shenjiarun/qwen3-c4-megatron
qwen3-c4-megatron Megatron-LM indexed dataset files for Qwen3 tokenizer preprocessing on the C4 corpus. Directory layout qwen3-c4-megatron/ ├── README.md ├── metadata.json ├── manifest.sha256 └── data/ ├── c4_full.bin └── c4_full.idx Files data/c4_full.bin: Megatron-LM binary token data. data/c4_full.idx: Megatron-LM index file for c4_full.bin. metadata.json: Dataset metadata and preprocessing configuration. manifest.sha256: SHA256… See the full description on the dataset page: https://huggingface.co/datasets/shenjiarun/qwen3-c4-megatron.
qwen3-c4-megatron
Megatron-LM indexed dataset files for Qwen3 tokenizer preprocessing on the C4 corpus.
Directory layout
qwen3-c4-megatron/
├── README.md
├── metadata.json
├── manifest.sha256
└── data/
├── c4_full.bin
└── c4_full.idxFiles
data/c4_full.bin: Megatron-LM binary token data.data/c4_full.idx: Megatron-LM index file forc4_full.bin.metadata.json: Dataset metadata and preprocessing configuration.manifest.sha256: SHA256 checksums for all files except the manifest itself.
Usage
Use the dataset prefix without the file extension when passing it to Megatron-LM:
--data-path /225045001/Megatron-LM/qwen3-c4-megatron/data/c4_full