apepkuss79's picture
Update README.md
608247a verified
metadata
base_model: Qwen/Qwen2.5-14B-Instruct
license: apache-2.0
license_link: https://huggingface.co/Qwen/Qwen2.5-14B-Instruct/blob/main/LICENSE
model_creator: Qwen
model_name: Qwen2.5-14B-Instruct
quantized_by: Second State Inc.
language:
  - en
pipeline_tag: text-generation
tags:
  - chat

Qwen2.5-14B-Instruct-GGUF

Original Model

Qwen/Qwen2.5-14B-Instruct

Run with LlamaEdge

  • LlamaEdge version: v0.14.3

  • Prompt template

    • Prompt type: chatml

    • Prompt string

      <|im_start|>system
      {system_message}<|im_end|>
      <|im_start|>user
      {prompt}<|im_end|>
      <|im_start|>assistant
      
  • Context size: 128000

  • Run as LlamaEdge service

    wasmedge --dir .:. --nn-preload default:GGML:AUTO:Qwen2.5-14B-Instruct-Q5_K_M.gguf \
      llama-api-server.wasm \
      --model-name Qwen2.5-14B-Instruct \
      --prompt-template chatml \
      --ctx-size 128000
    
  • Run as LlamaEdge command app

    wasmedge --dir .:. --nn-preload default:GGML:AUTO:Qwen2.5-14B-Instruct-Q5_K_M.gguf \
      llama-chat.wasm \
      --prompt-template chatml \
      --ctx-size 128000
    

Quantized GGUF Models

Name Quant method Bits Size Use case
Qwen2.5-14B-Instruct-Q2_K.gguf Q2_K 2 5.77 GB smallest, significant quality loss - not recommended for most purposes
Qwen2.5-14B-Instruct-Q3_K_L.gguf Q3_K_L 3 7.92 GB small, substantial quality loss
Qwen2.5-14B-Instruct-Q3_K_M.gguf Q3_K_M 3 7.34 GB very small, high quality loss
Qwen2.5-14B-Instruct-Q3_K_S.gguf Q3_K_S 3 6.66 GB very small, high quality loss
Qwen2.5-14B-Instruct-Q4_0.gguf Q4_0 4 8.52 GB legacy; small, very high quality loss - prefer using Q3_K_M
Qwen2.5-14B-Instruct-Q4_K_M.gguf Q4_K_M 4 8.99 GB medium, balanced quality - recommended
Qwen2.5-14B-Instruct-Q4_K_S.gguf Q4_K_S 4 8.57 GB small, greater quality loss
Qwen2.5-14B-Instruct-Q5_0.gguf Q5_0 5 10.3 GB legacy; medium, balanced quality - prefer using Q4_K_M
Qwen2.5-14B-Instruct-Q5_K_M.gguf Q5_K_M 5 10.5 GB large, very low quality loss - recommended
Qwen2.5-14B-Instruct-Q5_K_S.gguf Q5_K_S 5 10.3 GB large, low quality loss - recommended
Qwen2.5-14B-Instruct-Q6_K.gguf Q6_K 6 12.1 GB very large, extremely low quality loss
Qwen2.5-14B-Instruct-Q8_0.gguf Q8_0 8 15.1 GB very large, extremely low quality loss - not recommended
Qwen2.5-14B-Instruct-f16.gguf f16 16 29.5 GB

Quantized with llama.cpp b3751