1.将basic_demo中的openai_api_server中的 engine_args = AsyncEngineArgs( model=MODEL_PATH, tokenizer=MODEL_PATH, # 如果你有多张显卡,可以在这里设置成你的显卡数量 tensor_parallel_size=1, dtype="bfloat16", trust_remote_code=True, # 占
Your current environment vllm version: '0.5.0.post1' 🐛 Describe the bug When I set tensor_parallel_size=1, it works well. But, if I set tensor_parallel_size>1, below error occurs: RuntimeError: Cannot re-initialize CUDA in forked subproc...
v0.7.3正式支持DeepSeek-AI多令牌预测模块,实测推理速度最高提升69%。只需在启动参数添加--num-speculative-tokens=1即可开启,还能选配--draft-tensor-parallel-size=1进一步优化。更惊人的是,在ShareGPT数据集测试中,该功能实现了81%-82.3%的预测接受率。这意味着在保持精度的同时,大幅缩短了推理耗时。生成式AI开...
导致gpu_p2p_access_cache_for_0,1.json文件不存在。
vllm+cpu 后端(无 gpu 硬件)时,tensor_parallel_size 应该默认设置成 1 而不是 cuda_count(等于 0) #3207 Sign in to view logs Summary Jobs issue_assign Run details Usage Workflow file Triggered via issue November 14, 2024 08:07 qinxuye commented on #2552 042eb5b Status Success ...
When settingtensor_parallel_sizeto a value greater than 1, the program gets stuck and raises aRuntimeErrorrelated to the bootstrapping phase of new processes. This issue does not occur when using version v0.4.3, but persists in versions v0.5.0.post1 and v0.5.0. ...
Your current environment PyTorch version: 2.3.0+cu121 Is debug build: False CUDA used to build PyTorch: 12.1 ROCM used to build PyTorch: N/A OS: Ubuntu 22.04.4 LTS (x86_64) GCC version: (Ubuntu 11.4.0-1ubuntu1~22.04) 11.4.0 Clang version...
Your current environment The output of `python collect_env.py` Your output of `python collect_env.py` here 🐛 Describe the bug When using VLLM_USE_MODELSCOPE and the tensor-parallel-size > 1, I found that vllm will download the model many...
Let's default the value ofspeculative_draft_tensor_parallel_sizeto 1 when we detect MLPSpeculator, since this is the only case that works right now. botcommentedAug 3, 2024 👋 Hi! Thank you for contributing to the vLLM project.
🐛 Describe the bug By using tensor parallel API, I expect we can disable TP by setting its mesh size to 1. But this does not work. Here is a reproduction case: import torch from torch.distributed.tensor.parallel import ColwiseParallel, p...