apepkuss79's picture
Update README.md
548e4f5 verified
metadata
base_model: baichuan-inc/Baichuan2-13B-Chat
inference: false
library_name: transformers
license: other
model_creator: Baichuan Intelligent Technology
model_name: Baichuan2 13B Chat
model_type: baichuan
tasks:
  - text-generation
quantized_by: Second State Inc.
language:
  - en
  - zh

Baichuan2-13B-Chat-GGUF

Original Model

baichuan-inc/Baichuan2-13B-Chat

Run with LlamaEdge

  • LlamaEdge version: v0.2.11 and above

  • Prompt template

    • Prompt type: baichuan-2

    • Prompt string

      以下内容为人类用户与与一位智能助手的对话。
      
      用户:你好!
      助手:
      
    • Reverse prompt: 用户:

  • Context size: 5120

  • Run as LlamaEdge service

    wasmedge --dir .:. --nn-preload default:GGML:AUTO:Baichuan2-13B-Chat-Q5_K_M.gguf llama-api-server.wasm -p baichuan-2 -r '用户:'
    
  • Run as LlamaEdge command app

    wasmedge --dir .:. --nn-preload default:GGML:AUTO:Baichuan2-13B-Chat-Q5_K_M.gguf llama-chat.wasm -p baichuan-2 -r '用户:'
    

Quantized GGUF Models

Name Quant method Bits Size Use case
Baichuan2-13B-Chat-Q2_K.gguf Q2_K 2 5.93 GB smallest, significant quality loss - not recommended for most purposes
Baichuan2-13B-Chat-Q3_K_L.gguf Q3_K_L 3 7.67 GB small, substantial quality loss
Baichuan2-13B-Chat-Q3_K_M.gguf Q3_K_M 3 7.24 GB very small, high quality loss
Baichuan2-13B-Chat-Q3_K_S.gguf Q3_K_S 3 6.77 GB very small, high quality loss
Baichuan2-13B-Chat-Q4_0.gguf Q4_0 4 7.99 GB legacy; small, very high quality loss - prefer using Q3_K_M
Baichuan2-13B-Chat-Q4_K_M.gguf Q4_K_M 4 9.00 GB medium, balanced quality - recommended
Baichuan2-13B-Chat-Q4_K_S.gguf Q4_K_S 4 8.37 GB small, greater quality loss
Baichuan2-13B-Chat-Q5_0.gguf Q5_0 5 9.64 GB legacy; medium, balanced quality - prefer using Q4_K_M
Baichuan2-13B-Chat-Q5_K_M.gguf Q5_K_M 5 10.3 GB large, very low quality loss - recommended
Baichuan2-13B-Chat-Q5_K_S.gguf Q5_K_S 5 9.82 GB large, low quality loss - recommended
Baichuan2-13B-Chat-Q6_K.gguf Q6_K 6 12.1 GB very large, extremely low quality loss
Baichuan2-13B-Chat-Q8_0.gguf Q8_0 8 14.8 GB very large, extremely low quality loss - not recommended

Quantized with llama.cpp b1953