config.yml

---
repo: TheBloke/wizard-vicuna-13B-GGML
file: wizard-vicuna-13B.ggml.q5_1.bin
# if the repo above doesn't include the tokenizer set the base repo it was based on with a valid tokenizer model
base_model: junelee/wizard-vicuna-13b
llama_cpp:
  n_ctx: 2048
  n_gpu_layers: 40  # llama 13b has 40 layers
chat:
  stop:
    - "</s>"
    - "<unk>"
    - "### User:"