Skip to content

GitLab

Commit e3d1dfdd authored by zhuwenwen's avatar zhuwenwen
Browse files

update fill + moe align interface of lightop

parent 971a47d2
Hide whitespace changes
Inline Side-by-side
Showing with 4 additions and 2 deletions
vllm/model_executor/layers/fused_moe/moe_align_block_size.py
View file @ e3d1dfdd
...@@ -240,7 +240,8 @@ def moe_align_block_size( ...@@ -240,7 +240,8 @@ def moe_align_block_size(
expert_ids, num_tokens_post_pad, expert_ids, num_tokens_post_pad,
expert_map = expert_map, expert_map = expert_map,
expert_mask = expert_mask, expert_mask = expert_mask,
num_local_tokens = None) num_local_tokens = None,
is_fuse_fill = False if not envs.VLLM_USE_LIGHTOP_FILL_MOE_ALIGN else True)
else: else:
if envs.VLLM_USE_LIGHTOP_MOE_ALIGN: if envs.VLLM_USE_LIGHTOP_MOE_ALIGN:
from lightop import op as op from lightop import op as op
...@@ -248,7 +249,8 @@ def moe_align_block_size( ...@@ -248,7 +249,8 @@ def moe_align_block_size(
expert_ids, num_tokens_post_pad, expert_ids, num_tokens_post_pad,
expert_map = None, expert_map = None,
expert_mask = None, expert_mask = None,
num_local_tokens = None) num_local_tokens = None,
is_fuse_fill = False if not envs.VLLM_USE_LIGHTOP_FILL_MOE_ALIGN else True)
else: else:
ops.moe_align_block_size(topk_ids, num_experts, block_size, sorted_ids, ops.moe_align_block_size(topk_ids, num_experts, block_size, sorted_ids,
expert_ids, num_tokens_post_pad) expert_ids, num_tokens_post_pad)
......
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