Open
Conversation
…rch64 hardware to resolve quantization paradox
…nlocking TensorRT/NPU hardware acceleration
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
1 participant
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Outdated suggestions cannot be applied.This suggestion has been applied or marked resolved.Suggestions cannot be applied from pending reviews.Suggestions cannot be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.
Added a small verification check in the [benchmark.py]
to catch the "Quantization Paradox".
Sometimes INT8 models actually run slower than FP32 on certain ARM targets due to missing dot-product extensions or ORT threading overhead.
The loop now tracks the mean latency for
fp32,fp16, andint8. If you pass--paradox_strict, the build will fail if the INT8 model regresses performance compared to FP32, preventing us from merging mathematically slower quantized models.Note: I also patched a silent bug in
Benchmark.runwhere_benchmark_results_briefkept accumulating across different models rather than resetting, which was mixing up the statss