Executing with 1 GPU raises "OutOfMemory Exception", executing with 2 GPUs "RuntimeError: CUDA error: invalid device ordinal"
nmerkle opened this issue · 2 comments
Hi,
I have tried to implement GPT2 from scratch according to the Video tutorial. However, if I try to execute the code on 2 GPUs with:
torchrun --standalone --nproc_per_node=2 GPT.py
My program fails with the following error message:
Device: cuda:1
Device Count: 1
[rank1]: Traceback (most recent call last):
[rank1]: File "/my_transformer/GPT.py", line 238, in <module>
[rank1]: torch.cuda.set_device(device)
[rank1]: File "/.local/lib/python3.9/site-packages/torch/cuda/__init__.py", line 399, in set_device
[rank1]: torch._C._cuda_setDevice(device)
[rank1]: RuntimeError: CUDA error: invalid device ordinal
[rank1]: CUDA kernel errors might be asynchronously reported at some other API call, so the stacktrace below might be incorrect.
[rank1]: For debugging consider passing CUDA_LAUNCH_BLOCKING=1.
[rank1]: Compile with `TORCH_USE_CUDA_DSA` to enable device-side assertions.
Device: cuda:0
Device Count: 1
Master-Process: True
Total desired batch size: 524288
Calculated gradient accumulation steps: 16.
loaded 338025 tokens.
W0626 10:12:11.821799 22703772874560 torch/distributed/elastic/multiprocessing/api.py:851] Sending process 2031187 closing signal SIGTERM
E0626 10:12:11.853472 22703772874560 torch/distributed/elastic/multiprocessing/api.py:826] failed (exitcode: 1) local_rank: 1 (pid: 2031188) of binary: ~/my_transformer/.venv/bin/python3.9
Traceback (most recent call last):
File "~/my_transformer/.venv/bin/torchrun", line 8, in <module>
sys.exit(main())
File "/.local/lib/python3.9/site-packages/torch/distributed/elastic/multiprocessing/errors/__init__.py", line 347, in wrapper
return f(*args, **kwargs)
File "~/.local/lib/python3.9/site-packages/torch/distributed/run.py", line 879, in main
run(args)
File "/.local/lib/python3.9/site-packages/torch/distributed/run.py", line 870, in run
elastic_launch(
File "~/.local/lib/python3.9/site-packages/torch/distributed/launcher/api.py", line 132, in __call__
return launch_agent(self._config, self._entrypoint, list(args))
File "~/.local/lib/python3.9/site-packages/torch/distributed/launcher/api.py", line 263, in launch_agent
raise ChildFailedError(
torch.distributed.elastic.multiprocessing.errors.ChildFailedError:
============================================================
GPT.py FAILED
------------------------------------------------------------
Failures:
<NO_OTHER_FAILURES>
------------------------------------------------------------
Root Cause (first observed failure):
[0]:
time : 2024-06-26_10:12:11
host : haicn01.localdomain
rank : 1 (local_rank: 1)
exitcode : 1 (pid: 2031188)
error_file: <N/A>
traceback : To enable traceback see: https://pytorch.org/docs/stable/elastic/errors.html
If I execute with just 1 GPU, I get another error:
[rank0]: OutOfMemoryError: CUDA out of memory. Tried to allocate 786.00 MiB. GPU
Any ideas what could be the reason? I exactly followed the video tutorial and also checked the code in the repository. I should have enough memory. According to nvidia-smi I get the following output:
Wed Jun 26 10:51:56 2024
+-----------------------------------------------------------------------------------------+
| NVIDIA-SMI 550.54.15 Driver Version: 550.54.15 CUDA Version: 12.4 |
|-----------------------------------------+------------------------+----------------------+
| 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 NVIDIA A100-SXM4-40GB On | 00000000:CA:00.0 Off | On |
| N/A 55C P0 165W / 400W | 612MiB / 40960MiB | N/A Default |
| | | Enabled |
+-----------------------------------------+------------------------+----------------------+
+-----------------------------------------------------------------------------------------+
| MIG devices: |
+------------------+----------------------------------+-----------+-----------------------+
| GPU GI CI MIG | Memory-Usage | Vol| Shared |
| ID ID Dev | BAR1-Usage | SM Unc| CE ENC DEC OFA JPG |
| | | ECC| |
|==================+==================================+===========+=======================|
| 0 8 0 0 | 12MiB / 4864MiB | 14 0 | 1 0 0 0 0 |
| | 0MiB / 8191MiB | | |
+------------------+----------------------------------+-----------+-----------------------+
| 0 9 0 1 | 12MiB / 4864MiB | 14 0 | 1 0 0 0 0 |
| | 0MiB / 8191MiB | | |
+------------------+----------------------------------+-----------+-----------------------+
+-----------------------------------------------------------------------------------------+
| Processes: |
| GPU GI CI PID Type Process name GPU Memory |
| ID ID Usage |
|=========================================================================================|
| No running processes found |
+-----------------------------------------------------------------------------------------+
Thanks in advance.
IIUC, @karpathy used A100-80G but you seem to have 40G - have you tried reducing the batch size B to say 16 or 32? https://github.com/karpathy/build-nanogpt/blob/master/train_gpt2.py#L325
@andytwigg Thank you for your answer. I used again 1 GPU with 40 GB and decreased the batch size to 2 and then I got another error:
[rank0]: x, y = data_loader.next_batch()
[rank0]: File "~/my_transformer/GPT.py", line 215, in next_batch
[rank0]: x = (buf[:-1]).view(B,T)
[rank0]: RuntimeError: shape '[2, 1024]' is invalid for input of size 104
I think the problem now is that in the "next_batch()" function (see line https://github.com/karpathy/build-nanogpt/blob/master/train_gpt2.py#L243), the reshaping fails because the token size does not match when the end of the buffer is reached. The code runs a while but then raises the error message mentioned above because just 104 tokens remain for processing:
buf = self.tokens[self.current_position : self.current_position+B*T+1]
x = (buf[:-1]).view(B, T)
Any idea how to address this? I was thinking to check with a modulo (%) operator whether the remaining tokens are divisble through (B*T+1). However, I think that would be a quick and dirty solution. Any other suggestions? I am wondering why it works in the tutorial. I guess I must have missed something.