Skip to content

GitLab

Unverified Commit afd089f2 authored by lailoo's avatar lailoo Committed by GitHub
Browse files

[Bugfix][Model] Fix Qwen3.5/Qwen3Next ignoring --dtype flag on older GPUs (#35617)

parent 3ecd0bf9
Hide whitespace changes
Inline Side-by-side
Showing with 0 additions and 5 deletions
vllm/model_executor/models/qwen3_5.py
View file @ afd089f2
......@@ -274,7 +274,6 @@ class Qwen3_5DecoderLayer(Qwen3NextDecoderLayer):
1,
1,
config.hidden_size,
dtype=config.dtype,
),
)
self.ffn_layer_scale = torch.nn.Parameter(
......@@ -282,7 +281,6 @@ class Qwen3_5DecoderLayer(Qwen3NextDecoderLayer):
1,
1,
config.hidden_size,
dtype=config.dtype,
),
)
......
vllm/model_executor/models/qwen3_next.py
View file @ afd089f2
......@@ -463,7 +463,6 @@ class Qwen3NextGatedDeltaNet(nn.Module, MambaBase):
group_size=None,
norm_before_gate=True,
device=current_platform.current_device(),
dtype=config.dtype,
)
self.out_proj = RowParallelLinear(
......@@ -1018,7 +1017,6 @@ class Qwen3NextDecoderLayer(nn.Module):
1,
1,
config.hidden_size,
dtype=config.dtype,
),
)
self.ffn_layer_scale = torch.nn.Parameter(
......@@ -1026,7 +1024,6 @@ class Qwen3NextDecoderLayer(nn.Module):
1,
1,
config.hidden_size,
dtype=config.dtype,
),
)
......
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