[Int4-AWQ] Torch Int-4 AWQ Dequantization and Configuration Options #146
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
hegemanjw4amd
force-pushed
the
hegeman/basic-sdpa-attention-int4-awq-interim
branch
3 times, most recently
from
0b78568 to
dd9a148
Compare
hegemanjw4amd
force-pushed
the
hegeman/basic-sdpa-attention-int4-awq-interim
branch
from
dd9a148 to
d4332ec
Compare
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Outdated suggestions cannot be applied.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.
This PR creates a fully general Int4-AWQ dequantization function which uses torch and adds environment options (flags) for controlling torch-vs-triton codepaths.
Testing: Two HuggingFace models quantized in Int4-AWQ format have been successfully run:
Qwen2-7B-Instruct-AWQ (Latency benchmarking)
Phi-3-mini-4k-instruct-AWQ (Input verification)
For the latter model, specific input prompts were supplied and the output examined, in order to provide a sanity check for correctness.
Unit testing is accomplished via tests/kernels/test_awq_triton.py.
Resolves: https://github.com/ROCm/FasterTransformer-Internal/issues/287