squall/vllm
1
0
Fork 0
Code Issues Pull Requests Actions 1 Packages Projects Releases Wiki Activity

[Kernel] Fix Flashinfer Correctness (#7284)

Browse Source
This commit is contained in:
Lily Liu 2024-08-07 16:26:52 -07:00 committed by GitHub
parent 6d94420246
commit e53dfd3eaf
No known key found for this signature in database
GPG Key ID: B5690EEEBB952194
1 changed files with 7 additions and 3 deletions
Show Stats Download Patch File Download Diff File Expand all files Collapse all files

6
vllm/attention/backends/flashinfer.py
View File

@ -127,6 +127,7 @@ class FlashInferMetadata(AttentionMetadata):
raise ValueError(
f"Only {supported_head_sizes} are supported for head_dim,",
f"received {self.head_dim}.")
self.is_profile_run = is_block_tables_empty(self.block_tables)
def begin_forward(self):
if self.num_prefill_tokens > 0:
@ -140,11 +141,14 @@ class FlashInferMetadata(AttentionMetadata):
assert self.paged_kv_last_page_len is not None
batch_size = self.query_start_loc.shape[0] - 1
assert batch_size >= 0
# The prefill stage does not read kv cache.
# The profile run does not read kv cache.
# Both paged_kv_indices and paged_kv_last_page_len are empty.
# paged_kv_indptr is a zero tensor with size batch_size + 1.
if self.is_profile_run:
self.paged_kv_indptr = torch.zeros(batch_size + 1,
device=self.device)
else:
self.paged_kv_indptr = self.paged_kv_indptr.to(self.device)
self.paged_kv_last_page_len = self.paged_kv_last_page_len.to(
self.device)
self.paged_kv_indices = self.paged_kv_indices.to(self.device)