Bowen Bao

[ONNX 4] Add ONNX diagnostics tutorial

1

Similar to previous PRs in the stack. This PR adds an introduction to ONNX diagnostics. Please review only the last commit. cc @thiagocrepaldi @titaiwangms

[Tracking] TorchDynamo ONNX exporter issues

1

# Export Error Summary Dashboard ## - This report is generated from branch of https://github.com/huggingface/optimum/pull/1712 - Produced by `RUN_SLOW=1 pytest tests/exporters/onnx -k "test_exporters_cli_pytorch_cpu" -s -vvvvv -n 8 --disable-warnings` - The...

[RFC][WIP] Introduce PT2 dynamo onnx exporter

8

Hello from PyTorch ONNX exporter team. Since the early announcement of PT2 we have been actively developing the next generation of PyTorch ONNX exporter based on Dynamo, the new Python...

Is there a way to select which subtest to run?

1

Like what `pytest -k ` does, but on subtest level.

Wrong output from ONNX speculative decoding (assistant_model)

1

### System Info ```shell python 3.10.14 transformers 4.37.2 optimum 1.20.0 onnxruntime 1.18.0 ``` ### Who can help? _No response_ ### Information - [ ] The official example scripts - [X]...

[Feature request] FlexAttention alike in ONNX

1

Large language models (LLMs) have been popular for many years, yet there is still no dedicated attention operator/function in the standard ONNX specification. Previous attempts to include an attention operator...

Llama 3.3 70B Instruct download hits Client error '403 Forbidden' for url

17

As title, the error occurred when downloading the model via `llama model download --source meta --model-id Llama3.3-70B-Instruct`, with the custom URL acquired just now. ``` Some downloads failed: - tokenizer.model:...