CUDA安装错误：找不到libnvJitLink.so.13
我正在尝试在Google Colab上加载Qwen。
同时在GitHub的 bitsandbytes foundation上提出了这个问题 -> https://github.com/bitsandbytes-foundation/bitsandbytes/issues/1905#issuecomment-4126552960
单元1：
!pip install -U "transformers>=4.44.0" "accelerate>=0.32.0" "bitsandbytes>=0.49.1" "torch>=2.3.0" "safetensors>=0.4.0" "torchvision" "torchaudio"
import os, sys
print("✅ Packages upgraded. Restarting runtime...")
os.execv(sys.executable, ['python'] + sys.argv)
单元1 的结果：
Requirement already satisfied: transformers>=4.44.0 in /usr/local/lib/python3.12/dist-packages (5.0.0)
Collecting transformers>=4.44.0
Downloading transformers-5.3.0-py3-none-any.whl.metadata (32 kB)
Requirement already satisfied: accelerate>=0.32.0 in /usr/local/lib/python3.12/dist-packages (1.13.0)
Collecting bitsandbytes>=0.49.1
Downloading bitsandbytes-0.49.2-py3-none-manylinux_2_24_x86_64.whl.metadata (10 kB)
Requirement already satisfied: torch>=2.3.0 in /usr/local/lib/python3.12/dist-packages (2.10.0+cu128)
Collecting torch>=2.3.0
Downloading torch-2.11.0-cp312-cp312-manylinux_2_28_x86_64.whl.metadata (29 kB)
Requirement already satisfied: safetensors>=0.4.0 in /usr/local/lib/python3.12/dist-packages (0.7.0)
Requirement already satisfied: torchvision in /usr/local/lib/python3.12/dist-packages (0.25.0+cu128)
Collecting torchvision
Downloading torchvision-0.26.0-cp312-cp312-manylinux_2_28_x86_64.whl.metadata (5.5 kB)
Requirement already satisfied: torchaudio in /usr/local/lib/python3.12/dist-packages (2.10.0+cu128)
Collecting torchaudio
Downloading torchaudio-2.11.0-cp312-cp312-manylinux_2_28_x86_64.whl.metadata (6.9 kB)
Requirement already satisfied: huggingface-hub<2.0,>=1.3.0 in /usr/local/lib/python3.12/dist-packages (from transformers>=4.44.0) (1.7.1)
Requirement already satisfied: numpy>=1.17 in /usr/local/lib/python3.12/dist-packages (from transformers>=4.44.0) (2.0.2)
Requirement already satisfied: packaging>=20.0 in /usr/local/lib/python3.12/dist-packages (from transformers>=4.44.0) (26.0)
Requirement already satisfied: pyyaml>=5.1 in /usr/local/lib/python3.12/dist-packages (from transformers>=4.44.0) (6.0.3)
Requirement already satisfied: regex!=2019.12.17 in /usr/local/lib/python3.12/dist-packages (from transformers>=4.44.0) (2025.11.3)
Requirement already satisfied: tokenizers<=0.23.0,>=0.22.0 in /usr/local/lib/python3.12/dist-packages (from transformers>=4.44.0) (0.22.2)
Requirement already satisfied: typer in /usr/local/lib/python3.12/dist-packages (from transformers>=4.44.0) (0.24.1)
Requirement already satisfied: tqdm>=4.27 in /usr/local/lib/python3.12/dist-packages (from transformers>=4.44.0) (4.67.3)
Requirement already satisfied: psutil in /usr/local/lib/python3.12/dist-packages (from accelerate>=0.32.0) (5.9.5)
Requirement already satisfied: filelock in /usr/local/lib/python3.12/dist-packages (from torch>=2.3.0) (3.25.2)
Requirement already satisfied: typing-extensions>=4.10.0 in /usr/local/lib/python3.12/dist-packages (from torch>=2.3.0) (4.15.0)
Requirement already satisfied: setuptools<82 in /usr/local/lib/python3.12/dist-packages (from torch>=2.3.0) (75.2.0)
Requirement already satisfied: sympy>=1.13.3 in /usr/local/lib/python3.12/dist-packages (from torch>=2.3.0) (1.14.0)
Requirement already satisfied: networkx>=2.5.1 in /usr/local/lib/python3.12/dist-packages (from torch>=2.3.0) (3.6.1)
Requirement already satisfied: jinja2 in /usr/local/lib/python3.12/dist-packages (from torch>=2.3.0) (3.1.6)
Requirement already satisfied: fsspec>=0.8.5 in /usr/local/lib/python3.12/dist-packages (from torch>=2.3.0) (2025.3.0)
Collecting cuda-toolkit==13.0.2 (from cuda-toolkit[cublas,cudart,cufft,cufile,cupti,curand,cusolver,cusparse,nvjitlink,nvrtc,nvtx]==13.0.2; platform_system == "Linux"->torch>=2.3.0)
Downloading cuda_toolkit-13.0.2-py2.py3-none-any.whl.metadata (9.4 kB)
Collecting cuda-bindings<14,>=13.0.3 (from torch>=2.3.0)
Downloading cuda_bindings-13.2.0-cp312-cp312-manylinux_2_24_x86_64.manylinux_2_28_x86_64.whl.metadata (2.3 kB)
Collecting nvidia-cudnn-cu13==9.19.0.56 (from torch>=2.3.0)
Downloading nvidia_cudnn_cu13-9.19.0.56-py3-none-manylinux_2_27_x86_64.whl.metadata (1.9 kB)
Collecting nvidia-cusparselt-cu13==0.8.0 (from torch>=2.3.0)
Downloading nvidia_cusparselt_cu13-0.8.0-py3-none-manylinux2014_x86_64.whl.metadata (12 kB)
Collecting nvidia-nccl-cu13==2.28.9 (from torch>=2.3.0)
Downloading nvidia_nccl_cu13-2.28.9-py3-none-manylinux_2_18_x86_64.whl.metadata (2.0 kB)
Collecting nvidia-nvshmem-cu13==3.4.5 (from torch>=2.3.0)
Downloading nvidia_nvshmem_cu13-3.4.5-py3-none-manylinux2014_x86_64.manylinux_2_17_x86_64.whl.metadata (2.1 kB)
Requirement already satisfied: triton==3.6.0 in /usr/local/lib/python3.12/dist-packages (from torch>=2.3.0) (3.6.0)
Collecting nvidia-cublas==13.1.0.3.* (from cuda-toolkit[cublas,cudart,cufft,cufile,cupti,curand,cusolver,cusparse,nvjitlink,nvrtc,nvtx]==13.0.2; platform_system == "Linux"->torch>=2.3.0)
Downloading nvidia_cublas-13.1.0.3-py3-none-manylinux_2_27_x86_64.whl.metadata (1.7 kB)
Collecting nvidia-cuda-runtime==13.0.96.* (from cuda-toolkit[cublas,cudart,cufft,cufile,cupti,curand,cusolver,cusparse,nvjitlink,nvrtc,nvtx]==13.0.2; platform_system == "Linux"->torch>=2.3.0)
Downloading nvidia_cuda_runtime-13.0.96-py3-none-manylinux2014_x86_64.manylinux_2_17_x86_64.whl.metadata (1.7 kB)
Collecting nvidia-cufft==12.0.0.61.* (from cuda-toolkit[cublas,cudart,cufft,cufile,cupti,curand,cusolver,cusparse,nvjitlink,nvrtc,nvtx]==13.0.2; platform_system == "Linux"->torch>=2.3.0)
Downloading nvidia_cufft-12.0.0.61-py3-none-manylinux2014_x86_64.manylinux_2_17_x86_64.whl.metadata (1.8 kB)
Collecting nvidia-cufile==1.15.1.6.* (from cuda-toolkit[cublas,cudart,cufft,cufile,cupti,curand,cusolver,cusparse,nvjitlink,nvrtc,nvtx]==13.0.2; platform_system == "Linux"->torch>=2.3.0)
Downloading nvidia_cufile-1.15.1.6-py3-none-manylinux2014_x86_64.manylinux_2_17_x86_64.whl.metadata (1.7 kB)
Collecting nvidia-cuda-cupti==13.0.85.* (from cuda-toolkit[cublas,cudart,cufft,cufile,cupti,curand,cusolver,cusparse,nvjitlink,nvrtc,nvtx]==13.0.2; platform_system == "Linux"->torch>=2.3.0)
Downloading nvidia_cuda_cupti-13.0.85-py3-none-manylinux_2_25_x86_64.whl.metadata (1.7 kB)
Collecting nvidia-curand==10.4.0.35.* (from cuda-toolkit[cublas,cudart,cufft,cufile,cupti,curand,cusolver,cusparse,nvjitlink,nvrtc,nvtx]==13.0.2; platform_system == "Linux"->torch>=2.3.0)
Downloading nvidia_curand-10.4.0.35-py3-none-manylinux_2_27_x86_64.whl.metadata (1.7 kB)
Collecting nvidia-cusolver==12.0.4.66.* (from cuda-toolkit[cublas,cudart,cufft,cufile,cupti,curand,cusolver,cusparse,nvjitlink,nvrtc,nvtx]==13.0.2; platform_system == "Linux"->torch>=2.3.0)
Downloading nvidia_cusolver-12.0.4.66-py3-none-manylinux_2_27_x86_64.whl.metadata (1.8 kB)
Collecting nvidia-cusparse==12.6.3.3.* (from cuda-toolkit[cublas,cudart,cufft,cufile,cupti,curand,cusolver,cusparse,nvjitlink,nvrtc,nvtx]==13.0.2; platform_system == "Linux"->torch>=2.3.0)
Downloading nvidia_cusparse-12.6.3.3-py3-none-manylinux2014_x86_64.manylinux_2_17_x86_64.whl.metadata (1.8 kB)
Collecting nvidia-nvjitlink==13.0.88.* (from cuda-toolkit[cublas,cudart,cufft,cufile,cupti,curand,cusolver,cusparse,nvjitlink,nvrtc,nvtx]==13.0.2; platform_system == "Linux"->torch>=2.3.0)
Downloading nvidia_nvjitlink-13.0.88-py3-none-manylinux2010_x86_64.manylinux_2_12_x86_64.whl.metadata (1.7 kB)
Collecting nvidia-cuda-nvrtc==13.0.88.* (from cuda-toolkit[cublas,cudart,cufft,cufile,cupti,curand,cusolver,cusparse,nvjitlink,nvrtc,nvtx]==13.0.2; platform_system == "Linux"->torch>=2.3.0)
Downloading nvidia_cuda_nvrtc-13.0.88-py3-none-manylinux2010_x86_64.manylinux_2_12_x86_64.whl.metadata (1.7 kB)
Collecting nvidia-nvtx==13.0.85.* (from cuda-toolkit[cublas,cudart,cufft,cufile,cupti,curand,cusolver,cusparse,nvjitlink,nvrtc,nvtx]==13.0.2; platform_system == "Linux"->torch>=2.3.0)
Downloading nvidia_nvtx-13.0.85-py3-none-manylinux1_x86_64.manylinux_2_5_x86_64.whl.metadata (1.8 kB)
Requirement already satisfied: pillow!=8.3.*,>=5.3.0 in /usr/local/lib/python3.12/dist-packages (from torchvision) (11.3.0)
Requirement already satisfied: cuda-pathfinder~=1.1 in /usr/local/lib/python3.12/dist-packages (from cuda-bindings<14,>=13.0.3->torch>=2.3.0) (1.4.3)
Requirement already satisfied: hf-xet<2.0.0,>=1.4.2 in /usr/local/lib/python3.12/dist-packages (from huggingface-hub<2.0,>=1.3.0->transformers>=4.44.0) (1.4.2)
Requirement already satisfied: httpx<1,>=0.23.0 in /usr/local/lib/python3.12/dist-packages (from huggingface-hub<2.0,>=1.3.0->transformers>=4.44.0) (0.28.1)
Requirement already satisfied: mpmath<1.4,>=1.1.0 in /usr/local/lib/python3.12/dist-packages (from sympy>=1.13.3->torch>=2.3.0) (1.3.0)
Requirement already satisfied: MarkupSafe>=2.0 in /usr/local/lib/python3.12/dist-packages (from jinja2->torch>=2.3.0) (3.0.3)
Requirement already satisfied: click>=8.2.1 in /usr/local/lib/python3.12/dist-packages (from typer->transformers>=4.44.0) (8.3.1)
Requirement already satisfied: shellingham>=1.3.0 in /usr/local/lib/python3.12/dist-packages (from typer->transformers>=4.44.0) (1.5.4)
Requirement already satisfied: rich>=12.3.0 in /usr/local/lib/python3.12/dist-packages (from typer->transformers>=4.44.0) (13.9.4)
Requirement already satisfied: annotated-doc>=0.0.2 in /usr/local/lib/python3.12/dist-packages (from typer->transformers>=4.44.0) (0.0.4)
Requirement already satisfied: anyio in /usr/local/lib/python3.12/dist-packages (from httpx<1,>=0.23.0->huggingface-hub<2.0,>=1.3.0->transformers>=4.44.0) (4.12.1)
Requirement already satisfied: certifi in /usr/local/lib/python3.12/dist-packages (from httpx<1,>=0.23.0->huggingface-hub<2.0,>=1.3.0->transformers>=4.44.0) (2026.2.25)
Requirement already satisfied: httpcore==1.* in /usr/local/lib/python3.12/dist-packages (from httpx<1,>=0.23.0->huggingface-hub<2.0,>=1.3.0->transformers>=4.44.0) (1.0.9)
Requirement already satisfied: idna in /usr/local/lib/python3.12/dist-packages (from httpx<1,>=0.23.0->huggingface-hub<2.0,>=1.3.0->transformers>=4.44.0) (3.11)
Requirement already satisfied: h11>=0.16 in /usr/local/lib/python3.12/dist-packages (from httpcore==1.*->httpx<1,>=0.23.0->huggingface-hub<2.0,>=1.3.0->transformers>=4.44.0) (0.16.0)
Requirement already satisfied: markdown-it-py>=2.2.0 in /usr/local/lib/python3.12/dist-packages (from rich>=12.3.0->typer->transformers>=4.44.0) (4.0.0)
Requirement already satisfied: pygments<3.0.0,>=2.13.0 in /usr/local/lib/python3.12/dist-packages (from rich>=12.3.0->typer->transformers>=4.44.0) (2.19.2)
Requirement already satisfied: mdurl~=0.1 in /usr/local/lib/python3.12/dist-packages (from markdown-it-py>=2.2.0->rich>=12.3.0->typer->transformers>=4.44.0) (0.1.2)
Downloading transformers-5.3.0-py3-none-any.whl (10.7 MB)
━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━ 10.7/10.7 MB 93.6 MB/s eta 0:00:00
Downloading bitsandbytes-0.49.2-py3-none-manylinux_2_24_x86_64.whl (60.7 MB)
━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━ 60.7/60.7 MB 44.8 MB/s eta 0:00:00
Downloading torch-2.11.0-cp312-cp312-manylinux_2_28_x86_64.whl (530.7 MB)
━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━ 530.7/530.7 MB 2.6 MB/s eta 0:00:00
Downloading cuda_toolkit-13.0.2-py2.py3-none-any.whl (2.4 kB)
Downloading nvidia_cudnn_cu13-9.19.0.56-py3-none-manylinux_2_27_x86_64.whl (366.1 MB)
━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━ 366.1/366.1 MB 3.1 MB/s eta 0:00:00
Downloading nvidia_cusparselt_cu13-0.8.0-py3-none-manylinux2014_x86_64.whl (169.9 MB)
━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━ 169.9/169.9 MB 10.0 MB/s eta 0:00:00
Downloading nvidia_nccl_cu13-2.28.9-py3-none-manylinux_2_18_x86_64.whl (196.5 MB)
━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━ 196.5/196.5 MB 4.8 MB/s eta 0:00:00
Downloading nvidia_nvshmem_cu13-3.4.5-py3-none-manylinux2014_x86_64.manylinux_2_17_x86_64.whl (60.4 MB)
━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━ 60.4/60.4 MB 43.2 MB/s eta 0:00:00
Downloading nvidia_cublas-13.1.0.3-py3-none-manylinux_2_27_x86_64.whl (423.1 MB)
━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━ 423.1/423.1 MB 2.8 MB/s eta 0:00:00
Downloading nvidia_cuda_cupti-13.0.85-py3-none-manylinux_2_25_x86_64.whl (10.7 MB)
━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━ 10.7/10.7 MB 132.0 MB/s eta 0:00:00
Downloading nvidia_cuda_nvrtc-13.0.88-py3-none-manylinux2010_x86_64.manylinux_2_12_x86_64.whl (90.2 MB)
━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━ 90.2/90.2 MB 28.9 MB/s eta 0:00:00
Downloading nvidia_cuda_runtime-13.0.96-py3-none-manylinux2014_x86_64.manylinux_2_17_x86_64.whl (2.2 MB)
━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━ 2.2/2.2 MB 90.9 MB/s eta 0:00:00
Downloading nvidia_cufft-12.0.0.61-py3-none-manylinux2014_x86_64.manylinux_2_17_x86_64.whl (214.1 MB)
━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━ 214.1/214.1 MB 4.6 MB/s eta 0:00:00
Downloading nvidia_cufile-1.15.1.6-py3-none-manylinux2014_x86_64.manylinux_2_17_x86_64.whl (1.2 MB)
━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━ 1.2/1.2 MB 72.8 MB/s eta 0:00:00
Downloading nvidia_curand-10.4.0.35-py3-none-manylinux_2_27_x86_64.whl (59.5 MB)
━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━ 59.5/59.5 MB 44.4 MB/s eta 0:00:00
Downloading nvidia_cusolver-12.0.4.66-py3-none-manylinux_2_27_x86_64.whl (200.9 MB)
━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━ 200.9/200.9 MB 4.8 MB/s eta 0:00:00
Downloading nvidia_cusparse-12.6.3.3-py3-none-manylinux2014_x86_64.manylinux_2_17_x86_64.whl (145.9 MB)
━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━ 145.9/145.9 MB 17.9 MB/s eta 0:00:00
Downloading nvidia_nvjitlink-13.0.88-py3-none-manylinux2010_x86_64.manylinux_2_12_x86_64.whl (40.7 MB)
━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━ 40.7/40.7 MB 63.0 MB/s eta 0:00:00
Downloading nvidia_nvtx-13.0.85-py3-none-manylinux1_x86_64.manylinux_2_5_x86_64.whl (148 kB)
━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━ 148.0/148.0 kB 14.9 MB/s eta 0:00:00
Downloading torchvision-0.26.0-cp312-cp312-manylinux_2_28_x86_64.whl (7.5 MB)
━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━ 7.5/7.5 MB 104.3 MB/s eta 0:00:00
Downloading torchaudio-2.11.0-cp312-cp312-manylinux_2_28_x86_64.whl (1.8 MB)
━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━ 1.8/1.8 MB 97.6 MB/s eta 0:00:00
Downloading cuda_bindings-13.2.0-cp312-cp312-manylinux_2_24_x86_64.manylinux_2_28_x86_64.whl (6.3 MB)
━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━ 6.3/6.3 MB 149.2 MB/s eta 0:00:00
Installing collected packages: torchaudio, nvidia-cusparselt-cu13, cuda-toolkit, nvidia-nvtx, nvidia-nvshmem-cu13, nvidia-nvjitlink, nvidia-nccl-cu13, nvidia-curand, nvidia-cufile, nvidia-cuda-runtime, nvidia-cuda-nvrtc, nvidia-cuda-cupti, nvidia-cublas, cuda-bindings, nvidia-cusparse, nvidia-cufft, nvidia-cudnn-cu13, nvidia-cusolver, torch, transformers, torchvision, bitsandbytes
Attempting uninstall: torchaudio
Found existing installation: torchaudio 2.10.0+cu128
Uninstalling torchaudio-2.10.0+cu128:
Successfully uninstalled torchaudio-2.10.0+cu128
Attempting uninstall: cuda-toolkit
Found existing installation: cuda-toolkit 12.8.1
Uninstalling cuda-toolkit-12.8.1:
Successfully uninstalled cuda-toolkit-12.8.1
Attempting uninstall: cuda-bindings
Found existing installation: cuda-bindings 12.9.4
Uninstalling cuda-bindings-12.9.4:
Successfully uninstalled cuda-bindings-12.9.4
Attempting uninstall: torch
Found existing installation: torch 2.10.0+cu128
Uninstalling torch-2.10.0+cu128:
Successfully uninstalled torch-2.10.0+cu128
Attempting uninstall: transformers
Found existing installation: transformers 5.0.0
Uninstalling transformers-5.0.0:
Successfully uninstalled transformers-5.0.0
Attempting uninstall: torchvision
Found existing installation: torchvision 0.25.0+cu128
Uninstalling torchvision-0.25.0+cu128:
Successfully uninstalled torchvision-0.25.0+cu128
ERROR: pip's dependency resolver does not currently take into account all the packages that are installed. This behaviour is the source of the following dependency conflicts.
cuda-python 12.9.4 requires cuda-bindings~=12.9.4, but you have cuda-bindings 13.2.0 which is incompatible.
libcuml-cu12 26.2.0 requires cuda-toolkit[cublas,cufft,curand,cusolver,cusparse]==12.*, but you have cuda-toolkit 13.0.2 which is incompatible.
libraft-cu12 26.2.0 requires cuda-toolkit[cublas,curand,cusolver,cusparse]==12.*, but you have cuda-toolkit 13.0.2 which is incompatible.
cudf-cu12 26.2.1 requires cuda-toolkit[nvcc,nvrtc]==12.*, but you have cuda-toolkit 13.0.2 which is incompatible.
libcuvs-cu12 26.2.0 requires cuda-toolkit[cublas,curand,cusolver,cusparse]==12.*, but you have cuda-toolkit 13.0.2 which is incompatible.
cuml-cu12 26.2.0 requires cuda-toolkit[cublas,cufft,curand,cusolver,cusparse]==12.*, but you have cuda-toolkit 13.0.2 which is incompatible.
Successfully installed bitsandbytes-0.49.2 cuda-bindings-13.2.0 cuda-toolkit-13.0.2 nvidia-cublas-13.1.0.3 nvidia-cuda-cupti-13.0.85 nvidia-cuda-nvrtc-13.0.88 nvidia-cuda-runtime-13.0.96 nvidia-cudnn-cu13-9.19.0.56 nvidia-cufft-12.0.0.61 nvidia-cufile-1.15.1.6 nvidia-curand-10.4.0.35 nvidia-cusolver-12.0.4.66 nvidia-cusparse-12.6.3.3 nvidia-cusparselt-cu13-0.8.0 nvidia-nccl-cu13-2.28.9 nvidia-nvjitlink-13.0.88 nvidia-nvshmem-cu13-3.4.5 nvidia-nvtx-13.0.85 torch-2.11.0 torchaudio-2.11.0 torchvision-0.26.0 transformers-5.3.0
单元2：
import torch, os, sys
print("torch:", torch.__version__)
print("cuda available:", torch.cuda.is_available())
print("torch cuda:", torch.version.cuda)
!python -m bitsandbytes
!nvidia-smi
单元2 的结果：
torch: 2.11.0+cu130
cuda available: True
torch cuda: 13.0
bitsandbytes library load error: libnvJitLink.so.13: cannot open shared object file: No such file or directory
Traceback (most recent call last):
File "/usr/local/lib/python3.12/dist-packages/bitsandbytes/cextension.py", line 320, in <module>
lib = get_native_library()
^^^^^^^^^^^^^^^^^^^^
File "/usr/local/lib/python3.12/dist-packages/bitsandbytes/cextension.py", line 298, in get_native_library
dll = ct.cdll.LoadLibrary(str(binary_path))
^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^
File "/usr/lib/python3.12/ctypes/__init__.py", line 460, in LoadLibrary
return self._dlltype(name)
^^^^^^^^^^^^^^^^^^^
File "/usr/lib/python3.12/ctypes/__init__.py", line 379, in __init__
self._handle = _dlopen(self._name, mode)
^^^^^^^^^^^^^^^^^^^^^^^^^
OSError: libnvJitLink.so.13: cannot open shared object file: No such file or directory
=================== bitsandbytes v0.49.2 ===================
Platform: Linux-6.6.113+-x86_64-with-glibc2.35
libc: glibc-2.35
Python: 3.12.12
PyTorch: 2.11.0+cu130
CUDA: 13.0
HIP: N/A
XPU: N/A
Related packages:
accelerate: 1.13.0
diffusers: 0.37.0
numpy: 2.0.2
pip: 24.1.2
peft: 0.18.1
safetensors: 0.7.0
transformers: 5.3.0
triton: 3.6.0
trl: not found
============================================================
PyTorch settings found: CUDA_VERSION=130, Highest Compute Capability: (7, 5).
Checking that the library is importable and CUDA is callable...
Traceback (most recent call last):
File "<frozen runpy>", line 198, in _run_module_as_main
File "<frozen runpy>", line 88, in _run_code
File "/usr/local/lib/python3.12/dist-packages/bitsandbytes/__main__.py", line 4, in <module>
main()
File "/usr/local/lib/python3.12/dist-packages/bitsandbytes/diagnostics/main.py", line 107, in main
raise e
File "/usr/local/lib/python3.12/dist-packages/bitsandbytes/diagnostics/main.py", line 96, in main
sanity_check()
File "/usr/local/lib/python3.12/dist-packages/bitsandbytes/diagnostics/main.py", line 40, in sanity_check
adam.step()
File "/usr/local/lib/python3.12/dist-packages/torch/optim/optimizer.py", line 533, in wrapper
out = func(*args, **kwargs)
^^^^^^^^^^^^^^^^^^^^^
File "/usr/local/lib/python3.12/dist-packages/torch/utils/_contextlib.py", line 124, in decorate_context
return func(*args, **kwargs)
^^^^^^^^^^^^^^^^^^^^^
File "/usr/local/lib/python3.12/dist-packages/bitsandbytes/optim/optimizer.py", line 328, in step
self.update_step(group, p, gindex, pindex)
File "/usr/local/lib/python3.12/dist-packages/torch/utils/_contextlib.py", line 124, in decorate_context
return func(*args, **kwargs)
^^^^^^^^^^^^^^^^^^^^^
File "/usr/local/lib/python3.12/dist-packages/bitsandbytes/optim/optimizer.py", line 557, in update_step
F.optimizer_update_32bit(
File "/usr/local/lib/python3.12/dist-packages/bitsandbytes/functional.py", line 1235, in optimizer_update_32bit
torch.ops.bitsandbytes.optimizer_update_32bit(
File "/usr/local/lib/python3.12/dist-packages/torch/_ops.py", line 1269, in __call__
return self._op(*args, **kwargs)
^^^^^^^^^^^^^^^^^^^^^^^^^
File "/usr/local/lib/python3.12/dist-packages/torch/_compile.py", line 54, in inner
return disable_fn(*args, **kwargs)
^^^^^^^^^^^^^^^^^^^^^^^^^^^
File "/usr/local/lib/python3.12/dist-packages/torch/_dynamo/eval_frame.py", line 1263, in _fn
return fn(*args, **kwargs)
^^^^^^^^^^^^^^^^^^^
File "/usr/local/lib/python3.12/dist-packages/torch/library.py", line 751, in func_no_dynamo
return func(*args, **kwargs)
^^^^^^^^^^^^^^^^^^^^^
File "/usr/local/lib/python3.12/dist-packages/bitsandbytes/backends/cuda/ops.py", line 654, in _optimizer_update_32bit_impl
optim_func(
File "/usr/local/lib/python3.12/dist-packages/bitsandbytes/cextension.py", line 269, in throw_on_call
raise RuntimeError(f"{self.formatted_error}Native code method attempted to call: lib.{name}()")
RuntimeError:
🚨 CUDA SETUP ERROR: Missing dependency: libnvJitLink.so.13 🚨
CUDA 13.x runtime libraries were not found in the LD_LIBRARY_PATH.
To fix this, make sure that:
1. You have installed CUDA 13.x toolkit on your system
2. The CUDA runtime libraries are in your LD_LIBRARY_PATH
You can add them with (and persist the change by adding the line to your .bashrc):
export LD_LIBRARY_PATH=$LD_LIBRARY_PATH:/path/to/cuda-13.x/ lib64
Original error: libnvJitLink.so.13: cannot open shared object file: No such file or directory
🔍 Run this command for detailed diagnostics:
python -m bitsandbytes
If you've tried everything and still have issues:
1. Include ALL version info (operating system, bitsandbytes, pytorch, cuda, python)
2. Describe what you've tried in detail
3. Open an issue with this information:
https://github.com/bitsandbytes-foundation/bitsandbytes/issues
Native code method attempted to call: lib.cadam32bit_grad_fp32()
Tue Mar 24 05:50:05 2026
+-----------------------------------------------------------------------------------------+
| NVIDIA-SMI 580.82.07 Driver Version: 580.82.07 CUDA Version: 13.0 |
+-----------------------------------------+------------------------+----------------------+
| GPU Name Persistence-M | Bus-Id Disp.A | Volatile Uncorr. ECC |
| Fan Temp Perf Pwr:Usage/Cap | Memory-Usage | GPU-Util Compute M. |
| | | MIG M. |
|=========================================+========================+======================|
| 0 Tesla T4 Off | 00000000:00:04.0 Off | 0 |
| N/A 46C P0 28W / 70W | 3MiB / 15360MiB | 3% Default |
| | | N/A |
+-----------------------------------------+------------------------+----------------------+
+-----------------------------------------------------------------------------------------+
| Processes: |
| GPU GI CI PID Type Process name GPU Memory |
| ID ID Usage |
|=========================================================================================|
| No running processes found |
+-----------------------------------------------------------------------------------------+
更多背景信息：
在此之前，Bitsandbytes一直有问题。 我起初遇到了版本不匹配的问题。我看到有评论建议使用bitsandbytes的最新版，但那并没有解决我的问题。我不得不经过多轮安装，才得到这个cell 1的版本，能使用一段时间（3个月）。不得不说，那个过程非常临时。
但现在我的环境又开始出现问题。我可能现在在一个不同的容器里，因此又遇到了这个关于bitsandbytes的新问题。
据我所知，这似乎是CUDA的底层问题，以及与bitsandbytes的接口问题。我对环境的控制并不多，所以我能做的只有版本变更，对吗？
我非常希望得到各位的帮助。提前感谢。
解决方案
以下内容来自
matthewdouglas
!pip uninstall -y torch torchvision torchaudio
!pip install -U \
"torch==2.11.0" \
"torchvision==0.26.0" \
"torchaudio==2.11.0" \
--index-url https://download.pytorch.org/whl/cu128
我安装了带有Torch 2.11的 CUDA 12.8 wheel，bitsandbytes已经可以工作了。
站内所有文章版权归属LeftHeroAI导航站，无授权禁止任何主体转载、抄袭、复制内容，亦不得私自架设镜像站点。一经侵权，本站将通过法律途径追责。
相关文章
汇集全球数千款 AI 工具，每日持续更新，配套 AI 开发学习资源，轻松用 AI 提升工作效率。快捷键 Ctrl+D/⌘+D 一键收藏本站。