Cortex-0 / tokenizer /tokenizer_config.json
Frankenstein-Labs's picture
feat(tokenizer): Phase 1 — BPE 32k from scratch
f17dddf verified
Raw History Blame Contribute Delete
679 Bytes
{
"tokenizer_class": "Cortex0BPETokenizer",
"model_type": "bpe",
"vocab_size": 32000,
"byte_fallback": true,
"pre_tokenizer": "ByteLevel",
"add_prefix_space": false,
"special_tokens": [
"<|pad|>",
"<|bos|>",
"<|eos|>",
"<|unk|>",
"<|think|>",
"<|plan|>",
"<|tool_call|>",
"<|tool_result|>",
"<|critic|>",
"<|moe_route|>",
"<|image_start|>",
"<|image_end|>",
"<|image_pad|>",
"<|audio_start|>",
"<|audio_end|>",
"<|audio_pad|>",
"<|video_start|>",
"<|video_end|>",
"<|video_pad|>"
],
"pad_token": "<|pad|>",
"bos_token": "<|bos|>",
"eos_token": "<|eos|>",
"unk_token": "<|unk|>"
}