biondizzle/vllm
1
0
Fork 0
Code Issues Pull Requests Actions 2 Packages Projects Releases Wiki Activity

nano_nemotron_vl: fix tensor device mismatch exception when video profiling (#39029)

Browse Source 
Signed-off-by: Netanel Haber <58652339+netanel-haber@users.noreply.github.com>
This commit is contained in:
Netanel Haber
2026-04-06 01:23:45 +03:00
committed by GitHub
parent 56de443db1
commit d56e952239
1 changed files with 3 additions and 2 deletions
Download Patch File Download Diff File Expand all files Collapse all files

5
vllm/model_executor/models/nano_nemotron_vl.py
View File

@@ -1239,12 +1239,13 @@ class NemotronH_Nano_VL_V2(
img_context_token_ids=self._img_context_token_ids,
video_temporal_patch_size=video_temporal_patch_size,
)
device = video_embeddings.device
# video_repl.full is a list of token IDs
repl_token_ids = torch.tensor(video_repl.full)
repl_token_ids = torch.tensor(video_repl.full, device=device)
# Get embedding token IDs for image context (use pre-tokenized version)
embed_token_ids = torch.tensor(self._img_context_token_ids)
embed_token_ids = torch.tensor(self._img_context_token_ids, device=device)
# Create mask for video embedding positions
is_video_embed = torch.isin(repl_token_ids, embed_token_ids)