mkd-ai/keural-stage1-binary

Keural Stage 1 — Binary Training Dataset

Pre-tokenized binary dataset for the Keural MoE Foundation Model (14.83B parameters). Ready to use directly for training — no tokenization step needed.

Dataset Contents

Property	Value
Format	Binary (.bin) + Index (.idx) + Metadata (.meta)
Total tokens	43.17 billion
Total sequences	10,067,450
Sequence length	4,096 tokens (fixed)
Shards	102 shards
Uncompressed size	~155GB
Archive	binary_backup.tar (154GB)
Tokenizer	Keural SentencePiece Unigram, vocab=131,072

Data Sources

Source	Language	Tokens (approx)
FineWeb	English	~20B
WanJuan Korean	Korean	~5B
Korean WebText	Korean	~4B
ArXiv	English Science	~4B
CC100 Korean	Korean	~3B
PubMed	English Medical	~3B
The Stack v1	Code	~8B
Wikipedia Korean	Korean	~1B
PG19 Literature	English	~1B

Binary Format Specification

File: keural_NNN.bin
─────────────────────────────────────────
HEADER (36 bytes — CRITICAL: not 32):
  [0:8]   magic   = b"KEURAL\x00\x00"   (8 bytes)
  [8:12]  version = 1                    (uint32 LE)
  [12:20] num_seq                        (uint64 LE)
  [20:28] seq_len = 4096                 (uint64 LE)
  [28:36] padding = 0                    (uint64 LE)

BODY:
  num_seq × 4096 × 4 bytes (uint32 LE tokens)

File: keural_NNN.idx
─────────────────────────────────────────
  [0:4]  num_seq  (uint32)
  [4:8]  seq_len  (uint32)
  per sequence: 8-byte offset + 4-byte length

File: keural_NNN.meta (JSON)
─────────────────────────────────────────
  {"num_sequences": N, "seq_length": 4096, "source": "keural_NNN"}

How to Use

import struct, mmap, torch

HEADER_FMT  = "<8sIQQQ"
HEADER_SIZE = struct.calcsize(HEADER_FMT)  # 36 bytes

with open("keural_001.bin", "rb") as f:
    raw = f.read(HEADER_SIZE)
    magic, ver, num_seqs, seq_len, _ = struct.unpack(HEADER_FMT, raw)
    print(f"Sequences: {num_seqs}, Length: {seq_len}")

# Or use directly with train_keural_v2.py:
# torchrun --nproc_per_node=2 train_keural_v2.py --data_dir ./binary

How to Extract

tar -xf binary_backup.tar
# Extracts to: binary/ directory with 102 shards

Related Resources

• Model Training: github.com/mkd-hossain/Keural-Model-Training
• Tokenizer: huggingface.co/mkd-ai/keural-tokenizer
• Organization: huggingface.co/mkd-ai

Author

Md Najmul Hossain / MKD CO., LTD. Keural Foundation Model — Stage 1 pretraining dataset, 2026