apepkuss79's picture
Update README.md
29d5392 verified
metadata
base_model: mistralai/Mistral-7B-Instruct-v0.1
inference: false
license: apache-2.0
model_creator: Mistral AI
model_name: Mistral 7B Instruct v0.1
model_type: mistral
pipeline_tag: text-generation
tags:
  - finetuned
quantized_by: Second State Inc.

Mistral-7B-Instruct-v0.1-GGUF

Original Model

mistralai/Mistral-7B-Instruct-v0.1

Run with LlamaEdge

  • LlamaEdge version: v0.2.8 and above

  • Prompt template

    • Prompt type: mistral-instruct

    • Prompt string

      <s>[INST] {prompt} [/INST]
      
  • Context size: 4096

  • Run as LlamaEdge service

    wasmedge --dir .:. --nn-preload default:GGML:AUTO:Mistral-7B-Instruct-v0.1-Q5_K_M.gguf llama-api-server.wasm -p mistral-instruct
    
  • Run as LlamaEdge command app

    wasmedge --dir .:. --nn-preload default:GGML:AUTO:Mistral-7B-Instruct-v0.1-Q5_K_M.gguf llama-chat.wasm -p mistral-instruct
    

Quantized GGUF Models

Name Quant method Bits Size Use case
Mistral-7B-Instruct-v0.1-Q2_K.gguf Q2_K 2 2.7 GB smallest, significant quality loss - not recommended for most purposes
Mistral-7B-Instruct-v0.1-Q3_K_L.gguf Q3_K_L 3 3.82 GB small, substantial quality loss
Mistral-7B-Instruct-v0.1-Q3_K_M.gguf Q3_K_M 3 3.52 GB very small, high quality loss
Mistral-7B-Instruct-v0.1-Q3_K_S.gguf Q3_K_S 3 3.16 GB very small, high quality loss
Mistral-7B-Instruct-v0.1-Q4_0.gguf Q4_0 4 4.11 GB legacy; small, very high quality loss - prefer using Q3_K_M
Mistral-7B-Instruct-v0.1-Q4_K_M.gguf Q4_K_M 4 4.37 GB medium, balanced quality - recommended
Mistral-7B-Instruct-v0.1-Q4_K_S.gguf Q4_K_S 4 4.14 GB small, greater quality loss
Mistral-7B-Instruct-v0.1-Q5_0.gguf Q5_0 5 5 GB legacy; medium, balanced quality - prefer using Q4_K_M
Mistral-7B-Instruct-v0.1-Q5_K_M.gguf Q5_K_M 5 5.13 GB large, very low quality loss - recommended
Mistral-7B-Instruct-v0.1-Q5_K_S.gguf Q5_K_S 5 5 GB large, low quality loss - recommended
Mistral-7B-Instruct-v0.1-Q6_K.gguf Q6_K 6 5.94 GB very large, extremely low quality loss
Mistral-7B-Instruct-v0.1-Q8_0.gguf Q8_0 8 7.7 GB very large, extremely low quality loss - not recommended