Merge 3 existing PR related to OnnxDiscrepancyCheck + llama.cpp integration#2546

Open

xadupre wants to merge 95 commits into

mainfrom

Commits on Jun 22, 2026

Commits on Jun 25, 2026

Commits on Jun 26, 2026

Commits on Jun 29, 2026

Commits on Jun 30, 2026

Commits on Jul 1, 2026

Fix test_metrics not saved in dry_run: always write timing_iterations explicitly
Copilot
authored
Fix missing ref_model_path in test calls and write timing_iterations explicitly on update
Copilot
authored
Add test_metrics parameter to OnnxDiscrepancyCheck and store it in generated config
Copilot
authored
Fix crash when formatting MAE threshold with test_metrics
Copilot
authored
Add attn_impl parameter to OnnxDiscrepancyCheck for configurable attention implementation
Copilot
authored
Rename Olive CLI config from config.json to olive_config.json to prevent collision with HF model config
Copilot
authored
Add logging to OnnxDiscrepancyCheck and switch attn_implementation default to sdpa
Copilot
authored
todo
xadupre
committed
fix
xadupre
committed
Reduce test model dimensions to fix CI artifact size check
Copilot
authored
Save CLI dry-run config as config.json instead of olive_config.json
Copilot
authored
Merge 3 existing PRs for OnnxDiscrepancyCheck + llama.cpp integration (with dedicated GGUF conversion pass) (#2548)
Copilot
authored
refactor
xadupre
committed
Keep only Olive config + test model in output_path for optimize --test
Copilot
authored
Keep optimized ONNX model in output_path/model instead of temp dir for --test
Copilot
authored
Add olive_config.json save + first_token_20/tft/tf5t generation metrics
Copilot
authored
Fix fast test: set GPTQ group_size=32 for tiny test model (hidden_size 64)
Copilot
authored
documentation
xadupre
committed
Merge branch 'xadupre/merged' of https://github.com/microsoft/Olive into xadupre/merged
xadupre
committed
Move attn_impl to SaveTestModelConfig; OnnxDiscrepancyCheck uses saved model config
Copilot
authored
Merge remote-tracking branch 'origin/xadupre/merged' into xadupre/merged
Copilot
authored
Fix int32 JSON serialization error in OnnxDiscrepancyCheck results
Copilot
authored

Commits on Jul 2, 2026