Skip to content

GitLab

Commit b8c7ba0a authored by zhuwenwen's avatar zhuwenwen
Browse files

skip SPLIT_K

parent 2703e2e9
Show whitespace changes
Inline Side-by-side
Showing with 4 additions and 4 deletions
vllm/model_executor/layers/fused_moe/fused_moe.py
View file @ b8c7ba0a
......@@ -1478,7 +1478,7 @@ def get_default_config(
"BLOCK_SIZE_N": 64,
"BLOCK_SIZE_K": 32,
"GROUP_SIZE_M": 8,
"SPLIT_K": 1,
# "SPLIT_K": 1,
}
return config
......@@ -1492,7 +1492,7 @@ def get_default_config(
"BLOCK_SIZE_N": block_shape[0],
"BLOCK_SIZE_K": block_shape[1],
"GROUP_SIZE_M": 32,
"SPLIT_K": 1,
# "SPLIT_K": 1,
"num_warps": 4,
"num_stages": 3 if not current_platform.is_rocm() else 2,
}
......@@ -1516,7 +1516,7 @@ def get_default_config(
"BLOCK_SIZE_N": 32,
"BLOCK_SIZE_K": 64,
"GROUP_SIZE_M": 1,
"SPLIT_K": 1,
# "SPLIT_K": 1,
}
else:
config = {
......@@ -1524,7 +1524,7 @@ def get_default_config(
"BLOCK_SIZE_N": 64,
"BLOCK_SIZE_K": 32,
"GROUP_SIZE_M": 8,
"SPLIT_K": 1,
# "SPLIT_K": 1,
}
if use_nn_moe:
......
Markdown is supported
0% or .
You are about to add 0 people to the discussion. Proceed with caution.
Finish editing this message first!
Please register or to comment