Open
Conversation
🔗 Helpful Links🧪 See artifacts and rendered test results at hud.pytorch.org/pr/pytorch/ao/4244
Note: Links to docs will display an error until the docs builds have been completed. This comment was automatically generated by Dr. CI and updates every 15 minutes. |
meta-cla
bot
added
the
CLA Signed
Contributor
Author
|
@pytorchbot label "module: inference" "topic: for developers" |
pytorch-bot
bot
added
module: inference
Contributor
|
cc @jainapurva maybe you can take a look? |
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
2 participants
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Outdated suggestions cannot be applied.This suggestion has been applied or marked resolved.Suggestions cannot be applied from pending reviews.Suggestions cannot be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.
Summary:
Quantization is fundamentally designed to reduce memory footprint. However, torchao doesn't have e2e gpu benchmark with a real model run. For example, measure_accuracy_and_performance.sh focus on accuracy and throughput, not memory. Without an explicit memory benchmark, it's difficult to quantify the allocator behavior and memory framentation — which many quantization frameworks suffers to resolve.
To handle this issue, this PR introduces gpu benchmark with real model run. It covers 3-variants: (1) BF16 (original), (2) W8A8-INT, (3) W8A8-INT + torch.compile.
Result on NVIDIA RTX 5090 device and Qwen3-8B model:
Future plan:
This PR only introduces
Int8DynamicActivationInt8WeightConfigfor minimal change and design confirm, all torchao subclass support is planed: