Training Issues

[Allvey]

Thank you for your excellent work! I have some trouble with training: I tried to install slurm for cluster job scheduling, but unfortunately many attempts failed. So, what we want to know is if there is any impact on training if we don't use the srun command, but execute the training script directly (for example perform ./distributed_pretrain.sh 8 '/path/to/your/dataset' ... in pre-training stage)?

[deepcs233]

Hi! Executing the training script without slurm is ok and doesn't affect the performance.

[CoderXuans]

@Allvey Hello, may I ask what command you are using to train the visual encoder? I am also using a single GPU.