AssertionError: Torch not compiled with CUDA enabled

[alexgit256]

System Info

AM 5 platform: ryzen 9 7950x, RTX 4090, 64 GB RAM, Windows 11, Python 3.10.11. nvcc: NVIDIA (R) Cuda compiler driver Copyright (c) 2005-2024 NVIDIA Corporation Built on Wed_Apr_17_19:36:51_Pacific_Daylight_Time_2024 Cuda compilation tools, release 12.5, V12.5.40 Build cuda_12.5.r12.5/compiler.34177558_0

import torch torch.cuda.is_available() False

Reproduction

python -m bitsandbytes

Expected behavior

A successful test. Cuda is found. Gpu support present. 8-bit optimizers, 8-bit multiplication, and GPU quantization are available. (sd3lora) venv output.txt

[matthewdouglas]

Hi @alexgit256, Make sure to install NumPy 1.x, using pip install numpy<2. Additionally, make sure you have a CUDA-enabled version of PyTorch installed. On Windows the default is a CPU wheel. You can follow the instructions for PyTorch here: https://pytorch.org/get-started/locally/

[alexgit256]

Hello, matthewdouglas, After proceeding through the instructions, all worked perfectly well! Thank you!