EngineX/xc-llm-ascend
3
0
Fork 0
Code Issues Pull Requests Actions Projects Releases Wiki Activity

[Bugfix] Fix qwen2.5-vl positon input bug (#639)

Browse Source 
### What this PR does / why we need it?
Fix qwen2.5-vl positon input bug, fix #625 `TypeError: 'NoneType' object
is not iterable`

Signed-off-by: wangli <wangli858794774@gmail.com>
This commit is contained in:
Li Wang
2025-04-24 15:21:57 +08:00
committed by GitHub
parent 866ce7168c
commit a9c6b52205
1 changed files with 2 additions and 1 deletions
Download Patch File Download Diff File Expand all files Collapse all files

3
vllm_ascend/worker/model_runner.py
View File

@@ -556,7 +556,8 @@ class ModelInputForNPUBuilder(ModelRunnerInputBuilderBase[ModelInputForNPU]):
#print(f"before tensor input_positions: {input_positions}")
#print(f"before list seq_lens: {seq_lens}")
input_tokens = flatten_2d_lists(input_tokens)
input_positions = flatten_2d_lists(input_positions)
if input_positions:
input_positions = flatten_2d_lists(input_positions)
if graph_pad_size != -1 and not is_prompt:
input_tokens.extend(itertools.repeat(0, graph_pad_size))
input_positions.extend( # type: ignore