root commited on
Commit
239c833
·
1 Parent(s): 69ad7bf

distribute gpu

Browse files
Files changed (1) hide show
  1. levo_inference.py +1 -1
levo_inference.py CHANGED
@@ -47,7 +47,7 @@ class LeVoInference(torch.nn.Module):
47
  tensor_parallel_size=self.cfg.vllm.device_num,
48
  enforce_eager=True,
49
  dtype="bfloat16",
50
- gpu_memory_utilization=self.cfg.vllm.gpu_memory_utilization,
51
  max_num_seqs=4,
52
  tokenizer=None,
53
  skip_tokenizer_init=True,
 
47
  tensor_parallel_size=self.cfg.vllm.device_num,
48
  enforce_eager=True,
49
  dtype="bfloat16",
50
+ gpu_memory_utilization=0.7,
51
  max_num_seqs=4,
52
  tokenizer=None,
53
  skip_tokenizer_init=True,