add time to first token in OnnxDiscrepancyCheck#2535
Open
xadupre wants to merge 6 commits into
Open
Conversation
Contributor
There was a problem hiding this comment.
Pull request overview
Note
Copilot couldn't run its full agentic review because no GitHub Actions runner was available. Make sure your repository has a runner available to run Copilot's review, or add a copilot-setup-steps.yml file specifying one with the runs-on attribute. See the docs for more details.
Adds time-to-first-token and time-to-first-N-tokens latency metrics to OnnxDiscrepancyCheck’s generation comparison output, and updates tests accordingly.
Changes:
- Introduces a new
time_to_first_n_tokensconfig option and surfaces latency metrics for both Transformers and ORT GenAI generation. - Changes
compare_generationto return a results dictionary (instead of anint) and updates result aggregation/logging. - Updates ONNX discrepancy check tests to validate the new return shape and metric keys.
Reviewed changes
Copilot reviewed 2 out of 2 changed files in this pull request and generated 4 comments.
| File | Description |
|---|---|
| test/passes/onnx/test_discrepancy_check.py | Adjusts assertions for the new dict return type and validates presence of new latency metric fields. |
| olive/passes/onnx/discrepancy_check.py | Adds latency measurement/reporting for TTFT/TTFN and changes generation comparison to return structured results. |
5 tasks
5 tasks
xadupre
pushed a commit
that referenced
this pull request
Jul 1, 2026
… (with dedicated GGUF conversion pass) (#2548) ## Describe your changes Merges #2536, #2535, #2534. Additionally adds llama.cpp integration and other improvements to `OnnxDiscrepancyCheck` and test-mode workflow handling: - **New `llama_cpp` flag** (`bool`, default `False`) on `OnnxDiscrepancyCheck` — when enabled, compares inference with llama.cpp. - **New `llama_cpp_env_path` parameter** (`Optional[str]`) — path to the `llama_env` virtual environment where `llama-cpp-python` and `convert_hf_to_gguf.py` are installed (defaults to `"llama_env"` relative to cwd). - **New `--test_llama_path` CLI option** — specifies the path to the `llama_env` virtual environment when running with `--test`. Using `--test_llama_path` without `--test` emits a warning. - **New `ConvertHfToGGUF` pass** (`olive/passes/pytorch/convert_hf_to_gguf.py`) — injected when `--test_llama_path` is provided. This pass converts the test HF model to GGUF ahead of discrepancy checking and stores the GGUF path in model attributes for downstream reuse. - **`compare_llama_cpp()` updates** — now reuses a preconverted GGUF when available; otherwise it falls back to in-method HF→GGUF conversion. llama.cpp comparison failures are captured in discrepancy results (status/failures) instead of aborting the whole run, so ONNX generation can still complete. - **Improved `--test_metrics` parsing** — now accepts both space-separated (`--test_metrics mae speedup`) and comma-separated (`--test_metrics mae,speedup`) forms. - **Fixed `add_discrepancy_check_pass` update-in-place** — existing discrepancy-pass config generated by dry-run is updated in-place so current `--test_metrics`, `--output_path`, and llama settings are applied. - **Fixed test model persistence across engine cache hits** — `ModelBuilder` stores a reference HF copy (`reference_hf_model/`) alongside cached ONNX outputs; discrepancy check falls back to this copy if the original test model path is missing. - **New `SaveTestModelConfig` pass** (`olive/passes/pytorch/save_test_model_config.py`) — injected at the start of passes for `--test`; ensures test model config/marker (and random test model persistence path usage) is set up before downstream passes. - **CI workflow** (`test-model-fast.yml`) — includes setup of a llama environment and llama.cpp conversion script dependencies. - **Updated documentation** (`cli-fast-test.md`) — clarifies where layer reduction happens, when test-model directories are created, cache fallback behavior, and llama.cpp test flow including the dedicated GGUF conversion pass. ## Checklist before requesting a review - [ ] Add unit tests for this change. - [ ] Make sure all tests can pass. - [ ] Update documents if necessary. - [ ] Lint and apply fixes to your code by running `lintrunner -a` - [ ] Is this a user-facing change? If yes, give a description of this change to be included in the release notes. ## (Optional) Issue link --------- Co-authored-by: copilot-swe-agent[bot] <198982749+Copilot@users.noreply.github.com>
This file contains hidden or 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
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.
Describe your changes