Skip to content

[Model][QwenVL] Simplify cos/sin rotary embedding indexing #28962

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
Isotr0py merged 3 commits into from
Nov 19, 2025
Merged

[Model][QwenVL] Simplify cos/sin rotary embedding indexing #28962

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
c4b221c
[Model][QwenVL] Simplify cos/sin rotary embedding indexing
lgeiger Nov 18, 2025
cc1f0c5
[Model][Qwen3VL] Prevent synchronous CPU-GPU copy
lgeiger Nov 18, 2025
f3cd40f
Merge branch 'main' into qwen-cos-sin-indexing
lgeiger Nov 18, 2025
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
9 changes: 2 additions & 7 deletions vllm/model_executor/models/glm4_1v.py
View file
Open in desktop
Original file line number Diff line number Diff line change
Expand Up @@ -797,13 +797,8 @@ def rot_pos_emb(
# Use pre-computed cos_sin_cache from RotaryEmbedding
cos, sin = self.rotary_pos_emb.get_cos_sin(max_grid_size)

cos_h = cos[pos_ids[:, 0]] # (num_tokens, rotary_dim // 2)
cos_w = cos[pos_ids[:, 1]]
sin_h = sin[pos_ids[:, 0]]
sin_w = sin[pos_ids[:, 1]]

cos_combined = torch.cat([cos_h, cos_w], dim=-1)
sin_combined = torch.cat([sin_h, sin_w], dim=-1)
cos_combined = cos[pos_ids].flatten(1)
sin_combined = sin[pos_ids].flatten(1)
return cos_combined, sin_combined, pos_ids

def compute_attn_mask_seqlen(
Expand Down
9 changes: 2 additions & 7 deletions vllm/model_executor/models/qwen2_5_vl.py
View file
Open in desktop
Original file line number Diff line number Diff line change
Expand Up @@ -738,13 +738,8 @@ def rotary_pos_emb_thw(self, t, h, w):
# Use pre-computed cos_sin_cache from RotaryEmbedding
cos, sin = self.rotary_pos_emb.get_cos_sin(max_size)

cos_h = cos[pos_ids[:, 0]] # (num_tokens, rotary_dim // 2)
cos_w = cos[pos_ids[:, 1]]
sin_h = sin[pos_ids[:, 0]]
sin_w = sin[pos_ids[:, 1]]

cos_combined = torch.cat([cos_h, cos_w], dim=-1)
sin_combined = torch.cat([sin_h, sin_w], dim=-1)
cos_combined = cos[pos_ids].flatten(1)
sin_combined = sin[pos_ids].flatten(1)

cos_combined = cos_combined.reshape(
cos_combined.shape[0] // self.spatial_merge_unit,
Expand Down
9 changes: 2 additions & 7 deletions vllm/model_executor/models/qwen2_vl.py
View file
Open in desktop
Original file line number Diff line number Diff line change
Expand Up @@ -724,13 +724,8 @@ def rot_pos_emb(
# Use pre-computed cos_sin_cache from RotaryEmbedding
cos, sin = self.rotary_pos_emb.get_cos_sin(max_grid_size)

cos_h = cos[pos_ids[:, 0]] # (num_tokens, rotary_dim // 2)
cos_w = cos[pos_ids[:, 1]]
sin_h = sin[pos_ids[:, 0]]
sin_w = sin[pos_ids[:, 1]]

cos_combined = torch.cat([cos_h, cos_w], dim=-1)
sin_combined = torch.cat([sin_h, sin_w], dim=-1)
cos_combined = cos[pos_ids].flatten(1)
sin_combined = sin[pos_ids].flatten(1)
return cos_combined, sin_combined

def compute_attn_mask_seqlen(
Expand Down
9 changes: 2 additions & 7 deletions vllm/model_executor/models/qwen3_omni_moe_thinker.py
View file
Open in desktop
Original file line number Diff line number Diff line change
Expand Up @@ -428,13 +428,8 @@ def rot_pos_emb(self, grid_thw):
# Use pre-computed cos_sin_cache from RotaryEmbedding
cos, sin = self.rotary_pos_emb.get_cos_sin(max_grid_size)

cos_h = cos[pos_ids[:, 0]] # (num_tokens, rotary_dim // 2)
cos_w = cos[pos_ids[:, 1]]
sin_h = sin[pos_ids[:, 0]]
sin_w = sin[pos_ids[:, 1]]

cos_combined = torch.cat([cos_h, cos_w], dim=-1)
sin_combined = torch.cat([sin_h, sin_w], dim=-1)
cos_combined = cos[pos_ids].flatten(1)
sin_combined = sin[pos_ids].flatten(1)

return cos_combined, sin_combined

Expand Down
17 changes: 3 additions & 14 deletions vllm/model_executor/models/qwen3_vl.py
View file
Open in desktop
Original file line number Diff line number Diff line change
Expand Up @@ -459,18 +459,13 @@ def rot_pos_emb(self, grid_thw: list[list[int]]):
else self.rot_pos_ids(h, w, self.spatial_merge_size).repeat(t, 1)
for t, h, w in grid_thw
]
pos_ids = torch.cat(pos_ids, dim=0)
pos_ids = torch.cat(pos_ids, dim=0).to(self.device, non_blocking=True)

# Use pre-computed cos_sin_cache from RotaryEmbedding
cos, sin = self.rotary_pos_emb.get_cos_sin(max_grid_size)

cos_h = cos[pos_ids[:, 0]] # (num_tokens, rotary_dim // 2)
cos_w = cos[pos_ids[:, 1]]
sin_h = sin[pos_ids[:, 0]]
sin_w = sin[pos_ids[:, 1]]

cos_combined = torch.cat([cos_h, cos_w], dim=-1)
sin_combined = torch.cat([sin_h, sin_w], dim=-1)
cos_combined = cos[pos_ids].flatten(1)
sin_combined = sin[pos_ids].flatten(1)

return cos_combined, sin_combined

Expand Down Expand Up @@ -566,12 +561,6 @@ def forward(
pos_embeds = self.fast_pos_embed_interpolate(grid_thw_list)
hidden_states = hidden_states + pos_embeds
rotary_pos_emb_cos, rotary_pos_emb_sin = self.rot_pos_emb(grid_thw_list)
rotary_pos_emb_cos = rotary_pos_emb_cos.to(
hidden_states.device, non_blocking=True
)
rotary_pos_emb_sin = rotary_pos_emb_sin.to(
hidden_states.device, non_blocking=True
)

cu_seqlens = torch.repeat_interleave(
grid_thw[:, 1] * grid_thw[:, 2], grid_thw[:, 0]
Expand Down