Bennison J comments

Repositories
Issues
Comments

Results 2 comments of


                                            Bennison J

trafficstars

Failed to detect a default CUDA architecture

> @arthurwolf You can try building using the following, it worked for me. > > `CUDACXX=/usr/local/cuda-12/bin/nvcc CMAKE_ARGS="-DLLAMA_CUBLAS=on -DCMAKE_CUDA_ARCHITECTURES=native" FORCE_CMAKE=1 pip install llama-cpp-python --no-cache-dir --force-reinstall --upgrade` Thanks, it worked

Support Mistral's new visual model: Pixtral-12b-240910

Is it available at Ollama now?