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 15, 08:30 UTC
fix(regression): clone from reference items (#18662)
Signed-off-by: Aaron Pham <contact@aarnphm.xyz>
Aaron Pham committed
May 24, 2025 at 11:25 UTC
9ea7f1abf3c015c5e560da099ad347e88b4a517d
1 file changed
+1
-1
vllm/sequence.py
+1
-1
@@ -1494,7 +1494,7 @@ class ParallelSampleSequenceGroup(SequenceGroupBase):
1494
for i in range(original_params.n):
1495
request_id_i = f"{request_id}_parallel_sample_{i}"
1496
group.seq_id_to_index[request_id_i] = i
1497
- params = params.clone()
1497
+ params = original_params.clone()
1498
params.n = 1
1499
if params.seed is not None:
1500
params.seed += i