Signed-off-by: arpitkh101 <arpit5khandelwal@gmail.com>
Co-authored-by: Luka Govedič <ProExpertProg@users.noreply.github.com>
(cherry picked from commit d7284a2604)
This commit is contained in:
committed by
Kevin H. Luu
parent
a1d627e40f
commit
4fd9d6a85c
@@ -295,7 +295,7 @@ def all_reduce_fusion_pass_on_test_model(
|
||||
)
|
||||
)
|
||||
vllm_config.compilation_config.pass_config = PassConfig(
|
||||
enable_fi_allreduce_fusion=True, enable_noop=True
|
||||
fuse_allreduce_rms=True, eliminate_noops=True
|
||||
)
|
||||
vllm_config.device_config = DeviceConfig(device=torch.device("cuda"))
|
||||
vllm_config.parallel_config.rank = local_rank # Setup rank for debug path
|
||||
|
||||
Reference in New Issue
Block a user