qihqi

Make __deepcopy__ of GraphModule able to handle circular reference.

1

Summary: One of such places where circular reference can occur is: _load_state_dict_pre_hooks contains a _WrappedHook, _WrappedHook has a weakref to the same module. Test Plan: Reviewers: Subscribers: Tasks: Tags: Fixes...

Update flatbuffer test models to match pkl models

3

Also regenerate upgrader with ``` python torchgen/operator_versions/gen_mobile_upgraders.py ``` Summary: Test Plan: Reviewers: Subscribers: Tasks: Tags: Fixes #ISSUE_NUMBER

Also add a unit test to repro an error on HLFB conversion.

## 🐛 Bug If I add a int32 tensor, to a int64 scalar tensor (of rank 0); then the result cannot be casted to int64. The error message is ```...

[export] Support capturing of non-tensor inputs

4

### 🚀 The feature, motivation and pitch Currently, `torch.export.export` would burn-in non-tensor inputs: Example: ```python import torch def logsoftmax(x, dim): return torch.nn.LogSoftmax(dim)(x) exp = torch.export.export(logsoftmax, ( torch.randn(10, 10), 1)) print(exp.graph_module.code)...

Refactor jestream to allow different tokenizers

1

## Issue Currently we assume few things in jetstream which hinders it's generalization: 1. tokenizer is SentencePiece based. 2. pad_id is 0 3. after encode, we pad to nearest power...

Update docker file for pytorch

1

when using jetstream-pytorch, one should not install Jetstream again. `install-everything.sh` script will install a particular version of Jetstream, please don't overwrite it.

Op info test for `constant_pad_nd`

1

## Fix the Op info test for `constant_pad_nd` 1. Find the line 33 of [test_ops.py](test/test_ops.py) and remove `constant_pad_nd` from `skip_list` 2. Run op_info test with `pytest test/test_ops.py` 3. Fix the...

Context for googlers: cl/643872265