got Unexpected error from cudaGetDeviceCount() after calling torch.cuda.is_available()

[Wait845]

🐛 Describe the bug

I am trying to use PyTorch on a Hyper-V VM. After installing the GPU driver, I created a new Conda environment and installed PyTorch using the following command:

conda install pytorch torchvision torchaudio pytorch-cuda=12.1 -c pytorch -c nvidia

However, when I started Python and called torch.cuda.is_available(), I received an error message. According to nvidia-smi, I have sufficient memory when calling is_available().

Can you please help me resolve this issue?

Versions

python3 collect_env.py
Collecting environment information...
***/miniconda3/envs/test_gpu/lib/python3.10/site-packages/torch/cuda/init.py:118: UserWarning: CUDA initialization: Unexpected error from cudaGetDeviceCount(). Did you run some cuda functions before calling NumCudaDevices() that might have already set an error? Error 2: out of memory (Triggered internally at /opt/conda/conda-bld/pytorch_1712608935911/work/c10/cuda/CUDAFunctions.cpp:108.)
return torch._C._cuda_getDeviceCount() > 0
PyTorch version: 2.3.0
Is debug build: False
CUDA used to build PyTorch: 12.1
ROCM used to build PyTorch: N/A

OS: Ubuntu 22.04.3 LTS (x86_64)
GCC version: (Ubuntu 11.4.0-1ubuntu1~22.04) 11.4.0
Clang version: Could not collect
CMake version: Could not collect
Libc version: glibc-2.35

Python version: 3.10.14 (main, Mar 21 2024, 16:24:04) [GCC 11.2.0] (64-bit runtime)
Python platform: Linux-5.15.0-105-generic-x86_64-with-glibc2.35
Is CUDA available: False
CUDA runtime version: 12.1.66
CUDA_MODULE_LOADING set to: N/A
GPU models and configuration: GPU 0: NVIDIA GeForce RTX 4060
Nvidia driver version: 552.22
cuDNN version: Could not collect
HIP runtime version: N/A
MIOpen runtime version: N/A
Is XNNPACK available: True

CPU:
Architecture: x86_64
CPU op-mode(s): 32-bit, 64-bit
Address sizes: 39 bits physical, 48 bits virtual
Byte Order: Little Endian
CPU(s): 6
On-line CPU(s) list: 0-5
Vendor ID: GenuineIntel
Model name: 12th Gen Intel(R) Core(TM) i5-12490F
CPU family: 6
Model: 151
Thread(s) per core: 2
Core(s) per socket: 3
Socket(s): 1
Stepping: 2
BogoMIPS: 5990.39
Flags: fpu vme de pse tsc msr pae mce cx8 apic sep mtrr pge mca cmov pat pse36 clflush mmx fxsr sse sse2 ss ht syscall nx pdpe1gb rdtscp lm constant_tsc rep_good nopl xtopology tsc_reliable nonstop_tsc cpuid aperfmperf pni pclmulqdq ssse3 fma cx16 sse4_1 sse4_2 x2apic movbe popcnt tsc_deadline_timer aes xsave avx f16c rdrand hypervisor lahf_lm abm 3dnowprefetch ssbd ibrs ibpb stibp ibrs_enhanced fsgsbase tsc_adjust bmi1 avx2 smep bmi2 erms invpcid rdseed adx smap clflushopt clwb sha_ni xsaveopt xsavec xgetbv1 xsaves avx_vnni umip waitpkg gfni vaes vpclmulqdq rdpid movdiri movdir64b fsrm serialize flush_l1d arch_capabilities
Hypervisor vendor: Microsoft
Virtualization type: full
L1d cache: 144 KiB (3 instances)
L1i cache: 96 KiB (3 instances)
L2 cache: 3.8 MiB (3 instances)
L3 cache: 20 MiB (1 instance)
NUMA node(s): 1
NUMA node0 CPU(s): 0-5
Vulnerability Gather data sampling: Not affected
Vulnerability Itlb multihit: Not affected
Vulnerability L1tf: Not affected
Vulnerability Mds: Not affected
Vulnerability Meltdown: Not affected
Vulnerability Mmio stale data: Not affected
Vulnerability Retbleed: Mitigation; Enhanced IBRS
Vulnerability Spec rstack overflow: Not affected
Vulnerability Spec store bypass: Mitigation; Speculative Store Bypass disabled via prctl and seccomp
Vulnerability Spectre v1: Mitigation; usercopy/swapgs barriers and __user pointer sanitization
Vulnerability Spectre v2: Mitigation; Enhanced IBRS, IBPB conditional, RSB filling, PBRSB-eIBRS SW sequence
Vulnerability Srbds: Not affected
Vulnerability Tsx async abort: Not affected

Versions of relevant libraries:
[pip3] numpy==1.26.4
[pip3] torch==2.3.0
[pip3] torchaudio==2.3.0
[pip3] torchvision==0.18.0
[pip3] triton==2.3.0
[conda] blas 1.0 mkl
[conda] ffmpeg 4.3 hf484d3e_0 pytorch
[conda] libjpeg-turbo 2.0.0 h9bf148f_0 pytorch
[conda] mkl 2023.1.0 h213fc3f_46344
[conda] mkl-service 2.4.0 py310h5eee18b_1
[conda] mkl_fft 1.3.8 py310h5eee18b_0
[conda] mkl_random 1.2.4 py310hdb19cb5_0
[conda] numpy 1.26.4 py310h5f9d8c6_0
[conda] numpy-base 1.26.4 py310hb5e798b_0
[conda] pytorch 2.3.0 py3.10_cuda12.1_cudnn8.9.2_0 pytorch
[conda] pytorch-cuda 12.1 ha16c6d3_5 pytorch
[conda] pytorch-mutex 1.0 cuda pytorch
[conda] torchaudio 2.3.0 py310_cu121 pytorch
[conda] torchtriton 2.3.0 py310 pytorch
[conda] torchvision 0.18.0 py310_cu121 pytorch

cc @seemethere @malfet @osalpekar @atalman

[malfet]

Hmm, can you run any cuda program inside your container? Judging by the error it's very likely that GPU access from your container is misconfigured

[Wait845]

Thank you for your response. Here are my details:

> nvcc -V
nvcc: NVIDIA (R) Cuda compiler driver
Copyright (c) 2005-2023 NVIDIA Corporation
Built on Tue_Feb__7_19:32:13_PST_2023
Cuda compilation tools, release 12.1, V12.1.66
Build cuda_12.1.r12.1/compiler.32415258_0

Here’s a simple CUDA test script (test.cu):

#test.cu
#include <stdio.h>

__global__ void add(int *a, int *b, int *c) {
    c[blockIdx.x] = a[blockIdx.x] + b[blockIdx.x];
}

int main() {
    int a[5] = {1, 2, 3, 4, 5};
    int b[5] = {10, 20, 30, 40, 50};
    int c[5] = {0};

    int *d_a, *d_b, *d_c;
    int size = sizeof(int) * 5;

    cudaMalloc((void **)&d_a, size);
    cudaMalloc((void **)&d_b, size);
    cudaMalloc((void **)&d_c, size);

    cudaMemcpy(d_a, a, size, cudaMemcpyHostToDevice);
    cudaMemcpy(d_b, b, size, cudaMemcpyHostToDevice);

    add<<<5,1>>>(d_a, d_b, d_c);

    cudaMemcpy(c, d_c, size, cudaMemcpyDeviceToHost);

    printf("结果：");
    for (int i = 0; i < 5; i++)
        printf("%d ", c[i]);
    printf("\n");

    cudaFree(d_a);
    cudaFree(d_b);
    cudaFree(d_c);

    return 0;
}

When I compiled and ran it:

> nvcc test.cu -o test
> ./test 
结果：0 0 0 0 0 

---

After many attempts, I found a solution: setting an environment variable before calling torch.cuda.is_available():

os.environ['CUDA_VISIBLE_DEVICES'] = "0"

I suspect this is because I moved the VM location from a host Windows system and reassigned the GPU to the VM. This resulted in two PCI devices showing up, but only one working:

lspci -v
63b6:00:00.0 3D controller: Microsoft Corporation Device 008e
	Physical Slot: 3509819646
	Flags: bus master, fast devsel, latency 0, NUMA node 0
	Capabilities: <access denied>
	Kernel driver in use: dxgkrnl
	Kernel modules: dxgkrnl

b330:00:00.0 3D controller: Microsoft Corporation Device 008e
	Physical Slot: 3646467553
	Flags: bus master, fast devsel, latency 0, NUMA node 0
	Capabilities: <access denied>
	Kernel driver in use: dxgkrnl
	Kernel modules: dxgkrnl

[jbschlosser]

As this seems to be an environment issue, I encourage you to post on the PyTorch forums if additional assistance is needed, as you're more likely to receive help there. Note that the issues here are reserved for bugs within PyTorch itself. Closing for now but feel free to reopen if forum discussion points to a PyTorch bug.