3

when I build DCNv2 in my conda environment,I got this message.

I have checked cuda by nvidia-smi:

Tue Apr  6 20:03:13 2021       
+-----------------------------------------------------------------------------+
| NVIDIA-SMI 460.67       Driver Version: 460.67       CUDA Version: 11.2     |
|-------------------------------+----------------------+----------------------+
| GPU  Name        Persistence-M| Bus-Id        Disp.A | Volatile Uncorr. ECC |
| Fan  Temp  Perf  Pwr:Usage/Cap|         Memory-Usage | GPU-Util  Compute M. |
|                               |                      |               MIG M. |
|===============================+======================+======================|
|   0  GeForce RTX 3070    Off  | 00000000:01:00.0  On |                  N/A |
|  0%   45C    P8    17W / 220W |    448MiB /  7979MiB |     23%      Default |
|                               |                      |                  N/A |
+-------------------------------+----------------------+----------------------+
                                                                               
+-----------------------------------------------------------------------------+
| Processes:                                                                  |
|  GPU   GI   CI        PID   Type   Process name                  GPU Memory |
|        ID   ID                                                   Usage      |
|=============================================================================|
|    0   N/A  N/A       955      G   /usr/lib/xorg/Xorg                 53MiB |
|    0   N/A  N/A      1555      G   /usr/lib/xorg/Xorg                210MiB |
|    0   N/A  N/A      1690      G   /usr/bin/gnome-shell               61MiB |
|    0   N/A  N/A      3564      G   ...AAAAAAAAA= --shared-files      108MiB |

and nvcc -V:

nvcc: NVIDIA (R) Cuda compiler driver
Copyright (c) 2005-2019 NVIDIA Corporation
Built on Sun_Jul_28_19:07:16_PDT_2019
Cuda compilation tools, release 10.1, V10.1.243

Any help?

talonmies
  • 70,661
  • 34
  • 192
  • 269
David Zhuang
  • 31
  • 1
  • 1
  • 2

2 Answers2

5

Your GPU is "too new" for CUDA 10. Support for cards with compute capability 8.6 like yours was added in CUDA 11.1 (thank you @RobertCrovella for the correction). This means you'll need to use CUDA 11.1 or newer.

jodag
  • 19,885
  • 5
  • 47
  • 66
  • I upgraded my CUDA to 11.3.1 and still get the same error message about "CUDA arch (8.6)". Do you know why this would happen? – jtb Sep 23 '21 at 16:57
  • 2
    @JoshBone Assuming this is related to pytorch keep in mind that pytorch does *not* use your system's cuda. It installs it's own version along with the pytorch binaries. See [this answer](https://stackoverflow.com/a/61034368/2790047) and [this answer](https://stackoverflow.com/a/62361395/2790047) for more details on how to get pytorch working with CUDA. – jodag Sep 23 '21 at 17:23
-1

You can edit cpp_extension.py, eg "/usr/local/conda3/lib/python3.7/site-packages/torch/utils/cpp_extension.py", and add 8.6 to the list.

YesThatIsMyName
  • 1,585
  • 3
  • 23
  • 30
yufei wang
  • 1
  • 2