apepkuss79's picture
Update README.md
f46b15f verified
|
raw
history blame
3.68 kB
metadata
base_model: cloudyu/Yi-34Bx2-MoE-60B
license: cc-by-nc-4.0
model_creator: cloudyu
model_name: Yi 34Bx2 MoE 60B
model_type: mistral
quantized_by: Second State Inc.
tags:
  - moe

Yi-34Bx2-MoE-60B-GGUF

Original Model

cloudyu/Yi-34Bx2-MoE-60B

Run with LlamaEdge

  • LlamaEdge version: v0.2.8 and above

  • Prompt template

    • Prompt type: chatml

    • Prompt string

      <|im_start|>system
      {system_message}<|im_end|>
      <|im_start|>user
      {prompt}<|im_end|>
      <|im_start|>assistant
      
    • Reverse prompt: <|im_end|>

  • Context size: 7168

  • Run as LlamaEdge service

    wasmedge --dir .:. --nn-preload default:GGML:AUTO:Yi-34Bx2-MoE-60B-Q5_K_M.gguf llama-api-server.wasm -p chatml -r '<|im_end|>'
    
  • Run as LlamaEdge command app

    wasmedge --dir .:. --nn-preload default:GGML:AUTO:Yi-34Bx2-MoE-60B-Q5_K_M.gguf llama-chat.wasm -p chatml -r '<|im_end|>'
    

Quantized GGUF Models

Name Quant method Bits Size Use case
Yi-34Bx2-MoE-60B-Q2_K.gguf Q2_K 2 22.4 GB smallest, significant quality loss - not recommended for most purposes
Yi-34Bx2-MoE-60B-Q3_K_L.gguf Q3_K_L 3 31.8 GB small, substantial quality loss
Yi-34Bx2-MoE-60B-Q3_K_M.gguf Q3_K_M 3 29.2 GB very small, high quality loss
Yi-34Bx2-MoE-60B-Q3_K_S.gguf Q3_K_S 3 26.3 GB very small, high quality loss
Yi-34Bx2-MoE-60B-Q4_0.gguf Q4_0 4 34.3 GB legacy; small, very high quality loss - prefer using Q3_K_M
Yi-34Bx2-MoE-60B-Q4_K_M.gguf Q4_K_M 4 36.7 GB medium, balanced quality - recommended
Yi-34Bx2-MoE-60B-Q4_K_S.gguf Q4_K_S 4 34.6 GB small, greater quality loss
Yi-34Bx2-MoE-60B-Q5_0.gguf Q5_0 5 41.9 GB legacy; medium, balanced quality - prefer using Q4_K_M
Yi-34Bx2-MoE-60B-Q5_K_M.gguf Q5_K_M 5 43.1 GB large, very low quality loss - recommended
Yi-34Bx2-MoE-60B-Q5_K_S.gguf Q5_K_S 5 41.9 GB large, low quality loss - recommended
Yi-34Bx2-MoE-60B-Q6_K.gguf Q6_K 6 49.9 GB very large, extremely low quality loss