Skip to content

[#4745][fix] Pass lora_params through Qwen2/3 model forward #10174

New issue

Have a question about this project? Sign up for a free GitHub account to open an issue and contact its maintainers and the community.

Sign up for GitHub

By clicking “Sign up for GitHub”, you agree to our terms of service and privacy statement. We’ll occasionally send you account related emails.

Already on GitHub? Sign in to your account

Merged
karljang merged 3 commits into from
Jan 7, 2026
Merged

[#4745][fix] Pass lora_params through Qwen2/3 model forward #10174

Show file tree
Hide file tree
Changes from all commits
Commits
Show all changes
3 commits
Select commit Hold shift + click to select a range
6df0da2
fix(pytorch): pass lora_params through Qwen2/3 model forward
karljang Dec 20, 2025
7ea6f70
Pass kwargs to MLP in Qwen2/3 decoder layers
karljang Dec 24, 2025
3e5a3e0
Merge branch 'main' into fix-qwen-kwargs
karljang Jan 6, 2026
File filter

Filter by extension

Filter by extension
Conversations
Failed to load comments.
Loading
Jump to
Jump to file
Failed to load files.
Loading
Diff view
Diff view
8 changes: 5 additions & 3 deletions tensorrt_llm/_torch/models/modeling_qwen.py
View file
Open in desktop
Original file line number Diff line number Diff line change
Expand Up @@ -171,7 +171,7 @@ def forward(
# Fully Connected
hidden_states, residual = self.post_attention_layernorm(
hidden_states, residual)
hidden_states = self.mlp(hidden_states)
hidden_states = self.mlp(hidden_states, **kwargs)

if spec_metadata is not None:
spec_metadata.maybe_capture_hidden_states(self.layer_idx,
Expand Down Expand Up @@ -230,7 +230,8 @@ def forward(
attn_metadata=attn_metadata,
residual=residual,
mrope_config=mrope_config,
spec_metadata=spec_metadata)
spec_metadata=spec_metadata,
**kwargs)

hidden_states, _ = self.norm(hidden_states, residual)
return hidden_states
Expand Down Expand Up @@ -265,7 +266,8 @@ def forward(
position_ids=position_ids,
inputs_embeds=inputs_embeds,
mrope_config=mrope_config,
spec_metadata=spec_metadata)
spec_metadata=spec_metadata,
**kwargs)

return self.logits_processor.forward(
output,
Expand Down
6 changes: 5 additions & 1 deletion tensorrt_llm/_torch/models/modeling_qwen3.py
View file
Open in desktop
Original file line number Diff line number Diff line change
Expand Up @@ -152,6 +152,7 @@ def forward(
final_all_reduce_params=AllReduceParams(
enable_allreduce=not self.disable_allreduce),
cutlass_min_latency_mode=False,
**kwargs,
)
if deepstack_embeds is not None and self.layer_idx in range(
len(deepstack_embeds)):
Expand Down Expand Up @@ -221,7 +222,10 @@ def forward(
residual=residual,
spec_metadata=spec_metadata,
mrope_config=mrope_config,
deepstack_embeds=deepstack_embeds)
deepstack_embeds=deepstack_embeds,
**kwargs,
)

hidden_states, _ = self.norm(hidden_states, residual)
return hidden_states

Expand Down