Implement Save Tensor For Mxnet and Pytorch #291
Conversation
Codecov Report
@@ Coverage Diff @@
## save_inputs #291 +/- ##
===============================================
- Coverage 84.40% 79.91% -4.50%
===============================================
Files 86 86
Lines 6336 6363 +27
===============================================
- Hits 5348 5085 -263
- Misses 988 1278 +290
Continue to review full report at Codecov.
|
| @@ -0,0 +1,209 @@ | |||
| """ | |||
There was a problem hiding this comment.
Test:
numpy array.
regular array
smdebug/tensorflow/keras.py
Outdated
| self._save_tensor(tensor_name, tensor_value, collection_names) | ||
| self._save_tensor_to_file(tensor_name, tensor_value, collection_names) | ||
|
|
||
| def _save_tensor(self, tensor_name, tensor_value, collections_to_write): |
There was a problem hiding this comment.
is there a conflict for the name _save_tensor() ? or are you changing this to avoid confusion with hook.save_tensor()?
There was a problem hiding this comment.
I'm changing this name to avoid confusion
tests/pytorch/utils.py
Outdated
| model.train() | ||
| # for batch_idx, (data, target) in enumerate(train_loader): | ||
| if save_custom_tensor: | ||
| hook.save_tensor("custom_tensor", torch.tensor([[1.0, -1.0], [1.0, -1.0]])) |
There was a problem hiding this comment.
add save_tensor calls at different points in the training loop. before, during, after - to check if it is getting written at the correct step. similar to save_scalar
| hook.save_custom_tensor("custom_tensor_3", t3, include_collections) | ||
| hook.save_tensor("custom_tensor_1", t1, include_collections) | ||
| hook.save_tensor("custom_tensor_2", t2, include_collections) | ||
| hook.save_tensor("custom_tensor_3", t3, include_collections) |
There was a problem hiding this comment.
save_tensor after fit() too? how are we verifying if the tensor is written at the right step and correct mode?
Description of changes:
save_tensorapi for pytorch and mxnetStyle and formatting:
I have run
pre-commit installto ensure that auto-formatting happens with every commit.Issue number, if available
By submitting this pull request, I confirm that you can use, modify, copy, and redistribute this contribution, under the terms of your choice.