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, 19:45 UTC
[Bugfix] Fix torch_xla which can't handle None seed introduced in #14274 (#14459)
Signed-off-by: Yarong Mu <ymu@google.com>
yarongmu-google committed
Mar 7, 2025 at 15:17 UTC
66e16a038e9fe8bf04e133858621cd9803e7145b
1 file changed
+3
vllm/v1/worker/tpu_worker.py
+3
@@ -75,6 +75,9 @@ class TPUWorker:
75
self.profile_dir)
76
self.profiler = xp.start_server(9012)
77
78
+ if self.model_config.seed is None:
79
+ self.model_config.seed = 0
80
+
81
def init_device(self):
82
os.environ["PJRT_DEVICE"] = "TPU"
83
torch.set_grad_enabled(False)