rebase to main #446
Open
rebase to main #446
Conversation
This file contains bidirectional Unicode text that may be interpreted or compiled differently than what appears below. To review, open the file in an editor that reveals hidden Unicode characters.
Learn more about bidirectional Unicode characters
…unning (#424) This makes `--mode=cl-onnx-iree` the default mode since it is the most well-supported. A few code improvements are also made: 1. The flag `--iree-compile-args` now uses `-a` for shorthand, since `-ica` is not really valid. 2. Refactors the script running for `cl-onnx-iree` through a util function `run_command_and_log` in `logging_utils.py`. 3. Script running now uses `subprocess`, which more intelligently catches errors with `ret.returncode` rather than the existence of stage-specific output files. 4. reorganizes some imports for convenience and to allow case-by-case requirement dependency 5. Raises a warning if `torch-mlir-opt` is not found when trying to run with `preprocessing` (i.e., `--torchtolinalg`).
Updates actions and pins as suggested by OpenSSF Scorecard, see https://github.com/ossf/scorecard/blob/main/docs/checks.md#pinned-dependencies. In particular `actions/{download,upload}-artifact` was pointing to `master`, however, the current default branch is `main`, which might break the workflow after the announced deprecation date passed.
Actions are pinned with hashes as suggested by OpenSSF Scorecard, as soon as #438 lands. Those actions can be upgraded with Dependabot, see https://docs.github.com/en/code-security/dependabot/working-with-dependabot/keeping-your-actions-up-to-date-with-dependabot.
Updates tokenizer input registry for failing HuggingFace models with extra tokenizer attributes.
These models run successfully after adding the basic optimization
#444) If running ``` python run.py -t <some_test_1> -v -r test-run1 --get-metadata ``` and ``` python run.py -t <some_test_2> -v -r test-run2 --get-metadata ``` Then ``` python utils/find_duplicate_models.py -r . ``` from the `alt_e2eshark` directory will identify duplicates from both `./test-run1/` and `./test-run2/`.
Removes if conditions based on model name to decide Tokenizer class. Instead, directly uses AutoTokenizer class. Co-authored-by: Vinayak Dev <[email protected]> --------- Co-authored-by: Vivek Agrawal <[email protected]>
Adds extra input during construct_input phase to avoid input vs session_input size mismatch issue during native inference. --------- Co-authored-by: Vivek Agrawal <[email protected]>
Fixes the bug introduced in commit: "Fixes native inference input size mistmatch issue (#447)" tokenizer.model_input_names should be updated before tokenizer call.
Following testcases passes after fixing input issue at construct_input phase (which was causing failure at compiled_inference phase): ``` hf_ko-sroberta-multitask hf_robertuito-sentiment-analysis hf_sbert_large_nlu_ru hf_sentence-bert-base-ja-mean-tokens-v2 ```
- Fixes opset_version issue for following models: - hf_llama-7b - hf_oasst-sft-4-pythia-12b-epoch-3.5 - hf_Qwen2.5-1.5B-Instruct - hf_Qwen2.5-7B-Instruct - hf_Qwen2-7B-Instruct - hf_TinyLlama-1.1B-Chat-v1.0 - hf_vicuna-7b-v1.5 - Fixes OOM issue for hf_StableBeluga2 - removes hf_wasmai-7b-v1 model from the hf list --------- Co-authored-by: Vivek Agrawal <[email protected]>
Skip running the models that fails at setup stage with the following category of errors [ONNXRuntimeError] : 9 : NOT_IMPLEMENTED : Could not find an implementation Invalid source tar file [ONNXRuntimeError] : 10 : INVALID_GRAPH [ONNXRuntimeError] : 7 : INVALID_PROTOBUF :: Protobuf serialization failed [ONNXRuntimeError] : 1 : FAIL : Node
Fixes following setup issues: nod-ai/SHARK-ModelDev#919 nod-ai/SHARK-ModelDev#921
Sign up for free
to join this conversation on GitHub.
Already have an account?
Sign in to comment
Add this suggestion to a batch that can be applied as a single commit.
This suggestion is invalid because no changes were made to the code.
Suggestions cannot be applied while the pull request is closed.
Suggestions cannot be applied while viewing a subset of changes.
Only one suggestion per line can be applied in a batch.
Add this suggestion to a batch that can be applied as a single commit.
Applying suggestions on deleted lines is not supported.
You must change the existing code in this line in order to create a valid suggestion.
Outdated suggestions cannot be applied.
This suggestion has been applied or marked resolved.
Suggestions cannot be applied from pending reviews.
Suggestions cannot be applied on multi-line comments.
Suggestions cannot be applied while the pull request is queued to merge.
Suggestion cannot be applied right now. Please check back later.
rebase to main