pretraining-playground/pythia-pile-presampled
收藏Hugging Face2024-04-01 更新2025-11-03 收录
下载链接：
https://hf-mirror.com/datasets/pretraining-playground/pythia-pile-presampled
下载链接
链接失效反馈官方服务：
资源简介：
---
license:
- mit
language:
- en
dataset_info:
- config_name: full
splits:
- name: train
num_bytes: 600664064000
num_examples: 146432000
- config_name: checkpoints
splits:
- name: train
num_bytes: 6919004160
num_examples: 1683456
configs:
- config_name: full
data_files:
- split: train
path: data/shard*
- config_name: checkpoints
data_files:
- split: train
path: data/checkpoint_steps.parquet
pretty_name: Pythia Presampled Pile
---
提供机构：
pretraining-playground
