OSSForks/pytorch
1
0
Fork 0
mirror of https://github.com/zebrajr/pytorch.git synced 2025-12-07 12:21:27 +01:00
Code Issues Actions 38 Packages Projects Releases Wiki Activity
19,734 Commits 8,304 Branches 150 Tags 1.5 GiB
android_fbgemm_off
Commit Graph

594 Commits

Author SHA1 Message Date
Adam Paszke
80a827d3da Fix data_parallel bugs 2016-11-23 18:48:41 +01:00
Sam Gross
d40a7bf9eb Fix Scatter.backward() (#232) 2016-11-18 13:58:09 -05:00
Adam Paszke
e51d0bef97 Add cuDNN bindings for 2D transposed convolution 2016-11-17 14:34:40 -08:00
Adam Paszke
56fc639c9f Fix no bias mode of autogenerated THNN function 2016-11-16 15:32:18 -08:00
Adam Lerer
7f51af7cbc adding dropout, bidirection, etc. to RNN (#214) 2016-11-10 13:25:14 -05:00
Soumith Chintala
469dce4a2d skip test_scatter_gpu on no CUDA 2016-11-05 20:10:07 -04:00
Sam Gross
15377ac391 Copy Module._buffers in nn.parallel.replicate (#180) 2016-10-31 12:12:29 -04:00
Sam Gross
f2d7e94948 Use torch.Size for Tensor sizes and tuple for strides 
See issue #20

The torch.Size class is a tuple subclass which distinguishes sizes from
other tuples so that torch.Tensor(size) is interpreted as size instead
of data.
 2016-10-28 19:37:09 +02:00
Adam Paszke
e2458bce97 Add Parameter class to nn 2016-10-27 22:31:36 +02:00
Adam Lerer
b5d13296c6 addressing comments 2016-10-23 21:11:22 -07:00
Adam Lerer
86288265ad Adding rnn cell library 2016-10-23 20:23:48 -07:00
Adam Lerer
a559d94a44 docs and such 2016-10-23 20:23:48 -07:00
Adam Lerer
1eb6870853 add nobias option to rnn 2016-10-23 20:23:48 -07:00
Adam Lerer
942ca477a6 Copying weights for CUDNN 2016-10-23 20:23:48 -07:00
Sam Gross
98f67e90d5 Fix super call in Container.modules and Container.parameters (#142) 2016-10-19 13:21:03 -04:00
Sam Gross
fee67c2e1a Allow parameters and child modules to be assigned by attribute (#136) 
For example:
  self.linear = nn.Linear(10, 20)
  self.weight = torch.autograd.Variable(torch.Tensor(10, 20))
 2016-10-18 23:34:20 +02:00
soumith
59c628803a fixing padding_idx option 2016-10-14 15:05:21 -07:00
Adam Paszke
518cb6ec7c Allow specifying output size in MaxUnpooling 2016-10-10 20:51:15 -07:00
Adam Paszke
34bcd4c237 Rename FullConv to ConvTranspose and allow specifying output size 2016-10-10 20:51:15 -07:00
Adam Paszke
a22af69335 Add versioning and shared storage handling to autograd (#105) 2016-10-06 17:12:58 -04:00
Adam Lerer
1213149a2f add bias option to linear; allow modules to return nested lists/tuples of tensors (#106) 
* add bias option to linear; allow modules to return nested lists/tuples of tensors
 2016-10-06 15:59:12 -04:00
Adam Paszke
3cbe66ba8c Change requires_grad default to False 2016-10-05 08:46:34 -07:00
Adam Paszke
6efefac2df Add parameter_dict and load_parameter_dict methods for modules 2016-10-04 14:47:56 -07:00
Sam Gross
f4ebc65a12 Add Module.modules() and Module.children() (#90) 
modules(): returns an iterator over all modules in the network
 children(): returns an iterator over immediate children

Also fix __getitem__ in Sequential
 2016-10-01 21:18:53 -04:00
Adam Paszke
c8a4734b97 Add RReLU to both nn packages 2016-09-29 11:33:34 -07:00
Adam Paszke
2d8c2972ae Only allow leaf variables as module parameters 2016-09-29 11:31:26 -07:00
Sam Gross
f5a6a3b0e9 Fix torch.nn.Module._apply with None types (#66) 2016-09-28 19:31:07 -04:00
Sam Gross
bab7f89cdc Fix no_bias constructor for conv2d (#65) 2016-09-28 19:30:43 -04:00
Soumith Chintala
412019dbe4 fixing CPU builds by making cuda imports optional 2016-09-28 11:56:18 -04:00
Adam Paszke
7f4ff0e615 Fix type conversions in nn 2016-09-27 15:45:49 -07:00
Adam Paszke
3eac7164f4 Add data parallel functions to nn 2016-09-27 15:45:45 -07:00
Sam Gross
44481354fc Add back support for child=None in Container constructor (#55) 
It's often useful to have optional child modules, such as the
downsampling operation in ResNets. Add a test for this case:

  nn.Container(
    child=None,
  )
 2016-09-26 17:18:02 -04:00
Sam Gross
980300b381 Combine autograd.Leaf and autograd.Variable (#52) 
Prior to this change, there was a circular reference between Leaf and
Variable. This means that the objects (and referenced Tensors) are not
collected as soon as they go out of scope, which lead to higher memory
usage and out-of-memory errors.
 2016-09-25 20:21:14 -04:00
Adam Paszke
4cdeae3283 Return only unique variables from parameters() 2016-09-25 12:23:43 -07:00
Adam Paszke
8fdec15a55 Codemod to remove camel case method naming 2016-09-20 08:40:28 -07:00
Adam Paszke
d1fda539b7 Fix nn serialization errors 2016-09-15 19:28:34 -07:00
Adam Paszke
fb39971464 Add more modules to nn 2016-09-14 11:05:56 -07:00
Sam Gross
cd0929aa5e Use chainer-style constructor for Conv2d 
* Conv2d, MaxPool2d, and AvgPool2d have one argument for each of ksize,
   stride, and pad. This argument can be either a single number or a
   tuple of (h, w)
 2016-09-07 15:51:44 -07:00
Sam Gross
b738b09606 Clean up Module forward and __call__ (#14) 
* _forward is renamed forward since users should override it

 * some __call__ overrides are changed to forward

 * function which return a single variable are changed to return that
   variable instead of a one-element tuple
 2016-09-07 15:41:39 -04:00
Adam Paszke
774a6f1093 Add in-place operations to autograd and nn 2016-08-25 09:34:54 -07:00
Adam Paszke
24476090df Add volatile variables 2016-08-24 08:43:11 -07:00
Adam Paszke
ea93fb7ac0 Add more nn modules 2016-08-23 19:15:21 -07:00
Adam Paszke
2bf68e72d5 Add hook system to autograd and nn 2016-08-23 13:51:34 -07:00
Adam Paszke
d467a068c2 Add tests for new modules 2016-08-19 14:57:01 -07:00