搜索

按 Enter 键快速全文搜索，也可直接发起 AI 搜索
 首页 文库资料 文章资讯  上传文档  发布文章  登录账户

分类

全部综合其他(18)人工智能(17)云计算&大数据(3)机器学习(3)后端开发(2)Python(1)产品与服务(1)

语言

全部英语(16)中文（简体）(7)

格式

全部PDF文档 PDF(23)
 
本次搜索耗时 0.119 秒，为您找到相关结果约 23 个.

  • pdf文档 PyTorch Release Notes

    including Python 3.10 NVIDIA CUDA $ ^{®} $ 12.1.1 NVIDIA cuBLAS 12.1.3.1 NVIDIA cuDNN 8.9.3 NVIDIA NCCL 2.18.3 NVIDIA RAPIDS $ ^{TM} $ 23.06 Apex rdma-core 39.0 NVIDIA HPC-X 2.15 OpenMPI 4.1.4+ GDRCopy including Python 3.10 NVIDIA CUDA $ ^{®} $ 12.1.1 NVIDIA cuBLAS 12.1.3.1 NVIDIA cuDNN 8.9.2 NVIDIA NCCL 2.18.1 NVIDIA RAPIDS $ ^{TM} $ 23.04 Apex rdma-core 39.0 NVIDIA HPC-X 2.15 OpenMPI 4.1.4+ GDRCopy including Python 3.10 NVIDIA CUDA $ ^{®} $ 12.1.1 NVIDIA cuBLAS 12.1.3.1 NVIDIA cuDNN 8.9.1.23 NVIDIA NCCL 2.18.1 NVIDIA RAPIDS $ ^{™} $ 23.04 Apex rdma-core 36.0 NVIDIA HPC-X 2.14 OpenMPI 4.1.4+ GDRCopy
    0 码力 | 365 页 | 2.94 MB | 2 年前
    3

  • pdf文档 vLLM v0.6.2 Documentation

    install torch with separate library packages like NCCL, while conda installs torch with statically linked NCCL. This can cause issues when vLLM tries to use NCCL. See this issue for more details. Note: As of CUDA kernel is causing the trouble. - Set the environment variable export NCCL_DEBUG=TRACE to turn on more logging for NCCL. - Set the environment variable export VLLM_TRACE_FUNCTION=1. All the function setting the environment variable export VLLM_HOST_IP=your_ip_address. You might also need to set export NCCL_SOCKET_IFNAME=your_network_interface and export GLOO_SOCKET_IFNAME=your_network_interface to specify
    0 码力 | 227 页 | 1.33 MB | 3 月前
    3

  • pdf文档 vLLM v0.6.1.post2 Documentation

    install torch with separate library packages like NCCL, while conda installs torch with statically linked NCCL. This can cause issues when vLLM tries to use NCCL. See this issue for more details. Note: As of CUDA kernel is causing the trouble. - Set the environment variable export NCCL_DEBUG=TRACE to turn on more logging for NCCL. - Set the environment variable export VLLM_TRACE_FUNCTION=1. All the function setting the environment variable export VLLM_HOST_IP=your_ip_address. You might also need to set export NCCL_SOCKET_IFNAME=your_network_interface and export GLOO_SOCKET_IFNAME=your_network_interface to specify
    0 码力 | 215 页 | 1.29 MB | 3 月前
    3

  • pdf文档 vLLM v0.5.4 Documentation

    CUDA kernel is causing the trouble. - Set the environment variable export NCCL_DEBUG=TRACE to turn on more logging for NCCL. - Set the environment variable export VLLM_TRACE_FUNCTION=1. All the function backend=nccl. The IP address should be the correct one. If not, override the IP address by setting the environment variable export VLLM_HOST_IP=your_ip_address. You might also need to set export NCCL_SOCK ```python import torch import torch.distributed as dist dist.init_process_group(backend="nccl") local_rank = dist.get_rank() % torch.cuda.device_count() data = torch.FloatTensor([1,] * 128)
    0 码力 | 152 页 | 1.10 MB | 3 月前
    3

  • pdf文档 vLLM v0.5.5 Documentation

    CUDA kernel is causing the trouble. - Set the environment variable export NCCL_DEBUG=TRACE to turn on more logging for NCCL. - Set the environment variable export VLLM_TRACE_FUNCTION=1. All the function setting the environment variable export VLLM_HOST_IP=your_ip_address. You might also need to set export NCCL_SOCKET_IFNAME=your_network_interface and export GLOO_SOCKET_IFNAME=your_network_interface to specify communication is working correctly. ```python # Test PyTorch NCCL import torch import torch.distributed as dist dist.init_process_group(backend="nccl") local_rank = dist.get_rank() % torch.cuda.device_count()
    0 码力 | 193 页 | 1.22 MB | 3 月前
    5

  • pdf文档 vLLM v0.6.0 Documentation

    CUDA kernel is causing the trouble. - Set the environment variable export NCCL_DEBUG=TRACE to turn on more logging for NCCL. - Set the environment variable export VLLM_TRACE_FUNCTION=1. All the function setting the environment variable export VLLM_HOST_IP=your_ip_address. You might also need to set export NCCL_SOCKET_IFNAME=your_network_interface and export GLOO_SOCKET_IFNAME=your_network_interface to specify communication is working correctly. ```python # Test PyTorch NCCL import torch import torch.distributed as dist dist.init_process_group(backend="nccl") local_rank = dist.get_rank() % torch.cuda.device_count()
    0 码力 | 201 页 | 1.26 MB | 3 月前
    3

  • pdf文档 vLLM v0.6.1.post1 Documentation

    CUDA kernel is causing the trouble. - Set the environment variable export NCCL_DEBUG=TRACE to turn on more logging for NCCL. - Set the environment variable export VLLM_TRACE_FUNCTION=1. All the function backend=nccl. The IP address should be the correct one. If not, override the IP address by setting the environment variable export VLLM_HOST_IP=your_ip_address. You might also need to set export NCCL_SOCK communication is working correctly. ```python # Test PyTorch NCCL import torch import torch.distributed as dist dist.init_process_group(backend="nccl") local_rank = dist.get_rank() % torch.cuda.device_count()
    0 码力 | 215 页 | 1.28 MB | 3 月前
    3

  • pdf文档 vLLM v0.5.3.post1 Documentation

    which CUDA kernel is causing the trouble. - Set the environment variable export NCCL_DEBUG=TRACE to turn on more logging for NCCL. - Set the environment variable export VLLM_TRACE_FUNCTION=1. All the function backend=nccl. The IP address should be the correct one. If not, override the IP address by setting the environment variable export VLLM_HOST_IP=your_ip_address. You might also need to set export NCCL_SOCK ```python import torch import torch.distributed as dist dist.init_process_group(backend="nccl") local_rank = dist.get_rank() % torch.cuda.device_count() data = torch.FloatTensor([1,] * 128)
    0 码力 | 143 页 | 1.07 MB | 3 月前
    3

  • pdf文档 vLLM v0.5.3 Documentation

    which CUDA kernel is causing the trouble. - Set the environment variable export NCCL_DEBUG=TRACE to turn on more logging for NCCL. - Set the environment variable export VLLM_TRACE_FUNCTION=1. All the function backend=nccl. The IP address should be the correct one. If not, override the IP address by setting the environment variable export VLLM_HOST_IP=your_ip_address. You might also need to set export NCCL_SOCK ```python import torch import torch.distributed as dist dist.init_process_group(backend="nccl") local_rank = dist.get_rank() % torch.cuda.device_count() data = torch.FloatTensor([1,] * 128)
    0 码力 | 143 页 | 1.07 MB | 3 月前
    3

  • pdf文档 vLLM v0.6.1 Documentation

    CUDA kernel is causing the trouble. - Set the environment variable export NCCL_DEBUG=TRACE to turn on more logging for NCCL. - Set the environment variable export VLLM_TRACE_FUNCTION=1. All the function setting the environment variable export VLLM_HOST_IP=your_ip_address. You might also need to set export NCCL_SOCKET_IFNAME=your_network_interface and export GLOO_SOCKET_IFNAME=your_network_interface to specify communication is working correctly. ```python # Test PyTorch NCCL import torch import torch.distributed as dist dist.init_process_group(backend="nccl") local_rank = dist.get_rank() % torch.cuda.device_count()
    0 码力 | 215 页 | 1.29 MB | 3 月前
    3
共 23 条
  • 1
  • 2
  • 3
前往
页
相关搜索词
PyTorchCUDAcuDNNNCCLDALIvLLM量化模型多模态模型分布式推理OpenAI兼容服务器LoRA AdapterVision Language ModelsPerformance TuningSampling Parameterspaged attention多模态数据连续批量处理预emption性能基准测试KV缓存管理模型集成参数配置PagedAttention量化分批处理LoRA adapterVision Language Models (VLMs)multi_modal_datapreemptionchunked prefillLLMperformance tuningKV cacheLoRA