Download code_snapshot/run_train.sh from BlueSourceJY/SiT-Complementary: direct link, hf CLI and curl.
- Browser
- Download file 354 Bytes
-
https://huggingface.co/BlueSourceJY/SiT-Complementary/resolve/main/code_snapshot/run_train.sh
- Command line
-
hf download hf://BlueSourceJY/SiT-Complementary/code_snapshot/run_train.sh
-
curl -L -o run_train.sh https://huggingface.co/BlueSourceJY/SiT-Complementary/resolve/main/code_snapshot/run_train.sh
354 Bytes
| torchrun \ | |
| --nnodes=1 \ | |
| --nproc_per_node=8 \ | |
| train.py \ | |
| --model SiT-S/2 \ | |
| --epochs=800 \ | |
| --data-path /home/nvidia/datasets/imagenet-1k/train \ | |
| --wandb \ | |
| --global-batch-size=1024 \ | |
| --run-name 005-SiT-S-2-Linear-velocity-None \ | |
| --ckpt /home/nvidia/SiT-Complementary/results/005-SiT-S-2-Linear-velocity-None/checkpoints/0550000.pt | |
| # batch_size x 4, lr x 2 | |