kanshan/TensorRT-LLMs
1
0
Fork 0
mirror of https://github.com/NVIDIA/TensorRT-LLM.git synced 2026-02-04 18:21:52 +08:00
Code Issues Actions 1 Packages Projects Releases Wiki Activity

[None][fix] Add missing absolute pe in Qwen3-VL Vision Encoder (#11065)

Browse Source 
Signed-off-by: Necofish <liuxiangyang@mail.ustc.edu.cn>
This commit is contained in:
Necofish 2026-01-30 08:59:36 +08:00 committed by GitHub
parent 54ba056924
commit 144b61715f
No known key found for this signature in database
GPG Key ID: B5690EEEBB952194
1 changed files with 2 additions and 0 deletions
Show Stats Download Patch File Download Diff File Expand all files Collapse all files

2
tensorrt_llm/_torch/models/modeling_qwen3vl.py
View File

@ -739,9 +739,11 @@ class Qwen3VisionModel(torch.nn.Module):
# Getting positional embedding
rotary_pos_emb = self.rot_pos_emb(grid_thw)
pos_embeds = self.fast_pos_embed_interpolate(grid_thw)
# From this point, pure GPU operation
hidden_states = self.patch_embed(pixel_values)
hidden_states = hidden_states + pos_embeds
seq_len, _ = hidden_states.size()
hidden_states = hidden_states.reshape(seq_len, -1)