README.md

metadata

dataset_info:
  features:
    - name: input_ids
      sequence: int64
    - name: labels
      dtype: int64
    - name: source
      list:
        - name: end
          dtype: int64
        - name: source
          dtype: string
        - name: start
          dtype: int64
  splits:
    - name: train
      num_bytes: 40394155659
      num_examples: 5041
  download_size: 9203916526
  dataset_size: 40394155659
configs:
  - config_name: default
    data_files:
      - split: train
        path: data/train-*

Generated using https://github.com/FranxYao/Long-Context-Data-Engineering with the below command:

mkdir logs
mkdir data
mkdir data/slimpajama
mkdir data/slimpajama/per_source_downsample
cd data_engineering

PATH_TO_SLIMPAJAMA=rokset3/slim_pajama_chunk_1
nohup python -u slimpajama_packing.py\
    --dataset_size=5b\
    --print_interval=100 --num_process=200\
    --chunk_size=1000001 \
    --dataset_path=$PATH_TO_SLIMPAJAMA\
    --output_path=../data/slimpajama/per_source_downsample/ --down_sample_ratio=0.1 --down_sample_mode=per_source\
    > ../logs/slimpajama_packing_dist_per_source_downsample_0.1.log 2>&1 &
tail -f ../logs/slimpajama_packing_dist_per_source_downsample_0.1.log