metadata
license: apache-2.0
slimpajama_Llama2_Tokenizer
The original slimpajama_Llama2_Tokenizer.tar.gz archive (≈794 GB on Ubuntu) was split into smaller 40 GB chunks for easier upload to Hugging Face.
sudo apt install git-lfs
pip install -U huggingface_hub # `hf version`==1.1.4
tar cvf - slimpajama_Llama2_Tokenizer/ | pigz -p 16 > slimpajama_Llama2_Tokenizer.tar.gz
split -b 40G -d -a 3 slimpajama_Llama2_Tokenizer.tar.gz slimpajama_Llama2_Tokenizer/slimpajama_Llama2_Tokenizer_part_
# Upload files sequentially using a loop:
count=0
for file in slimpajama_Llama2_Tokenizer/slimpajama_Llama2_Tokenizer_part_*; do
count=$((count + 1))
# if [ $count -lt 3 ]; then
# continue
# fi
filename=$(basename "$file")
echo "Uploading $file (file #$count) to slimpajama_Llama2_Tokenizer/$filename ..."
hf upload jsun/slimpajama_Llama2_Tokenizer "$file" "slimpajama_Llama2_Tokenizer/$filename" --repo-type dataset
done
Once all parts are downloaded, you can reconstruct the original tar.gz file using:
cat slimpajama_Llama2_Tokenizer_part_* > slimpajama_Llama2_Tokenizer.tar.gz
Then extract the archive as usual:
pigz -dc -p 16 slimpajama_Llama2_Tokenizer.tar.gz | tar xvf -