Michael Benayoun

Fix the way the linear modules are removed from the parent module

1

# What does this PR do? Fixes # (issue) ## Before submitting - [ ] This PR fixes a typo or improves the docs (you can dismiss the other checks...

Quantization with torch FX [part 2]

5

# What does this PR do? This is the second PR in the process of adding PyTorch based quantization features to `optimum`. This PR continues what was started in the...

Community contribution - `optimum.exporters.onnx` support for new models!

41

Following what was done by @ChainYo in Transformers, in the [ONNXConfig: Add a configuration for all available models](https://github.com/huggingface/transformers/issues/16308) issue, the idea is to add support for exporting new models in...

Add support for Speech Encoder Decoder models in `optimum.exporters.onnx`

4

### Feature request Add support for [Speech Encoder Decoder Models](https://huggingface.co/docs/transformers/v4.25.1/en/model_doc/speech-encoder-decoder#speech-encoder-decoder-models) ### Your contribution Me or other members can implement it (cc @mht-sharma @fxmarty )

Uses torch.fx to parallelize and transform pipelined models

1

# What does this PR do? - Provides a custom tracer, built upon `transformers.utils.fx.HFTracer` allowing the trace and transform the models we support in `optimum-graphcore` - A set of pipelining...

- [ ] Greedy search - [ ] Sample - [ ] Beam search - [ ] Beam sample - [ ] Group beam search - [ ] Constrained beam...

Currently, during generation, the encoder outputs can only be computed after having compiled the encoder separately. It would be nice to be able to compute the encoder outputs directly from...

There are many tests that were temporarly disabled, either because they did not pass and were not critical, or because they could not be enabled until other PRs were merged...

- [ ] Add `test_generation_beam_search.py` - [ ] Add `test_generation_beam_constraints.py`

This PR enables Hyper Parameter Optimization with Optuna to optimize on latency and / or throughput. Both single and multi objective are supported. There are 3 tuning modes: - **Latency**:...