[BUG] Fix the logic error of lo #176
Open
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Outdated suggestions cannot be applied.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.
Currently running
pytest -rA gpt_oss/triton/attention.py
will report the following error:In the
_attn_fwd
kernel, there is a logical error in the calculation of the loop boundarylo
: whenstart_q > 0
(e.g.start_q = 5
),lo
is incorrectly restricted tostart_q
(i.e., 5). This causes key positions 0 tostart_q-1
(i.e., 0-4) to be skipped. However, these positions are valid under the causal mask (since query positionstart_q = 5
can attend to historical keys 0-4). In the reference implementation, these positions are correctly included. Therefore, regardless of whetherBANDWIDTH
is enabled, the loop starting pointlo
should begin from 0 (the absolute start of the key sequence) rather thanstart_q
.