常见方式:
ollama: command not found原因:没装或没加入 PATH
解决:
curl -fsSL https://ollama.com/install.sh | sh然后:
source ~/.bashrcpull model manifest: 404原因:
解决:
ollama run llama3或指定版本:
ollama run llama3:8b国内网络问题可:
报错示例:
CUDA out of memory解决:
ollama run llama3 --num_gpu 1OLLAMA_NUM_GPU=0 ollama run llama3Repository Not Found原因:
解决:
pip install huggingface_hub
huggingface-cli login然后去:
https://huggingface.co/meta-llama/Llama-3-8B
torch not compiled with CUDA解决:
pip install torch torchvision --index-url https://download.pytorch.org/whl/cu121llama_model_load: failedillegal instruction解决:
make LLAMA_CUDA=1make LLAMA_AVX2=0nvidia-smi请贴:
我可以一步一步带你修好。