Summary:
Originally, the tests for tensorboard writer are smoke tests only. This PR lets CI compare the output with expected results at low level. The randomness of the tensors in the test are also removed.
ps. I found that how protobuf serializes data differs between different python environment. One method to solve this is to write the data and then read it back instantly. (compare the data at a higher level)
For `add_custom_scalars`, the data to be written is a dictionary. and the serialized result might be different (not `ordereddict`). So only smoke test for that.
Pull Request resolved: https://github.com/pytorch/pytorch/pull/20987
Reviewed By: NarineK, lanpa
Differential Revision: D15804871
Pulled By: orionr
fbshipit-source-id: 69324c11ff823b19960d50def73adff36eb4a2ac
Summary:
The new implementation of tracing supports more module. So many error-handling code can be removed by placing the old one (LegacyTracedModule).
cc orionr
Pull Request resolved: https://github.com/pytorch/pytorch/pull/21339
Reviewed By: natalialunova
Differential Revision: D15695154
Pulled By: orionr
fbshipit-source-id: af7d35754e9f34bd1a0ad7b72a9ebe276ff8ab98
Summary:
Do tests with common models from torchvision.
Pull Request resolved: https://github.com/pytorch/pytorch/pull/20007
Differential Revision: D15251754
Pulled By: orionr
fbshipit-source-id: 9dc09bd407b3ccaaa310d2f4a8d53d5a7d12469d
Summary:
Add logging import and a failed MLP model that confirms that we don't fail `add_graph` when graph optimization fails.
This addresses part of https://github.com/pytorch/pytorch/issues/18903
cc lanpa ezyang natalialunova
Pull Request resolved: https://github.com/pytorch/pytorch/pull/20115
Reviewed By: natalialunova
Differential Revision: D15206765
Pulled By: orionr
fbshipit-source-id: c40b7e2671ef845a1529a2910ba030159f53f393
Summary:
This takes care of some outstanding review comments for https://github.com/pytorch/pytorch/pull/16196/
Specifically:
1. Add comment about kind
2. Add comment about GraphPy
3. Remove ONNX version comment
4. Remove scalar_dict from SummaryWriter and all history functions
cc lanpa ezyang
Pull Request resolved: https://github.com/pytorch/pytorch/pull/20038
Reviewed By: natalialunova
Differential Revision: D15177257
Pulled By: orionr
fbshipit-source-id: 218aa799d8b7dbb58f422a331236bba4959347de
Summary:
This PR adds TensorBoard logging support natively within PyTorch. It is based on the tensorboardX code developed by lanpa and relies on changes inside the tensorflow/tensorboard repo landing at https://github.com/tensorflow/tensorboard/pull/2065.
With these changes users can simply `pip install tensorboard; pip install torch` and then log PyTorch data directly to the TensorBoard protobuf format using
```
import torch
from torch.utils.tensorboard import SummaryWriter
writer = SummaryWriter()
s1 = torch.rand(1)
writer.add_scalar('data/scalar1', s1[0], 0)
writer.close()
```
Design:
- `EventFileWriter` and `RecordWriter` from tensorboardX now live in tensorflow/tensorboard
- `SummaryWriter` and PyTorch-specific conversion from tensors, nn modules, etc. now live in pytorch/pytorch. We also support Caffe2 blobs and nets.
Action items:
- [x] `from torch.utils.tensorboard import SummaryWriter`
- [x] rename functions
- [x] unittests
- [x] move actual writing function to tensorflow/tensorboard in https://github.com/tensorflow/tensorboard/pull/2065
Review:
- Please review for PyTorch standard formatting, code usage, etc.
- Please verify unittest usage is correct and executing in CI
Any significant changes made here will likely be synced back to github.com/lanpa/tensorboardX/ in the future.
cc orionr, ezyang
Pull Request resolved: https://github.com/pytorch/pytorch/pull/16196
Differential Revision: D15062901
Pulled By: orionr
fbshipit-source-id: 3812eb6aa07a2811979c5c7b70810261f9ea169e
