Hello,
I have 2 Quadro K4200 nvidia GPUs. They have a compute capability (CC) of 3.0. I’ve found many posting that indicate that I need to build PyTorch from source because a CC of 3.0 is too old. I have now built from source but am getting this common error:
RuntimeError: CUDA error: no kernel image is available for execution on the device
One posting indicated that I needed to modify, ~/torch/utils/cpp_extension.py, to specify my needed architecture. I found the following related code,
named_arches = collections.OrderedDict([
('Kepler+Tesla', '3.7'),
('Kepler', '3.5+PTX'),
('Maxwell+Tegra', '5.3'),
('Maxwell', '5.0;5.2+PTX'),
('Pascal', '6.0;6.1+PTX'),
('Volta', '7.0+PTX'),
('Turing', '7.5+PTX'),
('Ampere', '8.0+PTX'),
])
supported_arches = ['3.5', '3.7', '5.0', '5.2', '5.3', '6.0', '6.1', '6.2',
'7.0', '7.2', '7.5', '8.0']
but am unsure how to change this for my device.
Any advice here would be appreciated.
Some system information:
(SLLML-dev) joehays-local@sedacs17lx:~$ lspci | grep -i nvidia
03:00.0 VGA compatible controller: NVIDIA Corporation GK104GL [Quadro K4200] (rev a1)
03:00.1 Audio device: NVIDIA Corporation GK104 HDMI Audio Controller (rev a1)
04:00.0 VGA compatible controller: NVIDIA Corporation GK104GL [Quadro K4200] (rev a1)
04:00.1 Audio device: NVIDIA Corporation GK104 HDMI Audio Controller (rev a1)
(SLLML-dev) joehays-local@sedacs17lx:~$ nvcc --version
nvcc: NVIDIA (R) Cuda compiler driver
Copyright (c) 2005-2019 NVIDIA Corporation
Built on Wed_Oct_23_19:24:38_PDT_2019
Cuda compilation tools, release 10.2, V10.2.89
(SLLML-dev) joehays-local@sedacs17lx:~$ nvidia-smi
Wed Jul 8 18:26:37 2020
+-----------------------------------------------------------------------------+
| NVIDIA-SMI 440.33.01 Driver Version: 440.33.01 CUDA Version: 10.2 |
|-------------------------------+----------------------+----------------------+
| GPU Name Persistence-M| Bus-Id Disp.A | Volatile Uncorr. ECC |
| Fan Temp Perf Pwr:Usage/Cap| Memory-Usage | GPU-Util Compute M. |
|===============================+======================+======================|
| 0 Quadro K4200 Off | 00000000:03:00.0 Off | N/A |
| 30% 41C P8 14W / 110W | 164MiB / 4029MiB | 0% Default |
+-------------------------------+----------------------+----------------------+
| 1 Quadro K4200 Off | 00000000:04:00.0 Off | N/A |
| 30% 36C P8 13W / 110W | 1MiB / 4037MiB | 0% Default |
+-------------------------------+----------------------+----------------------+
+-----------------------------------------------------------------------------+
| Processes: GPU Memory |
| GPU PID Type Process name Usage |
|=============================================================================|
| 0 1692 G /usr/lib/xorg/Xorg 161MiB |
+-----------------------------------------------------------------------------+
(SLLML-dev) joehays-local@sedacs17lx:~$ echo $CUDA_HOME
/usr/local/cuda-10.2
(SLLML-dev) joehays-local@sedacs17lx:~$ echo $LD_LIBRARY_PATH
:/home/joehays-local/.mujoco/mujoco200/bin:/usr/local/cuda-10.2/lib64:/home/joehays-local/.mujoco/mujoco200/bin:/usr/local/cuda-10.2/lib64:/usr/lib/nvidia-440:/usr/lib/nvidia-440
(SLLML-dev) joehays-local@sedacs17lx:~$