DeepSeek-V4-Flash-W4A16-FP8 / tokenizer_config.json
pastapaul's picture
Phase 3b: AWQ-W4A16 quantization (FP8_BLOCK attn + W4A16 routed experts)
2e7ef6a verified
raw
history blame contribute delete
397 Bytes
{
"backend": "tokenizers",
"bos_token": "<|begin▁of▁sentence|>",
"clean_up_tokenization_spaces": false,
"eos_token": "<|end▁of▁sentence|>",
"is_local": true,
"legacy": true,
"local_files_only": false,
"model_max_length": 1048576,
"pad_token": "<|end▁of▁sentence|>",
"sp_model_kwargs": {},
"tokenizer_class": "TokenizersBackend",
"unk_token": null
}