Skip to content
New issue

Have a question about this project? Sign up for a free GitHub account to open an issue and contact its maintainers and the community.

Sign up for GitHub

By clicking “Sign up for GitHub”, you agree to our terms of service and privacy statement. We’ll occasionally send you account related emails.

Already on GitHub? Sign in to your account

Enable flashinfer when group_size == 6 #2124

Merged
merged 1 commit into from
Apr 12, 2024
Merged

Enable flashinfer when group_size == 6 #2124

Show file tree
Hide file tree
Changes from all commits
Commits
Show all changes
1 commit
Select commit
File filter

Filter by extension

Filter by extension
Conversations
Failed to load comments.
Loading
Jump to
Jump to file
Failed to load files.
Loading
Diff view
Diff view
2 changes: 1 addition & 1 deletion python/mlc_llm/compiler_pass/dispatch_kv_cache_creation.py
View file
Open in desktop
Original file line number Diff line number Diff line change
Expand Up @@ -155,7 +155,7 @@ def create_flashinfer_paged_kv_cache(
in self.metadata["model_type"]
)
# filter by attention group size
or kwargs["num_attention_heads"] // kwargs["num_key_value_heads"] not in [1, 4, 8]
or kwargs["num_attention_heads"] // kwargs["num_key_value_heads"] not in [1, 4, 6, 8]
):
return

Expand Down
4 changes: 2 additions & 2 deletions python/mlc_llm/op/attention.py
View file
Open in desktop
Original file line number Diff line number Diff line change
Expand Up @@ -103,12 +103,12 @@ def _fallback():
and k.dtype == "float16"
and v.dtype == "float16"
):
if group_size not in [1, 4, 8]:
if group_size not in [1, 4, 6, 8]:
global WARN_FLASHINFER_GROUP_SIZE # pylint: disable=global-statement
if not WARN_FLASHINFER_GROUP_SIZE:
WARN_FLASHINFER_GROUP_SIZE = True
logger.warning(
"FlashInfer only supports group size in [1, 4, 8], but got %d. Skip and "
"FlashInfer only supports group size in [1, 4, 6, 8], but got %d. Skip and "
"fallback to default implementation.",
group_size,
)
Expand Down
Loading