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

[Bugfix][TPU] Fix megacore setting for v5e-litepod (#6397)

Browse Source
This commit is contained in:
Woosuk Kwon 2024-07-12 15:59:47 -07:00 committed by GitHub
parent 6bc9710f6e
commit f8f9ff57ee
No known key found for this signature in database
GPG Key ID: B5690EEEBB952194
1 changed files with 1 additions and 1 deletions
Show Stats Download Patch File Download Diff File Expand all files Collapse all files

2
vllm/attention/backends/pallas.py
View File

@ -116,7 +116,7 @@ class PallasAttentionBackendImpl(AttentionImpl):
self.megacore_mode = None self.megacore_mode = None
tpu_type = torch_xla.tpu.get_tpu_env()["TYPE"].lower() tpu_type = torch_xla.tpu.get_tpu_env()["TYPE"].lower()
if not tpu_type.endswith("lite"): if "lite" not in tpu_type:
if self.num_kv_heads % 2 == 0: if self.num_kv_heads % 2 == 0:
self.megacore_mode = "kv_head" self.megacore_mode = "kv_head"
else: else: