Modelpublic

second-state/Codestral-22B-v0.1-GGUF

sourceHugging Faceotherupdated 2y agoView on Hugging Face

1likes743downloads

Model Card

<div style="width: auto; margin-left: auto; margin-right: auto"> <img src="https://github.com/LlamaEdge/LlamaEdge/raw/dev/assets/logo.svg" style="width: 100%; min-width: 400px; display: block; margin: auto;"> </div> <hr style="margin-top: 1.0em; margin-bottom: 1.0em;">

Codestral-22B-v0.1-GGUF

Original Model

mistralai/Codestral-22B-v0.1

Run with LlamaEdge

—LlamaEdge version: v0.11.2

—Prompt template

—Prompt type: mistral-instruct

—Prompt string

text

    <s>[INST] {prompt} [/INST]

—Context size: 32000

—Run as LlamaEdge service

bash

  wasmedge --dir .:. --nn-preload default:GGML:AUTO:Codestral-22B-v0.1-hf-Q5_K_M.gguf \
    llama-api-server.wasm \
    --prompt-template mistral-instruct \
    --ctx-size 32000 \
    --model-name Codestral-22B-v0.1

—Run as LlamaEdge command app

bash

  wasmedge --dir .:. --nn-preload default:GGML:AUTO:Codestral-22B-v0.1-hf-Q5_K_M.gguf \
    llama-chat.wasm \
    --prompt-template mistral-instruct \
    --ctx-size 32000

Quantized GGUF Models

Name	Quant method	Bits	Size	Use case
Codestral-22B-v0.1-hf-Q2_K.gguf	Q2_K	2	8.27 GB	smallest, significant quality loss - not recommended for most purposes
Codestral-22B-v0.1-hf-Q3_K_L.gguf	Q3KL	3	11.7 GB	small, substantial quality loss
Codestral-22B-v0.1-hf-Q3_K_M.gguf	Q3KM	3	10.8 GB	very small, high quality loss
Codestral-22B-v0.1-hf-Q3_K_S.gguf	Q3KS	3	9.64 GB	very small, high quality loss
Codestral-22B-v0.1-hf-Q4_0.gguf	Q4_0	4	12.6 GB	legacy; small, very high quality loss - prefer using Q3KM
Codestral-22B-v0.1-hf-Q4_K_M.gguf	Q4KM	4	13.3 GB	medium, balanced quality - recommended
Codestral-22B-v0.1-hf-Q4_K_S.gguf	Q4KS	4	12.7 GB	small, greater quality loss
Codestral-22B-v0.1-hf-Q5_0.gguf	Q5_0	5	15.3 GB	legacy; medium, balanced quality - prefer using Q4KM
Codestral-22B-v0.1-hf-Q5_K_M.gguf	Q5KM	5	15.7 GB	large, very low quality loss - recommended
Codestral-22B-v0.1-hf-Q5_K_S.gguf	Q5KS	5	15.3 GB	large, low quality loss - recommended
Codestral-22B-v0.1-hf-Q6_K.gguf	Q6_K	6	18.3 GB	very large, extremely low quality loss
Codestral-22B-v0.1-hf-Q8_0.gguf	Q8_0	8	23.6 GB	very large, extremely low quality loss - not recommended
Codestral-22B-v0.1-hf-f16.gguf	f16	16	44.5 GB

Quantized with llama.cpp b3030.