A high-throughput and memory-efficient inference and serving engine for LLMs
Mirrored from https://github.com/tduncan-lum/Nvidia-Optimization-Stack---VLLM.git
Synced Nov 14, 05:30 UTC
[BugFix] Fix logits repetition penalty cuda check (#22592)
Eugene Cheah committed
Aug 10, 2025 at 22:52 UTC
f919d4cb8faac8c869ab87ee705dbd340fae4679
1 file changed
+1
-1
vllm/_custom_ops.py
+1
-1
@@ -311,7 +311,7 @@ def apply_repetition_penalties(logits: torch.Tensor, prompt_mask: torch.Tensor,
311
output_mask: A boolean tensor indicating which tokens appear in the output.
312
repetition_penalties: The repetition penalties of shape (num_seqs, ).
313
"""
314
- if current_platform.is_cuda() and logits.is_contiguous():
314
+ if logits.is_cuda and logits.is_contiguous():
315
apply_repetition_penalties_cuda(logits, prompt_mask, output_mask,
316
repetition_penalties)
317
else: