OSError with multiple GPUs #86
Comments
|
Thanks for the detailed information! How many smiles are there in your
input file?
…---
"Where there is charity and wisdom, there is neither fear nor ignorance."
- Francis of Assisi (1182 – 1226) -
Zhen (Jack) Liu
Ph.D. candidate at the Isayev Lab
Department of Chemistry,
Carnegie Mellon University
4400 Fifth Avenue
Pittsburgh, PA 15213 USA
On Tue, Nov 5, 2024 at 6:14 PM VN-jsjeon ***@***.***> wrote:
*Describe the bug*
When running auto3d to generate 3D structures with multiple GPUs (e.g.,
using --gpu_idx 0,1), I encounter an OSError related to an invalid file,
arated.sdf. This error does not occur when using only a single GPU
(--gpu_idx 0), where the command executes successfully. It seems the
program is attempting to read a non-existent file when more than one GPU is
specified, likely due to how it splits tasks across GPUs.
*To Reproduce*
Steps to reproduce the behavior:
Prepare a SMILES file, e.g., abemaciclib.smi, with the desired structures.
Run auto3d with multiple GPUs:
auto3d abemaciclib.smi --k=100 --use_gpu=True --job_name auto3d_5
--gpu_idx 0,1
*Error*
The task will be divided into 2 jobs.
Job1, number of inputs: 1
Job2, number of inputs: 0
Isomer generation for job1
Enumerating cis/tran isomers for unspecified double bonds...
Enumerating R/S isomers for unspecified atomic centers...
Removing enantiomers...
Enumerating conformers/rotamers, removing duplicates...
100%|███████████████████████████████████████████████████████████████████████████████████
Isomer generation for job2
Optimizing on job1
Enumerating cis/tran isomers for unspecified double bonds...
Enumerating R/S isomers for unspecified atomic centers...
Removing enantiomers...
Enumerating conformers/rotamers, removing duplicates...
0it [00:00, ?it/s]
Optimizing on job2
Preparing for parallel optimizing... (Max optimization steps: 5000)
Preparing for parallel optimizing... (Max optimization steps: 5000)
Process Process-5:
Traceback (most recent call last):
File "/home/jeon/raid_new/miniconda3/envs/auto3D/lib/python3.11/multiprocessing/proces
self.run()
File "/home/jeon/raid_new/miniconda3/envs/auto3D/lib/python3.11/multiprocessing/proces
self._target(*self._args, **self._kwargs)
File "/home/jeon/raid_new/miniconda3/envs/auto3D/lib/python3.11/site-packages/Auto3D/aapper
optimizer.run()
File "/home/jeon/raid_new/miniconda3/envs/auto3D/lib/python3.11/site-packages/Auto3D/bun
mols = list(Chem.SDMolSupplier(self.in_f, removeHs=False))
^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^
OSError: File error: Invalid input file /home/jeon/raid_new/model/auto3d/example/files/arated.sdf
Total 3D conformers: 767
*System information:*
Operating System: ubuntu22.04
Auto3D version: 2.3.1
Python version: 3.11
RDKit version: 2024.09.2
PyTorch version: 2.5.1
*Additional context*
This issue appears to arise when auto3d creates empty jobs or fails to
handle cases where a job assigned to a GPU has zero inputs. When only a
single GPU is used, the program executes without issues, so it seems
specific to multi-GPU handling.
—
Reply to this email directly, view it on GitHub
<#86>, or unsubscribe
<https://github.com/notifications/unsubscribe-auth/AOK6RLNRDPOB3SWDQU7DDGTZ7F3Q3AVCNFSM6AAAAABRH2C6JWVHI2DSMVQWIX3LMV43ASLTON2WKOZSGYZTMOBWGUZTCOA>
.
You are receiving this because you are subscribed to this thread.Message
ID: ***@***.***>
|
Sign up for free
to join this conversation on GitHub.
Already have an account?
Sign in to comment
Describe the bug
When running auto3d to generate 3D structures with multiple GPUs (e.g., using --gpu_idx 0,1), I encounter an OSError related to an invalid file, arated.sdf. This error does not occur when using only a single GPU (--gpu_idx 0), where the command executes successfully. It seems the program is attempting to read a non-existent file when more than one GPU is specified, likely due to how it splits tasks across GPUs.
To Reproduce
Steps to reproduce the behavior:
Prepare a SMILES file, e.g., abemaciclib.smi, with the desired structures.
Run auto3d with multiple GPUs:
auto3d abemaciclib.smi --k=100 --use_gpu=True --job_name auto3d_5 --gpu_idx 0,1Error
System information:
Operating System: ubuntu22.04
Auto3D version: 2.3.1
Python version: 3.11
RDKit version: 2024.09.2
PyTorch version: 2.5.1
Additional context
The text was updated successfully, but these errors were encountered: