Wei Han (Henry)
Wei Han (Henry)
Hello. You can try different weights for contrastive loss. The results may vary on different GPUs for hardware conditions.
I remember torchvision is not imported in the project, so you can firstly try running without it.
> > I remember torchvision is not imported in the project, so you can firstly try running without it. > > Thank you for your reply, the problem has been...
> We have released the finetuning scripts, so let me know if you have any problem! Could you please tell me where it is? I found no finetuning scripts for...
> hi, how you run dp? on single-node or multi-node? single node, multiple GPUs. No model parallel.
> I am having the same issue on single node and vllm 0.8.4 I turned to tensor parallel with larger batch size as the alternative solution
> Hey, this seems like there's some stale shared memory. Just once, at the start of your training job, can you add a call to `streaming.base.util.clean_stale_shared_memory()` and see if that...
> Hey, this seems like there's some stale shared memory. Just once, at the start of your training job, can you add a call to `streaming.base.util.clean_stale_shared_memory()` and see if that...