Fabio Geraci
Fabio Geraci
Thanks for the reply
I have the same issue
> Or even more beneficial might be if you could please point me to an existing converter that might closely resemble the one that would need to be created for...
Hey, thank you for the quick response. I tried, as you suggested on a private tab as well as on firefox with no extensions and the outcome is the same...
thanks you. my data structure is different Path ---> train ---> images ---> masks ---> valid ---> images ---> masks def label_func(x): return data_dir/f'train/masks/{x.stem}_masks.png' my datablock db = DataBlock(blocks=(ImageBlock(),MaskBlock(codes)), batch_tfms=...
data_dir_str = f'/mnt/Datasets/img_sets/SV_masks/' data_dir = Path(data_dir_str)
Back to work on the 8/01, I will start then
I am using the configuration above to fine tuning 70B model on 2 nodes with 8 gpus each. the job took 75minutes to compile (is that usual?) I also noticed...
> Hey @fabiogeraci, just updating you on this. I'm waiting on a request for multi-node server (PyTorch has limited quantity). If I don't hear back today, I'll just rent one...
@joecummings any progress?