susie.sun@yz-amd1:~$ deepspeed --num_gpus 1 pipe.py
[2023-12-05 09:42:21,499] [INFO] [real_accelerator.py:161:get_accelerator] Setting ds_accelerator to cuda (auto detect)
[2023-12-05 09:42:21,780] [WARNING] [runner.py:203:fetch_hostfile] Unable to find hostfile, will proceed with training with local resources only.
Traceback (most recent call last):
File "/home/susie.sun/.local/bin/deepspeed", line 6, in
main()
File "/home/susie.sun/.local/lib/python3.10/site-packages/deepspeed/launcher/runner.py", line 422, in main
raise RuntimeError("Unable to proceed, no GPU resources available")
RuntimeError: Unable to proceed, no GPU resources available
Hi @sunpian1, our custom kernels do not currently support AMD GPUs. It is something we have considered, and I will share this with the other devs. Thank you!
Can deepspeed-MII run on AMD GPU?
susie.sun@yz-amd1:~$ deepspeed --num_gpus 1 pipe.py [2023-12-05 09:42:21,499] [INFO] [real_accelerator.py:161:get_accelerator] Setting ds_accelerator to cuda (auto detect) [2023-12-05 09:42:21,780] [WARNING] [runner.py:203:fetch_hostfile] Unable to find hostfile, will proceed with training with local resources only. Traceback (most recent call last): File "/home/susie.sun/.local/bin/deepspeed", line 6, in
main()
File "/home/susie.sun/.local/lib/python3.10/site-packages/deepspeed/launcher/runner.py", line 422, in main
raise RuntimeError("Unable to proceed, no GPU resources available")
RuntimeError: Unable to proceed, no GPU resources available
susie.sun@yz-amd1:~$ cat pipe.py import mii pipe = mii.pipeline("mistralai/Mistral-7B-v0.1") response = pipe(["DeepSpeed is", "Seattle is"], max_new_tokens=128) print(response)