config.production.yaml

HOST: "0.0.0.0"
PORT: 7860
STRATEGY: "cuda fp16"
RWKV_CUDA_ON: False
CHUNK_LEN: 256
MODELS:
  - SERVICE_NAME: "RWKV-x070-World-0.1B-v2.8-20241210-ctx4096"
    DOWNLOAD_MODEL_FILE_NAME: "RWKV-x070-World-0.1B-v2.8-20241210-ctx4096.pth"
    DOWNLOAD_MODEL_REPO_ID: "BlinkDL/rwkv-7-world"
    REASONING: False
    DEFAULT: True
    DEFAULT_SAMPLER:
      max_tokens: 512
      temperature: 1.0
      top_p: 0.3
      presence_penalty: 0.5
      count_penalty: 0.5
      penalty_decay: 0.996
      stop:
        - "\n\n"
  - SERVICE_NAME: "RWKV7-G1-0.1B-68%trained-20250303-ctx4k"
    DOWNLOAD_MODEL_FILE_NAME: "RWKV7-G1-0.1B-68%trained-20250303-ctx4k.pth"
    DOWNLOAD_MODEL_REPO_ID: "BlinkDL/temp-latest-training-models"
    REASONING: True
    DEFAULT: True
    DEFAULT_SAMPLER:
      max_tokens: 4096
      temperature: 1.0
      top_p: 0.3
      presence_penalty: 0.5
      count_penalty: 0.5
      penalty_decay: 0.996
      stop:
        - "\n\n"