8000 encountering error when converting resnet-50 torch model by facebook · Issue #45 · szagoruyko/imagine-nn · GitHub
[go: up one dir, main page]
More Web Proxy on the site http://driver.im/
Skip to content
encountering error when converting resnet-50 torch model by facebook #45
New issue
New issue
Open
Open
encountering error when converting resnet-50 torch model by facebook#45
@michaelhuang74

Description

@michaelhuang74
michaelhuang74
opened on Mar 28, 2017

When I used the script on https://gist.github.com/szagoruyko/8828e09cc4687afd324d, which used the utils.lua, to convert the facebook resnet-50 model file (https://github.com/facebook/fb.resnet.torch), I had the following error:

Couldnt fold these: {
1 :
{
gradBias : CudaTensor - size: 256
bias : CudaTensor - size: 256
output : CudaTensor - size: 10x256x56x56
gradInput : CudaTensor - empty
save_mean : CudaTensor - size: 256
running_var : CudaTensor - size: 256
momentum : 0.1
eps : 1e-05
weight : CudaTensor - size: 256
_type : "torch.CudaTensor"
affine : true
running_mean : CudaTensor - size: 256
gradWeight : CudaTensor - size: 256
save_std : CudaTensor - size: 256
train : false
}
2 :
{
gradBias : CudaTensor - size: 256
bias : CudaTensor - size: 256
output : CudaTensor - size: 10x256x56x56
gradInput : CudaTensor - empty
save_mean : CudaTensor - size: 256
running_var : CudaTensor - size: 256
momentum : 0.1
eps : 1e-05
weight : CudaTensor - size: 256
_type : "torch.CudaTensor"
affine : true
running_mean : CudaTensor - size: 256
gradWeight : CudaTensor - size: 256
save_std : CudaTensor - size: 256
train : false
}
3 :
{
gradBias : CudaTensor - size: 256
bias : CudaTensor - size: 256
output : CudaTensor - size: 10x256x56x56
gradInput : CudaTensor - empty
save_mean : CudaTensor - size: 256
running_var : CudaTensor - size: 256
momentum : 0.1
eps : 1e-05
weight : CudaTensor - size: 256
_type : "torch.CudaTensor"
affine : true
running_mean : CudaTensor - size: 256
gradWeight : CudaTensor - size: 256
save_std : CudaTensor - size: 256
train : false
}
4 :
{
gradBias : CudaTensor - size: 512
bias : CudaTensor - size: 512
output : CudaTensor - size: 10x512x28x28
gradInput : CudaTensor - empty
save_mean : CudaTensor - size: 512
running_var : CudaTensor - size: 512
momentum : 0.1
eps : 1e-05
weight : CudaTensor - size: 512
_type : "torch.CudaTensor"
affine : true
running_mean : CudaTensor - size: 512
gradWeight : CudaTensor - size: 512
save_std : CudaTensor - size: 512
train : false
}
5 :
{
gradBias : CudaTensor - size: 512
bias : CudaTensor - size: 512
output : CudaTensor - size: 10x512x28x28
gradInput : CudaTensor - empty
save_mean : CudaTensor - size: 512
running_var : CudaTensor - size: 512
momentum : 0.1
eps : 1e-05
weight : CudaTensor - size: 512
_type : "torch.CudaTensor"
affine : true
running_mean : CudaTensor - size: 512
gradWeight : CudaTensor - size: 512
save_std : CudaTensor - size: 512
train : false
}
6 :
{
gradBias : CudaTensor - size: 512
bias : CudaTensor - size: 512
output : CudaTensor - size: 10x512x28x28
gradInput : CudaTensor - empty
save_mean : CudaTensor - size: 512
running_var : CudaTensor - size: 512
momentum : 0.1
eps : 1e-05
weight : CudaTensor - size: 512
_type : "torch.CudaTensor"
affine : true
running_mean : CudaTensor - size: 512
gradWeight : CudaTensor - size: 512
save_std : CudaTensor - size: 512
train : false
}
7 :
{
gradBias : CudaTensor - size: 512
bias : CudaTensor - size: 512
output : CudaTensor - size: 10x512x28x28
gradInput : CudaTensor - empty
save_mean : CudaTensor - size: 512
running_var : CudaTensor - size: 512
momentum : 0.1
eps : 1e-05
weight : CudaTensor - size: 512
_type : "torch.CudaTensor"
affine : true
running_mean : CudaTensor - size: 512
gradWeight : CudaTensor - size: 512
save_std : CudaTensor - size: 512
train : false
}
8 :
{
gradBias : CudaTensor - size: 1024
bias : CudaTensor - size: 1024
output : CudaTensor - size: 10x1024x14x14
gradInput : CudaTensor - empty
save_mean : CudaTensor - size: 1024
running_var : CudaTensor - size: 1024
momentum : 0.1
eps : 1e-05
weight : CudaTensor - size: 1024
_type : "torch.CudaTensor"
affine : true
running_mean : CudaTensor - size: 1024
gradWeight : CudaTensor - size: 1024
save_std : CudaTensor - size: 1024
train : false
}
9 :
{
gradBias : CudaTensor - size: 1024
bias : CudaTensor - size: 1024
output : CudaTensor - size: 10x1024x14x14
gradInput : CudaTensor - empty
save_mean : CudaTensor - size: 1024
running_var : CudaTensor - size: 1024
momentum : 0.1
eps : 1e-05
weight : CudaTensor - size: 1024
_type : "torch.CudaTensor"
affine : true
running_mean : CudaTensor - size: 1024
gradWeight : CudaTensor - size: 1024
save_std : CudaTensor - size: 1024
train : false
}
10 :
{
gradBias : CudaTensor - size: 1024
bias : CudaTensor - size: 1024
output : CudaTensor - size: 10x1024x14x14
gradInput : CudaTensor - empty
save_mean : CudaTensor - size: 1024
running_var : CudaTensor - size: 1024
momentum : 0.1
eps : 1e-05
weight : CudaTensor - size: 1024
_type : "torch.CudaTensor"
affine : true
running_mean : CudaTensor - size: 1024
gradWeight : CudaTensor - size: 1024
save_std : CudaTensor - size: 1024
train : false
}
11 :
{
gradBias : CudaTensor - size: 1024
bias : CudaTensor - size: 1024
output : CudaTensor - size: 10x1024x14x14
gradInput : CudaTensor - empty
save_mean : CudaTensor - size: 1024
running_var : CudaTensor - size: 1024
momentum : 0.1
eps : 1e-05
weight : CudaTensor - size: 1024
_type : "torch.CudaTensor"
affine : true
running_mean : CudaTensor - size: 1024
gradWeight : CudaTensor - size: 1024
save_std : CudaTensor - size: 1024
train : false
}
12 :
{
gradBias : CudaTensor - size: 1024
bias : CudaTensor - size: 1024
output : CudaTensor - size: 10x1024x14x14
gradInput : CudaTensor - empty
save_mean : CudaTensor - size: 1024
running_var : CudaTensor - size: 1024
momentum : 0.1
eps : 1e-05
weight : CudaTensor - size: 1024
_type : "torch.CudaTensor"
affine : true
running_mean : CudaTensor - size: 1024
gradWeight : CudaTensor - size: 1024
save_std : CudaTensor - size: 1024
train : false
}
13 :
{
gradBias : CudaTensor - size: 1024
bias : CudaTensor - size: 1024
output : CudaTensor - size: 10x1024x14x14
gradInput : CudaTensor - empty
save_mean : CudaTensor - size: 1024
running_var : CudaTensor - size: 1024
momentum : 0.1
eps : 1e-05
weight : CudaTensor - size: 1024
_type : "torch.CudaTensor"
affine : true
running_mean : CudaTensor - size: 1024
gradWeight : CudaTensor - size: 1024
save_std : CudaTensor - size: 1024
train : false
}
14 :
{
gradBias : CudaTensor - size: 2048
bias : CudaTensor - size: 2048
output : CudaTensor - size: 10x2048x7x7
gradInput : CudaTensor - empty
save_mean : CudaTensor - size: 2048
running_var : CudaTensor - size: 2048
momentum : 0.1
eps : 1e-05
weight : CudaTensor - size: 2048
_type : "torch.CudaTensor"
affine : true
running_mean : CudaTensor - size: 2048
gradWeight : 5222 CudaTensor - size: 2048
save_std : CudaTensor - size: 2048
train : false
}
15 :
{
gradBias : CudaTensor - size: 2048
bias : CudaTensor - size: 2048
output : CudaTensor - size: 10x2048x7x7
gradInput : CudaTensor - empty
save_mean : CudaTensor - size: 2048
running_var : CudaTensor - size: 2048
momentum : 0.1
eps : 1e-05
weight : CudaTensor - size: 2048
_type : "torch.CudaTensor"
affine : true
running_mean : CudaTensor - size: 2048
gradWeight : CudaTensor - size: 2048
save_std : CudaTensor - size: 2048
train : false
}
16 :
{
gradBias : CudaTensor - size: 2048
bias : CudaTensor - size: 2048
output : CudaTensor - size: 10x2048x7x7
gradInput : CudaTensor - empty
save_mean : CudaTensor - size: 2048
running_var : CudaTensor - size: 2048
momentum : 0.1
eps : 1e-05
weight : CudaTensor - size: 2048
_type : "torch.CudaTensor"
affine : true
running_mean : CudaTensor - size: 2048
gradWeight : CudaTensor - size: 2048
save_std : CudaTensor - size: 2048
train : false
}
}

Any idea how to resolve it?

Metadata

Metadata

Assignees

No one assigned

    Labels

    No labels
    No labels

    Projects

    No projects

    Milestone

    No milestone

    Relationships

    None yet

    Development

    No branches or pull requests

    Issue actions
      0