mtasic85's picture
compress pretrain dataset
8a0641e
raw
history blame
247 Bytes
# pip install torch torchvision torchaudio --index-url https://download.pytorch.org/whl/cpu
tqdm
datasets
jinja2
transformers
wandb
# litgpt[all]
litgpt[all] @ git+https://github.com/Lightning-AI/litgpt.git
litdata
grokadamw
# bitsandbytes
pyzstd