Skip to content

Fix for logits issue in Qwen3 VL MOE #1043

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

Closed
tv-karthikeya wants to merge 3 commits into from
+34 −10
Closed

Fix for logits issue in Qwen3 VL MOE #1043

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
38143e4
Fix for logits issue in Qwen3 VL
tv-karthikeya Jun 5, 2026
5d1847a
Added minnor fix
abhishek-singh591 Jun 6, 2026
b0e2f80
Update modeling_qwen3_5_moe.py
abhishek-singh591 Jun 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
24 changes: 18 additions & 6 deletions QEfficient/transformers/models/qwen3_5_moe/modeling_qwen3_5_moe.py
View file
Open in desktop
Original file line number Diff line number Diff line change
Expand Up @@ -1013,7 +1013,15 @@ def forward(
# position_embeddings = None
all_hidden_states = () if output_hidden_states else None
layer_indices_to_run = kwargs.get("layer_indices_to_run", None)


layer_idx = 0
if QEffQwen3_5MoeTextModel._end == 0:
total_layers = len(self.layers)
end = total_layers
QEffQwen3_5MoeTextModel._end = total_layers
QEffQwen3_5MoeTextModel._total_layers = total_layers
layer_indices_to_run = kwargs.get("layer_indices_to_run", None)

for layer_idx, decoder_layer in enumerate(self.layers):
if layer_idx < start or layer_idx >= end:
continue
Expand Down Expand Up @@ -1498,12 +1506,16 @@ def forward(
batch_index=batch_index,
use_cache=True,
)
logit_index = position_ids[0].to(torch.int32).argmax(1, keepdim=True)
if outputs.last_hidden_state.shape[1] > 1:
hidden_states = outputs.last_hidden_state
if QEffQwen3_5MoeTextModel._end == 0:
logit_index = position_ids[0].to(torch.int32).argmax(1, keepdim=True)
hidden_states = outputs.last_hidden_state[torch.arange(position_ids[0].shape[0]).view(-1, 1), logit_index]
logits = self.model.lm_head(hidden_states)
else:
hidden_states = outputs.last_hidden_state[:, -1:, :]
logits = hidden_states
if outputs.last_hidden_state.shape[1] > 1:
hidden_states = outputs.last_hidden_state
else:
hidden_states = outputs.last_hidden_state[:, -1:, :]
logits = hidden_states
image_idx = (indices1.max() + 1).unsqueeze(0).unsqueeze(0)
return logits, vision_embeds, image_idx, outputs.past_key_values

Expand Down
20 changes: 16 additions & 4 deletions QEfficient/transformers/models/qwen3_vl_moe/modeling_qwen3_vl_moe.py
View file
Open in desktop
Original file line number Diff line number Diff line change
Expand Up @@ -578,6 +578,11 @@ def forward(
layer_idx = 0
start = QEffQwen3VLMoeTextModel._start
end = QEffQwen3VLMoeTextModel._end
if QEffQwen3VLMoeTextModel._end == 0:
total_layers = len(self.layers)
end = total_layers
QEffQwen3VLMoeTextModel._end = total_layers
QEffQwen3VLMoeTextModel._total_layers = total_layers
layer_indices_to_run = kwargs.get("layer_indices_to_run", None)

for layer_idx, decoder_layer in enumerate(self.layers):
Expand Down Expand Up @@ -849,11 +854,18 @@ def forward(
visual_pos_masks=visual_pos_masks,
deepstack_visual_embeds=deepstack_visual_embeds,
)
if outputs.last_hidden_state.shape[1] > 1:
hidden_states = outputs.last_hidden_state
if QEffQwen3VLMoeTextModel._end == 0:
logit_index = position_ids[0].to(torch.int32).argmax(1, keepdim=True)
hidden_states = outputs.last_hidden_state[
torch.arange(position_ids[0].shape[0]).view(-1, 1), logit_index
]
logits = self.model.lm_head(hidden_states)
else:
hidden_states = outputs.last_hidden_state[:, -1:, :]
logits = hidden_states
if outputs.last_hidden_state.shape[1] > 1:
hidden_states = outputs.last_hidden_state
else:
hidden_states = outputs.last_hidden_state[:, -1:, :]
logits = hidden_states
image_idx = (indices1.max() + 1).unsqueeze(0).unsqueeze(0)
return logits, vision_embeds, deepstack_features, image_idx, outputs.past_key_values

Expand Down
Loading