DiegoEmilio/OMP

Pre-trained weights for the OMP model.

Usage and additional configurations:

torchrun --standalone --nproc_per_node=6 train-6gpus.py \
  --batch_size 2 \
  --crop_length 1024 \
  --model 650M \
  --val_check_interval 12900 \
  --val_examples 32000 \
  --max_steps 8000000 \
  --accumulate_grad 43 \
  --run_name run \
  --use_glu \
  --parallel \
  --wandb \
  --gpu_change 0 \
  --dataset_split filtered \
  --print_freq 300 \