[Quantization] per tensor quantization kernel #42560
Merged
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
|
The docs for this PR live here. All of your documentation changes will be reflected on that endpoint. The docs are available until 30 days after the last update. |
SunMarc
approved these changes
Dec 2, 2025
Comment on lines
182
to
262
| assert len(block_size) == 2 | ||
| block_n, block_k = block_size[0], block_size[1] | ||
|
|
||
| # if we have per-tensor quantization, we use 128x128 block size for tiled matmul multiplication | ||
| if block_n == B.shape[-2] and block_k == B.shape[-1]: | ||
| block_n = 128 | ||
| block_k = 128 | ||
|
|
Member
There was a problem hiding this comment.
it doesn't make sense before to set blocks to something else than None when doing per tensor in the FP8Linear. Can we change that so that we fix it here also ?
Comment on lines
187
to
190
| """Triton-accelerated function used to perform linear operations (dot | ||
| product) on input tensors `A` and `B` with block-wise quantization, and | ||
| store the result in output tensor `C`. | ||
| """ |
ArthurZucker
approved these changes
Dec 2, 2025
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Outdated suggestions cannot be applied.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.
What does this PR do?
Adds a simple kernel for per tensor quantization, where the matmul is done per blocks of 128x128, and the weights scales, and activation scales are expected to be scalars