Babajaan
/

nanochat-ar

Text Generation

Model card Files Files and versions

nanochat-ar / reports /tokenizer-training.md

Babajaan's picture

Upload nanochat-ar Aziz HPC proof-run checkpoints

b6839ad verified 3 months ago

|

Raw History Blame Contribute Delete

265 Bytes

Tokenizer training

timestamp: 2026-07-01 07:54:22

max_chars: 2,000,000,000
doc_cap: 10,000
vocab_size: 32,768
train_time: 143.4839
num_special_tokens: 9
token_bytes_min: 1
token_bytes_max: 31
token_bytes_mean: 10.1118
token_bytes_std: 4.1123