21:29:34 +13:00
2
9
gradInput
4
Convert model files; Add new pretrained model - Add new pretrained model to ./models/upconv_7 - Move old models to ./models/vgg_7 - Use nn.LeakyReLU instead of w2nn.LeakyReLU - Add useful attribute to .json New JSON attribute: The first layer has `model_config` attribute. It contains: model_arch: architecture name of model. see `lib/srcnn.lua` scale_factor: if scale_factor > 1, model:forward() changes image resolution with scale_factor. channels: input/output channels. if channels == 3, model is RGB model. offset: pixel size that is to be removed from output. for example: (scale_factor=1, offset=7, input=100x100) => output=(100-7)x(100-7) (scale_factor=2, offset=12, input=100x100) => output=(100*2-12)x(100*2-12) And each layer has `class_name` attribute.
2016-05-15 06:04:08 +12:00
19
Add 2x model for photo scaling
2015-11-30 21:29:34 +13:00
3
V 1
16
torch.CudaTensor
0
1
0
2
6
output
4
Convert model files; Add new pretrained model - Add new pretrained model to ./models/upconv_7 - Move old models to ./models/vgg_7 - Use nn.LeakyReLU instead of w2nn.LeakyReLU - Add useful attribute to .json New JSON attribute: The first layer has `model_config` attribute. It contains: model_arch: architecture name of model. see `lib/srcnn.lua` scale_factor: if scale_factor > 1, model:forward() changes image resolution with scale_factor. channels: input/output channels. if channels == 3, model is RGB model. offset: pixel size that is to be removed from output. for example: (scale_factor=1, offset=7, input=100x100) => output=(100-7)x(100-7) (scale_factor=2, offset=12, input=100x100) => output=(100*2-12)x(100*2-12) And each layer has `class_name` attribute.
2016-05-15 06:04:08 +12:00
20
Add 2x model for photo scaling
2015-11-30 21:29:34 +13:00
3
V 1
16
torch.CudaTensor
0
1
0
2
Convert model files; Add new pretrained model - Add new pretrained model to ./models/upconv_7 - Move old models to ./models/vgg_7 - Use nn.LeakyReLU instead of w2nn.LeakyReLU - Add useful attribute to .json New JSON attribute: The first layer has `model_config` attribute. It contains: model_arch: architecture name of model. see `lib/srcnn.lua` scale_factor: if scale_factor > 1, model:forward() changes image resolution with scale_factor. channels: input/output channels. if channels == 3, model is RGB model. offset: pixel size that is to be removed from output. for example: (scale_factor=1, offset=7, input=100x100) => output=(100-7)x(100-7) (scale_factor=2, offset=12, input=100x100) => output=(100*2-12)x(100*2-12) And each layer has `class_name` attribute.
2016-05-15 06:04:08 +12:00
5
train
5
0
Add 2x model for photo scaling
2015-11-30 21:29:34 +13:00
1
3
4
Convert model files; Add new pretrained model - Add new pretrained model to ./models/upconv_7 - Move old models to ./models/vgg_7 - Use nn.LeakyReLU instead of w2nn.LeakyReLU - Add useful attribute to .json New JSON attribute: The first layer has `model_config` attribute. It contains: model_arch: architecture name of model. see `lib/srcnn.lua` scale_factor: if scale_factor > 1, model:forward() changes image resolution with scale_factor. channels: input/output channels. if channels == 3, model is RGB model. offset: pixel size that is to be removed from output. for example: (scale_factor=1, offset=7, input=100x100) => output=(100-7)x(100-7) (scale_factor=2, offset=12, input=100x100) => output=(100*2-12)x(100*2-12) And each layer has `class_name` attribute.
2016-05-15 06:04:08 +12:00
21
Add 2x model for photo scaling
2015-11-30 21:29:34 +13:00
3
V 1
23
nn.SpatialConvolutionMM
3
Convert model files; Add new pretrained model - Add new pretrained model to ./models/upconv_7 - Move old models to ./models/vgg_7 - Use nn.LeakyReLU instead of w2nn.LeakyReLU - Add useful attribute to .json New JSON attribute: The first layer has `model_config` attribute. It contains: model_arch: architecture name of model. see `lib/srcnn.lua` scale_factor: if scale_factor > 1, model:forward() changes image resolution with scale_factor. channels: input/output channels. if channels == 3, model is RGB model. offset: pixel size that is to be removed from output. for example: (scale_factor=1, offset=7, input=100x100) => output=(100-7)x(100-7) (scale_factor=2, offset=12, input=100x100) => output=(100*2-12)x(100*2-12) And each layer has `class_name` attribute.
2016-05-15 06:04:08 +12:00
22
16
Add 2x model for photo scaling
2015-11-30 21:29:34 +13:00
2
Convert model files; Add new pretrained model - Add new pretrained model to ./models/upconv_7 - Move old models to ./models/vgg_7 - Use nn.LeakyReLU instead of w2nn.LeakyReLU - Add useful attribute to .json New JSON attribute: The first layer has `model_config` attribute. It contains: model_arch: architecture name of model. see `lib/srcnn.lua` scale_factor: if scale_factor > 1, model:forward() changes image resolution with scale_factor. channels: input/output channels. if channels == 3, model is RGB model. offset: pixel size that is to be removed from output. for example: (scale_factor=1, offset=7, input=100x100) => output=(100-7)x(100-7) (scale_factor=2, offset=12, input=100x100) => output=(100*2-12)x(100*2-12) And each layer has `class_name` attribute.
2016-05-15 06:04:08 +12:00
2
dH
1
1
2
2
dW
1
Add 2x model for photo scaling
2015-11-30 21:29:34 +13:00
1
2
11
nInputPlane
1
32
2
6
output
4
Convert model files; Add new pretrained model - Add new pretrained model to ./models/upconv_7 - Move old models to ./models/vgg_7 - Use nn.LeakyReLU instead of w2nn.LeakyReLU - Add useful attribute to .json New JSON attribute: The first layer has `model_config` attribute. It contains: model_arch: architecture name of model. see `lib/srcnn.lua` scale_factor: if scale_factor > 1, model:forward() changes image resolution with scale_factor. channels: input/output channels. if channels == 3, model is RGB model. offset: pixel size that is to be removed from output. for example: (scale_factor=1, offset=7, input=100x100) => output=(100-7)x(100-7) (scale_factor=2, offset=12, input=100x100) => output=(100*2-12)x(100*2-12) And each layer has `class_name` attribute.
2016-05-15 06:04:08 +12:00
23
Add 2x model for photo scaling
2015-11-30 21:29:34 +13:00
3
V 1
16
torch.CudaTensor
0
1
0
2
Convert model files; Add new pretrained model - Add new pretrained model to ./models/upconv_7 - Move old models to ./models/vgg_7 - Use nn.LeakyReLU instead of w2nn.LeakyReLU - Add useful attribute to .json New JSON attribute: The first layer has `model_config` attribute. It contains: model_arch: architecture name of model. see `lib/srcnn.lua` scale_factor: if scale_factor > 1, model:forward() changes image resolution with scale_factor. channels: input/output channels. if channels == 3, model is RGB model. offset: pixel size that is to be removed from output. for example: (scale_factor=1, offset=7, input=100x100) => output=(100-7)x(100-7) (scale_factor=2, offset=12, input=100x100) => output=(100*2-12)x(100*2-12) And each layer has `class_name` attribute.
2016-05-15 06:04:08 +12:00
2
kH
1
3
2
Add 2x model for photo scaling
2015-11-30 21:29:34 +13:00
9
gradInput
4
Convert model files; Add new pretrained model - Add new pretrained model to ./models/upconv_7 - Move old models to ./models/vgg_7 - Use nn.LeakyReLU instead of w2nn.LeakyReLU - Add useful attribute to .json New JSON attribute: The first layer has `model_config` attribute. It contains: model_arch: architecture name of model. see `lib/srcnn.lua` scale_factor: if scale_factor > 1, model:forward() changes image resolution with scale_factor. channels: input/output channels. if channels == 3, model is RGB model. offset: pixel size that is to be removed from output. for example: (scale_factor=1, offset=7, input=100x100) => output=(100-7)x(100-7) (scale_factor=2, offset=12, input=100x100) => output=(100*2-12)x(100*2-12) And each layer has `class_name` attribute.
2016-05-15 06:04:08 +12:00
24
Add 2x model for photo scaling
2015-11-30 21:29:34 +13:00
3
V 1
16
torch.CudaTensor
0
1
0
2
2
Convert model files; Add new pretrained model - Add new pretrained model to ./models/upconv_7 - Move old models to ./models/vgg_7 - Use nn.LeakyReLU instead of w2nn.LeakyReLU - Add useful attribute to .json New JSON attribute: The first layer has `model_config` attribute. It contains: model_arch: architecture name of model. see `lib/srcnn.lua` scale_factor: if scale_factor > 1, model:forward() changes image resolution with scale_factor. channels: input/output channels. if channels == 3, model is RGB model. offset: pixel size that is to be removed from output. for example: (scale_factor=1, offset=7, input=100x100) => output=(100-7)x(100-7) (scale_factor=2, offset=12, input=100x100) => output=(100*2-12)x(100*2-12) And each layer has `class_name` attribute.
2016-05-15 06:04:08 +12:00
kW
Add 2x model for photo scaling
2015-11-30 21:29:34 +13:00
1
Convert model files; Add new pretrained model - Add new pretrained model to ./models/upconv_7 - Move old models to ./models/vgg_7 - Use nn.LeakyReLU instead of w2nn.LeakyReLU - Add useful attribute to .json New JSON attribute: The first layer has `model_config` attribute. It contains: model_arch: architecture name of model. see `lib/srcnn.lua` scale_factor: if scale_factor > 1, model:forward() changes image resolution with scale_factor. channels: input/output channels. if channels == 3, model is RGB model. offset: pixel size that is to be removed from output. for example: (scale_factor=1, offset=7, input=100x100) => output=(100-7)x(100-7) (scale_factor=2, offset=12, input=100x100) => output=(100*2-12)x(100*2-12) And each layer has `class_name` attribute.
2016-05-15 06:04:08 +12:00
3
Add 2x model for photo scaling
2015-11-30 21:29:34 +13:00
2
12
nOutputPlane
1
32
2
4
Convert model files; Add new pretrained model - Add new pretrained model to ./models/upconv_7 - Move old models to ./models/vgg_7 - Use nn.LeakyReLU instead of w2nn.LeakyReLU - Add useful attribute to .json New JSON attribute: The first layer has `model_config` attribute. It contains: model_arch: architecture name of model. see `lib/srcnn.lua` scale_factor: if scale_factor > 1, model:forward() changes image resolution with scale_factor. channels: input/output channels. if channels == 3, model is RGB model. offset: pixel size that is to be removed from output. for example: (scale_factor=1, offset=7, input=100x100) => output=(100-7)x(100-7) (scale_factor=2, offset=12, input=100x100) => output=(100*2-12)x(100*2-12) And each layer has `class_name` attribute.
2016-05-15 06:04:08 +12:00
padW
Add 2x model for photo scaling
2015-11-30 21:29:34 +13:00
1
0
2
6
weight
4
Convert model files; Add new pretrained model - Add new pretrained model to ./models/upconv_7 - Move old models to ./models/vgg_7 - Use nn.LeakyReLU instead of w2nn.LeakyReLU - Add useful attribute to .json New JSON attribute: The first layer has `model_config` attribute. It contains: model_arch: architecture name of model. see `lib/srcnn.lua` scale_factor: if scale_factor > 1, model:forward() changes image resolution with scale_factor. channels: input/output channels. if channels == 3, model is RGB model. offset: pixel size that is to be removed from output. for example: (scale_factor=1, offset=7, input=100x100) => output=(100-7)x(100-7) (scale_factor=2, offset=12, input=100x100) => output=(100*2-12)x(100*2-12) And each layer has `class_name` attribute.
2016-05-15 06:04:08 +12:00
25
Add 2x model for photo scaling
2015-11-30 21:29:34 +13:00
3
V 1
16
torch.CudaTensor
2
32 288
288 1
1
4
Convert model files; Add new pretrained model - Add new pretrained model to ./models/upconv_7 - Move old models to ./models/vgg_7 - Use nn.LeakyReLU instead of w2nn.LeakyReLU - Add useful attribute to .json New JSON attribute: The first layer has `model_config` attribute. It contains: model_arch: architecture name of model. see `lib/srcnn.lua` scale_factor: if scale_factor > 1, model:forward() changes image resolution with scale_factor. channels: input/output channels. if channels == 3, model is RGB model. offset: pixel size that is to be removed from output. for example: (scale_factor=1, offset=7, input=100x100) => output=(100-7)x(100-7) (scale_factor=2, offset=12, input=100x100) => output=(100*2-12)x(100*2-12) And each layer has `class_name` attribute.
2016-05-15 06:04:08 +12:00
26
Add 2x model for photo scaling
2015-11-30 21:29:34 +13:00
3
V 1
17
torch.CudaStorage
9216
-0.0209273864 0.100784712 -0.0376805924 0.0296244584 -0.0388070717 0.0535670556 -0.0562614501 -0.0726919249 -0.068171382 0.052393645 -0.0472772233 0.00922184717 0.0246529412 -0.0787168071 0.0803164095 0.0368224829 0.00537091773 -0.0072160312 0.0505270325 0.0222731847 0.019749511 0.0677835718 0.341889471 0.210179389 0.22403346 0.318408221 0.214213654 -0.00284464681 -0.0406171009 0.080896318 0.109615356 0.0142217558 0.193845883 -0.149881944 -0.273067832 -0.0235034451 -0.0923726782 0.119294524 0.0790113136 -0.0282414556 -0.0191942099 0.0980937183 -0.016728403 -0.0291197784 -0.0135406181 0.0789235011 0.0692683458 0.091105856 0.156770244 0.12325979 0.224705994 0.119213149 0.221781269 -0.0966617763 -0.122169174 -0.163423315 0.0275717694 -0.00640750909 0.0162411928 0.0440846011 0.0254129041 0.0100754807 0.0120621612 0.000655566808 0.0211740006 0.11964751 -0.0206685401 -0.0740386546 0.069221966 0.0388211235 0.0553292222 -0.0168767553 -0.00335001433 0.131833702 0.104615621 0.0225449651 0.0273399819 0.283489197 -0.0812887624 0.0677725524 -0.16896309 -0.0255892593 -0.0494218431 -0.0158815999 0.0365488976 0.000659360841 0.0238984264 -0.00556499697 -0.0223928038 0.0835790932 0.0431613252 -0.0950226486 0.138132483 0.244161353 0.110620417 0.25031364 -0.0314353965 0.0984130502 0.0804312676 -0.0949323997 0.0327388868 -0.0332029723 -0.0799899548 0.10149128 0.164759457 -0.122655183 -0.104392476 -0.00296235713 0.0574206226 0.0971534476 0.0486663692 0.0819783956 -0.0235940479 0.100565806 0.0209038723 0.00932763889 0.111967891 0.108910911 -0.0179201774 0.0554009937 0.0865619481 -0.144485697 0.00417221757 -0.0660190508 -0.0558729209 -0.0346676297 -0.0161496662 0.0535534546 -0.0670778081 -0.0208141766 0.0296293814 0.0354784839 0.00999607146 0.054940518 -0.16630064 0.0491140559 0.0151749998 0.050601352 -0.0958439559 -0.119363688 0.0572280325 0.022067029 0.0182099175 0.030712653 -0.0775896907 0.00775478641 0.117012575 -0.0969435871 0.128972515 0.0109945545 0.00398015277 0.0188820306 0.0294697676 0.0778955817 0.177904785 -0.0637549907 -0.0196519848 0.127551168 0.0249881595 0.0548668616 -0.11531835 0.0469377488 -0.00430031214 -0.0810509771 -0.0681153536 0.0387130566 -0.00411570026 0.125483185 -0.136481375 0.104615912 0.12028531 -0.00609394163 -0.0364181735 0.0321536399 0.0832785368 0.00604579598 -0.027922336 -0.0566285439 0.0647696331 0.0202638675 0.0171959046 -0.0773412958 -0.101464324 0.0135008367 0.0577888303 -0.0648008734 0.049588982 -0.0275376122 0.00388561049 -0.0972890779 0.0358217768 0.167219669 0.137880042 -0.0214744937 0.0341431573 0.167158604 -0.163271338 -0.197542191 -0.0646010265 -0.0249606911 0.0476673022 -0.00606096955 0.0357332192 0.065678753 -0.0248843282 -0.0113018462 0.0631250963 -0.0534148701 -0.102016613 0.101836823 -0.0394178778 0.0795968473 0.0130256778 -0.121010453 0.135572731 0.0324839093 0.0390856043 0.139970377 0.052734185 0.00534330634 0.0388596281 0.132347599 -0.0231753215 0.0450428389 -0.164957359 0.0446061417 0.0800377578 -0.0799509659 0.0150066866 -0.0344706103 0.142222658 0.0150639862 -0.0432454459 0.223395348 -0.0122748418 0.00192308065 0.0203653201 -0.0498774387 0.000865237031 -0.00241121091 -0.0171664525 0.0392480306 0.0262295641 -0.0434569567 -0.00379386963 0.000510787882 -0.0397930518 0.0960251167 0.00593118742 0.0206693262 0.0345879681 -0.0620849729 -0.0158110186 0.0696216151 0.00341766467 0.0274607651 0.0325004309 -0.0387920514 -0.0505673289 0.00462431414 0.0184641201 -0.0320697054 -0.00062843092 -0.0305197667 -0.0267850328 0.17220442 0.112546906 0.0316710509 0.104615316 -0.141219199 0.0917976871 0.188978121 -0.024177039 -0.07635317 -0.0500751734 -0.0788859203 -0.134227931 0.0162007622 -0.0967114046 -0.0921706185 -0.0247860793 0.00498846034 -0.116762325 -0.00754553545 0.0656336322 -0.0238237884 0.0151935397 0.168568879 0.000514879648 -0.417712182 -0.0498028621 0.310529977 0.0748673379 -0.114912562 0.0425042138 0.336377144 -0.315983832 -0.0206585303 0.117428951 -0.0645330548 0.207359597 0.193399757 -0.135356158 -0.182384536 0.00970120355 -0.0441468507 -0.243005499 -0.152016342 -0.0626423806 0.0423752815 0.176
2
5
train
5
0
2
10
gradWeight
4
Convert model files; Add new pretrained model - Add new pretrained model to ./models/upconv_7 - Move old models to ./models/vgg_7 - Use nn.LeakyReLU instead of w2nn.LeakyReLU - Add useful attribute to .json New JSON attribute: The first layer has `model_config` attribute. It contains: model_arch: architecture name of model. see `lib/srcnn.lua` scale_factor: if scale_factor > 1, model:forward() changes image resolution with scale_factor. channels: input/output channels. if channels == 3, model is RGB model. offset: pixel size that is to be removed from output. for example: (scale_factor=1, offset=7, input=100x100) => output=(100-7)x(100-7) (scale_factor=2, offset=12, input=100x100) => output=(100*2-12)x(100*2-12) And each layer has `class_name` attribute.
2016-05-15 06:04:08 +12:00
27
Add 2x model for photo scaling
2015-11-30 21:29:34 +13:00
3
V 1
16
torch.CudaTensor
2
32 288
288 1
1
4
Convert model files; Add new pretrained model - Add new pretrained model to ./models/upconv_7 - Move old models to ./models/vgg_7 - Use nn.LeakyReLU instead of w2nn.LeakyReLU - Add useful attribute to .json New JSON attribute: The first layer has `model_config` attribute. It contains: model_arch: architecture name of model. see `lib/srcnn.lua` scale_factor: if scale_factor > 1, model:forward() changes image resolution with scale_factor. channels: input/output channels. if channels == 3, model is RGB model. offset: pixel size that is to be removed from output. for example: (scale_factor=1, offset=7, input=100x100) => output=(100-7)x(100-7) (scale_factor=2, offset=12, input=100x100) => output=(100*2-12)x(100*2-12) And each layer has `class_name` attribute.
2016-05-15 06:04:08 +12:00
28
Add 2x model for photo scaling
2015-11-30 21:29:34 +13:00
3
V 1
17
torch.CudaStorage
9216
Convert model files; Add new pretrained model - Add new pretrained model to ./models/upconv_7 - Move old models to ./models/vgg_7 - Use nn.LeakyReLU instead of w2nn.LeakyReLU - Add useful attribute to .json New JSON attribute: The first layer has `model_config` attribute. It contains: model_arch: architecture name of model. see `lib/srcnn.lua` scale_factor: if scale_factor > 1, model:forward() changes image resolution with scale_factor. channels: input/output channels. if channels == 3, model is RGB model. offset: pixel size that is to be removed from output. for example: (scale_factor=1, offset=7, input=100x100) => output=(100-7)x(100-7) (scale_factor=2, offset=12, input=100x100) => output=(100*2-12)x(100*2-12) And each layer has `class_name` attribute.
2016-05-15 06:04:08 +12:00
0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0
2
5
_type
2
16
torch.CudaTensor
Add 2x model for photo scaling
2015-11-30 21:29:34 +13:00
2
4
Convert model files; Add new pretrained model - Add new pretrained model to ./models/upconv_7 - Move old models to ./models/vgg_7 - Use nn.LeakyReLU instead of w2nn.LeakyReLU - Add useful attribute to .json New JSON attribute: The first layer has `model_config` attribute. It contains: model_arch: architecture name of model. see `lib/srcnn.lua` scale_factor: if scale_factor > 1, model:forward() changes image resolution with scale_factor. channels: input/output channels. if channels == 3, model is RGB model. offset: pixel size that is to be removed from output. for example: (scale_factor=1, offset=7, input=100x100) => output=(100-7)x(100-7) (scale_factor=2, offset=12, input=100x100) => output=(100*2-12)x(100*2-12) And each layer has `class_name` attribute.
2016-05-15 06:04:08 +12:00
bias
4
29
Add 2x model for photo scaling
2015-11-30 21:29:34 +13:00
3
V 1
16
torch.CudaTensor
1
Convert model files; Add new pretrained model - Add new pretrained model to ./models/upconv_7 - Move old models to ./models/vgg_7 - Use nn.LeakyReLU instead of w2nn.LeakyReLU - Add useful attribute to .json New JSON attribute: The first layer has `model_config` attribute. It contains: model_arch: architecture name of model. see `lib/srcnn.lua` scale_factor: if scale_factor > 1, model:forward() changes image resolution with scale_factor. channels: input/output channels. if channels == 3, model is RGB model. offset: pixel size that is to be removed from output. for example: (scale_factor=1, offset=7, input=100x100) => output=(100-7)x(100-7) (scale_factor=2, offset=12, input=100x100) => output=(100*2-12)x(100*2-12) And each layer has `class_name` attribute.
2016-05-15 06:04:08 +12:00
32
1
1
4
30
3
V 1
17
torch.CudaStorage
32
0.106073655 0.00352948415 0.00938062835 0.00164515455 -0.0872725397 0.0215106644 -0.0392723829 0.0113751888 0.000382447557 -0.00502240984 -0.0137665709 -0.199852228 -0.0187911727 0.0455863923 0.017438177 0.0262026098 0.0192497429 -0.00395555329 -0.0166453496 0.0394801013 0.0168405361 -0.00667148782 -0.00754785677 0.0089894468 -0.0228441805 -0.0118729901 0.03073439 -0.120279565 -0.032817848 -0.0145876091 -0.0387578644 0.0143209379
Add 2x model for photo scaling
2015-11-30 21:29:34 +13:00
2
Convert model files; Add new pretrained model - Add new pretrained model to ./models/upconv_7 - Move old models to ./models/vgg_7 - Use nn.LeakyReLU instead of w2nn.LeakyReLU - Add useful attribute to .json New JSON attribute: The first layer has `model_config` attribute. It contains: model_arch: architecture name of model. see `lib/srcnn.lua` scale_factor: if scale_factor > 1, model:forward() changes image resolution with scale_factor. channels: input/output channels. if channels == 3, model is RGB model. offset: pixel size that is to be removed from output. for example: (scale_factor=1, offset=7, input=100x100) => output=(100-7)x(100-7) (scale_factor=2, offset=12, input=100x100) => output=(100*2-12)x(100*2-12) And each layer has `class_name` attribute.
2016-05-15 06:04:08 +12:00
8
gradBias
Add 2x model for photo scaling
2015-11-30 21:29:34 +13:00
4
Convert model files; Add new pretrained model - Add new pretrained model to ./models/upconv_7 - Move old models to ./models/vgg_7 - Use nn.LeakyReLU instead of w2nn.LeakyReLU - Add useful attribute to .json New JSON attribute: The first layer has `model_config` attribute. It contains: model_arch: architecture name of model. see `lib/srcnn.lua` scale_factor: if scale_factor > 1, model:forward() changes image resolution with scale_factor. channels: input/output channels. if channels == 3, model is RGB model. offset: pixel size that is to be removed from output. for example: (scale_factor=1, offset=7, input=100x100) => output=(100-7)x(100-7) (scale_factor=2, offset=12, input=100x100) => output=(100*2-12)x(100*2-12) And each layer has `class_name` attribute.
2016-05-15 06:04:08 +12:00
31
3
V 1
16
torch.CudaTensor
Add 2x model for photo scaling
2015-11-30 21:29:34 +13:00
1
Convert model files; Add new pretrained model - Add new pretrained model to ./models/upconv_7 - Move old models to ./models/vgg_7 - Use nn.LeakyReLU instead of w2nn.LeakyReLU - Add useful attribute to .json New JSON attribute: The first layer has `model_config` attribute. It contains: model_arch: architecture name of model. see `lib/srcnn.lua` scale_factor: if scale_factor > 1, model:forward() changes image resolution with scale_factor. channels: input/output channels. if channels == 3, model is RGB model. offset: pixel size that is to be removed from output. for example: (scale_factor=1, offset=7, input=100x100) => output=(100-7)x(100-7) (scale_factor=2, offset=12, input=100x100) => output=(100*2-12)x(100*2-12) And each layer has `class_name` attribute.
2016-05-15 06:04:08 +12:00
32
Add 2x model for photo scaling
2015-11-30 21:29:34 +13:00
1
1
Convert model files; Add new pretrained model - Add new pretrained model to ./models/upconv_7 - Move old models to ./models/vgg_7 - Use nn.LeakyReLU instead of w2nn.LeakyReLU - Add useful attribute to .json New JSON attribute: The first layer has `model_config` attribute. It contains: model_arch: architecture name of model. see `lib/srcnn.lua` scale_factor: if scale_factor > 1, model:forward() changes image resolution with scale_factor. channels: input/output channels. if channels == 3, model is RGB model. offset: pixel size that is to be removed from output. for example: (scale_factor=1, offset=7, input=100x100) => output=(100-7)x(100-7) (scale_factor=2, offset=12, input=100x100) => output=(100*2-12)x(100*2-12) And each layer has `class_name` attribute.
2016-05-15 06:04:08 +12:00
4
32
3
V 1
17
torch.CudaStorage
32
Add upconv_7/photo models. Update vgg_7/photo/noise* models
2016-07-04 07:51:22 +12:00
0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0
Add 2x model for photo scaling
2015-11-30 21:29:34 +13:00
2
Convert model files; Add new pretrained model - Add new pretrained model to ./models/upconv_7 - Move old models to ./models/vgg_7 - Use nn.LeakyReLU instead of w2nn.LeakyReLU - Add useful attribute to .json New JSON attribute: The first layer has `model_config` attribute. It contains: model_arch: architecture name of model. see `lib/srcnn.lua` scale_factor: if scale_factor > 1, model:forward() changes image resolution with scale_factor. channels: input/output channels. if channels == 3, model is RGB model. offset: pixel size that is to be removed from output. for example: (scale_factor=1, offset=7, input=100x100) => output=(100-7)x(100-7) (scale_factor=2, offset=12, input=100x100) => output=(100*2-12)x(100*2-12) And each layer has `class_name` attribute.
2016-05-15 06:04:08 +12:00
4
padH
Add 2x model for photo scaling
2015-11-30 21:29:34 +13:00
1
Convert model files; Add new pretrained model - Add new pretrained model to ./models/upconv_7 - Move old models to ./models/vgg_7 - Use nn.LeakyReLU instead of w2nn.LeakyReLU - Add useful attribute to .json New JSON attribute: The first layer has `model_config` attribute. It contains: model_arch: architecture name of model. see `lib/srcnn.lua` scale_factor: if scale_factor > 1, model:forward() changes image resolution with scale_factor. channels: input/output channels. if channels == 3, model is RGB model. offset: pixel size that is to be removed from output. for example: (scale_factor=1, offset=7, input=100x100) => output=(100-7)x(100-7) (scale_factor=2, offset=12, input=100x100) => output=(100*2-12)x(100*2-12) And each layer has `class_name` attribute.
2016-05-15 06:04:08 +12:00
0
Add 2x model for photo scaling
2015-11-30 21:29:34 +13:00
1
4
4
Convert model files; Add new pretrained model - Add new pretrained model to ./models/upconv_7 - Move old models to ./models/vgg_7 - Use nn.LeakyReLU instead of w2nn.LeakyReLU - Add useful attribute to .json New JSON attribute: The first layer has `model_config` attribute. It contains: model_arch: architecture name of model. see `lib/srcnn.lua` scale_factor: if scale_factor > 1, model:forward() changes image resolution with scale_factor. channels: input/output channels. if channels == 3, model is RGB model. offset: pixel size that is to be removed from output. for example: (scale_factor=1, offset=7, input=100x100) => output=(100-7)x(100-7) (scale_factor=2, offset=12, input=100x100) => output=(100*2-12)x(100*2-12) And each layer has `class_name` attribute.
2016-05-15 06:04:08 +12:00
33
Add 2x model for photo scaling
2015-11-30 21:29:34 +13:00
3
V 1
Convert model files; Add new pretrained model - Add new pretrained model to ./models/upconv_7 - Move old models to ./models/vgg_7 - Use nn.LeakyReLU instead of w2nn.LeakyReLU - Add useful attribute to .json New JSON attribute: The first layer has `model_config` attribute. It contains: model_arch: architecture name of model. see `lib/srcnn.lua` scale_factor: if scale_factor > 1, model:forward() changes image resolution with scale_factor. channels: input/output channels. if channels == 3, model is RGB model. offset: pixel size that is to be removed from output. for example: (scale_factor=1, offset=7, input=100x100) => output=(100-7)x(100-7) (scale_factor=2, offset=12, input=100x100) => output=(100*2-12)x(100*2-12) And each layer has `class_name` attribute.
2016-05-15 06:04:08 +12:00
12
nn.LeakyReLU
Add 2x model for photo scaling
2015-11-30 21:29:34 +13:00
3
Convert model files; Add new pretrained model - Add new pretrained model to ./models/upconv_7 - Move old models to ./models/vgg_7 - Use nn.LeakyReLU instead of w2nn.LeakyReLU - Add useful attribute to .json New JSON attribute: The first layer has `model_config` attribute. It contains: model_arch: architecture name of model. see `lib/srcnn.lua` scale_factor: if scale_factor > 1, model:forward() changes image resolution with scale_factor. channels: input/output channels. if channels == 3, model is RGB model. offset: pixel size that is to be removed from output. for example: (scale_factor=1, offset=7, input=100x100) => output=(100-7)x(100-7) (scale_factor=2, offset=12, input=100x100) => output=(100*2-12)x(100*2-12) And each layer has `class_name` attribute.
2016-05-15 06:04:08 +12:00
34
6
Add 2x model for photo scaling
2015-11-30 21:29:34 +13:00
2
Convert model files; Add new pretrained model - Add new pretrained model to ./models/upconv_7 - Move old models to ./models/vgg_7 - Use nn.LeakyReLU instead of w2nn.LeakyReLU - Add useful attribute to .json New JSON attribute: The first layer has `model_config` attribute. It contains: model_arch: architecture name of model. see `lib/srcnn.lua` scale_factor: if scale_factor > 1, model:forward() changes image resolution with scale_factor. channels: input/output channels. if channels == 3, model is RGB model. offset: pixel size that is to be removed from output. for example: (scale_factor=1, offset=7, input=100x100) => output=(100-7)x(100-7) (scale_factor=2, offset=12, input=100x100) => output=(100*2-12)x(100*2-12) And each layer has `class_name` attribute.
2016-05-15 06:04:08 +12:00
7
inplace
Add 2x model for photo scaling
2015-11-30 21:29:34 +13:00
5
Convert model files; Add new pretrained model - Add new pretrained model to ./models/upconv_7 - Move old models to ./models/vgg_7 - Use nn.LeakyReLU instead of w2nn.LeakyReLU - Add useful attribute to .json New JSON attribute: The first layer has `model_config` attribute. It contains: model_arch: architecture name of model. see `lib/srcnn.lua` scale_factor: if scale_factor > 1, model:forward() changes image resolution with scale_factor. channels: input/output channels. if channels == 3, model is RGB model. offset: pixel size that is to be removed from output. for example: (scale_factor=1, offset=7, input=100x100) => output=(100-7)x(100-7) (scale_factor=2, offset=12, input=100x100) => output=(100*2-12)x(100*2-12) And each layer has `class_name` attribute.
2016-05-15 06:04:08 +12:00
1
2
Add 2x model for photo scaling
2015-11-30 21:29:34 +13:00
5
Convert model files; Add new pretrained model - Add new pretrained model to ./models/upconv_7 - Move old models to ./models/vgg_7 - Use nn.LeakyReLU instead of w2nn.LeakyReLU - Add useful attribute to .json New JSON attribute: The first layer has `model_config` attribute. It contains: model_arch: architecture name of model. see `lib/srcnn.lua` scale_factor: if scale_factor > 1, model:forward() changes image resolution with scale_factor. channels: input/output channels. if channels == 3, model is RGB model. offset: pixel size that is to be removed from output. for example: (scale_factor=1, offset=7, input=100x100) => output=(100-7)x(100-7) (scale_factor=2, offset=12, input=100x100) => output=(100*2-12)x(100*2-12) And each layer has `class_name` attribute.
2016-05-15 06:04:08 +12:00
_type
Add 2x model for photo scaling
2015-11-30 21:29:34 +13:00
2
16
torch.CudaTensor
Convert model files; Add new pretrained model - Add new pretrained model to ./models/upconv_7 - Move old models to ./models/vgg_7 - Use nn.LeakyReLU instead of w2nn.LeakyReLU - Add useful attribute to .json New JSON attribute: The first layer has `model_config` attribute. It contains: model_arch: architecture name of model. see `lib/srcnn.lua` scale_factor: if scale_factor > 1, model:forward() changes image resolution with scale_factor. channels: input/output channels. if channels == 3, model is RGB model. offset: pixel size that is to be removed from output. for example: (scale_factor=1, offset=7, input=100x100) => output=(100-7)x(100-7) (scale_factor=2, offset=12, input=100x100) => output=(100*2-12)x(100*2-12) And each layer has `class_name` attribute.
2016-05-15 06:04:08 +12:00
2
6
negval
Add 2x model for photo scaling
2015-11-30 21:29:34 +13:00
1
Convert model files; Add new pretrained model - Add new pretrained model to ./models/upconv_7 - Move old models to ./models/vgg_7 - Use nn.LeakyReLU instead of w2nn.LeakyReLU - Add useful attribute to .json New JSON attribute: The first layer has `model_config` attribute. It contains: model_arch: architecture name of model. see `lib/srcnn.lua` scale_factor: if scale_factor > 1, model:forward() changes image resolution with scale_factor. channels: input/output channels. if channels == 3, model is RGB model. offset: pixel size that is to be removed from output. for example: (scale_factor=1, offset=7, input=100x100) => output=(100-7)x(100-7) (scale_factor=2, offset=12, input=100x100) => output=(100*2-12)x(100*2-12) And each layer has `class_name` attribute.
2016-05-15 06:04:08 +12:00
0.10000000000000001
Add 2x model for photo scaling
2015-11-30 21:29:34 +13:00
2
9
gradInput
4
Convert model files; Add new pretrained model - Add new pretrained model to ./models/upconv_7 - Move old models to ./models/vgg_7 - Use nn.LeakyReLU instead of w2nn.LeakyReLU - Add useful attribute to .json New JSON attribute: The first layer has `model_config` attribute. It contains: model_arch: architecture name of model. see `lib/srcnn.lua` scale_factor: if scale_factor > 1, model:forward() changes image resolution with scale_factor. channels: input/output channels. if channels == 3, model is RGB model. offset: pixel size that is to be removed from output. for example: (scale_factor=1, offset=7, input=100x100) => output=(100-7)x(100-7) (scale_factor=2, offset=12, input=100x100) => output=(100*2-12)x(100*2-12) And each layer has `class_name` attribute.
2016-05-15 06:04:08 +12:00
35
Add 2x model for photo scaling
2015-11-30 21:29:34 +13:00
3
V 1
16
torch.CudaTensor
0
1
0
2
6
output
4
Convert model files; Add new pretrained model - Add new pretrained model to ./models/upconv_7 - Move old models to ./models/vgg_7 - Use nn.LeakyReLU instead of w2nn.LeakyReLU - Add useful attribute to .json New JSON attribute: The first layer has `model_config` attribute. It contains: model_arch: architecture name of model. see `lib/srcnn.lua` scale_factor: if scale_factor > 1, model:forward() changes image resolution with scale_factor. channels: input/output channels. if channels == 3, model is RGB model. offset: pixel size that is to be removed from output. for example: (scale_factor=1, offset=7, input=100x100) => output=(100-7)x(100-7) (scale_factor=2, offset=12, input=100x100) => output=(100*2-12)x(100*2-12) And each layer has `class_name` attribute.
2016-05-15 06:04:08 +12:00
36
Add 2x model for photo scaling
2015-11-30 21:29:34 +13:00
3
V 1
16
torch.CudaTensor
0
1
0
2
Convert model files; Add new pretrained model - Add new pretrained model to ./models/upconv_7 - Move old models to ./models/vgg_7 - Use nn.LeakyReLU instead of w2nn.LeakyReLU - Add useful attribute to .json New JSON attribute: The first layer has `model_config` attribute. It contains: model_arch: architecture name of model. see `lib/srcnn.lua` scale_factor: if scale_factor > 1, model:forward() changes image resolution with scale_factor. channels: input/output channels. if channels == 3, model is RGB model. offset: pixel size that is to be removed from output. for example: (scale_factor=1, offset=7, input=100x100) => output=(100-7)x(100-7) (scale_factor=2, offset=12, input=100x100) => output=(100*2-12)x(100*2-12) And each layer has `class_name` attribute.
2016-05-15 06:04:08 +12:00
5
train
5
0
Add 2x model for photo scaling
2015-11-30 21:29:34 +13:00
1
5
4
Convert model files; Add new pretrained model - Add new pretrained model to ./models/upconv_7 - Move old models to ./models/vgg_7 - Use nn.LeakyReLU instead of w2nn.LeakyReLU - Add useful attribute to .json New JSON attribute: The first layer has `model_config` attribute. It contains: model_arch: architecture name of model. see `lib/srcnn.lua` scale_factor: if scale_factor > 1, model:forward() changes image resolution with scale_factor. channels: input/output channels. if channels == 3, model is RGB model. offset: pixel size that is to be removed from output. for example: (scale_factor=1, offset=7, input=100x100) => output=(100-7)x(100-7) (scale_factor=2, offset=12, input=100x100) => output=(100*2-12)x(100*2-12) And each layer has `class_name` attribute.
2016-05-15 06:04:08 +12:00
37
Add 2x model for photo scaling
2015-11-30 21:29:34 +13:00
3
V 1
23
nn.SpatialConvolutionMM
3
Convert model files; Add new pretrained model - Add new pretrained model to ./models/upconv_7 - Move old models to ./models/vgg_7 - Use nn.LeakyReLU instead of w2nn.LeakyReLU - Add useful attribute to .json New JSON attribute: The first layer has `model_config` attribute. It contains: model_arch: architecture name of model. see `lib/srcnn.lua` scale_factor: if scale_factor > 1, model:forward() changes image resolution with scale_factor. channels: input/output channels. if channels == 3, model is RGB model. offset: pixel size that is to be removed from output. for example: (scale_factor=1, offset=7, input=100x100) => output=(100-7)x(100-7) (scale_factor=2, offset=12, input=100x100) => output=(100*2-12)x(100*2-12) And each layer has `class_name` attribute.
2016-05-15 06:04:08 +12:00
38
16
Add 2x model for photo scaling
2015-11-30 21:29:34 +13:00
2
Convert model files; Add new pretrained model - Add new pretrained model to ./models/upconv_7 - Move old models to ./models/vgg_7 - Use nn.LeakyReLU instead of w2nn.LeakyReLU - Add useful attribute to .json New JSON attribute: The first layer has `model_config` attribute. It contains: model_arch: architecture name of model. see `lib/srcnn.lua` scale_factor: if scale_factor > 1, model:forward() changes image resolution with scale_factor. channels: input/output channels. if channels == 3, model is RGB model. offset: pixel size that is to be removed from output. for example: (scale_factor=1, offset=7, input=100x100) => output=(100-7)x(100-7) (scale_factor=2, offset=12, input=100x100) => output=(100*2-12)x(100*2-12) And each layer has `class_name` attribute.
2016-05-15 06:04:08 +12:00
2
dH
1
1
2
2
dW
1
Add 2x model for photo scaling
2015-11-30 21:29:34 +13:00
1
2
11
nInputPlane
1
32
2
6
output
4
Convert model files; Add new pretrained model - Add new pretrained model to ./models/upconv_7 - Move old models to ./models/vgg_7 - Use nn.LeakyReLU instead of w2nn.LeakyReLU - Add useful attribute to .json New JSON attribute: The first layer has `model_config` attribute. It contains: model_arch: architecture name of model. see `lib/srcnn.lua` scale_factor: if scale_factor > 1, model:forward() changes image resolution with scale_factor. channels: input/output channels. if channels == 3, model is RGB model. offset: pixel size that is to be removed from output. for example: (scale_factor=1, offset=7, input=100x100) => output=(100-7)x(100-7) (scale_factor=2, offset=12, input=100x100) => output=(100*2-12)x(100*2-12) And each layer has `class_name` attribute.
2016-05-15 06:04:08 +12:00
39
Add 2x model for photo scaling
2015-11-30 21:29:34 +13:00
3
V 1
16
torch.CudaTensor
0
1
0
2
Convert model files; Add new pretrained model - Add new pretrained model to ./models/upconv_7 - Move old models to ./models/vgg_7 - Use nn.LeakyReLU instead of w2nn.LeakyReLU - Add useful attribute to .json New JSON attribute: The first layer has `model_config` attribute. It contains: model_arch: architecture name of model. see `lib/srcnn.lua` scale_factor: if scale_factor > 1, model:forward() changes image resolution with scale_factor. channels: input/output channels. if channels == 3, model is RGB model. offset: pixel size that is to be removed from output. for example: (scale_factor=1, offset=7, input=100x100) => output=(100-7)x(100-7) (scale_factor=2, offset=12, input=100x100) => output=(100*2-12)x(100*2-12) And each layer has `class_name` attribute.
2016-05-15 06:04:08 +12:00
2
kH
1
3
2
Add 2x model for photo scaling
2015-11-30 21:29:34 +13:00
9
gradInput
4
Convert model files; Add new pretrained model - Add new pretrained model to ./models/upconv_7 - Move old models to ./models/vgg_7 - Use nn.LeakyReLU instead of w2nn.LeakyReLU - Add useful attribute to .json New JSON attribute: The first layer has `model_config` attribute. It contains: model_arch: architecture name of model. see `lib/srcnn.lua` scale_factor: if scale_factor > 1, model:forward() changes image resolution with scale_factor. channels: input/output channels. if channels == 3, model is RGB model. offset: pixel size that is to be removed from output. for example: (scale_factor=1, offset=7, input=100x100) => output=(100-7)x(100-7) (scale_factor=2, offset=12, input=100x100) => output=(100*2-12)x(100*2-12) And each layer has `class_name` attribute.
2016-05-15 06:04:08 +12:00
40
Add 2x model for photo scaling
2015-11-30 21:29:34 +13:00
3
V 1
16
torch.CudaTensor
0
1
0
2
2
Convert model files; Add new pretrained model - Add new pretrained model to ./models/upconv_7 - Move old models to ./models/vgg_7 - Use nn.LeakyReLU instead of w2nn.LeakyReLU - Add useful attribute to .json New JSON attribute: The first layer has `model_config` attribute. It contains: model_arch: architecture name of model. see `lib/srcnn.lua` scale_factor: if scale_factor > 1, model:forward() changes image resolution with scale_factor. channels: input/output channels. if channels == 3, model is RGB model. offset: pixel size that is to be removed from output. for example: (scale_factor=1, offset=7, input=100x100) => output=(100-7)x(100-7) (scale_factor=2, offset=12, input=100x100) => output=(100*2-12)x(100*2-12) And each layer has `class_name` attribute.
2016-05-15 06:04:08 +12:00
kW
Add 2x model for photo scaling
2015-11-30 21:29:34 +13:00
1
Convert model files; Add new pretrained model - Add new pretrained model to ./models/upconv_7 - Move old models to ./models/vgg_7 - Use nn.LeakyReLU instead of w2nn.LeakyReLU - Add useful attribute to .json New JSON attribute: The first layer has `model_config` attribute. It contains: model_arch: architecture name of model. see `lib/srcnn.lua` scale_factor: if scale_factor > 1, model:forward() changes image resolution with scale_factor. channels: input/output channels. if channels == 3, model is RGB model. offset: pixel size that is to be removed from output. for example: (scale_factor=1, offset=7, input=100x100) => output=(100-7)x(100-7) (scale_factor=2, offset=12, input=100x100) => output=(100*2-12)x(100*2-12) And each layer has `class_name` attribute.
2016-05-15 06:04:08 +12:00
3
Add 2x model for photo scaling
2015-11-30 21:29:34 +13:00
2
12
nOutputPlane
1
64
2
4
Convert model files; Add new pretrained model - Add new pretrained model to ./models/upconv_7 - Move old models to ./models/vgg_7 - Use nn.LeakyReLU instead of w2nn.LeakyReLU - Add useful attribute to .json New JSON attribute: The first layer has `model_config` attribute. It contains: model_arch: architecture name of model. see `lib/srcnn.lua` scale_factor: if scale_factor > 1, model:forward() changes image resolution with scale_factor. channels: input/output channels. if channels == 3, model is RGB model. offset: pixel size that is to be removed from output. for example: (scale_factor=1, offset=7, input=100x100) => output=(100-7)x(100-7) (scale_factor=2, offset=12, input=100x100) => output=(100*2-12)x(100*2-12) And each layer has `class_name` attribute.
2016-05-15 06:04:08 +12:00
padW
Add 2x model for photo scaling
2015-11-30 21:29:34 +13:00
1
0
2
6
weight
4
Convert model files; Add new pretrained model - Add new pretrained model to ./models/upconv_7 - Move old models to ./models/vgg_7 - Use nn.LeakyReLU instead of w2nn.LeakyReLU - Add useful attribute to .json New JSON attribute: The first layer has `model_config` attribute. It contains: model_arch: architecture name of model. see `lib/srcnn.lua` scale_factor: if scale_factor > 1, model:forward() changes image resolution with scale_factor. channels: input/output channels. if channels == 3, model is RGB model. offset: pixel size that is to be removed from output. for example: (scale_factor=1, offset=7, input=100x100) => output=(100-7)x(100-7) (scale_factor=2, offset=12, input=100x100) => output=(100*2-12)x(100*2-12) And each layer has `class_name` attribute.
2016-05-15 06:04:08 +12:00
41
Add 2x model for photo scaling
2015-11-30 21:29:34 +13:00
3
V 1
16
torch.CudaTensor
2
64 288
288 1
1
4
Convert model files; Add new pretrained model - Add new pretrained model to ./models/upconv_7 - Move old models to ./models/vgg_7 - Use nn.LeakyReLU instead of w2nn.LeakyReLU - Add useful attribute to .json New JSON attribute: The first layer has `model_config` attribute. It contains: model_arch: architecture name of model. see `lib/srcnn.lua` scale_factor: if scale_factor > 1, model:forward() changes image resolution with scale_factor. channels: input/output channels. if channels == 3, model is RGB model. offset: pixel size that is to be removed from output. for example: (scale_factor=1, offset=7, input=100x100) => output=(100-7)x(100-7) (scale_factor=2, offset=12, input=100x100) => output=(100*2-12)x(100*2-12) And each layer has `class_name` attribute.
2016-05-15 06:04:08 +12:00
42
Add 2x model for photo scaling
2015-11-30 21:29:34 +13:00
3
V 1
17
torch.CudaStorage
18432
-0.0276309941 -0.0586161129 -0.0887865946 -0.0532789603 -0.00374814426 0.198013127 -0.100600235 0.012298421 0.0651289523 -0.189471155 -0.0906496495 -0.247982234 0.0753483698 -0.0928719863 0.179731905 -0.0998375565 0.0157173742 0.431987435 -0.0729753897 -0.0293409359 0.00339784985 -0.211053967 0.0214618407 0.14716588 0.0189302769 -0.035040468 -0.0494439937 0.273902595 -0.255677998 -0.0440003388 0.0544384234 -0.125089794 0.00227564666 -0.173053026 -0.0540445894 0.0585548393 0.0887554437 0.12215738 -0.118877068 0.0181112532 0.126286104 -0.0917192325 -0.00472670421 -0.0759346858 -0.028143676 -0.0606454909 -0.0807257518 -0.0972565562 0.318595201 0.187754199 0.210003361 -0.218315855 -0.064540945 -0.325068623 -0.0621750876 0.0454025939 0.126243129 -0.130169004 -0.255289882 -0.0893236324 0.0769107863 0.00513626728 0.109366231 0.194957167 -0.0771846175 -0.201756865 -0.171381429 0.0186360702 0.0751977488 -0.0153208356 0.00574518507 0.162145227 -0.627090454 -0.00848865788 -0.0404033102 0.116635486 0.0755242109 0.133656055 -0.458168298 0.221540555 -0.296565473 -0.201613769 -0.0131607577 0.0828816667 -0.0991613865 0.0836545303 0.0174577776 0.200580344 -0.246851891 0.193628997 -0.235789165 -0.189250782 0.0211475622 0.0354787521 0.0488631055 -0.0435919948 -0.097922802 0.11809431 0.0692063645 -0.0647612512 -0.0270988569 -0.0206139199 -0.125112802 -0.0436020158 0.112927094 -0.106271595 0.0429576449 0.0510717146 -0.112378933 -0.054704532 0.369289368 -0.275667936 -0.106195197 -0.135608345 0.281118542 -0.2447422 0.0924625695 0.0474622883 0.0918400586 -0.0120755611 -0.0833284184 -0.0976762548 0.0984101743 -0.0525610819 0.101084128 -0.0103873461 0.45269832 0.0750562847 0.180776805 0.0317995884 -0.0572376586 0.173665673 -0.195553333 -0.14158459 0.293485582 0.0334790535 -0.042417109 -0.0299745314 -0.0675877035 -0.211988673 -0.0923139006 0.102009609 0.0983943492 0.0745906457 -0.15683347 0.0630007237 0.442542851 -0.106544122 0.00182953267 -0.0724341348 0.233303949 0.0136983395 0.0196113624 -0.167656481 -0.099353753 -0.0332951508 0.023816064 -0.278693706 0.105481841 0.0141938142 0.195703283 -0.0128477039 -0.517166615 -0.147580504 0.203752726 -0.0448254757 -0.0939865336 0.238104954 0.124011353 0.146549091 0.236243561 0.0500036143 -0.116362616 -0.0312547646 0.0484602563 -0.00321704498 0.0135049876 -0.0459395982 -0.0356993116 0.099820666 -0.129382685 0.0998002216 -0.00961756334 -0.241549924 0.0822927654 0.0918138027 0.025272036 0.00897740945 0.0561862104 -0.237187073 0.187458977 0.299138159 -0.145822033 -0.0615501925 0.224989951 -0.243618175 -0.268578619 -0.0870819986 -0.0209991448 0.113212414 0.148045257 -0.163854331 0.0303207301 -0.00996947009 0.0768868402 -0.031376861 -0.115536027 -0.0871950835 -0.127398998 -0.644163668 0.0320090353 -0.061769031 -0.0268949624 0.116538547 -0.0305443052 0.274329603 0.478363782 0.216715291 -0.0629276484 0.378880262 0.0101088826 -0.493420362 0.55304426 -0.00407175161 -0.400824875 -0.00893691648 0.111067526 0.078747794 0.0903691947 -0.0409923382 0.0471123494 -0.212105826 -0.0109627862 0.052429907 -0.0858319923 0.132179201 -0.0697147772 -0.146467924 -0.0233582687 0.0393332094 0.0865327865 -0.116955407 -0.0158479065 0.161025509 -0.232603207 -0.148209631 -0.115449041 0.125631183 0.338107795 0.0510702468 -0.0195464101 -0.0597160049 -0.141018987 -0.242795885 0.0201560389 0.0333536416 -0.131087974 -0.114499606 -0.0558729023 0.256458104 0.0656251237 0.560753107 -0.111046784 0.0235738177 0.764862359 0.114537716 0.142225012 0.0913827717 -0.026524812 -0.103057548 -0.0215435829 0.105929434 0.0197637025 -0.113508843 -0.0575285442 -0.0161594022 -0.242408782 0.0474074073 -0.0248990431 -0.100733921 0.0504216924 -0.192643613 0.00120002183 -0.0585549176 0.0883375704 0.052020289 0.0186457392 0.163310543 0.101340577 -0.0905186161 -0.0816826448 0.0577676371 -0.046995759 0.0640654415 -0.03522598 0.0742966905 -0.0540665127 -0.140344709 -0.168153569 -0.018025171 0.0238754619 0.00209902134 0.173265845 -0.0590758398 -0.0521284901 0.140324444 0.0104652252 0.204921544 0.0678840727 0.0124606099 0.0943572596 -0.509817421 0.0927596912 0.0224270467
2
5
train
5
0
2
10
gradWeight
4
Convert model files; Add new pretrained model - Add new pretrained model to ./models/upconv_7 - Move old models to ./models/vgg_7 - Use nn.LeakyReLU instead of w2nn.LeakyReLU - Add useful attribute to .json New JSON attribute: The first layer has `model_config` attribute. It contains: model_arch: architecture name of model. see `lib/srcnn.lua` scale_factor: if scale_factor > 1, model:forward() changes image resolution with scale_factor. channels: input/output channels. if channels == 3, model is RGB model. offset: pixel size that is to be removed from output. for example: (scale_factor=1, offset=7, input=100x100) => output=(100-7)x(100-7) (scale_factor=2, offset=12, input=100x100) => output=(100*2-12)x(100*2-12) And each layer has `class_name` attribute.
2016-05-15 06:04:08 +12:00
43
Add 2x model for photo scaling
2015-11-30 21:29:34 +13:00
3
V 1
16
torch.CudaTensor
2
64 288
288 1
1
4
Convert model files; Add new pretrained model - Add new pretrained model to ./models/upconv_7 - Move old models to ./models/vgg_7 - Use nn.LeakyReLU instead of w2nn.LeakyReLU - Add useful attribute to .json New JSON attribute: The first layer has `model_config` attribute. It contains: model_arch: architecture name of model. see `lib/srcnn.lua` scale_factor: if scale_factor > 1, model:forward() changes image resolution with scale_factor. channels: input/output channels. if channels == 3, model is RGB model. offset: pixel size that is to be removed from output. for example: (scale_factor=1, offset=7, input=100x100) => output=(100-7)x(100-7) (scale_factor=2, offset=12, input=100x100) => output=(100*2-12)x(100*2-12) And each layer has `class_name` attribute.
2016-05-15 06:04:08 +12:00
44
Add 2x model for photo scaling
2015-11-30 21:29:34 +13:00
3
V 1
17
torch.CudaStorage
18432
Convert model files; Add new pretrained model - Add new pretrained model to ./models/upconv_7 - Move old models to ./models/vgg_7 - Use nn.LeakyReLU instead of w2nn.LeakyReLU - Add useful attribute to .json New JSON attribute: The first layer has `model_config` attribute. It contains: model_arch: architecture name of model. see `lib/srcnn.lua` scale_factor: if scale_factor > 1, model:forward() changes image resolution with scale_factor. channels: input/output channels. if channels == 3, model is RGB model. offset: pixel size that is to be removed from output. for example: (scale_factor=1, offset=7, input=100x100) => output=(100-7)x(100-7) (scale_factor=2, offset=12, input=100x100) => output=(100*2-12)x(100*2-12) And each layer has `class_name` attribute.
2016-05-15 06:04:08 +12:00
0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0
2
5
_type
2
16
torch.CudaTensor
Add 2x model for photo scaling
2015-11-30 21:29:34 +13:00
2
4
Convert model files; Add new pretrained model - Add new pretrained model to ./models/upconv_7 - Move old models to ./models/vgg_7 - Use nn.LeakyReLU instead of w2nn.LeakyReLU - Add useful attribute to .json New JSON attribute: The first layer has `model_config` attribute. It contains: model_arch: architecture name of model. see `lib/srcnn.lua` scale_factor: if scale_factor > 1, model:forward() changes image resolution with scale_factor. channels: input/output channels. if channels == 3, model is RGB model. offset: pixel size that is to be removed from output. for example: (scale_factor=1, offset=7, input=100x100) => output=(100-7)x(100-7) (scale_factor=2, offset=12, input=100x100) => output=(100*2-12)x(100*2-12) And each layer has `class_name` attribute.
2016-05-15 06:04:08 +12:00
bias
4
45
Add 2x model for photo scaling
2015-11-30 21:29:34 +13:00
3
V 1
16
torch.CudaTensor
1
Convert model files; Add new pretrained model - Add new pretrained model to ./models/upconv_7 - Move old models to ./models/vgg_7 - Use nn.LeakyReLU instead of w2nn.LeakyReLU - Add useful attribute to .json New JSON attribute: The first layer has `model_config` attribute. It contains: model_arch: architecture name of model. see `lib/srcnn.lua` scale_factor: if scale_factor > 1, model:forward() changes image resolution with scale_factor. channels: input/output channels. if channels == 3, model is RGB model. offset: pixel size that is to be removed from output. for example: (scale_factor=1, offset=7, input=100x100) => output=(100-7)x(100-7) (scale_factor=2, offset=12, input=100x100) => output=(100*2-12)x(100*2-12) And each layer has `class_name` attribute.
2016-05-15 06:04:08 +12:00
64
Add 2x model for photo scaling
2015-11-30 21:29:34 +13:00
1
1
4
Convert model files; Add new pretrained model - Add new pretrained model to ./models/upconv_7 - Move old models to ./models/vgg_7 - Use nn.LeakyReLU instead of w2nn.LeakyReLU - Add useful attribute to .json New JSON attribute: The first layer has `model_config` attribute. It contains: model_arch: architecture name of model. see `lib/srcnn.lua` scale_factor: if scale_factor > 1, model:forward() changes image resolution with scale_factor. channels: input/output channels. if channels == 3, model is RGB model. offset: pixel size that is to be removed from output. for example: (scale_factor=1, offset=7, input=100x100) => output=(100-7)x(100-7) (scale_factor=2, offset=12, input=100x100) => output=(100*2-12)x(100*2-12) And each layer has `class_name` attribute.
2016-05-15 06:04:08 +12:00
46
Add 2x model for photo scaling
2015-11-30 21:29:34 +13:00
3
V 1
Convert model files; Add new pretrained model - Add new pretrained model to ./models/upconv_7 - Move old models to ./models/vgg_7 - Use nn.LeakyReLU instead of w2nn.LeakyReLU - Add useful attribute to .json New JSON attribute: The first layer has `model_config` attribute. It contains: model_arch: architecture name of model. see `lib/srcnn.lua` scale_factor: if scale_factor > 1, model:forward() changes image resolution with scale_factor. channels: input/output channels. if channels == 3, model is RGB model. offset: pixel size that is to be removed from output. for example: (scale_factor=1, offset=7, input=100x100) => output=(100-7)x(100-7) (scale_factor=2, offset=12, input=100x100) => output=(100*2-12)x(100*2-12) And each layer has `class_name` attribute.
2016-05-15 06:04:08 +12:00
17
torch.CudaStorage
64
0.0215742197 0.0090412423 0.0105380267 -0.000352880452 -0.248228177 0.000625068438 0.0339393504 -0.0109324092 0.0192658454 0.00998621713 -0.00773774972 -0.0502326936 -0.00141308259 0.00651591225 0.00204963447 -0.0034551206 -0.0294672642 0.0361296907 -0.0391665697 -0.0184348579 -0.0196739808 -0.0173255596 0.00223238161 -0.0119746188 -0.00448895711 0.00507306075 -0.00443085004 0.00634975918 0.00161403534 0.0163667053 -0.0142647307 -0.0243762564 -0.00304970331 -0.010091166 -0.00825821422 0.00805984251 -0.00289951311 0.00545240054 -0.00337123568 0.0253688525 0.0021856688 -0.0154370219 0.0223934781 -0.0163220651 -0.0339340679 -0.0202156473 0.019086346 -0.0315149873 -0.0148248486 0.0142913591 -0.00373896025 -0.054801181 -0.00323036802 0.0291093942 0.023505291 -0.0271944534 0.000158945753 0.00952955708 0.0175566599 0.000590868469 0.010338529 0.00786158442 -0.0104221348 -0.00877797417
Add 2x model for photo scaling
2015-11-30 21:29:34 +13:00
2
8
Convert model files; Add new pretrained model - Add new pretrained model to ./models/upconv_7 - Move old models to ./models/vgg_7 - Use nn.LeakyReLU instead of w2nn.LeakyReLU - Add useful attribute to .json New JSON attribute: The first layer has `model_config` attribute. It contains: model_arch: architecture name of model. see `lib/srcnn.lua` scale_factor: if scale_factor > 1, model:forward() changes image resolution with scale_factor. channels: input/output channels. if channels == 3, model is RGB model. offset: pixel size that is to be removed from output. for example: (scale_factor=1, offset=7, input=100x100) => output=(100-7)x(100-7) (scale_factor=2, offset=12, input=100x100) => output=(100*2-12)x(100*2-12) And each layer has `class_name` attribute.
2016-05-15 06:04:08 +12:00
gradBias
Add 2x model for photo scaling
2015-11-30 21:29:34 +13:00
4
Convert model files; Add new pretrained model - Add new pretrained model to ./models/upconv_7 - Move old models to ./models/vgg_7 - Use nn.LeakyReLU instead of w2nn.LeakyReLU - Add useful attribute to .json New JSON attribute: The first layer has `model_config` attribute. It contains: model_arch: architecture name of model. see `lib/srcnn.lua` scale_factor: if scale_factor > 1, model:forward() changes image resolution with scale_factor. channels: input/output channels. if channels == 3, model is RGB model. offset: pixel size that is to be removed from output. for example: (scale_factor=1, offset=7, input=100x100) => output=(100-7)x(100-7) (scale_factor=2, offset=12, input=100x100) => output=(100*2-12)x(100*2-12) And each layer has `class_name` attribute.
2016-05-15 06:04:08 +12:00
47
Add 2x model for photo scaling
2015-11-30 21:29:34 +13:00
3
V 1
16
torch.CudaTensor
Convert model files; Add new pretrained model - Add new pretrained model to ./models/upconv_7 - Move old models to ./models/vgg_7 - Use nn.LeakyReLU instead of w2nn.LeakyReLU - Add useful attribute to .json New JSON attribute: The first layer has `model_config` attribute. It contains: model_arch: architecture name of model. see `lib/srcnn.lua` scale_factor: if scale_factor > 1, model:forward() changes image resolution with scale_factor. channels: input/output channels. if channels == 3, model is RGB model. offset: pixel size that is to be removed from output. for example: (scale_factor=1, offset=7, input=100x100) => output=(100-7)x(100-7) (scale_factor=2, offset=12, input=100x100) => output=(100*2-12)x(100*2-12) And each layer has `class_name` attribute.
2016-05-15 06:04:08 +12:00
1
64
1
1
4
48
3
V 1
17
torch.CudaStorage
64
Add upconv_7/photo models. Update vgg_7/photo/noise* models
2016-07-04 07:51:22 +12:00
0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0
Convert model files; Add new pretrained model - Add new pretrained model to ./models/upconv_7 - Move old models to ./models/vgg_7 - Use nn.LeakyReLU instead of w2nn.LeakyReLU - Add useful attribute to .json New JSON attribute: The first layer has `model_config` attribute. It contains: model_arch: architecture name of model. see `lib/srcnn.lua` scale_factor: if scale_factor > 1, model:forward() changes image resolution with scale_factor. channels: input/output channels. if channels == 3, model is RGB model. offset: pixel size that is to be removed from output. for example: (scale_factor=1, offset=7, input=100x100) => output=(100-7)x(100-7) (scale_factor=2, offset=12, input=100x100) => output=(100*2-12)x(100*2-12) And each layer has `class_name` attribute.
2016-05-15 06:04:08 +12:00
2
4
padH
Add 2x model for photo scaling
2015-11-30 21:29:34 +13:00
1
0
Convert model files; Add new pretrained model - Add new pretrained model to ./models/upconv_7 - Move old models to ./models/vgg_7 - Use nn.LeakyReLU instead of w2nn.LeakyReLU - Add useful attribute to .json New JSON attribute: The first layer has `model_config` attribute. It contains: model_arch: architecture name of model. see `lib/srcnn.lua` scale_factor: if scale_factor > 1, model:forward() changes image resolution with scale_factor. channels: input/output channels. if channels == 3, model is RGB model. offset: pixel size that is to be removed from output. for example: (scale_factor=1, offset=7, input=100x100) => output=(100-7)x(100-7) (scale_factor=2, offset=12, input=100x100) => output=(100*2-12)x(100*2-12) And each layer has `class_name` attribute.
2016-05-15 06:04:08 +12:00
1
6
4
49
3
V 1
12
nn.LeakyReLU
3
50
6
2
7
inplace
5
1
2
5
_type
2
16
torch.CudaTensor
2
6
negval
1
0.10000000000000001
Add 2x model for photo scaling
2015-11-30 21:29:34 +13:00
2
9
gradInput
4
Convert model files; Add new pretrained model - Add new pretrained model to ./models/upconv_7 - Move old models to ./models/vgg_7 - Use nn.LeakyReLU instead of w2nn.LeakyReLU - Add useful attribute to .json New JSON attribute: The first layer has `model_config` attribute. It contains: model_arch: architecture name of model. see `lib/srcnn.lua` scale_factor: if scale_factor > 1, model:forward() changes image resolution with scale_factor. channels: input/output channels. if channels == 3, model is RGB model. offset: pixel size that is to be removed from output. for example: (scale_factor=1, offset=7, input=100x100) => output=(100-7)x(100-7) (scale_factor=2, offset=12, input=100x100) => output=(100*2-12)x(100*2-12) And each layer has `class_name` attribute.
2016-05-15 06:04:08 +12:00
51
Add 2x model for photo scaling
2015-11-30 21:29:34 +13:00
3
V 1
16
torch.CudaTensor
0
1
0
2
6
output
4
Convert model files; Add new pretrained model - Add new pretrained model to ./models/upconv_7 - Move old models to ./models/vgg_7 - Use nn.LeakyReLU instead of w2nn.LeakyReLU - Add useful attribute to .json New JSON attribute: The first layer has `model_config` attribute. It contains: model_arch: architecture name of model. see `lib/srcnn.lua` scale_factor: if scale_factor > 1, model:forward() changes image resolution with scale_factor. channels: input/output channels. if channels == 3, model is RGB model. offset: pixel size that is to be removed from output. for example: (scale_factor=1, offset=7, input=100x100) => output=(100-7)x(100-7) (scale_factor=2, offset=12, input=100x100) => output=(100*2-12)x(100*2-12) And each layer has `class_name` attribute.
2016-05-15 06:04:08 +12:00
52
Add 2x model for photo scaling
2015-11-30 21:29:34 +13:00
3
V 1
16
torch.CudaTensor
0
1
0
2
Convert model files; Add new pretrained model - Add new pretrained model to ./models/upconv_7 - Move old models to ./models/vgg_7 - Use nn.LeakyReLU instead of w2nn.LeakyReLU - Add useful attribute to .json New JSON attribute: The first layer has `model_config` attribute. It contains: model_arch: architecture name of model. see `lib/srcnn.lua` scale_factor: if scale_factor > 1, model:forward() changes image resolution with scale_factor. channels: input/output channels. if channels == 3, model is RGB model. offset: pixel size that is to be removed from output. for example: (scale_factor=1, offset=7, input=100x100) => output=(100-7)x(100-7) (scale_factor=2, offset=12, input=100x100) => output=(100*2-12)x(100*2-12) And each layer has `class_name` attribute.
2016-05-15 06:04:08 +12:00
5
train
5
0
Add 2x model for photo scaling
2015-11-30 21:29:34 +13:00
1
7
4
Convert model files; Add new pretrained model - Add new pretrained model to ./models/upconv_7 - Move old models to ./models/vgg_7 - Use nn.LeakyReLU instead of w2nn.LeakyReLU - Add useful attribute to .json New JSON attribute: The first layer has `model_config` attribute. It contains: model_arch: architecture name of model. see `lib/srcnn.lua` scale_factor: if scale_factor > 1, model:forward() changes image resolution with scale_factor. channels: input/output channels. if channels == 3, model is RGB model. offset: pixel size that is to be removed from output. for example: (scale_factor=1, offset=7, input=100x100) => output=(100-7)x(100-7) (scale_factor=2, offset=12, input=100x100) => output=(100*2-12)x(100*2-12) And each layer has `class_name` attribute.
2016-05-15 06:04:08 +12:00
53
Add 2x model for photo scaling
2015-11-30 21:29:34 +13:00
3
V 1
23
nn.SpatialConvolutionMM
3
Convert model files; Add new pretrained model - Add new pretrained model to ./models/upconv_7 - Move old models to ./models/vgg_7 - Use nn.LeakyReLU instead of w2nn.LeakyReLU - Add useful attribute to .json New JSON attribute: The first layer has `model_config` attribute. It contains: model_arch: architecture name of model. see `lib/srcnn.lua` scale_factor: if scale_factor > 1, model:forward() changes image resolution with scale_factor. channels: input/output channels. if channels == 3, model is RGB model. offset: pixel size that is to be removed from output. for example: (scale_factor=1, offset=7, input=100x100) => output=(100-7)x(100-7) (scale_factor=2, offset=12, input=100x100) => output=(100*2-12)x(100*2-12) And each layer has `class_name` attribute.
2016-05-15 06:04:08 +12:00
54
16
Add 2x model for photo scaling
2015-11-30 21:29:34 +13:00
2
Convert model files; Add new pretrained model - Add new pretrained model to ./models/upconv_7 - Move old models to ./models/vgg_7 - Use nn.LeakyReLU instead of w2nn.LeakyReLU - Add useful attribute to .json New JSON attribute: The first layer has `model_config` attribute. It contains: model_arch: architecture name of model. see `lib/srcnn.lua` scale_factor: if scale_factor > 1, model:forward() changes image resolution with scale_factor. channels: input/output channels. if channels == 3, model is RGB model. offset: pixel size that is to be removed from output. for example: (scale_factor=1, offset=7, input=100x100) => output=(100-7)x(100-7) (scale_factor=2, offset=12, input=100x100) => output=(100*2-12)x(100*2-12) And each layer has `class_name` attribute.
2016-05-15 06:04:08 +12:00
2
dH
1
1
2
2
dW
1
Add 2x model for photo scaling
2015-11-30 21:29:34 +13:00
1
2
11
nInputPlane
1
64
2
6
output
4
Convert model files; Add new pretrained model - Add new pretrained model to ./models/upconv_7 - Move old models to ./models/vgg_7 - Use nn.LeakyReLU instead of w2nn.LeakyReLU - Add useful attribute to .json New JSON attribute: The first layer has `model_config` attribute. It contains: model_arch: architecture name of model. see `lib/srcnn.lua` scale_factor: if scale_factor > 1, model:forward() changes image resolution with scale_factor. channels: input/output channels. if channels == 3, model is RGB model. offset: pixel size that is to be removed from output. for example: (scale_factor=1, offset=7, input=100x100) => output=(100-7)x(100-7) (scale_factor=2, offset=12, input=100x100) => output=(100*2-12)x(100*2-12) And each layer has `class_name` attribute.
2016-05-15 06:04:08 +12:00
55
Add 2x model for photo scaling
2015-11-30 21:29:34 +13:00
3
V 1
16
torch.CudaTensor
0
1
0
2
Convert model files; Add new pretrained model - Add new pretrained model to ./models/upconv_7 - Move old models to ./models/vgg_7 - Use nn.LeakyReLU instead of w2nn.LeakyReLU - Add useful attribute to .json New JSON attribute: The first layer has `model_config` attribute. It contains: model_arch: architecture name of model. see `lib/srcnn.lua` scale_factor: if scale_factor > 1, model:forward() changes image resolution with scale_factor. channels: input/output channels. if channels == 3, model is RGB model. offset: pixel size that is to be removed from output. for example: (scale_factor=1, offset=7, input=100x100) => output=(100-7)x(100-7) (scale_factor=2, offset=12, input=100x100) => output=(100*2-12)x(100*2-12) And each layer has `class_name` attribute.
2016-05-15 06:04:08 +12:00
2
kH
1
3
2
Add 2x model for photo scaling
2015-11-30 21:29:34 +13:00
9
gradInput
4
Convert model files; Add new pretrained model - Add new pretrained model to ./models/upconv_7 - Move old models to ./models/vgg_7 - Use nn.LeakyReLU instead of w2nn.LeakyReLU - Add useful attribute to .json New JSON attribute: The first layer has `model_config` attribute. It contains: model_arch: architecture name of model. see `lib/srcnn.lua` scale_factor: if scale_factor > 1, model:forward() changes image resolution with scale_factor. channels: input/output channels. if channels == 3, model is RGB model. offset: pixel size that is to be removed from output. for example: (scale_factor=1, offset=7, input=100x100) => output=(100-7)x(100-7) (scale_factor=2, offset=12, input=100x100) => output=(100*2-12)x(100*2-12) And each layer has `class_name` attribute.
2016-05-15 06:04:08 +12:00
56
Add 2x model for photo scaling
2015-11-30 21:29:34 +13:00
3
V 1
16
torch.CudaTensor
0
1
0
2
2
Convert model files; Add new pretrained model - Add new pretrained model to ./models/upconv_7 - Move old models to ./models/vgg_7 - Use nn.LeakyReLU instead of w2nn.LeakyReLU - Add useful attribute to .json New JSON attribute: The first layer has `model_config` attribute. It contains: model_arch: architecture name of model. see `lib/srcnn.lua` scale_factor: if scale_factor > 1, model:forward() changes image resolution with scale_factor. channels: input/output channels. if channels == 3, model is RGB model. offset: pixel size that is to be removed from output. for example: (scale_factor=1, offset=7, input=100x100) => output=(100-7)x(100-7) (scale_factor=2, offset=12, input=100x100) => output=(100*2-12)x(100*2-12) And each layer has `class_name` attribute.
2016-05-15 06:04:08 +12:00
kW
Add 2x model for photo scaling
2015-11-30 21:29:34 +13:00
1
Convert model files; Add new pretrained model - Add new pretrained model to ./models/upconv_7 - Move old models to ./models/vgg_7 - Use nn.LeakyReLU instead of w2nn.LeakyReLU - Add useful attribute to .json New JSON attribute: The first layer has `model_config` attribute. It contains: model_arch: architecture name of model. see `lib/srcnn.lua` scale_factor: if scale_factor > 1, model:forward() changes image resolution with scale_factor. channels: input/output channels. if channels == 3, model is RGB model. offset: pixel size that is to be removed from output. for example: (scale_factor=1, offset=7, input=100x100) => output=(100-7)x(100-7) (scale_factor=2, offset=12, input=100x100) => output=(100*2-12)x(100*2-12) And each layer has `class_name` attribute.
2016-05-15 06:04:08 +12:00
3
Add 2x model for photo scaling
2015-11-30 21:29:34 +13:00
2
12
nOutputPlane
1
64
2
4
Convert model files; Add new pretrained model - Add new pretrained model to ./models/upconv_7 - Move old models to ./models/vgg_7 - Use nn.LeakyReLU instead of w2nn.LeakyReLU - Add useful attribute to .json New JSON attribute: The first layer has `model_config` attribute. It contains: model_arch: architecture name of model. see `lib/srcnn.lua` scale_factor: if scale_factor > 1, model:forward() changes image resolution with scale_factor. channels: input/output channels. if channels == 3, model is RGB model. offset: pixel size that is to be removed from output. for example: (scale_factor=1, offset=7, input=100x100) => output=(100-7)x(100-7) (scale_factor=2, offset=12, input=100x100) => output=(100*2-12)x(100*2-12) And each layer has `class_name` attribute.
2016-05-15 06:04:08 +12:00
padW
Add 2x model for photo scaling
2015-11-30 21:29:34 +13:00
1
0
2
6
weight
4
Convert model files; Add new pretrained model - Add new pretrained model to ./models/upconv_7 - Move old models to ./models/vgg_7 - Use nn.LeakyReLU instead of w2nn.LeakyReLU - Add useful attribute to .json New JSON attribute: The first layer has `model_config` attribute. It contains: model_arch: architecture name of model. see `lib/srcnn.lua` scale_factor: if scale_factor > 1, model:forward() changes image resolution with scale_factor. channels: input/output channels. if channels == 3, model is RGB model. offset: pixel size that is to be removed from output. for example: (scale_factor=1, offset=7, input=100x100) => output=(100-7)x(100-7) (scale_factor=2, offset=12, input=100x100) => output=(100*2-12)x(100*2-12) And each layer has `class_name` attribute.
2016-05-15 06:04:08 +12:00
57
Add 2x model for photo scaling
2015-11-30 21:29:34 +13:00
3
V 1
16
torch.CudaTensor
2
64 576
576 1
1
4
Convert model files; Add new pretrained model - Add new pretrained model to ./models/upconv_7 - Move old models to ./models/vgg_7 - Use nn.LeakyReLU instead of w2nn.LeakyReLU - Add useful attribute to .json New JSON attribute: The first layer has `model_config` attribute. It contains: model_arch: architecture name of model. see `lib/srcnn.lua` scale_factor: if scale_factor > 1, model:forward() changes image resolution with scale_factor. channels: input/output channels. if channels == 3, model is RGB model. offset: pixel size that is to be removed from output. for example: (scale_factor=1, offset=7, input=100x100) => output=(100-7)x(100-7) (scale_factor=2, offset=12, input=100x100) => output=(100*2-12)x(100*2-12) And each layer has `class_name` attribute.
2016-05-15 06:04:08 +12:00
58
Add 2x model for photo scaling
2015-11-30 21:29:34 +13:00
3
V 1
17
torch.CudaStorage
36864
-0.294924229 -0.204114184 0.138147831 -0.270741373 0.126082197 -0.0178978704 0.00409492012 0.247113183 -0.0790794194 -0.308719426 0.21912618 0.0362051949 0.148285866 0.180347651 -0.0860397071 -0.217363983 -0.028386347 -0.130117372 -0.161580011 0.0210670549 -0.0614774488 -0.0460210331 0.210208341 0.051141616 -0.107838094 0.0772857293 -0.174258605 -0.208767533 -0.0699473396 -0.104498841 0.0897797495 0.0356469974 0.0775092468 0.0138271982 0.0790980831 0.103566132 -0.295947969 -0.03968665 -0.356974363 0.102388903 -0.0234534144 0.150572807 -0.0388604216 -0.177958235 0.2777161 0.00773475179 -0.165719077 -0.0614288002 -0.0460439362 -0.159793645 0.0376371741 -0.0384638719 -0.0979657993 -0.0500645898 0.121706441 0.121319935 -0.358875811 -0.11077477 0.21433033 -0.172303692 -0.0655637607 0.139115691 -0.0729209036 -0.297193676 -0.135454789 -0.252382904 -0.0215883888 -0.00778541341 -0.061534822 -0.00984365027 -0.052027449 -0.0704909414 0.205667302 0.0768471584 0.280405521 0.105862468 0.0471622907 0.00620007934 0.0487682931 0.0742151514 -0.193558395 0.0375033133 -0.131287053 0.0937591568 -0.0749890134 0.11959216 -0.10090524 -0.203210652 0.213794276 0.116699703 -0.222571194 -0.0549192503 -0.0765276551 -0.0111150304 0.11058896 0.235930011 0.175889194 0.0855109617 0.110142656 0.0226084627 -0.0105470438 -0.149713293 -0.148435295 -0.0207430571 -0.300825149 0.0461848751 0.19787091 0.0954660103 0.13959682 0.164662927 -0.00354395155 -0.0864380598 -0.0644707084 0.0555157065 -0.331485927 -0.00104717119 -0.224193111 -0.112174049 -0.268620431 -0.130828038 0.191224217 0.222958922 0.197556838 0.0444055311 -0.104734361 -0.272629321 0.190435871 -0.126535237 -0.505708516 0.0728621408 0.0981846228 -0.176411584 -0.040994823 -0.158220068 0.0633931831 0.0882036909 -0.386216462 -0.0632110313 0.0278356951 0.130986586 -0.14878884 0.0864660963 0.0645789281 -0.0971944332 0.0494845547 0.10006284 -0.186903119 -0.0747788623 -0.156304911 -0.446843415 0.0762374625 -0.0228137206 -0.0921951607 -0.225026652 0.167033464 -0.104677506 -0.184935853 0.117793001 -0.152568668 0.024697477 0.378492862 -0.00876558665 0.117194921 -0.23655352 0.371409893 0.0177774578 -0.0109268464 -0.176212698 -0.161602333 0.132028818 -0.256152153 0.279314578 0.0288759917 0.201912194 0.316685319 -0.0192215145 0.127975151 0.108052649 0.0372396596 0.0787585601 0.166379005 0.0125067988 -0.241400838 -0.0318718404 -0.198375925 -0.202551931 0.137162745 0.18687436 -0.221244678 0.0550274961 0.285653234 0.246166617 -0.122911312 0.273986459 -0.0558218844 0.0877074003 0.276288271 0.230519682 -0.0331284925 -0.119121455 0.0240772795 0.0648028329 -0.119300351 0.0442560613 0.0233814288 -0.0680696219 0.0298832227 -0.00483219 -0.245989233 0.0231061801 0.200544357 -0.0422711335 0.0951148421 0.0591043569 -0.0211096238 -0.0376205668 -0.0755366161 -0.262458056 -0.0896945074 -0.130449831 -0.159926191 -0.0950840935 0.0463864766 -0.152367905 0.254049957 -0.274011135 -0.0861207172 -0.0149266608 0.221833363 0.133900985 0.095540449 -0.0772062317 -0.131706804 -0.0659582913 -0.0235453155 0.124864727 -0.385188848 0.00733976578 0.141924277 -0.128529787 0.171712056 0.0055503021 -0.209412113 0.269412369 0.108136572 -0.114097729 -0.0863100216 -0.369878411 0.0993738994 0.0508586131 0.0601719879 0.039817486 -0.0026290277 -0.0167008545 0.151709974 0.126753196 0.0528845005 0.0821200162 0.233412087 0.075548023 0.138215914 -0.17773813 -0.0449593291 0.112909168 -0.0254901201 0.00492251106 0.108768247 -0.0519566946 -0.38608852 0.305131525 0.0131367249 0.138917789 0.121562816 -0.0967461839 -0.144712016 0.268788218 0.156948537 -0.131203264 0.0546574742 0.194796428 0.139096335 0.211193055 0.155852899 0.325675189 0.306947649 0.0274126604 -0.0518334284 0.000712117937 0.0599953569 0.202243447 0.0114413928 0.170114949 0.163487151 0.00939543732 -0.163645148 -0.306215733 -0.21526219 -0.187088966 0.0434449837 -0.188002184 0.335722864 0.259481519 -0.0660416111 -0.129535526 -0.168799892 -0.28390345 0.00139613741 -0.363267154 -0.294439912 -0.0520256571 -0.0819733441 -0.0469310209 0.0994663015 0.0843334123 0.0577550046 -0.128044128 0.0710603967 -0.29
2
5
train
5
0
2
10
gradWeight
4
Convert model files; Add new pretrained model - Add new pretrained model to ./models/upconv_7 - Move old models to ./models/vgg_7 - Use nn.LeakyReLU instead of w2nn.LeakyReLU - Add useful attribute to .json New JSON attribute: The first layer has `model_config` attribute. It contains: model_arch: architecture name of model. see `lib/srcnn.lua` scale_factor: if scale_factor > 1, model:forward() changes image resolution with scale_factor. channels: input/output channels. if channels == 3, model is RGB model. offset: pixel size that is to be removed from output. for example: (scale_factor=1, offset=7, input=100x100) => output=(100-7)x(100-7) (scale_factor=2, offset=12, input=100x100) => output=(100*2-12)x(100*2-12) And each layer has `class_name` attribute.
2016-05-15 06:04:08 +12:00
59
Add 2x model for photo scaling
2015-11-30 21:29:34 +13:00
3
V 1
16
torch.CudaTensor
2
64 576
576 1
1
4
Convert model files; Add new pretrained model - Add new pretrained model to ./models/upconv_7 - Move old models to ./models/vgg_7 - Use nn.LeakyReLU instead of w2nn.LeakyReLU - Add useful attribute to .json New JSON attribute: The first layer has `model_config` attribute. It contains: model_arch: architecture name of model. see `lib/srcnn.lua` scale_factor: if scale_factor > 1, model:forward() changes image resolution with scale_factor. channels: input/output channels. if channels == 3, model is RGB model. offset: pixel size that is to be removed from output. for example: (scale_factor=1, offset=7, input=100x100) => output=(100-7)x(100-7) (scale_factor=2, offset=12, input=100x100) => output=(100*2-12)x(100*2-12) And each layer has `class_name` attribute.
2016-05-15 06:04:08 +12:00
60
Add 2x model for photo scaling
2015-11-30 21:29:34 +13:00
3
V 1
17
torch.CudaStorage
36864
Convert model files; Add new pretrained model - Add new pretrained model to ./models/upconv_7 - Move old models to ./models/vgg_7 - Use nn.LeakyReLU instead of w2nn.LeakyReLU - Add useful attribute to .json New JSON attribute: The first layer has `model_config` attribute. It contains: model_arch: architecture name of model. see `lib/srcnn.lua` scale_factor: if scale_factor > 1, model:forward() changes image resolution with scale_factor. channels: input/output channels. if channels == 3, model is RGB model. offset: pixel size that is to be removed from output. for example: (scale_factor=1, offset=7, input=100x100) => output=(100-7)x(100-7) (scale_factor=2, offset=12, input=100x100) => output=(100*2-12)x(100*2-12) And each layer has `class_name` attribute.
2016-05-15 06:04:08 +12:00
0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0
2
5
_type
2
16
torch.CudaTensor
Add 2x model for photo scaling
2015-11-30 21:29:34 +13:00
2
4
Convert model files; Add new pretrained model - Add new pretrained model to ./models/upconv_7 - Move old models to ./models/vgg_7 - Use nn.LeakyReLU instead of w2nn.LeakyReLU - Add useful attribute to .json New JSON attribute: The first layer has `model_config` attribute. It contains: model_arch: architecture name of model. see `lib/srcnn.lua` scale_factor: if scale_factor > 1, model:forward() changes image resolution with scale_factor. channels: input/output channels. if channels == 3, model is RGB model. offset: pixel size that is to be removed from output. for example: (scale_factor=1, offset=7, input=100x100) => output=(100-7)x(100-7) (scale_factor=2, offset=12, input=100x100) => output=(100*2-12)x(100*2-12) And each layer has `class_name` attribute.
2016-05-15 06:04:08 +12:00
bias
4
61
Add 2x model for photo scaling
2015-11-30 21:29:34 +13:00
3
V 1
16
torch.CudaTensor
1
Convert model files; Add new pretrained model - Add new pretrained model to ./models/upconv_7 - Move old models to ./models/vgg_7 - Use nn.LeakyReLU instead of w2nn.LeakyReLU - Add useful attribute to .json New JSON attribute: The first layer has `model_config` attribute. It contains: model_arch: architecture name of model. see `lib/srcnn.lua` scale_factor: if scale_factor > 1, model:forward() changes image resolution with scale_factor. channels: input/output channels. if channels == 3, model is RGB model. offset: pixel size that is to be removed from output. for example: (scale_factor=1, offset=7, input=100x100) => output=(100-7)x(100-7) (scale_factor=2, offset=12, input=100x100) => output=(100*2-12)x(100*2-12) And each layer has `class_name` attribute.
2016-05-15 06:04:08 +12:00
64
1
1
4
62
3
V 1
17
torch.CudaStorage
64
-0.0128027974 0.000780455011 0.0167234335 -0.0495668724 -0.0301821344 0.000416165887 -0.0154441977 -0.0127612744 0.0104534151 -0.27753222 0.0134043247 -0.0130837569 -0.00346644386 -0.00737886131 -0.0102475602 -0.0397284962 0.00773614086 0.00965277199 -0.00135415955 0.000929481292 -0.030949574 -0.0249426719 -0.000392071379 -0.0118755931 0.00114599906 -0.0155703677 -0.00808453746 -0.00864954013 -0.0518773347 -0.0280516949 -0.0287742149 0.01914501 -0.0123233246 -0.00287509151 -0.00263790251 -0.0136038885 -0.00547038345 0.00589828519 -0.0147454096 -0.0117058782 0.00427318364 -0.0198211893 -0.0128154103 0.00511160959 0.0056190053 -0.00926772784 -0.00422329735 0.0052212528 -0.0154713616 -0.024439536 -0.0326975733 -0.0229648016 -0.0424627587 -0.0174505301 -0.00639187125 -0.00637968164 -0.0324530602 0.00838527456 -0.0110613266 0.00526939845 -0.0217877179 0.00595240388 -0.0337144099 -0.148745432
Add 2x model for photo scaling
2015-11-30 21:29:34 +13:00
2
Convert model files; Add new pretrained model - Add new pretrained model to ./models/upconv_7 - Move old models to ./models/vgg_7 - Use nn.LeakyReLU instead of w2nn.LeakyReLU - Add useful attribute to .json New JSON attribute: The first layer has `model_config` attribute. It contains: model_arch: architecture name of model. see `lib/srcnn.lua` scale_factor: if scale_factor > 1, model:forward() changes image resolution with scale_factor. channels: input/output channels. if channels == 3, model is RGB model. offset: pixel size that is to be removed from output. for example: (scale_factor=1, offset=7, input=100x100) => output=(100-7)x(100-7) (scale_factor=2, offset=12, input=100x100) => output=(100*2-12)x(100*2-12) And each layer has `class_name` attribute.
2016-05-15 06:04:08 +12:00
8
gradBias
Add 2x model for photo scaling
2015-11-30 21:29:34 +13:00
4
Convert model files; Add new pretrained model - Add new pretrained model to ./models/upconv_7 - Move old models to ./models/vgg_7 - Use nn.LeakyReLU instead of w2nn.LeakyReLU - Add useful attribute to .json New JSON attribute: The first layer has `model_config` attribute. It contains: model_arch: architecture name of model. see `lib/srcnn.lua` scale_factor: if scale_factor > 1, model:forward() changes image resolution with scale_factor. channels: input/output channels. if channels == 3, model is RGB model. offset: pixel size that is to be removed from output. for example: (scale_factor=1, offset=7, input=100x100) => output=(100-7)x(100-7) (scale_factor=2, offset=12, input=100x100) => output=(100*2-12)x(100*2-12) And each layer has `class_name` attribute.
2016-05-15 06:04:08 +12:00
63
3
V 1
16
torch.CudaTensor
Add 2x model for photo scaling
2015-11-30 21:29:34 +13:00
1
Convert model files; Add new pretrained model - Add new pretrained model to ./models/upconv_7 - Move old models to ./models/vgg_7 - Use nn.LeakyReLU instead of w2nn.LeakyReLU - Add useful attribute to .json New JSON attribute: The first layer has `model_config` attribute. It contains: model_arch: architecture name of model. see `lib/srcnn.lua` scale_factor: if scale_factor > 1, model:forward() changes image resolution with scale_factor. channels: input/output channels. if channels == 3, model is RGB model. offset: pixel size that is to be removed from output. for example: (scale_factor=1, offset=7, input=100x100) => output=(100-7)x(100-7) (scale_factor=2, offset=12, input=100x100) => output=(100*2-12)x(100*2-12) And each layer has `class_name` attribute.
2016-05-15 06:04:08 +12:00
64
Add 2x model for photo scaling
2015-11-30 21:29:34 +13:00
1
1
Convert model files; Add new pretrained model - Add new pretrained model to ./models/upconv_7 - Move old models to ./models/vgg_7 - Use nn.LeakyReLU instead of w2nn.LeakyReLU - Add useful attribute to .json New JSON attribute: The first layer has `model_config` attribute. It contains: model_arch: architecture name of model. see `lib/srcnn.lua` scale_factor: if scale_factor > 1, model:forward() changes image resolution with scale_factor. channels: input/output channels. if channels == 3, model is RGB model. offset: pixel size that is to be removed from output. for example: (scale_factor=1, offset=7, input=100x100) => output=(100-7)x(100-7) (scale_factor=2, offset=12, input=100x100) => output=(100*2-12)x(100*2-12) And each layer has `class_name` attribute.
2016-05-15 06:04:08 +12:00
4
64
3
V 1
17
torch.CudaStorage
64
Add upconv_7/photo models. Update vgg_7/photo/noise* models
2016-07-04 07:51:22 +12:00
0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0
Add 2x model for photo scaling
2015-11-30 21:29:34 +13:00
2
Convert model files; Add new pretrained model - Add new pretrained model to ./models/upconv_7 - Move old models to ./models/vgg_7 - Use nn.LeakyReLU instead of w2nn.LeakyReLU - Add useful attribute to .json New JSON attribute: The first layer has `model_config` attribute. It contains: model_arch: architecture name of model. see `lib/srcnn.lua` scale_factor: if scale_factor > 1, model:forward() changes image resolution with scale_factor. channels: input/output channels. if channels == 3, model is RGB model. offset: pixel size that is to be removed from output. for example: (scale_factor=1, offset=7, input=100x100) => output=(100-7)x(100-7) (scale_factor=2, offset=12, input=100x100) => output=(100*2-12)x(100*2-12) And each layer has `class_name` attribute.
2016-05-15 06:04:08 +12:00
4
padH
Add 2x model for photo scaling
2015-11-30 21:29:34 +13:00
1
Convert model files; Add new pretrained model - Add new pretrained model to ./models/upconv_7 - Move old models to ./models/vgg_7 - Use nn.LeakyReLU instead of w2nn.LeakyReLU - Add useful attribute to .json New JSON attribute: The first layer has `model_config` attribute. It contains: model_arch: architecture name of model. see `lib/srcnn.lua` scale_factor: if scale_factor > 1, model:forward() changes image resolution with scale_factor. channels: input/output channels. if channels == 3, model is RGB model. offset: pixel size that is to be removed from output. for example: (scale_factor=1, offset=7, input=100x100) => output=(100-7)x(100-7) (scale_factor=2, offset=12, input=100x100) => output=(100*2-12)x(100*2-12) And each layer has `class_name` attribute.
2016-05-15 06:04:08 +12:00
0
Add 2x model for photo scaling
2015-11-30 21:29:34 +13:00
1
8
4
Convert model files; Add new pretrained model - Add new pretrained model to ./models/upconv_7 - Move old models to ./models/vgg_7 - Use nn.LeakyReLU instead of w2nn.LeakyReLU - Add useful attribute to .json New JSON attribute: The first layer has `model_config` attribute. It contains: model_arch: architecture name of model. see `lib/srcnn.lua` scale_factor: if scale_factor > 1, model:forward() changes image resolution with scale_factor. channels: input/output channels. if channels == 3, model is RGB model. offset: pixel size that is to be removed from output. for example: (scale_factor=1, offset=7, input=100x100) => output=(100-7)x(100-7) (scale_factor=2, offset=12, input=100x100) => output=(100*2-12)x(100*2-12) And each layer has `class_name` attribute.
2016-05-15 06:04:08 +12:00
65
Add 2x model for photo scaling
2015-11-30 21:29:34 +13:00
3
V 1
Convert model files; Add new pretrained model - Add new pretrained model to ./models/upconv_7 - Move old models to ./models/vgg_7 - Use nn.LeakyReLU instead of w2nn.LeakyReLU - Add useful attribute to .json New JSON attribute: The first layer has `model_config` attribute. It contains: model_arch: architecture name of model. see `lib/srcnn.lua` scale_factor: if scale_factor > 1, model:forward() changes image resolution with scale_factor. channels: input/output channels. if channels == 3, model is RGB model. offset: pixel size that is to be removed from output. for example: (scale_factor=1, offset=7, input=100x100) => output=(100-7)x(100-7) (scale_factor=2, offset=12, input=100x100) => output=(100*2-12)x(100*2-12) And each layer has `class_name` attribute.
2016-05-15 06:04:08 +12:00
12
nn.LeakyReLU
Add 2x model for photo scaling
2015-11-30 21:29:34 +13:00
3
Convert model files; Add new pretrained model - Add new pretrained model to ./models/upconv_7 - Move old models to ./models/vgg_7 - Use nn.LeakyReLU instead of w2nn.LeakyReLU - Add useful attribute to .json New JSON attribute: The first layer has `model_config` attribute. It contains: model_arch: architecture name of model. see `lib/srcnn.lua` scale_factor: if scale_factor > 1, model:forward() changes image resolution with scale_factor. channels: input/output channels. if channels == 3, model is RGB model. offset: pixel size that is to be removed from output. for example: (scale_factor=1, offset=7, input=100x100) => output=(100-7)x(100-7) (scale_factor=2, offset=12, input=100x100) => output=(100*2-12)x(100*2-12) And each layer has `class_name` attribute.
2016-05-15 06:04:08 +12:00
66
6
Add 2x model for photo scaling
2015-11-30 21:29:34 +13:00
2
Convert model files; Add new pretrained model - Add new pretrained model to ./models/upconv_7 - Move old models to ./models/vgg_7 - Use nn.LeakyReLU instead of w2nn.LeakyReLU - Add useful attribute to .json New JSON attribute: The first layer has `model_config` attribute. It contains: model_arch: architecture name of model. see `lib/srcnn.lua` scale_factor: if scale_factor > 1, model:forward() changes image resolution with scale_factor. channels: input/output channels. if channels == 3, model is RGB model. offset: pixel size that is to be removed from output. for example: (scale_factor=1, offset=7, input=100x100) => output=(100-7)x(100-7) (scale_factor=2, offset=12, input=100x100) => output=(100*2-12)x(100*2-12) And each layer has `class_name` attribute.
2016-05-15 06:04:08 +12:00
7
inplace
Add 2x model for photo scaling
2015-11-30 21:29:34 +13:00
5
Convert model files; Add new pretrained model - Add new pretrained model to ./models/upconv_7 - Move old models to ./models/vgg_7 - Use nn.LeakyReLU instead of w2nn.LeakyReLU - Add useful attribute to .json New JSON attribute: The first layer has `model_config` attribute. It contains: model_arch: architecture name of model. see `lib/srcnn.lua` scale_factor: if scale_factor > 1, model:forward() changes image resolution with scale_factor. channels: input/output channels. if channels == 3, model is RGB model. offset: pixel size that is to be removed from output. for example: (scale_factor=1, offset=7, input=100x100) => output=(100-7)x(100-7) (scale_factor=2, offset=12, input=100x100) => output=(100*2-12)x(100*2-12) And each layer has `class_name` attribute.
2016-05-15 06:04:08 +12:00
1
2
Add 2x model for photo scaling
2015-11-30 21:29:34 +13:00
5
Convert model files; Add new pretrained model - Add new pretrained model to ./models/upconv_7 - Move old models to ./models/vgg_7 - Use nn.LeakyReLU instead of w2nn.LeakyReLU - Add useful attribute to .json New JSON attribute: The first layer has `model_config` attribute. It contains: model_arch: architecture name of model. see `lib/srcnn.lua` scale_factor: if scale_factor > 1, model:forward() changes image resolution with scale_factor. channels: input/output channels. if channels == 3, model is RGB model. offset: pixel size that is to be removed from output. for example: (scale_factor=1, offset=7, input=100x100) => output=(100-7)x(100-7) (scale_factor=2, offset=12, input=100x100) => output=(100*2-12)x(100*2-12) And each layer has `class_name` attribute.
2016-05-15 06:04:08 +12:00
_type
Add 2x model for photo scaling
2015-11-30 21:29:34 +13:00
2
16
torch.CudaTensor
Convert model files; Add new pretrained model - Add new pretrained model to ./models/upconv_7 - Move old models to ./models/vgg_7 - Use nn.LeakyReLU instead of w2nn.LeakyReLU - Add useful attribute to .json New JSON attribute: The first layer has `model_config` attribute. It contains: model_arch: architecture name of model. see `lib/srcnn.lua` scale_factor: if scale_factor > 1, model:forward() changes image resolution with scale_factor. channels: input/output channels. if channels == 3, model is RGB model. offset: pixel size that is to be removed from output. for example: (scale_factor=1, offset=7, input=100x100) => output=(100-7)x(100-7) (scale_factor=2, offset=12, input=100x100) => output=(100*2-12)x(100*2-12) And each layer has `class_name` attribute.
2016-05-15 06:04:08 +12:00
2
6
negval
Add 2x model for photo scaling
2015-11-30 21:29:34 +13:00
1
Convert model files; Add new pretrained model - Add new pretrained model to ./models/upconv_7 - Move old models to ./models/vgg_7 - Use nn.LeakyReLU instead of w2nn.LeakyReLU - Add useful attribute to .json New JSON attribute: The first layer has `model_config` attribute. It contains: model_arch: architecture name of model. see `lib/srcnn.lua` scale_factor: if scale_factor > 1, model:forward() changes image resolution with scale_factor. channels: input/output channels. if channels == 3, model is RGB model. offset: pixel size that is to be removed from output. for example: (scale_factor=1, offset=7, input=100x100) => output=(100-7)x(100-7) (scale_factor=2, offset=12, input=100x100) => output=(100*2-12)x(100*2-12) And each layer has `class_name` attribute.
2016-05-15 06:04:08 +12:00
0.10000000000000001
Add 2x model for photo scaling
2015-11-30 21:29:34 +13:00
2
9
gradInput
4
Convert model files; Add new pretrained model - Add new pretrained model to ./models/upconv_7 - Move old models to ./models/vgg_7 - Use nn.LeakyReLU instead of w2nn.LeakyReLU - Add useful attribute to .json New JSON attribute: The first layer has `model_config` attribute. It contains: model_arch: architecture name of model. see `lib/srcnn.lua` scale_factor: if scale_factor > 1, model:forward() changes image resolution with scale_factor. channels: input/output channels. if channels == 3, model is RGB model. offset: pixel size that is to be removed from output. for example: (scale_factor=1, offset=7, input=100x100) => output=(100-7)x(100-7) (scale_factor=2, offset=12, input=100x100) => output=(100*2-12)x(100*2-12) And each layer has `class_name` attribute.
2016-05-15 06:04:08 +12:00
67
Add 2x model for photo scaling
2015-11-30 21:29:34 +13:00
3
V 1
16
torch.CudaTensor
0
1
0
2
6
output
4
Convert model files; Add new pretrained model - Add new pretrained model to ./models/upconv_7 - Move old models to ./models/vgg_7 - Use nn.LeakyReLU instead of w2nn.LeakyReLU - Add useful attribute to .json New JSON attribute: The first layer has `model_config` attribute. It contains: model_arch: architecture name of model. see `lib/srcnn.lua` scale_factor: if scale_factor > 1, model:forward() changes image resolution with scale_factor. channels: input/output channels. if channels == 3, model is RGB model. offset: pixel size that is to be removed from output. for example: (scale_factor=1, offset=7, input=100x100) => output=(100-7)x(100-7) (scale_factor=2, offset=12, input=100x100) => output=(100*2-12)x(100*2-12) And each layer has `class_name` attribute.
2016-05-15 06:04:08 +12:00
68
Add 2x model for photo scaling
2015-11-30 21:29:34 +13:00
3
V 1
16
torch.CudaTensor
0
1
0
2
Convert model files; Add new pretrained model - Add new pretrained model to ./models/upconv_7 - Move old models to ./models/vgg_7 - Use nn.LeakyReLU instead of w2nn.LeakyReLU - Add useful attribute to .json New JSON attribute: The first layer has `model_config` attribute. It contains: model_arch: architecture name of model. see `lib/srcnn.lua` scale_factor: if scale_factor > 1, model:forward() changes image resolution with scale_factor. channels: input/output channels. if channels == 3, model is RGB model. offset: pixel size that is to be removed from output. for example: (scale_factor=1, offset=7, input=100x100) => output=(100-7)x(100-7) (scale_factor=2, offset=12, input=100x100) => output=(100*2-12)x(100*2-12) And each layer has `class_name` attribute.
2016-05-15 06:04:08 +12:00
5
train
5
0
Add 2x model for photo scaling
2015-11-30 21:29:34 +13:00
1
9
4
Convert model files; Add new pretrained model - Add new pretrained model to ./models/upconv_7 - Move old models to ./models/vgg_7 - Use nn.LeakyReLU instead of w2nn.LeakyReLU - Add useful attribute to .json New JSON attribute: The first layer has `model_config` attribute. It contains: model_arch: architecture name of model. see `lib/srcnn.lua` scale_factor: if scale_factor > 1, model:forward() changes image resolution with scale_factor. channels: input/output channels. if channels == 3, model is RGB model. offset: pixel size that is to be removed from output. for example: (scale_factor=1, offset=7, input=100x100) => output=(100-7)x(100-7) (scale_factor=2, offset=12, input=100x100) => output=(100*2-12)x(100*2-12) And each layer has `class_name` attribute.
2016-05-15 06:04:08 +12:00
69
Add 2x model for photo scaling
2015-11-30 21:29:34 +13:00
3
V 1
23
nn.SpatialConvolutionMM
3
Convert model files; Add new pretrained model - Add new pretrained model to ./models/upconv_7 - Move old models to ./models/vgg_7 - Use nn.LeakyReLU instead of w2nn.LeakyReLU - Add useful attribute to .json New JSON attribute: The first layer has `model_config` attribute. It contains: model_arch: architecture name of model. see `lib/srcnn.lua` scale_factor: if scale_factor > 1, model:forward() changes image resolution with scale_factor. channels: input/output channels. if channels == 3, model is RGB model. offset: pixel size that is to be removed from output. for example: (scale_factor=1, offset=7, input=100x100) => output=(100-7)x(100-7) (scale_factor=2, offset=12, input=100x100) => output=(100*2-12)x(100*2-12) And each layer has `class_name` attribute.
2016-05-15 06:04:08 +12:00
70
16
Add 2x model for photo scaling
2015-11-30 21:29:34 +13:00
2
Convert model files; Add new pretrained model - Add new pretrained model to ./models/upconv_7 - Move old models to ./models/vgg_7 - Use nn.LeakyReLU instead of w2nn.LeakyReLU - Add useful attribute to .json New JSON attribute: The first layer has `model_config` attribute. It contains: model_arch: architecture name of model. see `lib/srcnn.lua` scale_factor: if scale_factor > 1, model:forward() changes image resolution with scale_factor. channels: input/output channels. if channels == 3, model is RGB model. offset: pixel size that is to be removed from output. for example: (scale_factor=1, offset=7, input=100x100) => output=(100-7)x(100-7) (scale_factor=2, offset=12, input=100x100) => output=(100*2-12)x(100*2-12) And each layer has `class_name` attribute.
2016-05-15 06:04:08 +12:00
2
dH
1
1
2
2
dW
1
Add 2x model for photo scaling
2015-11-30 21:29:34 +13:00
1
2
11
nInputPlane
1
64
2
6
output
4
Convert model files; Add new pretrained model - Add new pretrained model to ./models/upconv_7 - Move old models to ./models/vgg_7 - Use nn.LeakyReLU instead of w2nn.LeakyReLU - Add useful attribute to .json New JSON attribute: The first layer has `model_config` attribute. It contains: model_arch: architecture name of model. see `lib/srcnn.lua` scale_factor: if scale_factor > 1, model:forward() changes image resolution with scale_factor. channels: input/output channels. if channels == 3, model is RGB model. offset: pixel size that is to be removed from output. for example: (scale_factor=1, offset=7, input=100x100) => output=(100-7)x(100-7) (scale_factor=2, offset=12, input=100x100) => output=(100*2-12)x(100*2-12) And each layer has `class_name` attribute.
2016-05-15 06:04:08 +12:00
71
Add 2x model for photo scaling
2015-11-30 21:29:34 +13:00
3
V 1
16
torch.CudaTensor
0
1
0
2
Convert model files; Add new pretrained model - Add new pretrained model to ./models/upconv_7 - Move old models to ./models/vgg_7 - Use nn.LeakyReLU instead of w2nn.LeakyReLU - Add useful attribute to .json New JSON attribute: The first layer has `model_config` attribute. It contains: model_arch: architecture name of model. see `lib/srcnn.lua` scale_factor: if scale_factor > 1, model:forward() changes image resolution with scale_factor. channels: input/output channels. if channels == 3, model is RGB model. offset: pixel size that is to be removed from output. for example: (scale_factor=1, offset=7, input=100x100) => output=(100-7)x(100-7) (scale_factor=2, offset=12, input=100x100) => output=(100*2-12)x(100*2-12) And each layer has `class_name` attribute.
2016-05-15 06:04:08 +12:00
2
kH
1
3
2
Add 2x model for photo scaling
2015-11-30 21:29:34 +13:00
9
gradInput
4
Convert model files; Add new pretrained model - Add new pretrained model to ./models/upconv_7 - Move old models to ./models/vgg_7 - Use nn.LeakyReLU instead of w2nn.LeakyReLU - Add useful attribute to .json New JSON attribute: The first layer has `model_config` attribute. It contains: model_arch: architecture name of model. see `lib/srcnn.lua` scale_factor: if scale_factor > 1, model:forward() changes image resolution with scale_factor. channels: input/output channels. if channels == 3, model is RGB model. offset: pixel size that is to be removed from output. for example: (scale_factor=1, offset=7, input=100x100) => output=(100-7)x(100-7) (scale_factor=2, offset=12, input=100x100) => output=(100*2-12)x(100*2-12) And each layer has `class_name` attribute.
2016-05-15 06:04:08 +12:00
72
Add 2x model for photo scaling
2015-11-30 21:29:34 +13:00
3
V 1
16
torch.CudaTensor
0
1
0
2
2
Convert model files; Add new pretrained model - Add new pretrained model to ./models/upconv_7 - Move old models to ./models/vgg_7 - Use nn.LeakyReLU instead of w2nn.LeakyReLU - Add useful attribute to .json New JSON attribute: The first layer has `model_config` attribute. It contains: model_arch: architecture name of model. see `lib/srcnn.lua` scale_factor: if scale_factor > 1, model:forward() changes image resolution with scale_factor. channels: input/output channels. if channels == 3, model is RGB model. offset: pixel size that is to be removed from output. for example: (scale_factor=1, offset=7, input=100x100) => output=(100-7)x(100-7) (scale_factor=2, offset=12, input=100x100) => output=(100*2-12)x(100*2-12) And each layer has `class_name` attribute.
2016-05-15 06:04:08 +12:00
kW
Add 2x model for photo scaling
2015-11-30 21:29:34 +13:00
1
Convert model files; Add new pretrained model - Add new pretrained model to ./models/upconv_7 - Move old models to ./models/vgg_7 - Use nn.LeakyReLU instead of w2nn.LeakyReLU - Add useful attribute to .json New JSON attribute: The first layer has `model_config` attribute. It contains: model_arch: architecture name of model. see `lib/srcnn.lua` scale_factor: if scale_factor > 1, model:forward() changes image resolution with scale_factor. channels: input/output channels. if channels == 3, model is RGB model. offset: pixel size that is to be removed from output. for example: (scale_factor=1, offset=7, input=100x100) => output=(100-7)x(100-7) (scale_factor=2, offset=12, input=100x100) => output=(100*2-12)x(100*2-12) And each layer has `class_name` attribute.
2016-05-15 06:04:08 +12:00
3
Add 2x model for photo scaling
2015-11-30 21:29:34 +13:00
2
12
nOutputPlane
1
128
2
4
Convert model files; Add new pretrained model - Add new pretrained model to ./models/upconv_7 - Move old models to ./models/vgg_7 - Use nn.LeakyReLU instead of w2nn.LeakyReLU - Add useful attribute to .json New JSON attribute: The first layer has `model_config` attribute. It contains: model_arch: architecture name of model. see `lib/srcnn.lua` scale_factor: if scale_factor > 1, model:forward() changes image resolution with scale_factor. channels: input/output channels. if channels == 3, model is RGB model. offset: pixel size that is to be removed from output. for example: (scale_factor=1, offset=7, input=100x100) => output=(100-7)x(100-7) (scale_factor=2, offset=12, input=100x100) => output=(100*2-12)x(100*2-12) And each layer has `class_name` attribute.
2016-05-15 06:04:08 +12:00
padW
Add 2x model for photo scaling
2015-11-30 21:29:34 +13:00
1
0
2
6
weight
4
Convert model files; Add new pretrained model - Add new pretrained model to ./models/upconv_7 - Move old models to ./models/vgg_7 - Use nn.LeakyReLU instead of w2nn.LeakyReLU - Add useful attribute to .json New JSON attribute: The first layer has `model_config` attribute. It contains: model_arch: architecture name of model. see `lib/srcnn.lua` scale_factor: if scale_factor > 1, model:forward() changes image resolution with scale_factor. channels: input/output channels. if channels == 3, model is RGB model. offset: pixel size that is to be removed from output. for example: (scale_factor=1, offset=7, input=100x100) => output=(100-7)x(100-7) (scale_factor=2, offset=12, input=100x100) => output=(100*2-12)x(100*2-12) And each layer has `class_name` attribute.
2016-05-15 06:04:08 +12:00
73
Add 2x model for photo scaling
2015-11-30 21:29:34 +13:00
3
V 1
16
torch.CudaTensor
2
128 576
576 1
1
4
Convert model files; Add new pretrained model - Add new pretrained model to ./models/upconv_7 - Move old models to ./models/vgg_7 - Use nn.LeakyReLU instead of w2nn.LeakyReLU - Add useful attribute to .json New JSON attribute: The first layer has `model_config` attribute. It contains: model_arch: architecture name of model. see `lib/srcnn.lua` scale_factor: if scale_factor > 1, model:forward() changes image resolution with scale_factor. channels: input/output channels. if channels == 3, model is RGB model. offset: pixel size that is to be removed from output. for example: (scale_factor=1, offset=7, input=100x100) => output=(100-7)x(100-7) (scale_factor=2, offset=12, input=100x100) => output=(100*2-12)x(100*2-12) And each layer has `class_name` attribute.
2016-05-15 06:04:08 +12:00
74
Add 2x model for photo scaling
2015-11-30 21:29:34 +13:00
3
V 1
17
torch.CudaStorage
73728
-0.0943586081 0.019499762 0.023095822 -0.000800619018 0.0755086094 -0.0336215161 -0.0414616428 0.041240178 -0.0460994914 0.0374755934 0.000253639824 -0.0440291986 0.0333939083 0.0695288703 -0.1244758 -0.00553223817 0.0846068934 -0.0891164318 -0.0308638215 -0.00941293873 -0.00107267313 0.00117364072 0.000805024116 0.076463595 0.00863153953 -0.00665325392 -0.0102843707 0.01067184 -0.0256755035 -0.0366768353 0.00580847636 0.0176789258 0.010094516 0.0189092774 -0.0233778935 0.062834695 -0.0284771509 -0.0798706561 0.0241675507 -0.012149075 -0.113624446 0.0408481434 0.0518860817 0.112903066 0.118575111 0.0598139092 0.043427445 0.0615716726 -0.0344669223 -0.0708110556 -0.102991216 0.0357807651 0.0449316576 -0.0566231534 -0.0201409888 0.0194259379 -0.00175588683 0.0319221094 0.0931309536 -0.0842079967 -0.00123800163 0.0402820185 -0.0631977916 0.038784299 0.0739782155 -0.0446495526 0.0103821186 0.0197541993 0.00266122562 0.0190007892 -0.0636778921 -0.00947718509 -0.052817829 -0.0069494443 0.00283947843 0.0564627126 0.00882084481 0.0262560248 0.00750482967 0.015056341 -0.0143902767 0.0360130817 0.0722780153 -0.00963022746 0.0317510515 -0.0066606882 0.0222265963 0.0654032677 0.0375376046 0.0501593277 -0.045364745 -0.00440466637 0.0124009456 0.0198031794 -0.00817656238 -0.0144253196 0.0262356158 -0.017588675 -0.0223657172 0.0268110484 0.0245214086 0.00410552882 -0.00627128314 -0.00662661297 0.033387918 -0.0208718684 0.0468191579 -0.0312740132 0.0249238797 0.0705383942 0.0326002315 0.00868578162 0.0172572248 0.0771144331 -0.0141034536 0.0431126654 -0.048406139 0.00449959701 0.0150529109 0.031346336 0.0667427629 -0.00877507683 -0.0528880954 0.0009667587 -0.0211361982 -0.0343600512 0.0314690843 -0.0232675765 -0.023980232 0.03836748 0.0197818112 -0.0267554689 0.0361036211 0.0161598809 -0.0463947691 -0.0148060322 0.0707234889 0.087499164 -0.0146282278 -0.0557430834 0.0897580907 -0.0890710205 -0.0974710807 -0.0545954295 0.00107912859 -0.0111010242 -0.0142222755 -0.0507334284 -0.00143139751 -0.0181834381 0.0192198977 0.047798451 0.0384377539 -0.160971805 -0.189945847 -0.16595386 -0.180890083 -0.262700737 -0.243870229 -0.121893309 -0.139433712 -0.112355083 0.0288754106 -0.0263262577 0.00223155459 0.0427518226 0.0209620446 -0.0671411976 -0.0111222127 -0.00431432435 0.0181217492 0.0055930214 -0.0315829292 -0.00165171234 0.0227250829 0.0415608846 -0.0222412627 0.00961099099 0.000514824002 -0.00438605808 0.0114871114 -0.0544265322 -0.0103127165 0.00452340115 -0.0294795893 0.00075229269 -0.0604572706 0.0487442054 0.0247837305 0.0896804929 -0.080007717 -0.022688916 -0.0162110068 0.0116906725 0.0129637402 -0.0271738525 -0.00713643245 -0.00371619337 0.0434050262 -0.059731748 0.0221709739 -0.0517403409 -0.050838545 0.0259031132 0.0936316475 0.010752786 -0.0143914046 -0.0132499505 0.038936004 -0.071731396 0.019724641 0.0138212927 0.0263636429 0.0155670093 0.00481642736 0.0168627966 -0.00858930033 -0.00398861663 -0.0287291743 0.033857327 0.0119787222 -0.0280276295 -0.0242050644 -0.00950287562 0.0400480442 0.0254904404 -0.000199845614 -0.00867438968 0.00266420539 0.00369523 0.0148419607 -0.0308940504 0.018805813 0.0106449239 0.0910648182 0.16745311 0.102860853 0.116073966 0.151873127 0.188976452 0.0807103664 0.0771497265 0.119579621 0.0183243956 0.106890112 0.0425131805 -0.0666430444 -0.114995994 0.0129723353 -0.0304903537 -0.0512079261 0.091719456 0.0209242739 0.0632917657 0.0271397512 0.0197107345 -0.0426132455 -0.0199036226 0.00804463588 -0.0569729209 0.0323083475 0.0472799763 -0.0411936752 -0.019348396 0.034163747 -0.0695596784 0.00360810221 0.0940521508 -0.00604892056 0.0268778075 -0.0799693614 -0.0309172217 -0.0497773364 0.0296074189 -0.0518239662 -0.0178907998 0.109944873 0.0403750651 0.0941347852 0.0442286022 0.045230601 0.0269576907 -0.0194989387 -0.0590851083 -0.00541532179 0.0274457335 -0.024385564 -0.0504632518 -0.0278968289 -0.0148732988 0.0225526653 -0.0257976465 -0.0303106103 -0.0155137125 -0.0423982777 0.0701675639 0.0172311962 0.00168346893 -0.00192381721 -0.00669956114 -0.00122172211 -0.0362566002 0.0152124641 0.0422232561 -0.0277965
2
5
train
5
0
2
10
gradWeight
4
Convert model files; Add new pretrained model - Add new pretrained model to ./models/upconv_7 - Move old models to ./models/vgg_7 - Use nn.LeakyReLU instead of w2nn.LeakyReLU - Add useful attribute to .json New JSON attribute: The first layer has `model_config` attribute. It contains: model_arch: architecture name of model. see `lib/srcnn.lua` scale_factor: if scale_factor > 1, model:forward() changes image resolution with scale_factor. channels: input/output channels. if channels == 3, model is RGB model. offset: pixel size that is to be removed from output. for example: (scale_factor=1, offset=7, input=100x100) => output=(100-7)x(100-7) (scale_factor=2, offset=12, input=100x100) => output=(100*2-12)x(100*2-12) And each layer has `class_name` attribute.
2016-05-15 06:04:08 +12:00
75
Add 2x model for photo scaling
2015-11-30 21:29:34 +13:00
3
V 1
16
torch.CudaTensor
2
128 576
576 1
1
4
Convert model files; Add new pretrained model - Add new pretrained model to ./models/upconv_7 - Move old models to ./models/vgg_7 - Use nn.LeakyReLU instead of w2nn.LeakyReLU - Add useful attribute to .json New JSON attribute: The first layer has `model_config` attribute. It contains: model_arch: architecture name of model. see `lib/srcnn.lua` scale_factor: if scale_factor > 1, model:forward() changes image resolution with scale_factor. channels: input/output channels. if channels == 3, model is RGB model. offset: pixel size that is to be removed from output. for example: (scale_factor=1, offset=7, input=100x100) => output=(100-7)x(100-7) (scale_factor=2, offset=12, input=100x100) => output=(100*2-12)x(100*2-12) And each layer has `class_name` attribute.
2016-05-15 06:04:08 +12:00
76
Add 2x model for photo scaling
2015-11-30 21:29:34 +13:00
3
V 1
17
torch.CudaStorage
73728
0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0
2
Convert model files; Add new pretrained model - Add new pretrained model to ./models/upconv_7 - Move old models to ./models/vgg_7 - Use nn.LeakyReLU instead of w2nn.LeakyReLU - Add useful attribute to .json New JSON attribute: The first layer has `model_config` attribute. It contains: model_arch: architecture name of model. see `lib/srcnn.lua` scale_factor: if scale_factor > 1, model:forward() changes image resolution with scale_factor. channels: input/output channels. if channels == 3, model is RGB model. offset: pixel size that is to be removed from output. for example: (scale_factor=1, offset=7, input=100x100) => output=(100-7)x(100-7) (scale_factor=2, offset=12, input=100x100) => output=(100*2-12)x(100*2-12) And each layer has `class_name` attribute.
2016-05-15 06:04:08 +12:00
5
_type
2
16
torch.CudaTensor
2
Add 2x model for photo scaling
2015-11-30 21:29:34 +13:00
4
Convert model files; Add new pretrained model - Add new pretrained model to ./models/upconv_7 - Move old models to ./models/vgg_7 - Use nn.LeakyReLU instead of w2nn.LeakyReLU - Add useful attribute to .json New JSON attribute: The first layer has `model_config` attribute. It contains: model_arch: architecture name of model. see `lib/srcnn.lua` scale_factor: if scale_factor > 1, model:forward() changes image resolution with scale_factor. channels: input/output channels. if channels == 3, model is RGB model. offset: pixel size that is to be removed from output. for example: (scale_factor=1, offset=7, input=100x100) => output=(100-7)x(100-7) (scale_factor=2, offset=12, input=100x100) => output=(100*2-12)x(100*2-12) And each layer has `class_name` attribute.
2016-05-15 06:04:08 +12:00
bias
4
77
Add 2x model for photo scaling
2015-11-30 21:29:34 +13:00
3
V 1
16
torch.CudaTensor
1
Convert model files; Add new pretrained model - Add new pretrained model to ./models/upconv_7 - Move old models to ./models/vgg_7 - Use nn.LeakyReLU instead of w2nn.LeakyReLU - Add useful attribute to .json New JSON attribute: The first layer has `model_config` attribute. It contains: model_arch: architecture name of model. see `lib/srcnn.lua` scale_factor: if scale_factor > 1, model:forward() changes image resolution with scale_factor. channels: input/output channels. if channels == 3, model is RGB model. offset: pixel size that is to be removed from output. for example: (scale_factor=1, offset=7, input=100x100) => output=(100-7)x(100-7) (scale_factor=2, offset=12, input=100x100) => output=(100*2-12)x(100*2-12) And each layer has `class_name` attribute.
2016-05-15 06:04:08 +12:00
128
1
1
4
78
3
V 1
17
torch.CudaStorage
128
-0.0230254401 -0.0320929736 -0.183469772 -0.0479027107 -0.0396506898 -0.0122689111 -0.0319250859 -0.0225100126 -0.0619807988 -0.00928188767 -0.0647423938 -0.0154011166 -0.0159912277 -0.0378874876 -0.0579101741 -0.0389927514 -0.0293316822 -0.0349555574 -0.024290327 -0.0520616323 -0.033595968 -0.030756088 -0.0231666993 -0.0292511135 -0.00460516661 -0.0452847742 0.00432133768 -0.061889112 -0.00359596009 -0.028602099 -0.0282280911 -0.0478429832 -0.0171499029 -0.0376362093 -0.0823942348 -0.00914665684 -0.0321313404 -0.0615626909 -0.0321626551 -0.0521960445 -0.0246308874 -0.00461728452 -0.0657208264 -0.0543198548 -0.0495360866 -0.0462905169 -0.039081715 -0.0455454141 -0.0422659963 -0.0141413929 -0.011633466 -0.00453525875 -0.0485552661 -0.020942362 -0.0560742058 -0.0451447442 -0.00590724451 -0.0275886953 -0.0275057554 -0.0297113787 -0.021590706 -0.0311509296 -0.0362407304 -0.0307147205 -0.0534982271 -0.150198281 -0.0181510616 0.0298388842 -0.0643438622 -0.0409783572 -0.0393072404 -0.0085968338 -0.0152782546 -0.0228686891 -0.0270226412 -0.0223976709 -0.0179830492 -0.0308989603 -0.0191027988 -0.0287246127 -0.00568204606 -0.0385414287 -0.0296798535 -0.00500963395 -0.0119575812 -0.0453072302 -0.0167921521 0.0404605418 -0.037318144 -0.231546491 -0.139222711 -0.0170493778 0.00201331964 -0.0462610237 -0.0374037214 -0.0487893261 -0.0614479184 -0.0363025591 -0.00511131855 -0.0144408531 -0.062918447 -0.0531623513 -0.0377137661 -0.00863071252 -0.111626908 -0.0238150675 -0.0113299945 -0.0543570444 -0.024570886 -0.0263978764 -0.0422146656 -0.0261044763 -0.0290812049 -0.054231029 -0.0308624711 -0.0262798015 -0.041432485 -0.0304430611 0.00194422307 -0.0115850046 -0.0309145898 -0.0404561162 -0.0337276384 -0.0270799026 -0.0417871587 -0.0583283864 -0.032396961 -0.0424592383
Add 2x model for photo scaling
2015-11-30 21:29:34 +13:00
2
Convert model files; Add new pretrained model - Add new pretrained model to ./models/upconv_7 - Move old models to ./models/vgg_7 - Use nn.LeakyReLU instead of w2nn.LeakyReLU - Add useful attribute to .json New JSON attribute: The first layer has `model_config` attribute. It contains: model_arch: architecture name of model. see `lib/srcnn.lua` scale_factor: if scale_factor > 1, model:forward() changes image resolution with scale_factor. channels: input/output channels. if channels == 3, model is RGB model. offset: pixel size that is to be removed from output. for example: (scale_factor=1, offset=7, input=100x100) => output=(100-7)x(100-7) (scale_factor=2, offset=12, input=100x100) => output=(100*2-12)x(100*2-12) And each layer has `class_name` attribute.
2016-05-15 06:04:08 +12:00
8
gradBias
Add 2x model for photo scaling
2015-11-30 21:29:34 +13:00
4
Convert model files; Add new pretrained model - Add new pretrained model to ./models/upconv_7 - Move old models to ./models/vgg_7 - Use nn.LeakyReLU instead of w2nn.LeakyReLU - Add useful attribute to .json New JSON attribute: The first layer has `model_config` attribute. It contains: model_arch: architecture name of model. see `lib/srcnn.lua` scale_factor: if scale_factor > 1, model:forward() changes image resolution with scale_factor. channels: input/output channels. if channels == 3, model is RGB model. offset: pixel size that is to be removed from output. for example: (scale_factor=1, offset=7, input=100x100) => output=(100-7)x(100-7) (scale_factor=2, offset=12, input=100x100) => output=(100*2-12)x(100*2-12) And each layer has `class_name` attribute.
2016-05-15 06:04:08 +12:00
79
3
V 1
16
torch.CudaTensor
Add 2x model for photo scaling
2015-11-30 21:29:34 +13:00
1
Convert model files; Add new pretrained model - Add new pretrained model to ./models/upconv_7 - Move old models to ./models/vgg_7 - Use nn.LeakyReLU instead of w2nn.LeakyReLU - Add useful attribute to .json New JSON attribute: The first layer has `model_config` attribute. It contains: model_arch: architecture name of model. see `lib/srcnn.lua` scale_factor: if scale_factor > 1, model:forward() changes image resolution with scale_factor. channels: input/output channels. if channels == 3, model is RGB model. offset: pixel size that is to be removed from output. for example: (scale_factor=1, offset=7, input=100x100) => output=(100-7)x(100-7) (scale_factor=2, offset=12, input=100x100) => output=(100*2-12)x(100*2-12) And each layer has `class_name` attribute.
2016-05-15 06:04:08 +12:00
128
Add 2x model for photo scaling
2015-11-30 21:29:34 +13:00
1
1
Convert model files; Add new pretrained model - Add new pretrained model to ./models/upconv_7 - Move old models to ./models/vgg_7 - Use nn.LeakyReLU instead of w2nn.LeakyReLU - Add useful attribute to .json New JSON attribute: The first layer has `model_config` attribute. It contains: model_arch: architecture name of model. see `lib/srcnn.lua` scale_factor: if scale_factor > 1, model:forward() changes image resolution with scale_factor. channels: input/output channels. if channels == 3, model is RGB model. offset: pixel size that is to be removed from output. for example: (scale_factor=1, offset=7, input=100x100) => output=(100-7)x(100-7) (scale_factor=2, offset=12, input=100x100) => output=(100*2-12)x(100*2-12) And each layer has `class_name` attribute.
2016-05-15 06:04:08 +12:00
4
80
3
V 1
17
torch.CudaStorage
128
Add upconv_7/photo models. Update vgg_7/photo/noise* models
2016-07-04 07:51:22 +12:00
0 0 0 0 2.84337311e-11 9.92269849e-17 3.12762438e-07 -1.89571612e-16 2.54548414e-24 5.01777994e-11 -4.37371307e-15 -5.669754e-17 2.19924162e-10 0 0 0 0 0 -6.68768981e-24 2.96778505e-24 1.28804113e-13 1.81370625e-13 -1.21761133e-07 -3.4570663e-10 -3.5163491e-12 1.19891155e-07 2.26362351e-09 -9.74424074e-06 1.78421971e-07 -2.88187973e-10 -2.78318721e-15 4.61104889e-07 8.82231267e-12 -7.61160801e-10 -1.22663137e-13 8.62322509e-07 7.55929302e-15 9.25246582e-07 -1.50714775e-14 1.39040061e-16 3.569137e-19 -1.33681833e-11 1.14548695e-18 -2.72392569e-12 -1.35365543e-17 5.87259907e-10 -1.24997244e-22 -2.66378716e-12 -1.06234202e-10 5.0158369e-06 -2.05852585e-10 -2.27069421e-13 -5.34770082e-17 -1.96842225e-16 -3.4682877e-08 -4.10390541e-11 -5.69285839e-05 4.42034391e-08 -3.9451e-13 -1.71890258e-09 1.11751786e-17 -1.21019427e-11 8.1851861e-20 -1.3248671e-09 3.7215403e-10 -1.14045842e-10 -7.67874986e-10 -1.97528998e-08 1.0911222e-12 2.15262287e-11 -2.13065503e-11 2.08921734e-11 6.1688441e-09 -4.01501723e-07 6.62606769e-09 6.3913411e-16 -2.44490157e-11 -3.65240449e-10 7.2488524e-14 -2.52732649e-31 -4.62989576e-08 -6.16322358e-12 -2.60538002e-07 -2.07835018e-16 -4.79979512e-19 -4.78922971e-07 1.81517056e-19 3.07741777e-10 1.50545465e-09 -8.97746655e-10 -3.46676203e-12 1.84175165e-16 5.50125809e-13 2.06271396e-16 -6.57223299e-13 -8.71506778e-14 4.41939109e-17 1.71635572e-09 3.0116041e-13 -1.18550884e-06 1.01984102e-10 -3.65324404e-24 -5.67517474e-16 2.69477496e-10 7.87495072e-19 8.49985499e-11 -9.90991822e-08 1.06279663e-13 1.85418081e-08 2.11086509e-10 -4.19540594e-21 -1.81273862e-07 -1.00296872e-11 -9.6594629e-13 1.02251072e-11 -3.76353097e-16 4.52699076e-12 5.37381073e-09 -1.47770844e-08 -1.43383552e-11 -1.97442019e-12 5.3871112e-08 -4.30798341e-10 -1.3083304e-12 3.9942174e-09 -7.92288368e-09 4.23331148e-09 8.94143158e-19
Add 2x model for photo scaling
2015-11-30 21:29:34 +13:00
2
Convert model files; Add new pretrained model - Add new pretrained model to ./models/upconv_7 - Move old models to ./models/vgg_7 - Use nn.LeakyReLU instead of w2nn.LeakyReLU - Add useful attribute to .json New JSON attribute: The first layer has `model_config` attribute. It contains: model_arch: architecture name of model. see `lib/srcnn.lua` scale_factor: if scale_factor > 1, model:forward() changes image resolution with scale_factor. channels: input/output channels. if channels == 3, model is RGB model. offset: pixel size that is to be removed from output. for example: (scale_factor=1, offset=7, input=100x100) => output=(100-7)x(100-7) (scale_factor=2, offset=12, input=100x100) => output=(100*2-12)x(100*2-12) And each layer has `class_name` attribute.
2016-05-15 06:04:08 +12:00
4
padH
Add 2x model for photo scaling
2015-11-30 21:29:34 +13:00
1
Convert model files; Add new pretrained model - Add new pretrained model to ./models/upconv_7 - Move old models to ./models/vgg_7 - Use nn.LeakyReLU instead of w2nn.LeakyReLU - Add useful attribute to .json New JSON attribute: The first layer has `model_config` attribute. It contains: model_arch: architecture name of model. see `lib/srcnn.lua` scale_factor: if scale_factor > 1, model:forward() changes image resolution with scale_factor. channels: input/output channels. if channels == 3, model is RGB model. offset: pixel size that is to be removed from output. for example: (scale_factor=1, offset=7, input=100x100) => output=(100-7)x(100-7) (scale_factor=2, offset=12, input=100x100) => output=(100*2-12)x(100*2-12) And each layer has `class_name` attribute.
2016-05-15 06:04:08 +12:00
0
Add 2x model for photo scaling
2015-11-30 21:29:34 +13:00
1
10
4
Convert model files; Add new pretrained model - Add new pretrained model to ./models/upconv_7 - Move old models to ./models/vgg_7 - Use nn.LeakyReLU instead of w2nn.LeakyReLU - Add useful attribute to .json New JSON attribute: The first layer has `model_config` attribute. It contains: model_arch: architecture name of model. see `lib/srcnn.lua` scale_factor: if scale_factor > 1, model:forward() changes image resolution with scale_factor. channels: input/output channels. if channels == 3, model is RGB model. offset: pixel size that is to be removed from output. for example: (scale_factor=1, offset=7, input=100x100) => output=(100-7)x(100-7) (scale_factor=2, offset=12, input=100x100) => output=(100*2-12)x(100*2-12) And each layer has `class_name` attribute.
2016-05-15 06:04:08 +12:00
81
Add 2x model for photo scaling
2015-11-30 21:29:34 +13:00
3
V 1
Convert model files; Add new pretrained model - Add new pretrained model to ./models/upconv_7 - Move old models to ./models/vgg_7 - Use nn.LeakyReLU instead of w2nn.LeakyReLU - Add useful attribute to .json New JSON attribute: The first layer has `model_config` attribute. It contains: model_arch: architecture name of model. see `lib/srcnn.lua` scale_factor: if scale_factor > 1, model:forward() changes image resolution with scale_factor. channels: input/output channels. if channels == 3, model is RGB model. offset: pixel size that is to be removed from output. for example: (scale_factor=1, offset=7, input=100x100) => output=(100-7)x(100-7) (scale_factor=2, offset=12, input=100x100) => output=(100*2-12)x(100*2-12) And each layer has `class_name` attribute.
2016-05-15 06:04:08 +12:00
12
nn.LeakyReLU
Add 2x model for photo scaling
2015-11-30 21:29:34 +13:00
3
Convert model files; Add new pretrained model - Add new pretrained model to ./models/upconv_7 - Move old models to ./models/vgg_7 - Use nn.LeakyReLU instead of w2nn.LeakyReLU - Add useful attribute to .json New JSON attribute: The first layer has `model_config` attribute. It contains: model_arch: architecture name of model. see `lib/srcnn.lua` scale_factor: if scale_factor > 1, model:forward() changes image resolution with scale_factor. channels: input/output channels. if channels == 3, model is RGB model. offset: pixel size that is to be removed from output. for example: (scale_factor=1, offset=7, input=100x100) => output=(100-7)x(100-7) (scale_factor=2, offset=12, input=100x100) => output=(100*2-12)x(100*2-12) And each layer has `class_name` attribute.
2016-05-15 06:04:08 +12:00
82
6
Add 2x model for photo scaling
2015-11-30 21:29:34 +13:00
2
Convert model files; Add new pretrained model - Add new pretrained model to ./models/upconv_7 - Move old models to ./models/vgg_7 - Use nn.LeakyReLU instead of w2nn.LeakyReLU - Add useful attribute to .json New JSON attribute: The first layer has `model_config` attribute. It contains: model_arch: architecture name of model. see `lib/srcnn.lua` scale_factor: if scale_factor > 1, model:forward() changes image resolution with scale_factor. channels: input/output channels. if channels == 3, model is RGB model. offset: pixel size that is to be removed from output. for example: (scale_factor=1, offset=7, input=100x100) => output=(100-7)x(100-7) (scale_factor=2, offset=12, input=100x100) => output=(100*2-12)x(100*2-12) And each layer has `class_name` attribute.
2016-05-15 06:04:08 +12:00
7
inplace
Add 2x model for photo scaling
2015-11-30 21:29:34 +13:00
5
Convert model files; Add new pretrained model - Add new pretrained model to ./models/upconv_7 - Move old models to ./models/vgg_7 - Use nn.LeakyReLU instead of w2nn.LeakyReLU - Add useful attribute to .json New JSON attribute: The first layer has `model_config` attribute. It contains: model_arch: architecture name of model. see `lib/srcnn.lua` scale_factor: if scale_factor > 1, model:forward() changes image resolution with scale_factor. channels: input/output channels. if channels == 3, model is RGB model. offset: pixel size that is to be removed from output. for example: (scale_factor=1, offset=7, input=100x100) => output=(100-7)x(100-7) (scale_factor=2, offset=12, input=100x100) => output=(100*2-12)x(100*2-12) And each layer has `class_name` attribute.
2016-05-15 06:04:08 +12:00
1
2
Add 2x model for photo scaling
2015-11-30 21:29:34 +13:00
5
Convert model files; Add new pretrained model - Add new pretrained model to ./models/upconv_7 - Move old models to ./models/vgg_7 - Use nn.LeakyReLU instead of w2nn.LeakyReLU - Add useful attribute to .json New JSON attribute: The first layer has `model_config` attribute. It contains: model_arch: architecture name of model. see `lib/srcnn.lua` scale_factor: if scale_factor > 1, model:forward() changes image resolution with scale_factor. channels: input/output channels. if channels == 3, model is RGB model. offset: pixel size that is to be removed from output. for example: (scale_factor=1, offset=7, input=100x100) => output=(100-7)x(100-7) (scale_factor=2, offset=12, input=100x100) => output=(100*2-12)x(100*2-12) And each layer has `class_name` attribute.
2016-05-15 06:04:08 +12:00
_type
Add 2x model for photo scaling
2015-11-30 21:29:34 +13:00
2
16
torch.CudaTensor
Convert model files; Add new pretrained model - Add new pretrained model to ./models/upconv_7 - Move old models to ./models/vgg_7 - Use nn.LeakyReLU instead of w2nn.LeakyReLU - Add useful attribute to .json New JSON attribute: The first layer has `model_config` attribute. It contains: model_arch: architecture name of model. see `lib/srcnn.lua` scale_factor: if scale_factor > 1, model:forward() changes image resolution with scale_factor. channels: input/output channels. if channels == 3, model is RGB model. offset: pixel size that is to be removed from output. for example: (scale_factor=1, offset=7, input=100x100) => output=(100-7)x(100-7) (scale_factor=2, offset=12, input=100x100) => output=(100*2-12)x(100*2-12) And each layer has `class_name` attribute.
2016-05-15 06:04:08 +12:00
2
6
negval
Add 2x model for photo scaling
2015-11-30 21:29:34 +13:00
1
Convert model files; Add new pretrained model - Add new pretrained model to ./models/upconv_7 - Move old models to ./models/vgg_7 - Use nn.LeakyReLU instead of w2nn.LeakyReLU - Add useful attribute to .json New JSON attribute: The first layer has `model_config` attribute. It contains: model_arch: architecture name of model. see `lib/srcnn.lua` scale_factor: if scale_factor > 1, model:forward() changes image resolution with scale_factor. channels: input/output channels. if channels == 3, model is RGB model. offset: pixel size that is to be removed from output. for example: (scale_factor=1, offset=7, input=100x100) => output=(100-7)x(100-7) (scale_factor=2, offset=12, input=100x100) => output=(100*2-12)x(100*2-12) And each layer has `class_name` attribute.
2016-05-15 06:04:08 +12:00
0.10000000000000001
Add 2x model for photo scaling
2015-11-30 21:29:34 +13:00
2
9
gradInput
4
Convert model files; Add new pretrained model - Add new pretrained model to ./models/upconv_7 - Move old models to ./models/vgg_7 - Use nn.LeakyReLU instead of w2nn.LeakyReLU - Add useful attribute to .json New JSON attribute: The first layer has `model_config` attribute. It contains: model_arch: architecture name of model. see `lib/srcnn.lua` scale_factor: if scale_factor > 1, model:forward() changes image resolution with scale_factor. channels: input/output channels. if channels == 3, model is RGB model. offset: pixel size that is to be removed from output. for example: (scale_factor=1, offset=7, input=100x100) => output=(100-7)x(100-7) (scale_factor=2, offset=12, input=100x100) => output=(100*2-12)x(100*2-12) And each layer has `class_name` attribute.
2016-05-15 06:04:08 +12:00
83
Add 2x model for photo scaling
2015-11-30 21:29:34 +13:00
3
V 1
16
torch.CudaTensor
0
1
0
2
6
output
4
Convert model files; Add new pretrained model - Add new pretrained model to ./models/upconv_7 - Move old models to ./models/vgg_7 - Use nn.LeakyReLU instead of w2nn.LeakyReLU - Add useful attribute to .json New JSON attribute: The first layer has `model_config` attribute. It contains: model_arch: architecture name of model. see `lib/srcnn.lua` scale_factor: if scale_factor > 1, model:forward() changes image resolution with scale_factor. channels: input/output channels. if channels == 3, model is RGB model. offset: pixel size that is to be removed from output. for example: (scale_factor=1, offset=7, input=100x100) => output=(100-7)x(100-7) (scale_factor=2, offset=12, input=100x100) => output=(100*2-12)x(100*2-12) And each layer has `class_name` attribute.
2016-05-15 06:04:08 +12:00
84
Add 2x model for photo scaling
2015-11-30 21:29:34 +13:00
3
V 1
16
torch.CudaTensor
0
1
0
2
Convert model files; Add new pretrained model - Add new pretrained model to ./models/upconv_7 - Move old models to ./models/vgg_7 - Use nn.LeakyReLU instead of w2nn.LeakyReLU - Add useful attribute to .json New JSON attribute: The first layer has `model_config` attribute. It contains: model_arch: architecture name of model. see `lib/srcnn.lua` scale_factor: if scale_factor > 1, model:forward() changes image resolution with scale_factor. channels: input/output channels. if channels == 3, model is RGB model. offset: pixel size that is to be removed from output. for example: (scale_factor=1, offset=7, input=100x100) => output=(100-7)x(100-7) (scale_factor=2, offset=12, input=100x100) => output=(100*2-12)x(100*2-12) And each layer has `class_name` attribute.
2016-05-15 06:04:08 +12:00
5
train
5
0
Add 2x model for photo scaling
2015-11-30 21:29:34 +13:00
1
11
4
Convert model files; Add new pretrained model - Add new pretrained model to ./models/upconv_7 - Move old models to ./models/vgg_7 - Use nn.LeakyReLU instead of w2nn.LeakyReLU - Add useful attribute to .json New JSON attribute: The first layer has `model_config` attribute. It contains: model_arch: architecture name of model. see `lib/srcnn.lua` scale_factor: if scale_factor > 1, model:forward() changes image resolution with scale_factor. channels: input/output channels. if channels == 3, model is RGB model. offset: pixel size that is to be removed from output. for example: (scale_factor=1, offset=7, input=100x100) => output=(100-7)x(100-7) (scale_factor=2, offset=12, input=100x100) => output=(100*2-12)x(100*2-12) And each layer has `class_name` attribute.
2016-05-15 06:04:08 +12:00
85
Add 2x model for photo scaling
2015-11-30 21:29:34 +13:00
3
V 1
23
nn.SpatialConvolutionMM
3
Convert model files; Add new pretrained model - Add new pretrained model to ./models/upconv_7 - Move old models to ./models/vgg_7 - Use nn.LeakyReLU instead of w2nn.LeakyReLU - Add useful attribute to .json New JSON attribute: The first layer has `model_config` attribute. It contains: model_arch: architecture name of model. see `lib/srcnn.lua` scale_factor: if scale_factor > 1, model:forward() changes image resolution with scale_factor. channels: input/output channels. if channels == 3, model is RGB model. offset: pixel size that is to be removed from output. for example: (scale_factor=1, offset=7, input=100x100) => output=(100-7)x(100-7) (scale_factor=2, offset=12, input=100x100) => output=(100*2-12)x(100*2-12) And each layer has `class_name` attribute.
2016-05-15 06:04:08 +12:00
86
16
Add 2x model for photo scaling
2015-11-30 21:29:34 +13:00
2
Convert model files; Add new pretrained model - Add new pretrained model to ./models/upconv_7 - Move old models to ./models/vgg_7 - Use nn.LeakyReLU instead of w2nn.LeakyReLU - Add useful attribute to .json New JSON attribute: The first layer has `model_config` attribute. It contains: model_arch: architecture name of model. see `lib/srcnn.lua` scale_factor: if scale_factor > 1, model:forward() changes image resolution with scale_factor. channels: input/output channels. if channels == 3, model is RGB model. offset: pixel size that is to be removed from output. for example: (scale_factor=1, offset=7, input=100x100) => output=(100-7)x(100-7) (scale_factor=2, offset=12, input=100x100) => output=(100*2-12)x(100*2-12) And each layer has `class_name` attribute.
2016-05-15 06:04:08 +12:00
2
dH
1
1
2
2
dW
1
Add 2x model for photo scaling
2015-11-30 21:29:34 +13:00
1
2
11
nInputPlane
1
128
2
6
output
4
Convert model files; Add new pretrained model - Add new pretrained model to ./models/upconv_7 - Move old models to ./models/vgg_7 - Use nn.LeakyReLU instead of w2nn.LeakyReLU - Add useful attribute to .json New JSON attribute: The first layer has `model_config` attribute. It contains: model_arch: architecture name of model. see `lib/srcnn.lua` scale_factor: if scale_factor > 1, model:forward() changes image resolution with scale_factor. channels: input/output channels. if channels == 3, model is RGB model. offset: pixel size that is to be removed from output. for example: (scale_factor=1, offset=7, input=100x100) => output=(100-7)x(100-7) (scale_factor=2, offset=12, input=100x100) => output=(100*2-12)x(100*2-12) And each layer has `class_name` attribute.
2016-05-15 06:04:08 +12:00
87
Add 2x model for photo scaling
2015-11-30 21:29:34 +13:00
3
V 1
16
torch.CudaTensor
0
1
0
2
Convert model files; Add new pretrained model - Add new pretrained model to ./models/upconv_7 - Move old models to ./models/vgg_7 - Use nn.LeakyReLU instead of w2nn.LeakyReLU - Add useful attribute to .json New JSON attribute: The first layer has `model_config` attribute. It contains: model_arch: architecture name of model. see `lib/srcnn.lua` scale_factor: if scale_factor > 1, model:forward() changes image resolution with scale_factor. channels: input/output channels. if channels == 3, model is RGB model. offset: pixel size that is to be removed from output. for example: (scale_factor=1, offset=7, input=100x100) => output=(100-7)x(100-7) (scale_factor=2, offset=12, input=100x100) => output=(100*2-12)x(100*2-12) And each layer has `class_name` attribute.
2016-05-15 06:04:08 +12:00
2
kH
1
3
2
Add 2x model for photo scaling
2015-11-30 21:29:34 +13:00
9
gradInput
4
Convert model files; Add new pretrained model - Add new pretrained model to ./models/upconv_7 - Move old models to ./models/vgg_7 - Use nn.LeakyReLU instead of w2nn.LeakyReLU - Add useful attribute to .json New JSON attribute: The first layer has `model_config` attribute. It contains: model_arch: architecture name of model. see `lib/srcnn.lua` scale_factor: if scale_factor > 1, model:forward() changes image resolution with scale_factor. channels: input/output channels. if channels == 3, model is RGB model. offset: pixel size that is to be removed from output. for example: (scale_factor=1, offset=7, input=100x100) => output=(100-7)x(100-7) (scale_factor=2, offset=12, input=100x100) => output=(100*2-12)x(100*2-12) And each layer has `class_name` attribute.
2016-05-15 06:04:08 +12:00
88
Add 2x model for photo scaling
2015-11-30 21:29:34 +13:00
3
V 1
16
torch.CudaTensor
0
1
0
2
2
Convert model files; Add new pretrained model - Add new pretrained model to ./models/upconv_7 - Move old models to ./models/vgg_7 - Use nn.LeakyReLU instead of w2nn.LeakyReLU - Add useful attribute to .json New JSON attribute: The first layer has `model_config` attribute. It contains: model_arch: architecture name of model. see `lib/srcnn.lua` scale_factor: if scale_factor > 1, model:forward() changes image resolution with scale_factor. channels: input/output channels. if channels == 3, model is RGB model. offset: pixel size that is to be removed from output. for example: (scale_factor=1, offset=7, input=100x100) => output=(100-7)x(100-7) (scale_factor=2, offset=12, input=100x100) => output=(100*2-12)x(100*2-12) And each layer has `class_name` attribute.
2016-05-15 06:04:08 +12:00
kW
Add 2x model for photo scaling
2015-11-30 21:29:34 +13:00
1
Convert model files; Add new pretrained model - Add new pretrained model to ./models/upconv_7 - Move old models to ./models/vgg_7 - Use nn.LeakyReLU instead of w2nn.LeakyReLU - Add useful attribute to .json New JSON attribute: The first layer has `model_config` attribute. It contains: model_arch: architecture name of model. see `lib/srcnn.lua` scale_factor: if scale_factor > 1, model:forward() changes image resolution with scale_factor. channels: input/output channels. if channels == 3, model is RGB model. offset: pixel size that is to be removed from output. for example: (scale_factor=1, offset=7, input=100x100) => output=(100-7)x(100-7) (scale_factor=2, offset=12, input=100x100) => output=(100*2-12)x(100*2-12) And each layer has `class_name` attribute.
2016-05-15 06:04:08 +12:00
3
Add 2x model for photo scaling
2015-11-30 21:29:34 +13:00
2
12
nOutputPlane
1
128
2
4
Convert model files; Add new pretrained model - Add new pretrained model to ./models/upconv_7 - Move old models to ./models/vgg_7 - Use nn.LeakyReLU instead of w2nn.LeakyReLU - Add useful attribute to .json New JSON attribute: The first layer has `model_config` attribute. It contains: model_arch: architecture name of model. see `lib/srcnn.lua` scale_factor: if scale_factor > 1, model:forward() changes image resolution with scale_factor. channels: input/output channels. if channels == 3, model is RGB model. offset: pixel size that is to be removed from output. for example: (scale_factor=1, offset=7, input=100x100) => output=(100-7)x(100-7) (scale_factor=2, offset=12, input=100x100) => output=(100*2-12)x(100*2-12) And each layer has `class_name` attribute.
2016-05-15 06:04:08 +12:00
padW
Add 2x model for photo scaling
2015-11-30 21:29:34 +13:00
1
0
2
6
weight
4
Convert model files; Add new pretrained model - Add new pretrained model to ./models/upconv_7 - Move old models to ./models/vgg_7 - Use nn.LeakyReLU instead of w2nn.LeakyReLU - Add useful attribute to .json New JSON attribute: The first layer has `model_config` attribute. It contains: model_arch: architecture name of model. see `lib/srcnn.lua` scale_factor: if scale_factor > 1, model:forward() changes image resolution with scale_factor. channels: input/output channels. if channels == 3, model is RGB model. offset: pixel size that is to be removed from output. for example: (scale_factor=1, offset=7, input=100x100) => output=(100-7)x(100-7) (scale_factor=2, offset=12, input=100x100) => output=(100*2-12)x(100*2-12) And each layer has `class_name` attribute.
2016-05-15 06:04:08 +12:00
89
Add 2x model for photo scaling
2015-11-30 21:29:34 +13:00
3
V 1
16
torch.CudaTensor
2
128 1152
1152 1
1
4
Convert model files; Add new pretrained model - Add new pretrained model to ./models/upconv_7 - Move old models to ./models/vgg_7 - Use nn.LeakyReLU instead of w2nn.LeakyReLU - Add useful attribute to .json New JSON attribute: The first layer has `model_config` attribute. It contains: model_arch: architecture name of model. see `lib/srcnn.lua` scale_factor: if scale_factor > 1, model:forward() changes image resolution with scale_factor. channels: input/output channels. if channels == 3, model is RGB model. offset: pixel size that is to be removed from output. for example: (scale_factor=1, offset=7, input=100x100) => output=(100-7)x(100-7) (scale_factor=2, offset=12, input=100x100) => output=(100*2-12)x(100*2-12) And each layer has `class_name` attribute.
2016-05-15 06:04:08 +12:00
90
Add 2x model for photo scaling
2015-11-30 21:29:34 +13:00
3
V 1
17
torch.CudaStorage
147456
0.0209450889 -0.0680612698 0.0995793417 0.0606686324 0.0781204849 -0.0118010882 -0.0667781457 -0.0248299967 -0.0375632271 -0.0729407072 -0.119927682 -0.133100197 -0.066974856 -0.00271027931 0.00345312455 0.0443625301 0.0744522139 -0.0457487702 0.070975475 0.101454012 0.039276734 0.00609189784 0.0221349783 -0.0111012645 -0.0296104345 -0.0508304201 0.0786588341 -0.0369297676 0.0895602852 0.0290802233 0.112912014 0.0276085883 0.0352013148 -0.0171577353 -0.0186377726 -0.042302195 -0.156927824 -0.0626143292 0.0135121522 0.110128716 0.154390767 0.228338972 0.0646578223 0.244581103 0.181153908 0.028365545 -0.0539438613 -0.0584402494 0.0135140959 -0.0468157642 -0.0387293734 -0.0281136893 -0.0448254459 -0.0386305153 0.0616219081 0.0927674398 0.0222622789 -0.0671736002 -0.0109353531 0.000167718434 -0.0651241168 0.0665301681 -0.0155721614 0.100278847 0.201317191 0.161036387 -0.00957399793 0.0219791178 0.0264200065 0.111782931 0.0716942102 0.131853878 0.0264479928 0.0155562581 0.0413758084 0.000657072233 -0.0109138712 -0.173427716 0.0997290164 -0.00651620794 0.0336814076 -0.0353060961 0.0987005383 0.0234957505 0.108313479 -0.0588228442 0.0455007814 -0.0196255017 -0.0426573455 0.00680462737 -0.0194061566 -0.0169357639 -0.0334721096 0.0329716951 -0.0192986559 0.00368339475 0.120016031 0.0677292943 0.087065056 0.0671332702 0.125541076 0.0742571652 0.0463878512 -0.0117257731 0.0112270098 0.117838979 -0.0834989622 -0.0366101153 0.0565867908 -0.0219243877 0.0152381631 -0.00377073465 -0.0660851449 -0.089033328 -0.0715203509 -0.0665586889 -0.0876455009 -0.0385540351 0.0471786596 0.0219816826 0.0875851437 0.0714678541 0.0924535468 0.0376763605 0.0688811317 0.0441766568 0.0574014708 0.067547448 0.0111806421 0.118498169 0.100391932 0.0112185366 0.0490432903 0.10557995 0.0364885405 0.0480024666 0.0496687032 0.0266213845 -0.00976179261 0.0833884478 -0.0193608012 0.0255650692 -0.0134196961 -0.014939012 0.00373313902 -0.011477272 -0.00102845719 0.0452211834 0.00692166155 -0.12049944 -0.0107782604 -0.0488071442 -0.0317309201 0.0540469922 -0.0306088682 0.0257801004 -0.0259699542 -0.0295664705 -0.0304162446 -0.0241553392 -0.121525027 -0.113990866 -0.0041331742 -0.0365197174 -0.0110384524 0.0683944151 0.010585689 0.12713334 0.0466411151 0.134186342 0.00339263747 -0.0484464914 0.0260125343 0.109310985 0.0848336816 -0.0135357175 0.0112314373 -0.0484677218 0.025997011 -0.0365852453 0.0384349748 0.0263503827 0.0177860577 -0.0531170778 -0.018202439 -0.00503918063 -0.0908312574 -0.0328697748 -0.0909673795 -0.0955293551 -0.0283295698 -0.0054457169 -0.050709825 -0.069436267 -0.0166110005 -0.0834354982 -0.0562432781 0.01278097 -0.0205118898 -0.0142334336 -0.0691833049 -0.0321115144 -0.0255934522 -0.084406212 -0.143022418 -0.0962692872 -0.0568019375 -0.0590399727 -0.0743435249 0.0267024841 0.0391814858 -0.0875503272 -0.0420785099 -0.0371035077 0.232473254 0.140685961 -0.0669303238 -0.0466331169 0.0227085054 -0.0245551504 0.015664639 -0.0730091855 -0.0664128512 0.00249818573 -0.051219698 0.119953744 0.0692355856 0.0316974521 0.0799955279 0.0948783532 0.0476122573 0.0197886098 0.117258176 -0.0228830371 -0.0408276245 -0.0383763239 -0.0874627531 0.0425309427 0.00329649285 -0.0107126981 -0.081404455 -0.041870378 -0.0158664212 0.163562223 -0.000999161741 0.0626195744 -0.0275047515 -0.0580612756 0.0124451667 -0.0924901366 -0.0582129508 -0.119301811 -0.0533644706 0.011512165 0.054929588 -0.00889949221 0.0502929427 0.0848515108 0.0197472423 0.0314259231 0.0912784263 -0.00791361276 0.059108492 0.0737384111 -0.0834869146 -0.0889248624 0.0404304154 -0.0053745294 0.0340068638 0.0666274652 -0.0843068883 -0.137163296 -0.0276424177 0.0612729192 -0.0479097515 0.0232102722 0.131889746 0.101950623 0.0327859484 -0.0615075082 -0.0259109754 0.00276945601 -0.0287455767 -0.0388553478 -0.0333706997 -0.0500964746 -0.0221340861 0.0130602261 0.0480656214 0.0724817291 0.113888741 -0.0143577401 -0.0997292325 -0.0508677214 -0.102572046 -0.0749782771 -0.105266288 -0.0651893541 0.0632002503 0.0369376503 -0.00356589444 -0.00545265945 -0.00169220474 -0.0103195291 -0.091439873 0.0520201065 -0.03
2
5
train
5
0
2
10
gradWeight
4
Convert model files; Add new pretrained model - Add new pretrained model to ./models/upconv_7 - Move old models to ./models/vgg_7 - Use nn.LeakyReLU instead of w2nn.LeakyReLU - Add useful attribute to .json New JSON attribute: The first layer has `model_config` attribute. It contains: model_arch: architecture name of model. see `lib/srcnn.lua` scale_factor: if scale_factor > 1, model:forward() changes image resolution with scale_factor. channels: input/output channels. if channels == 3, model is RGB model. offset: pixel size that is to be removed from output. for example: (scale_factor=1, offset=7, input=100x100) => output=(100-7)x(100-7) (scale_factor=2, offset=12, input=100x100) => output=(100*2-12)x(100*2-12) And each layer has `class_name` attribute.
2016-05-15 06:04:08 +12:00
91
Add 2x model for photo scaling
2015-11-30 21:29:34 +13:00
3
V 1
16
torch.CudaTensor
2
128 1152
1152 1
1
4
Convert model files; Add new pretrained model - Add new pretrained model to ./models/upconv_7 - Move old models to ./models/vgg_7 - Use nn.LeakyReLU instead of w2nn.LeakyReLU - Add useful attribute to .json New JSON attribute: The first layer has `model_config` attribute. It contains: model_arch: architecture name of model. see `lib/srcnn.lua` scale_factor: if scale_factor > 1, model:forward() changes image resolution with scale_factor. channels: input/output channels. if channels == 3, model is RGB model. offset: pixel size that is to be removed from output. for example: (scale_factor=1, offset=7, input=100x100) => output=(100-7)x(100-7) (scale_factor=2, offset=12, input=100x100) => output=(100*2-12)x(100*2-12) And each layer has `class_name` attribute.
2016-05-15 06:04:08 +12:00
92
Add 2x model for photo scaling
2015-11-30 21:29:34 +13:00
3
V 1
17
torch.CudaStorage
147456
0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0
2
Convert model files; Add new pretrained model - Add new pretrained model to ./models/upconv_7 - Move old models to ./models/vgg_7 - Use nn.LeakyReLU instead of w2nn.LeakyReLU - Add useful attribute to .json New JSON attribute: The first layer has `model_config` attribute. It contains: model_arch: architecture name of model. see `lib/srcnn.lua` scale_factor: if scale_factor > 1, model:forward() changes image resolution with scale_factor. channels: input/output channels. if channels == 3, model is RGB model. offset: pixel size that is to be removed from output. for example: (scale_factor=1, offset=7, input=100x100) => output=(100-7)x(100-7) (scale_factor=2, offset=12, input=100x100) => output=(100*2-12)x(100*2-12) And each layer has `class_name` attribute.
2016-05-15 06:04:08 +12:00
5
_type
2
16
torch.CudaTensor
2
Add 2x model for photo scaling
2015-11-30 21:29:34 +13:00
4
Convert model files; Add new pretrained model - Add new pretrained model to ./models/upconv_7 - Move old models to ./models/vgg_7 - Use nn.LeakyReLU instead of w2nn.LeakyReLU - Add useful attribute to .json New JSON attribute: The first layer has `model_config` attribute. It contains: model_arch: architecture name of model. see `lib/srcnn.lua` scale_factor: if scale_factor > 1, model:forward() changes image resolution with scale_factor. channels: input/output channels. if channels == 3, model is RGB model. offset: pixel size that is to be removed from output. for example: (scale_factor=1, offset=7, input=100x100) => output=(100-7)x(100-7) (scale_factor=2, offset=12, input=100x100) => output=(100*2-12)x(100*2-12) And each layer has `class_name` attribute.
2016-05-15 06:04:08 +12:00
bias
4
93
Add 2x model for photo scaling
2015-11-30 21:29:34 +13:00
3
V 1
16
torch.CudaTensor
1
Convert model files; Add new pretrained model - Add new pretrained model to ./models/upconv_7 - Move old models to ./models/vgg_7 - Use nn.LeakyReLU instead of w2nn.LeakyReLU - Add useful attribute to .json New JSON attribute: The first layer has `model_config` attribute. It contains: model_arch: architecture name of model. see `lib/srcnn.lua` scale_factor: if scale_factor > 1, model:forward() changes image resolution with scale_factor. channels: input/output channels. if channels == 3, model is RGB model. offset: pixel size that is to be removed from output. for example: (scale_factor=1, offset=7, input=100x100) => output=(100-7)x(100-7) (scale_factor=2, offset=12, input=100x100) => output=(100*2-12)x(100*2-12) And each layer has `class_name` attribute.
2016-05-15 06:04:08 +12:00
128
1
1
4
94
3
V 1
17
torch.CudaStorage
128
-0.00379276159 -0.0808914602 -0.0759049132 0.0129945539 -0.110297836 -0.0174011104 -0.0445540212 -0.0444122106 -0.0824233592 -0.0133075006 -0.0534983762 -0.00101549714 0.00250795437 -0.0165236518 -0.041742418 -0.011781794 0.0194729213 -0.0396952815 -0.0305467993 0.0153996376 0.0901807174 -0.0169309415 -0.037387725 0.0116884476 0.00130114146 -0.0729780644 -0.0519642197 0.015921399 0.00115926366 -0.0239332039 -0.00140202767 -0.0305543244 -0.0766137466 -0.0211149566 -0.00630454626 -0.0360609703 0.0109216589 0.0236540958 0.0215929374 0.0170814861 -0.0293169636 -0.0156904012 -0.0979253426 -0.0144581757 0.00229684659 0.021335993 0.0212035608 -0.00257379352 -0.00606245361 0.00455210544 -0.0100096688 -0.000223904441 -0.0276127271 -0.0954380855 0.0176942423 -0.0327945687 0.109508641 -0.0183703862 0.0139431646 0.00407862803 -0.111115374 -0.0927176997 -0.00943775941 -0.0413400047 -0.0044927639 -0.0307077542 -0.0864248723 -0.0899788737 -0.341241002 -0.0267093945 0.0634841025 0.0124915969 -0.0175637789 -0.0415981524 -0.0317761339 -0.0394420698 0.00159198535 -0.089641653 0.00133293599 -0.00417169742 0.0331192389 -0.0120884087 -0.000298509025 -0.0231348164 -0.0446313545 -0.0868544728 0.0321557447 0.0134520261 -0.103432551 0.000787280675 -0.0554386936 0.0258215498 -0.0444543511 -0.0803946853 -0.0894080997 0.010607616 -0.0321705826 0.00500020664 -0.0349902064 -0.0445353314 -0.0130315218 -0.0408871956 0.040136423 -0.115254611 0.00990788545 -0.130081117 -0.104747251 -0.00865302421 -0.00541918678 -0.025852656 0.00420172745 -0.118859418 -0.00447401358 0.0151451482 -0.012214425 -0.0045935181 -0.021698758 -0.0726601258 -0.060676083 -0.0234303493 -0.0597750507 -0.0113571985 -0.00645807199 0.0134687889 0.00161582499 -0.0109407594 -0.0241621006 0.0173461568
Add 2x model for photo scaling
2015-11-30 21:29:34 +13:00
2
Convert model files; Add new pretrained model - Add new pretrained model to ./models/upconv_7 - Move old models to ./models/vgg_7 - Use nn.LeakyReLU instead of w2nn.LeakyReLU - Add useful attribute to .json New JSON attribute: The first layer has `model_config` attribute. It contains: model_arch: architecture name of model. see `lib/srcnn.lua` scale_factor: if scale_factor > 1, model:forward() changes image resolution with scale_factor. channels: input/output channels. if channels == 3, model is RGB model. offset: pixel size that is to be removed from output. for example: (scale_factor=1, offset=7, input=100x100) => output=(100-7)x(100-7) (scale_factor=2, offset=12, input=100x100) => output=(100*2-12)x(100*2-12) And each layer has `class_name` attribute.
2016-05-15 06:04:08 +12:00
8
gradBias
Add 2x model for photo scaling
2015-11-30 21:29:34 +13:00
4
Convert model files; Add new pretrained model - Add new pretrained model to ./models/upconv_7 - Move old models to ./models/vgg_7 - Use nn.LeakyReLU instead of w2nn.LeakyReLU - Add useful attribute to .json New JSON attribute: The first layer has `model_config` attribute. It contains: model_arch: architecture name of model. see `lib/srcnn.lua` scale_factor: if scale_factor > 1, model:forward() changes image resolution with scale_factor. channels: input/output channels. if channels == 3, model is RGB model. offset: pixel size that is to be removed from output. for example: (scale_factor=1, offset=7, input=100x100) => output=(100-7)x(100-7) (scale_factor=2, offset=12, input=100x100) => output=(100*2-12)x(100*2-12) And each layer has `class_name` attribute.
2016-05-15 06:04:08 +12:00
95
3
V 1
16
torch.CudaTensor
Add 2x model for photo scaling
2015-11-30 21:29:34 +13:00
1
Convert model files; Add new pretrained model - Add new pretrained model to ./models/upconv_7 - Move old models to ./models/vgg_7 - Use nn.LeakyReLU instead of w2nn.LeakyReLU - Add useful attribute to .json New JSON attribute: The first layer has `model_config` attribute. It contains: model_arch: architecture name of model. see `lib/srcnn.lua` scale_factor: if scale_factor > 1, model:forward() changes image resolution with scale_factor. channels: input/output channels. if channels == 3, model is RGB model. offset: pixel size that is to be removed from output. for example: (scale_factor=1, offset=7, input=100x100) => output=(100-7)x(100-7) (scale_factor=2, offset=12, input=100x100) => output=(100*2-12)x(100*2-12) And each layer has `class_name` attribute.
2016-05-15 06:04:08 +12:00
128
Add 2x model for photo scaling
2015-11-30 21:29:34 +13:00
1
1
Convert model files; Add new pretrained model - Add new pretrained model to ./models/upconv_7 - Move old models to ./models/vgg_7 - Use nn.LeakyReLU instead of w2nn.LeakyReLU - Add useful attribute to .json New JSON attribute: The first layer has `model_config` attribute. It contains: model_arch: architecture name of model. see `lib/srcnn.lua` scale_factor: if scale_factor > 1, model:forward() changes image resolution with scale_factor. channels: input/output channels. if channels == 3, model is RGB model. offset: pixel size that is to be removed from output. for example: (scale_factor=1, offset=7, input=100x100) => output=(100-7)x(100-7) (scale_factor=2, offset=12, input=100x100) => output=(100*2-12)x(100*2-12) And each layer has `class_name` attribute.
2016-05-15 06:04:08 +12:00
4
96
3
V 1
17
torch.CudaStorage
128
Add upconv_7/photo models. Update vgg_7/photo/noise* models
2016-07-04 07:51:22 +12:00
0 0 0 0 5.03309949e-10 -1.63593944e-10 3.22839325e-11 4.15044721e-10 -7.82364822e-08 5.17749797e-13 3.24385718e-09 9.98239413e-18 1.01044678e-14 -4.75474867e-22 -2.72812025e-12 -2.47818051e-11 5.87924067e-12 1.58995132e-13 -5.23499537e-08 4.7778892e-09 6.81021181e-15 -6.28725627e-10 4.34394812e-11 1.58809961e-08 -3.59061858e-07 -4.21433299e-07 2.22281642e-06 2.69688542e-07 5.67835969e-12 9.85988445e-05 -1.21970934e-05 2.20464678e-13 9.24063159e-10 -6.81736879e-07 4.94425704e-22 -2.73634707e-19 -1.26924347e-14 -1.39903955e-06 -3.52404541e-08 2.99384453e-07 2.65956521e-14 2.23662998e-11 -7.76655758e-13 9.11770104e-10 -5.55966883e-17 1.43151608e-18 8.03676613e-16 1.30341306e-07 -2.3704164e-08 2.48510652e-08 -5.44737031e-06 6.63248016e-12 -2.47281093e-07 3.79535948e-07 2.80561185e-10 8.93121514e-12 1.16249635e-10 5.41629155e-12 -2.35573025e-19 -4.29963443e-12 -1.49970626e-11 -9.42782771e-14 9.52708017e-21 -1.18524548e-11 -5.56833601e-09 2.08647899e-09 1.60879239e-07 7.0593444e-07 1.00097239e-07 2.00263065e-13 -8.2819372e-07 4.95139307e-09 -8.35563583e-07 6.52099766e-14 4.23663493e-10 6.8073118e-18 -2.19634944e-12 -3.40888207e-08 -1.21376111e-14 -1.15290266e-08 4.72123118e-10 -6.74992966e-14 -4.9710718e-07 2.3865848e-08 -1.38151512e-09 -2.09910382e-07 -4.44277987e-07 -3.43432191e-07 -6.66388047e-08 7.7798419e-25 7.36081233e-07 8.15616801e-08 -1.43729831e-05 6.32270676e-05 -1.66824407e-06 -8.07901142e-12 -1.63925478e-07 -2.06005742e-12 1.41021556e-10 -1.27317406e-07 3.10525599e-14 -2.4900902e-07 -1.48124446e-08 -6.23606979e-07 -2.43752128e-14 6.62747059e-16 4.83977565e-05 -3.03172461e-12 5.22730113e-30 3.22293614e-09 -2.69607363e-21 -1.57890606e-12 -1.52438756e-10 -3.29291983e-10 7.46069695e-09 -1.85623071e-06 8.45901447e-14 2.26286005e-07 -5.49706489e-16 2.30182508e-07 1.98609081e-11 -1.19102315e-11 -1.42537604e-09 -1.00274181e-11 4.20633806e-10 -8.7957211e-11 2.23603941e-10 2.87458342e-08
Add 2x model for photo scaling
2015-11-30 21:29:34 +13:00
2
Convert model files; Add new pretrained model - Add new pretrained model to ./models/upconv_7 - Move old models to ./models/vgg_7 - Use nn.LeakyReLU instead of w2nn.LeakyReLU - Add useful attribute to .json New JSON attribute: The first layer has `model_config` attribute. It contains: model_arch: architecture name of model. see `lib/srcnn.lua` scale_factor: if scale_factor > 1, model:forward() changes image resolution with scale_factor. channels: input/output channels. if channels == 3, model is RGB model. offset: pixel size that is to be removed from output. for example: (scale_factor=1, offset=7, input=100x100) => output=(100-7)x(100-7) (scale_factor=2, offset=12, input=100x100) => output=(100*2-12)x(100*2-12) And each layer has `class_name` attribute.
2016-05-15 06:04:08 +12:00
4
padH
Add 2x model for photo scaling
2015-11-30 21:29:34 +13:00
1
Convert model files; Add new pretrained model - Add new pretrained model to ./models/upconv_7 - Move old models to ./models/vgg_7 - Use nn.LeakyReLU instead of w2nn.LeakyReLU - Add useful attribute to .json New JSON attribute: The first layer has `model_config` attribute. It contains: model_arch: architecture name of model. see `lib/srcnn.lua` scale_factor: if scale_factor > 1, model:forward() changes image resolution with scale_factor. channels: input/output channels. if channels == 3, model is RGB model. offset: pixel size that is to be removed from output. for example: (scale_factor=1, offset=7, input=100x100) => output=(100-7)x(100-7) (scale_factor=2, offset=12, input=100x100) => output=(100*2-12)x(100*2-12) And each layer has `class_name` attribute.
2016-05-15 06:04:08 +12:00
0
Add 2x model for photo scaling
2015-11-30 21:29:34 +13:00
1
12
4
Convert model files; Add new pretrained model - Add new pretrained model to ./models/upconv_7 - Move old models to ./models/vgg_7 - Use nn.LeakyReLU instead of w2nn.LeakyReLU - Add useful attribute to .json New JSON attribute: The first layer has `model_config` attribute. It contains: model_arch: architecture name of model. see `lib/srcnn.lua` scale_factor: if scale_factor > 1, model:forward() changes image resolution with scale_factor. channels: input/output channels. if channels == 3, model is RGB model. offset: pixel size that is to be removed from output. for example: (scale_factor=1, offset=7, input=100x100) => output=(100-7)x(100-7) (scale_factor=2, offset=12, input=100x100) => output=(100*2-12)x(100*2-12) And each layer has `class_name` attribute.
2016-05-15 06:04:08 +12:00
97
Add 2x model for photo scaling
2015-11-30 21:29:34 +13:00
3
V 1
Convert model files; Add new pretrained model - Add new pretrained model to ./models/upconv_7 - Move old models to ./models/vgg_7 - Use nn.LeakyReLU instead of w2nn.LeakyReLU - Add useful attribute to .json New JSON attribute: The first layer has `model_config` attribute. It contains: model_arch: architecture name of model. see `lib/srcnn.lua` scale_factor: if scale_factor > 1, model:forward() changes image resolution with scale_factor. channels: input/output channels. if channels == 3, model is RGB model. offset: pixel size that is to be removed from output. for example: (scale_factor=1, offset=7, input=100x100) => output=(100-7)x(100-7) (scale_factor=2, offset=12, input=100x100) => output=(100*2-12)x(100*2-12) And each layer has `class_name` attribute.
2016-05-15 06:04:08 +12:00
12
nn.LeakyReLU
Add 2x model for photo scaling
2015-11-30 21:29:34 +13:00
3
Convert model files; Add new pretrained model - Add new pretrained model to ./models/upconv_7 - Move old models to ./models/vgg_7 - Use nn.LeakyReLU instead of w2nn.LeakyReLU - Add useful attribute to .json New JSON attribute: The first layer has `model_config` attribute. It contains: model_arch: architecture name of model. see `lib/srcnn.lua` scale_factor: if scale_factor > 1, model:forward() changes image resolution with scale_factor. channels: input/output channels. if channels == 3, model is RGB model. offset: pixel size that is to be removed from output. for example: (scale_factor=1, offset=7, input=100x100) => output=(100-7)x(100-7) (scale_factor=2, offset=12, input=100x100) => output=(100*2-12)x(100*2-12) And each layer has `class_name` attribute.
2016-05-15 06:04:08 +12:00
98
6
Add 2x model for photo scaling
2015-11-30 21:29:34 +13:00
2
Convert model files; Add new pretrained model - Add new pretrained model to ./models/upconv_7 - Move old models to ./models/vgg_7 - Use nn.LeakyReLU instead of w2nn.LeakyReLU - Add useful attribute to .json New JSON attribute: The first layer has `model_config` attribute. It contains: model_arch: architecture name of model. see `lib/srcnn.lua` scale_factor: if scale_factor > 1, model:forward() changes image resolution with scale_factor. channels: input/output channels. if channels == 3, model is RGB model. offset: pixel size that is to be removed from output. for example: (scale_factor=1, offset=7, input=100x100) => output=(100-7)x(100-7) (scale_factor=2, offset=12, input=100x100) => output=(100*2-12)x(100*2-12) And each layer has `class_name` attribute.
2016-05-15 06:04:08 +12:00
7
inplace
Add 2x model for photo scaling
2015-11-30 21:29:34 +13:00
5
Convert model files; Add new pretrained model - Add new pretrained model to ./models/upconv_7 - Move old models to ./models/vgg_7 - Use nn.LeakyReLU instead of w2nn.LeakyReLU - Add useful attribute to .json New JSON attribute: The first layer has `model_config` attribute. It contains: model_arch: architecture name of model. see `lib/srcnn.lua` scale_factor: if scale_factor > 1, model:forward() changes image resolution with scale_factor. channels: input/output channels. if channels == 3, model is RGB model. offset: pixel size that is to be removed from output. for example: (scale_factor=1, offset=7, input=100x100) => output=(100-7)x(100-7) (scale_factor=2, offset=12, input=100x100) => output=(100*2-12)x(100*2-12) And each layer has `class_name` attribute.
2016-05-15 06:04:08 +12:00
1
2
Add 2x model for photo scaling
2015-11-30 21:29:34 +13:00
5
Convert model files; Add new pretrained model - Add new pretrained model to ./models/upconv_7 - Move old models to ./models/vgg_7 - Use nn.LeakyReLU instead of w2nn.LeakyReLU - Add useful attribute to .json New JSON attribute: The first layer has `model_config` attribute. It contains: model_arch: architecture name of model. see `lib/srcnn.lua` scale_factor: if scale_factor > 1, model:forward() changes image resolution with scale_factor. channels: input/output channels. if channels == 3, model is RGB model. offset: pixel size that is to be removed from output. for example: (scale_factor=1, offset=7, input=100x100) => output=(100-7)x(100-7) (scale_factor=2, offset=12, input=100x100) => output=(100*2-12)x(100*2-12) And each layer has `class_name` attribute.
2016-05-15 06:04:08 +12:00
_type
Add 2x model for photo scaling
2015-11-30 21:29:34 +13:00
2
16
torch.CudaTensor
Convert model files; Add new pretrained model - Add new pretrained model to ./models/upconv_7 - Move old models to ./models/vgg_7 - Use nn.LeakyReLU instead of w2nn.LeakyReLU - Add useful attribute to .json New JSON attribute: The first layer has `model_config` attribute. It contains: model_arch: architecture name of model. see `lib/srcnn.lua` scale_factor: if scale_factor > 1, model:forward() changes image resolution with scale_factor. channels: input/output channels. if channels == 3, model is RGB model. offset: pixel size that is to be removed from output. for example: (scale_factor=1, offset=7, input=100x100) => output=(100-7)x(100-7) (scale_factor=2, offset=12, input=100x100) => output=(100*2-12)x(100*2-12) And each layer has `class_name` attribute.
2016-05-15 06:04:08 +12:00
2
6
negval
Add 2x model for photo scaling
2015-11-30 21:29:34 +13:00
1
Convert model files; Add new pretrained model - Add new pretrained model to ./models/upconv_7 - Move old models to ./models/vgg_7 - Use nn.LeakyReLU instead of w2nn.LeakyReLU - Add useful attribute to .json New JSON attribute: The first layer has `model_config` attribute. It contains: model_arch: architecture name of model. see `lib/srcnn.lua` scale_factor: if scale_factor > 1, model:forward() changes image resolution with scale_factor. channels: input/output channels. if channels == 3, model is RGB model. offset: pixel size that is to be removed from output. for example: (scale_factor=1, offset=7, input=100x100) => output=(100-7)x(100-7) (scale_factor=2, offset=12, input=100x100) => output=(100*2-12)x(100*2-12) And each layer has `class_name` attribute.
2016-05-15 06:04:08 +12:00
0.10000000000000001
Add 2x model for photo scaling
2015-11-30 21:29:34 +13:00
2
9
gradInput
4
Convert model files; Add new pretrained model - Add new pretrained model to ./models/upconv_7 - Move old models to ./models/vgg_7 - Use nn.LeakyReLU instead of w2nn.LeakyReLU - Add useful attribute to .json New JSON attribute: The first layer has `model_config` attribute. It contains: model_arch: architecture name of model. see `lib/srcnn.lua` scale_factor: if scale_factor > 1, model:forward() changes image resolution with scale_factor. channels: input/output channels. if channels == 3, model is RGB model. offset: pixel size that is to be removed from output. for example: (scale_factor=1, offset=7, input=100x100) => output=(100-7)x(100-7) (scale_factor=2, offset=12, input=100x100) => output=(100*2-12)x(100*2-12) And each layer has `class_name` attribute.
2016-05-15 06:04:08 +12:00
99
Add 2x model for photo scaling
2015-11-30 21:29:34 +13:00
3
V 1
16
torch.CudaTensor
0
1
0
2
6
output
4
Convert model files; Add new pretrained model - Add new pretrained model to ./models/upconv_7 - Move old models to ./models/vgg_7 - Use nn.LeakyReLU instead of w2nn.LeakyReLU - Add useful attribute to .json New JSON attribute: The first layer has `model_config` attribute. It contains: model_arch: architecture name of model. see `lib/srcnn.lua` scale_factor: if scale_factor > 1, model:forward() changes image resolution with scale_factor. channels: input/output channels. if channels == 3, model is RGB model. offset: pixel size that is to be removed from output. for example: (scale_factor=1, offset=7, input=100x100) => output=(100-7)x(100-7) (scale_factor=2, offset=12, input=100x100) => output=(100*2-12)x(100*2-12) And each layer has `class_name` attribute.
2016-05-15 06:04:08 +12:00
100
Add 2x model for photo scaling
2015-11-30 21:29:34 +13:00
3
V 1
16
torch.CudaTensor
0
1
0
2
Convert model files; Add new pretrained model - Add new pretrained model to ./models/upconv_7 - Move old models to ./models/vgg_7 - Use nn.LeakyReLU instead of w2nn.LeakyReLU - Add useful attribute to .json New JSON attribute: The first layer has `model_config` attribute. It contains: model_arch: architecture name of model. see `lib/srcnn.lua` scale_factor: if scale_factor > 1, model:forward() changes image resolution with scale_factor. channels: input/output channels. if channels == 3, model is RGB model. offset: pixel size that is to be removed from output. for example: (scale_factor=1, offset=7, input=100x100) => output=(100-7)x(100-7) (scale_factor=2, offset=12, input=100x100) => output=(100*2-12)x(100*2-12) And each layer has `class_name` attribute.
2016-05-15 06:04:08 +12:00
5
train
5
0
Add 2x model for photo scaling
2015-11-30 21:29:34 +13:00
1
13
4
Convert model files; Add new pretrained model - Add new pretrained model to ./models/upconv_7 - Move old models to ./models/vgg_7 - Use nn.LeakyReLU instead of w2nn.LeakyReLU - Add useful attribute to .json New JSON attribute: The first layer has `model_config` attribute. It contains: model_arch: architecture name of model. see `lib/srcnn.lua` scale_factor: if scale_factor > 1, model:forward() changes image resolution with scale_factor. channels: input/output channels. if channels == 3, model is RGB model. offset: pixel size that is to be removed from output. for example: (scale_factor=1, offset=7, input=100x100) => output=(100-7)x(100-7) (scale_factor=2, offset=12, input=100x100) => output=(100*2-12)x(100*2-12) And each layer has `class_name` attribute.
2016-05-15 06:04:08 +12:00
101
Add 2x model for photo scaling
2015-11-30 21:29:34 +13:00
3
V 1
23
nn.SpatialConvolutionMM
3
Convert model files; Add new pretrained model - Add new pretrained model to ./models/upconv_7 - Move old models to ./models/vgg_7 - Use nn.LeakyReLU instead of w2nn.LeakyReLU - Add useful attribute to .json New JSON attribute: The first layer has `model_config` attribute. It contains: model_arch: architecture name of model. see `lib/srcnn.lua` scale_factor: if scale_factor > 1, model:forward() changes image resolution with scale_factor. channels: input/output channels. if channels == 3, model is RGB model. offset: pixel size that is to be removed from output. for example: (scale_factor=1, offset=7, input=100x100) => output=(100-7)x(100-7) (scale_factor=2, offset=12, input=100x100) => output=(100*2-12)x(100*2-12) And each layer has `class_name` attribute.
2016-05-15 06:04:08 +12:00
102
16
Add 2x model for photo scaling
2015-11-30 21:29:34 +13:00
2
Convert model files; Add new pretrained model - Add new pretrained model to ./models/upconv_7 - Move old models to ./models/vgg_7 - Use nn.LeakyReLU instead of w2nn.LeakyReLU - Add useful attribute to .json New JSON attribute: The first layer has `model_config` attribute. It contains: model_arch: architecture name of model. see `lib/srcnn.lua` scale_factor: if scale_factor > 1, model:forward() changes image resolution with scale_factor. channels: input/output channels. if channels == 3, model is RGB model. offset: pixel size that is to be removed from output. for example: (scale_factor=1, offset=7, input=100x100) => output=(100-7)x(100-7) (scale_factor=2, offset=12, input=100x100) => output=(100*2-12)x(100*2-12) And each layer has `class_name` attribute.
2016-05-15 06:04:08 +12:00
2
dH
1
1
2
2
dW
1
Add 2x model for photo scaling
2015-11-30 21:29:34 +13:00
1
2
11
nInputPlane
1
128
2
6
output
4
Convert model files; Add new pretrained model - Add new pretrained model to ./models/upconv_7 - Move old models to ./models/vgg_7 - Use nn.LeakyReLU instead of w2nn.LeakyReLU - Add useful attribute to .json New JSON attribute: The first layer has `model_config` attribute. It contains: model_arch: architecture name of model. see `lib/srcnn.lua` scale_factor: if scale_factor > 1, model:forward() changes image resolution with scale_factor. channels: input/output channels. if channels == 3, model is RGB model. offset: pixel size that is to be removed from output. for example: (scale_factor=1, offset=7, input=100x100) => output=(100-7)x(100-7) (scale_factor=2, offset=12, input=100x100) => output=(100*2-12)x(100*2-12) And each layer has `class_name` attribute.
2016-05-15 06:04:08 +12:00
103
Add 2x model for photo scaling
2015-11-30 21:29:34 +13:00
3
V 1
16
torch.CudaTensor
0
1
0
2
Convert model files; Add new pretrained model - Add new pretrained model to ./models/upconv_7 - Move old models to ./models/vgg_7 - Use nn.LeakyReLU instead of w2nn.LeakyReLU - Add useful attribute to .json New JSON attribute: The first layer has `model_config` attribute. It contains: model_arch: architecture name of model. see `lib/srcnn.lua` scale_factor: if scale_factor > 1, model:forward() changes image resolution with scale_factor. channels: input/output channels. if channels == 3, model is RGB model. offset: pixel size that is to be removed from output. for example: (scale_factor=1, offset=7, input=100x100) => output=(100-7)x(100-7) (scale_factor=2, offset=12, input=100x100) => output=(100*2-12)x(100*2-12) And each layer has `class_name` attribute.
2016-05-15 06:04:08 +12:00
2
kH
1
3
2
Add 2x model for photo scaling
2015-11-30 21:29:34 +13:00
9
gradInput
4
Convert model files; Add new pretrained model - Add new pretrained model to ./models/upconv_7 - Move old models to ./models/vgg_7 - Use nn.LeakyReLU instead of w2nn.LeakyReLU - Add useful attribute to .json New JSON attribute: The first layer has `model_config` attribute. It contains: model_arch: architecture name of model. see `lib/srcnn.lua` scale_factor: if scale_factor > 1, model:forward() changes image resolution with scale_factor. channels: input/output channels. if channels == 3, model is RGB model. offset: pixel size that is to be removed from output. for example: (scale_factor=1, offset=7, input=100x100) => output=(100-7)x(100-7) (scale_factor=2, offset=12, input=100x100) => output=(100*2-12)x(100*2-12) And each layer has `class_name` attribute.
2016-05-15 06:04:08 +12:00
104
Add 2x model for photo scaling
2015-11-30 21:29:34 +13:00
3
V 1
16
torch.CudaTensor
0
1
0
2
2
Convert model files; Add new pretrained model - Add new pretrained model to ./models/upconv_7 - Move old models to ./models/vgg_7 - Use nn.LeakyReLU instead of w2nn.LeakyReLU - Add useful attribute to .json New JSON attribute: The first layer has `model_config` attribute. It contains: model_arch: architecture name of model. see `lib/srcnn.lua` scale_factor: if scale_factor > 1, model:forward() changes image resolution with scale_factor. channels: input/output channels. if channels == 3, model is RGB model. offset: pixel size that is to be removed from output. for example: (scale_factor=1, offset=7, input=100x100) => output=(100-7)x(100-7) (scale_factor=2, offset=12, input=100x100) => output=(100*2-12)x(100*2-12) And each layer has `class_name` attribute.
2016-05-15 06:04:08 +12:00
kW
Add 2x model for photo scaling
2015-11-30 21:29:34 +13:00
1
Convert model files; Add new pretrained model - Add new pretrained model to ./models/upconv_7 - Move old models to ./models/vgg_7 - Use nn.LeakyReLU instead of w2nn.LeakyReLU - Add useful attribute to .json New JSON attribute: The first layer has `model_config` attribute. It contains: model_arch: architecture name of model. see `lib/srcnn.lua` scale_factor: if scale_factor > 1, model:forward() changes image resolution with scale_factor. channels: input/output channels. if channels == 3, model is RGB model. offset: pixel size that is to be removed from output. for example: (scale_factor=1, offset=7, input=100x100) => output=(100-7)x(100-7) (scale_factor=2, offset=12, input=100x100) => output=(100*2-12)x(100*2-12) And each layer has `class_name` attribute.
2016-05-15 06:04:08 +12:00
3
Add 2x model for photo scaling
2015-11-30 21:29:34 +13:00
2
12
nOutputPlane
1
3
2
4
Convert model files; Add new pretrained model - Add new pretrained model to ./models/upconv_7 - Move old models to ./models/vgg_7 - Use nn.LeakyReLU instead of w2nn.LeakyReLU - Add useful attribute to .json New JSON attribute: The first layer has `model_config` attribute. It contains: model_arch: architecture name of model. see `lib/srcnn.lua` scale_factor: if scale_factor > 1, model:forward() changes image resolution with scale_factor. channels: input/output channels. if channels == 3, model is RGB model. offset: pixel size that is to be removed from output. for example: (scale_factor=1, offset=7, input=100x100) => output=(100-7)x(100-7) (scale_factor=2, offset=12, input=100x100) => output=(100*2-12)x(100*2-12) And each layer has `class_name` attribute.
2016-05-15 06:04:08 +12:00
padW
Add 2x model for photo scaling
2015-11-30 21:29:34 +13:00
1
0
2
6
weight
4
Convert model files; Add new pretrained model - Add new pretrained model to ./models/upconv_7 - Move old models to ./models/vgg_7 - Use nn.LeakyReLU instead of w2nn.LeakyReLU - Add useful attribute to .json New JSON attribute: The first layer has `model_config` attribute. It contains: model_arch: architecture name of model. see `lib/srcnn.lua` scale_factor: if scale_factor > 1, model:forward() changes image resolution with scale_factor. channels: input/output channels. if channels == 3, model is RGB model. offset: pixel size that is to be removed from output. for example: (scale_factor=1, offset=7, input=100x100) => output=(100-7)x(100-7) (scale_factor=2, offset=12, input=100x100) => output=(100*2-12)x(100*2-12) And each layer has `class_name` attribute.
2016-05-15 06:04:08 +12:00
105
Add 2x model for photo scaling
2015-11-30 21:29:34 +13:00
3
V 1
16
torch.CudaTensor
2
3 1152
1152 1
1
4
Convert model files; Add new pretrained model - Add new pretrained model to ./models/upconv_7 - Move old models to ./models/vgg_7 - Use nn.LeakyReLU instead of w2nn.LeakyReLU - Add useful attribute to .json New JSON attribute: The first layer has `model_config` attribute. It contains: model_arch: architecture name of model. see `lib/srcnn.lua` scale_factor: if scale_factor > 1, model:forward() changes image resolution with scale_factor. channels: input/output channels. if channels == 3, model is RGB model. offset: pixel size that is to be removed from output. for example: (scale_factor=1, offset=7, input=100x100) => output=(100-7)x(100-7) (scale_factor=2, offset=12, input=100x100) => output=(100*2-12)x(100*2-12) And each layer has `class_name` attribute.
2016-05-15 06:04:08 +12:00
106
Add 2x model for photo scaling
2015-11-30 21:29:34 +13:00
3
V 1
17
torch.CudaStorage
3456
-0.19085139 -0.0184137002 -0.182110876 0.113014735 -0.155278683 0.473543435 0.011965286 -0.165696725 0.150411919 0.271408975 -0.211291164 0.0187345613 -0.241021812 -0.447341353 -0.379978806 0.631171048 0.132019013 0.0244705398 -0.212493598 0.302716583 -0.137326136 -0.0133078899 0.0763861462 -0.182719216 0.0491056032 -0.0989028141 0.1916271 0.0442547575 -0.109021269 -0.0792020187 0.0432557426 0.228624448 -0.108919211 0.0418880321 -0.378917128 0.270588219 0.133542344 -0.324867338 -0.56265831 0.583552539 -0.04758678 -0.177757844 -0.00649042847 0.113410309 0.318607777 0.0579827316 -0.0998779833 0.0808827132 -0.137765855 0.318014532 -0.117159687 -0.508262396 0.0483766533 0.340340436 0.389551669 0.445028305 0.13305597 -0.140111566 0.262988448 -0.101140738 -0.319074363 -0.518285155 -0.196214467 -0.0973687693 0.0316034183 -0.106237084 0.0702388138 -0.0252760984 -0.0120307328 0.352421343 -0.0968269259 -0.119003072 -0.0649613589 -0.179567873 0.255397707 0.0860874131 -0.145871028 -0.0591907054 -0.0291446 -0.453487396 -0.466689527 0.00719899964 0.122836381 -0.181092039 -0.0988057181 -0.189330354 0.401081204 0.164695948 0.0215882491 -0.168801472 0.0136945816 0.176803097 -0.134591892 0.0459877104 0.155588061 -0.155099973 0.0178145226 0.113147438 -0.101905823 0.385712653 -0.135218754 0.271985471 0.230298758 -0.0390889905 -0.0844831765 -0.466184795 0.049445264 -0.16084224 0.374087989 -0.156480744 -0.0397642814 -0.387561768 -0.202743426 0.429548353 -0.0613380261 0.0710524321 -0.0291327704 0.120054573 0.202270225 -0.236220688 -0.0254056584 0.103033923 -0.0541062765 -0.0317116976 -0.16395466 0.111141488 0.131322131 0.106220566 -0.128366545 0.494216353 -0.315725476 0.28125 0.0893280432 -0.266592979 -0.347400784 0.202805266 -0.412183225 0.0182501469 0.0849432051 -0.0714995489 0.0201458037 0.142200813 0.231589094 -0.223064572 -0.0952850282 0.109182432 -0.070515722 0.309621751 -0.411704481 0.220151052 -0.170136809 0.281448275 -0.122363903 -0.187048182 0.018418856 -0.28516376 0.0361551382 0.12675713 0.169704378 0.158983588 -0.181022838 0.168010384 0.0529835187 -0.235270202 0.00395936007 0.0400758758 0.204429179 -0.0595274791 -0.284934819 0.161461502 0.136893272 0.0407942645 -0.0495131649 -0.118199162 -0.0461021289 0.0622461177 0.216805845 0.00999687426 -0.0494212918 -0.133616224 -0.149409294 0.130909994 -0.0182547104 -0.087780498 0.0893906876 -0.166994631 -0.0834396482 -0.0808297843 0.28166765 0.0603851005 0.274373174 -0.0978926271 -0.179588005 -0.250352204 0.269907176 -0.207301408 0.241873696 -0.0967822075 -0.230192006 0.163457423 0.0795969293 -0.114325933 0.165544212 -0.07569056 -0.0875888169 0.0588253178 0.055978369 -0.0972488075 -0.0190365855 0.147133991 0.0393671803 -0.0972448736 0.147886679 0.0746462867 0.00044133974 -0.212941617 -0.0741248578 0.0703103468 -0.0821608976 -0.273380339 0.0333085097 -0.0593674779 0.0525348708 0.330823988 0.00490457844 -0.600237489 -7.04831618e-05 0.156319961 0.314076066 0.683730125 0.129411697 0.253684461 -0.206975713 -0.011121382 -0.261627197 0.21874547 0.214098826 0.296544552 -0.211864725 -0.115861468 -0.306154132 0.371288627 -0.194954813 0.0952872783 0.0998558402 0.123912349 -0.244161323 -0.142009065 0.190776467 0.248105422 -0.106862396 -0.317825109 -0.063610442 -0.189209357 0.16718553 0.00106821419 0.305210382 -0.279836565 0.148617819 -0.0394165292 0.0579071306 0.0978405625 0.540911019 -0.592735887 -0.272935331 0.474419087 0.0659710467 0.277104646 -0.087191917 0.137616217 -0.0857446268 0.144483104 0.0126244733 0.0141675547 0.0926944911 -0.137823924 0.00137415307 0.033433605 -0.065914005 -0.0166800879 -0.0436222292 -0.37836346 0.0321108438 -0.105350897 0.0166966654 0.353476316 -0.3473306 0.0298720952 0.0012913798 -0.0166476145 -0.389914453 0.0422025286 -0.247361764 0.00967587903 0.0483959876 0.136403829 0.423518062 -0.318536341 0.0442246646 -0.139123708 -0.0160183068 -0.153850913 0.357548088 -0.159372807 -0.403626829 0.425236106 0.0402078293 0.103275537 -0.280537903 0.000183147902 0.122124098 0.0147417206 -0.0240629036 -0.0730133951 0.0829578787 0.0362329371 -0.0440827981 -0.00147571694 -0.117842071 -0.37
2
5
train
5
0
2
10
gradWeight
4
Convert model files; Add new pretrained model - Add new pretrained model to ./models/upconv_7 - Move old models to ./models/vgg_7 - Use nn.LeakyReLU instead of w2nn.LeakyReLU - Add useful attribute to .json New JSON attribute: The first layer has `model_config` attribute. It contains: model_arch: architecture name of model. see `lib/srcnn.lua` scale_factor: if scale_factor > 1, model:forward() changes image resolution with scale_factor. channels: input/output channels. if channels == 3, model is RGB model. offset: pixel size that is to be removed from output. for example: (scale_factor=1, offset=7, input=100x100) => output=(100-7)x(100-7) (scale_factor=2, offset=12, input=100x100) => output=(100*2-12)x(100*2-12) And each layer has `class_name` attribute.
2016-05-15 06:04:08 +12:00
107
Add 2x model for photo scaling
2015-11-30 21:29:34 +13:00
3
V 1
16
torch.CudaTensor
2
3 1152
1152 1
1
4
Convert model files; Add new pretrained model - Add new pretrained model to ./models/upconv_7 - Move old models to ./models/vgg_7 - Use nn.LeakyReLU instead of w2nn.LeakyReLU - Add useful attribute to .json New JSON attribute: The first layer has `model_config` attribute. It contains: model_arch: architecture name of model. see `lib/srcnn.lua` scale_factor: if scale_factor > 1, model:forward() changes image resolution with scale_factor. channels: input/output channels. if channels == 3, model is RGB model. offset: pixel size that is to be removed from output. for example: (scale_factor=1, offset=7, input=100x100) => output=(100-7)x(100-7) (scale_factor=2, offset=12, input=100x100) => output=(100*2-12)x(100*2-12) And each layer has `class_name` attribute.
2016-05-15 06:04:08 +12:00
108
Add 2x model for photo scaling
2015-11-30 21:29:34 +13:00
3
V 1
17
torch.CudaStorage
3456
0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0
2
Convert model files; Add new pretrained model - Add new pretrained model to ./models/upconv_7 - Move old models to ./models/vgg_7 - Use nn.LeakyReLU instead of w2nn.LeakyReLU - Add useful attribute to .json New JSON attribute: The first layer has `model_config` attribute. It contains: model_arch: architecture name of model. see `lib/srcnn.lua` scale_factor: if scale_factor > 1, model:forward() changes image resolution with scale_factor. channels: input/output channels. if channels == 3, model is RGB model. offset: pixel size that is to be removed from output. for example: (scale_factor=1, offset=7, input=100x100) => output=(100-7)x(100-7) (scale_factor=2, offset=12, input=100x100) => output=(100*2-12)x(100*2-12) And each layer has `class_name` attribute.
2016-05-15 06:04:08 +12:00
5
_type
2
16
torch.CudaTensor
2
Add 2x model for photo scaling
2015-11-30 21:29:34 +13:00
4
Convert model files; Add new pretrained model - Add new pretrained model to ./models/upconv_7 - Move old models to ./models/vgg_7 - Use nn.LeakyReLU instead of w2nn.LeakyReLU - Add useful attribute to .json New JSON attribute: The first layer has `model_config` attribute. It contains: model_arch: architecture name of model. see `lib/srcnn.lua` scale_factor: if scale_factor > 1, model:forward() changes image resolution with scale_factor. channels: input/output channels. if channels == 3, model is RGB model. offset: pixel size that is to be removed from output. for example: (scale_factor=1, offset=7, input=100x100) => output=(100-7)x(100-7) (scale_factor=2, offset=12, input=100x100) => output=(100*2-12)x(100*2-12) And each layer has `class_name` attribute.
2016-05-15 06:04:08 +12:00
bias
4
109
Add 2x model for photo scaling
2015-11-30 21:29:34 +13:00
3
V 1
16
torch.CudaTensor
1
Convert model files; Add new pretrained model - Add new pretrained model to ./models/upconv_7 - Move old models to ./models/vgg_7 - Use nn.LeakyReLU instead of w2nn.LeakyReLU - Add useful attribute to .json New JSON attribute: The first layer has `model_config` attribute. It contains: model_arch: architecture name of model. see `lib/srcnn.lua` scale_factor: if scale_factor > 1, model:forward() changes image resolution with scale_factor. channels: input/output channels. if channels == 3, model is RGB model. offset: pixel size that is to be removed from output. for example: (scale_factor=1, offset=7, input=100x100) => output=(100-7)x(100-7) (scale_factor=2, offset=12, input=100x100) => output=(100*2-12)x(100*2-12) And each layer has `class_name` attribute.
2016-05-15 06:04:08 +12:00
3
1
1
4
110
3
V 1
17
torch.CudaStorage
3
0.0530770086 0.0594205558 0.0605555736
Add 2x model for photo scaling
2015-11-30 21:29:34 +13:00
2
Convert model files; Add new pretrained model - Add new pretrained model to ./models/upconv_7 - Move old models to ./models/vgg_7 - Use nn.LeakyReLU instead of w2nn.LeakyReLU - Add useful attribute to .json New JSON attribute: The first layer has `model_config` attribute. It contains: model_arch: architecture name of model. see `lib/srcnn.lua` scale_factor: if scale_factor > 1, model:forward() changes image resolution with scale_factor. channels: input/output channels. if channels == 3, model is RGB model. offset: pixel size that is to be removed from output. for example: (scale_factor=1, offset=7, input=100x100) => output=(100-7)x(100-7) (scale_factor=2, offset=12, input=100x100) => output=(100*2-12)x(100*2-12) And each layer has `class_name` attribute.
2016-05-15 06:04:08 +12:00
8
gradBias
Add 2x model for photo scaling
2015-11-30 21:29:34 +13:00
4
Convert model files; Add new pretrained model - Add new pretrained model to ./models/upconv_7 - Move old models to ./models/vgg_7 - Use nn.LeakyReLU instead of w2nn.LeakyReLU - Add useful attribute to .json New JSON attribute: The first layer has `model_config` attribute. It contains: model_arch: architecture name of model. see `lib/srcnn.lua` scale_factor: if scale_factor > 1, model:forward() changes image resolution with scale_factor. channels: input/output channels. if channels == 3, model is RGB model. offset: pixel size that is to be removed from output. for example: (scale_factor=1, offset=7, input=100x100) => output=(100-7)x(100-7) (scale_factor=2, offset=12, input=100x100) => output=(100*2-12)x(100*2-12) And each layer has `class_name` attribute.
2016-05-15 06:04:08 +12:00
111
3
V 1
16
torch.CudaTensor
Add 2x model for photo scaling
2015-11-30 21:29:34 +13:00
1
Convert model files; Add new pretrained model - Add new pretrained model to ./models/upconv_7 - Move old models to ./models/vgg_7 - Use nn.LeakyReLU instead of w2nn.LeakyReLU - Add useful attribute to .json New JSON attribute: The first layer has `model_config` attribute. It contains: model_arch: architecture name of model. see `lib/srcnn.lua` scale_factor: if scale_factor > 1, model:forward() changes image resolution with scale_factor. channels: input/output channels. if channels == 3, model is RGB model. offset: pixel size that is to be removed from output. for example: (scale_factor=1, offset=7, input=100x100) => output=(100-7)x(100-7) (scale_factor=2, offset=12, input=100x100) => output=(100*2-12)x(100*2-12) And each layer has `class_name` attribute.
2016-05-15 06:04:08 +12:00
3
Add 2x model for photo scaling
2015-11-30 21:29:34 +13:00
1
1
Convert model files; Add new pretrained model - Add new pretrained model to ./models/upconv_7 - Move old models to ./models/vgg_7 - Use nn.LeakyReLU instead of w2nn.LeakyReLU - Add useful attribute to .json New JSON attribute: The first layer has `model_config` attribute. It contains: model_arch: architecture name of model. see `lib/srcnn.lua` scale_factor: if scale_factor > 1, model:forward() changes image resolution with scale_factor. channels: input/output channels. if channels == 3, model is RGB model. offset: pixel size that is to be removed from output. for example: (scale_factor=1, offset=7, input=100x100) => output=(100-7)x(100-7) (scale_factor=2, offset=12, input=100x100) => output=(100*2-12)x(100*2-12) And each layer has `class_name` attribute.
2016-05-15 06:04:08 +12:00
4
112
3
V 1
17
torch.CudaStorage
3
0 0 0
Add 2x model for photo scaling
2015-11-30 21:29:34 +13:00
2
Convert model files; Add new pretrained model - Add new pretrained model to ./models/upconv_7 - Move old models to ./models/vgg_7 - Use nn.LeakyReLU instead of w2nn.LeakyReLU - Add useful attribute to .json New JSON attribute: The first layer has `model_config` attribute. It contains: model_arch: architecture name of model. see `lib/srcnn.lua` scale_factor: if scale_factor > 1, model:forward() changes image resolution with scale_factor. channels: input/output channels. if channels == 3, model is RGB model. offset: pixel size that is to be removed from output. for example: (scale_factor=1, offset=7, input=100x100) => output=(100-7)x(100-7) (scale_factor=2, offset=12, input=100x100) => output=(100*2-12)x(100*2-12) And each layer has `class_name` attribute.
2016-05-15 06:04:08 +12:00
4
padH
Add 2x model for photo scaling
2015-11-30 21:29:34 +13:00
1
Convert model files; Add new pretrained model - Add new pretrained model to ./models/upconv_7 - Move old models to ./models/vgg_7 - Use nn.LeakyReLU instead of w2nn.LeakyReLU - Add useful attribute to .json New JSON attribute: The first layer has `model_config` attribute. It contains: model_arch: architecture name of model. see `lib/srcnn.lua` scale_factor: if scale_factor > 1, model:forward() changes image resolution with scale_factor. channels: input/output channels. if channels == 3, model is RGB model. offset: pixel size that is to be removed from output. for example: (scale_factor=1, offset=7, input=100x100) => output=(100-7)x(100-7) (scale_factor=2, offset=12, input=100x100) => output=(100*2-12)x(100*2-12) And each layer has `class_name` attribute.
2016-05-15 06:04:08 +12:00
0
Add 2x model for photo scaling
2015-11-30 21:29:34 +13:00
1
14
4
Convert model files; Add new pretrained model - Add new pretrained model to ./models/upconv_7 - Move old models to ./models/vgg_7 - Use nn.LeakyReLU instead of w2nn.LeakyReLU - Add useful attribute to .json New JSON attribute: The first layer has `model_config` attribute. It contains: model_arch: architecture name of model. see `lib/srcnn.lua` scale_factor: if scale_factor > 1, model:forward() changes image resolution with scale_factor. channels: input/output channels. if channels == 3, model is RGB model. offset: pixel size that is to be removed from output. for example: (scale_factor=1, offset=7, input=100x100) => output=(100-7)x(100-7) (scale_factor=2, offset=12, input=100x100) => output=(100*2-12)x(100*2-12) And each layer has `class_name` attribute.
2016-05-15 06:04:08 +12:00
113
Add 2x model for photo scaling
2015-11-30 21:29:34 +13:00
3
V 1
7
nn.View
3
Convert model files; Add new pretrained model - Add new pretrained model to ./models/upconv_7 - Move old models to ./models/vgg_7 - Use nn.LeakyReLU instead of w2nn.LeakyReLU - Add useful attribute to .json New JSON attribute: The first layer has `model_config` attribute. It contains: model_arch: architecture name of model. see `lib/srcnn.lua` scale_factor: if scale_factor > 1, model:forward() changes image resolution with scale_factor. channels: input/output channels. if channels == 3, model is RGB model. offset: pixel size that is to be removed from output. for example: (scale_factor=1, offset=7, input=100x100) => output=(100-7)x(100-7) (scale_factor=2, offset=12, input=100x100) => output=(100*2-12)x(100*2-12) And each layer has `class_name` attribute.
2016-05-15 06:04:08 +12:00
114
7
Add 2x model for photo scaling
2015-11-30 21:29:34 +13:00
2
Convert model files; Add new pretrained model - Add new pretrained model to ./models/upconv_7 - Move old models to ./models/vgg_7 - Use nn.LeakyReLU instead of w2nn.LeakyReLU - Add useful attribute to .json New JSON attribute: The first layer has `model_config` attribute. It contains: model_arch: architecture name of model. see `lib/srcnn.lua` scale_factor: if scale_factor > 1, model:forward() changes image resolution with scale_factor. channels: input/output channels. if channels == 3, model is RGB model. offset: pixel size that is to be removed from output. for example: (scale_factor=1, offset=7, input=100x100) => output=(100-7)x(100-7) (scale_factor=2, offset=12, input=100x100) => output=(100*2-12)x(100*2-12) And each layer has `class_name` attribute.
2016-05-15 06:04:08 +12:00
11
numElements
1
Add 2x model for photo scaling
2015-11-30 21:29:34 +13:00
1
Convert model files; Add new pretrained model - Add new pretrained model to ./models/upconv_7 - Move old models to ./models/vgg_7 - Use nn.LeakyReLU instead of w2nn.LeakyReLU - Add useful attribute to .json New JSON attribute: The first layer has `model_config` attribute. It contains: model_arch: architecture name of model. see `lib/srcnn.lua` scale_factor: if scale_factor > 1, model:forward() changes image resolution with scale_factor. channels: input/output channels. if channels == 3, model is RGB model. offset: pixel size that is to be removed from output. for example: (scale_factor=1, offset=7, input=100x100) => output=(100-7)x(100-7) (scale_factor=2, offset=12, input=100x100) => output=(100*2-12)x(100*2-12) And each layer has `class_name` attribute.
2016-05-15 06:04:08 +12:00
2
5
_type
2
16
torch.CudaTensor
2
6
output
4
Add 2x model for photo scaling
2015-11-30 21:29:34 +13:00
3
2
Convert model files; Add new pretrained model - Add new pretrained model to ./models/upconv_7 - Move old models to ./models/vgg_7 - Use nn.LeakyReLU instead of w2nn.LeakyReLU - Add useful attribute to .json New JSON attribute: The first layer has `model_config` attribute. It contains: model_arch: architecture name of model. see `lib/srcnn.lua` scale_factor: if scale_factor > 1, model:forward() changes image resolution with scale_factor. channels: input/output channels. if channels == 3, model is RGB model. offset: pixel size that is to be removed from output. for example: (scale_factor=1, offset=7, input=100x100) => output=(100-7)x(100-7) (scale_factor=2, offset=12, input=100x100) => output=(100*2-12)x(100*2-12) And each layer has `class_name` attribute.
2016-05-15 06:04:08 +12:00
9
gradInput
4
115
3
V 1
16
torch.CudaTensor
0
1
0
2
Add 2x model for photo scaling
2015-11-30 21:29:34 +13:00
4
size
4
Convert model files; Add new pretrained model - Add new pretrained model to ./models/upconv_7 - Move old models to ./models/vgg_7 - Use nn.LeakyReLU instead of w2nn.LeakyReLU - Add useful attribute to .json New JSON attribute: The first layer has `model_config` attribute. It contains: model_arch: architecture name of model. see `lib/srcnn.lua` scale_factor: if scale_factor > 1, model:forward() changes image resolution with scale_factor. channels: input/output channels. if channels == 3, model is RGB model. offset: pixel size that is to be removed from output. for example: (scale_factor=1, offset=7, input=100x100) => output=(100-7)x(100-7) (scale_factor=2, offset=12, input=100x100) => output=(100*2-12)x(100*2-12) And each layer has `class_name` attribute.
2016-05-15 06:04:08 +12:00
116
Add 2x model for photo scaling
2015-11-30 21:29:34 +13:00
3
V 1
17
torch.LongStorage
1
-1
2
Convert model files; Add new pretrained model - Add new pretrained model to ./models/upconv_7 - Move old models to ./models/vgg_7 - Use nn.LeakyReLU instead of w2nn.LeakyReLU - Add useful attribute to .json New JSON attribute: The first layer has `model_config` attribute. It contains: model_arch: architecture name of model. see `lib/srcnn.lua` scale_factor: if scale_factor > 1, model:forward() changes image resolution with scale_factor. channels: input/output channels. if channels == 3, model is RGB model. offset: pixel size that is to be removed from output. for example: (scale_factor=1, offset=7, input=100x100) => output=(100-7)x(100-7) (scale_factor=2, offset=12, input=100x100) => output=(100*2-12)x(100*2-12) And each layer has `class_name` attribute.
2016-05-15 06:04:08 +12:00
12
numInputDims
1
3
2
Add 2x model for photo scaling
2015-11-30 21:29:34 +13:00
5
train
5
0
2
Convert model files; Add new pretrained model - Add new pretrained model to ./models/upconv_7 - Move old models to ./models/vgg_7 - Use nn.LeakyReLU instead of w2nn.LeakyReLU - Add useful attribute to .json New JSON attribute: The first layer has `model_config` attribute. It contains: model_arch: architecture name of model. see `lib/srcnn.lua` scale_factor: if scale_factor > 1, model:forward() changes image resolution with scale_factor. channels: input/output channels. if channels == 3, model is RGB model. offset: pixel size that is to be removed from output. for example: (scale_factor=1, offset=7, input=100x100) => output=(100-7)x(100-7) (scale_factor=2, offset=12, input=100x100) => output=(100*2-12)x(100*2-12) And each layer has `class_name` attribute.
2016-05-15 06:04:08 +12:00
14
w2nn_arch_name
2
5
vgg_7
Add 2x model for photo scaling
2015-11-30 21:29:34 +13:00
2
5
train
5
0
Convert model files; Add new pretrained model - Add new pretrained model to ./models/upconv_7 - Move old models to ./models/vgg_7 - Use nn.LeakyReLU instead of w2nn.LeakyReLU - Add useful attribute to .json New JSON attribute: The first layer has `model_config` attribute. It contains: model_arch: architecture name of model. see `lib/srcnn.lua` scale_factor: if scale_factor > 1, model:forward() changes image resolution with scale_factor. channels: input/output channels. if channels == 3, model is RGB model. offset: pixel size that is to be removed from output. for example: (scale_factor=1, offset=7, input=100x100) => output=(100-7)x(100-7) (scale_factor=2, offset=12, input=100x100) => output=(100*2-12)x(100*2-12) And each layer has `class_name` attribute.
2016-05-15 06:04:08 +12:00
2
11
w2nn_offset
1
7
2
5
_type
2
16
torch.CudaTensor
Reference in a new issue Copy permalink