Fix Weight loading for Qwen3.5-MTP and Qwen3-VL using runai_streamer (#42716)
Signed-off-by: weizhoublue <weizhou.lan@daocloud.io>
This commit is contained in:
@@ -175,8 +175,8 @@ class Qwen3_5MultiTokenPredictor(nn.Module):
|
||||
param,
|
||||
curr_expert_weight,
|
||||
name,
|
||||
shard_id,
|
||||
expert_id,
|
||||
shard_id=shard_id,
|
||||
expert_id=expert_id,
|
||||
return_success=True,
|
||||
)
|
||||
if success:
|
||||
|
||||
@@ -152,8 +152,8 @@ class Qwen3MoeLLMModel(Qwen3MoeModel):
|
||||
param,
|
||||
curr_expert_weight,
|
||||
name,
|
||||
shard_id,
|
||||
expert_id,
|
||||
shard_id=shard_id,
|
||||
expert_id=expert_id,
|
||||
return_success=True,
|
||||
)
|
||||
if success:
|
||||
|
||||
Reference in New Issue
Block a user