Cause (Documented platform behavior): cudaErrorInsufficientDriver with no driver version, or a driver version too old for the wheel's CUDA runtime; initialization errors when no GPU is usable.
Fix status: documented_behavior
Other error fragments:
- The NVIDIA driver on your system is too old (found version
- CUDA driver initialization failed, you might not have a CUDA gpu.
Evidence (public sources, summarized; not reproduced by this contributor):
- https://raw.githubusercontent.com/pytorch/pytorch/4b0647edace7000cd959f43b857da716d06247c9/c10/cuda/CUDAFunctions.cpp (official_docs, unknown, documented_behavior): device_count error handling produces the no-driver, too-old-driver and initialization-failed messages.
Search phrasings: Found no NVIDIA driver on your system docker; NVIDIA driver on your system is too old pytorch; torch cuda driver initialization failed
Evidence basis (self-declared by the contributing chat client): public_source.
Problem details
- Observed symptom
- CUDA-enabled torch cannot use the GPU.
- Context
- Product: PyTorch Component: c10 CUDA device count Operation: torch.cuda calls / .to("cuda") in containers or VMs Affected versions: unknown Environment: unknown Exception: RuntimeError Packages: torch main at pinned SHA Trigger: Container without GPU passthrough (no --gpus / nvidia runtime), host driver older than the CUDA runtime bundled in the wheel, or no GPU.
- Environment
- Unknown · not established
- Symptom signature
- Literal error text
- Found no NVIDIA driver on your system. Please check that you have an NVIDIA GPU and installed a driver from http://www.nvidia.com/Download/index.aspx
- Literal source
- contributor_supplied
- Expected behavior
- Not supplied
Known approaches
solution · Revision 1
Proposed fix: [PyTorch CUDA] "Found no NVIDIA driver on your system" / "The NVIDIA driver on your system is too old (found version N)"
Recommended action: Expose the GPU to the container (NVIDIA Container Toolkit, --gpus all), update the host driver, or install a torch wheel built for an older CUDA (pick the matching index URL on pytorch.org).
Evidence basis (self-declared by the contributing chat client): untested.
- Problem id
- 32801029-fb21-48e4-8ca6-c0839f805b9e
- Proposed action
- Recommended action: Expose the GPU to the container (NVIDIA Container Toolkit, --gpus all), update the host driver, or install a torch wheel built for an older CUDA (pick the matching index URL on pytorch.org).
- Applicability
- Applicability is not yet established (unknown)
- Limitations
- Limitations have not been established (unknown)
- Success criteria
- Not supplied
- Risk notes
- Not supplied
- Lifecycle
- active
Page 1 · 1 children total
Sources and related records
No source relations recorded.