ByteDance-Seed

/

Seed-X-PPO-7B

like 302

Follow

ByteDance Seed 3.45k

README recommends to use vllm==0.8.0 and transformers==4.51.3.
But this model can be run on modern versions of vLLM as well.
Tested on vLLM 0.22.0.
All this model requires is a tokenizer_config.json.
Just store this along side the model weights and it can be run on latest vLLM versions.

tokenizer_config.json

{
  "add_bos_token": true,
  "add_eos_token": false,
  "bos_token": "<s>",
  "clean_up_tokenization_spaces": false,
  "eos_token": "</s>",
  "model_max_length": 32768,
  "pad_token": "</s>",
  "tokenizer_class": "PreTrainedTokenizerFast",
  "unk_token": "<unk>"
}

Server version and Model

Inference