Static quant support for SmoothQuant #3089
Draft
Conversation
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
🔗 Helpful Links🧪 See artifacts and rendered test results at hud.pytorch.org/pr/pytorch/ao/3089
Note: Links to docs will display an error until the docs builds have been completed. ❌ 4 New FailuresAs of commit abfce41 with merge base 4013764 ( NEW FAILURES - The following jobs have failed:
This comment was automatically generated by Dr. CI and updates every 15 minutes. |
meta-cla
bot
added
the
CLA Signed
Xia-Weiwen
added
the
topic: new feature
|
@Xia-Weiwen I think it's better to wait until the Int8Tensor migration is done |
Thanks for the info |
namgyu-youn
reviewed
Oct 14, 2025
| tokenizer = AutoTokenizer.from_pretrained(model_id) | ||
| model = ( | ||
| AutoModelForCausalLM.from_pretrained(model_id, dtype=torch.bfloat16) | ||
| AutoModelForCausalLM.from_pretrained(model_id, torch_dtype=torch.bfloat16) |
There was a problem hiding this comment.
nit: torch_dtype is deprecated; please check #2982 for more info
| tokenizer = AutoTokenizer.from_pretrained(model_id) | ||
| model = ( | ||
| AutoModelForCausalLM.from_pretrained(model_id, dtype=torch.bfloat16) | ||
| AutoModelForCausalLM.from_pretrained(model_id, torch_dtype=torch.bfloat16) |
| torch.manual_seed(34) | ||
| w8a8_model = ( | ||
| AutoModelForCausalLM.from_pretrained(model_id, dtype=torch.bfloat16) | ||
| AutoModelForCausalLM.from_pretrained(model_id, torch_dtype=torch.bfloat16) |
| model_save_path: str, | ||
| model_save_hf_hub_path: str, | ||
| static_quant_act: bool, | ||
| compile: bool, |
There was a problem hiding this comment.
Could you share result for torch.compile with static-quant? Not sure for the reason, but it decreased Token/Sec within dynamic-quant, and discussed to remove at #2728 (comment) .
Sign up for free
to join this conversation on GitHub.
Already have an account?
Sign in to comment
3 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
This PR adds static quant support for SmoothQuant by adding a new
Int8StaticActivationInt8WeightConfigconfiguration. Static quantization will generally have better latency & throughput than dynamic quant as it saves the overhead of runtime qparam selection.In the implementation:
SmoothQuantObserverreturns act scale along with the smoothing factor.Int8StaticActivationInt8WeightConfigfor transformation of each linear layer.Int8StaticActivationInt8WeightConfigis not suitable for general static quantization (although it works), users should use PT2E in that case. It's because the act scale for the config are global instead of per-linear-layer, which is the same asFloat8StaticActivationFloat8WeightConfigTest plan
This PR also updates the test cases for SmoothQuant: