George S

🐛 [Bug] Issue in conversion when parameters/buffers are moved during compilation

1

## Bug Description ### Bug 1 ```python File "/root/.pyenv/versions/3.10.13/lib/python3.10/site-packages/torch_tensorrt/dynamo/conversion/converter_utils.py", line 491, in to_numpy output = value.cpu().detach().contiguous().numpy() RuntimeError: .numpy() is not supported for tensor subclasses. ``` ### Suggested Fix 1 Need...

- Formalize removal of detach operators pre-AOT via lowering pass

# Description - Add clear logging at the beginning and end of each phase of compilation - Reword logging in certain locations for clarity Addresses #2399 ## Type of change...

Follow-up to #2502

Windows CI for Python-Only Implementation

3

- Get CI up and running for Windows implementation using only the Python runtime

📖 [Story] Coverage for Core ATen Ops

2

## TL;DR Prioritze coverage for the [core ATen opset](https://pytorch.org/docs/main/torch.compiler_ir.html#core-aten-ir). ## Goal(s) - Determine priority order, based on criteria such as key model requirements, which operators from the core ATen opset...

# Description - Update `setup.py` to enable `bdist_wheel` build - Add GHA tooling for Windows Fixes #2489 ## Type of change - New CI Path # Checklist: - [ ]...

Refactor `truncate_long_and_double` in Dynamo

1

- Explicit considerations of truncation of inputs versus truncation of constants Sourced from https://github.com/pytorch/TensorRT/pull/2457#issuecomment-1889824984: - Avoid running PyTorch graphs with invalid casts - Refactor `repair_long_and_double` to consume output of type...

See the failures in the `test-py-dynamo-converters-cu121` [section here](https://github.com/pytorch/TensorRT/actions/runs/7350298746/job/20011958833?pr=2538). The bitwise converters are off by 1 in certain test cases, pointing to either implementation/functionality changes in TRT 9.x, or correctness/accuracy issues.