Hugging Face's logo

second-state
/
Qwen2-VL-72B-Instruct-GGUF

Image-Text-to-Text
Transformers
GGUF
English
multimodal
Inference Endpoints
conversational
Model card Files Community
Qwen2-VL-72B-Instruct-GGUF / README.md
apepkuss79's picture
apepkuss79
Update README.md
9b00931 verified
preview code
|
raw
history blame contribute delete
7.38 kB
metadata
base_model: Qwen/Qwen2-VL-72B-Instruct
license: other
license_name: tongyi-qianwen
license_link: https://huggingface.co/Qwen/Qwen2-VL-72B-Instruct/blob/main/LICENSE
model_creator: Qwen
model_name: Qwen2-VL-72B-Instruct
quantized_by: Second State Inc.
language:
  - en
pipeline_tag: image-text-to-text
tags:
  - multimodal
library_name: transformers

Qwen2-VL-72B-Instruct-GGUF

Original Model

Qwen/Qwen2-VL-72B-Instruct

Run with LlamaEdge

  • LlamaEdge version: v0.16.0

  • Prompt template

    • Prompt type: qwen2-vision

    • Prompt string

      <|im_start|>system
{system_prompt}<|im_end|>
<|im_start|>user
<|vision_start|>{image_placeholder}<|vision_end|>{user_prompt}<|im_end|>
<|im_start|>assistant

  • Context size: 128000

  • Run as LlamaEdge service

    wasmedge --dir .:. --nn-preload default:GGML:AUTO:Qwen2-VL-72B-Instruct-Q5_K_M.gguf \
  llama-api-server.wasm \
  --model-name Qwen2-VL-72B-Instruct \
  --prompt-template qwen2-vision \
  --llava-mmproj Qwen2-VL-72B-Instruct-vision-encoder.gguf \
  --ctx-size 128000

Quantized GGUF Models

Name Quant method Bits Size Use case
Qwen2-VL-72B-Instruct-Q2_K.gguf Q2_K 2 29.8 GB smallest, significant quality loss - not recommended for most purposes
Qwen2-VL-72B-Instruct-Q3_K_L.gguf Q3_K_L 3 39.5 GB small, substantial quality loss
Qwen2-VL-72B-Instruct-Q3_K_M.gguf Q3_K_M 3 37.7 GB very small, high quality loss
Qwen2-VL-72B-Instruct-Q3_K_S.gguf Q3_K_S 3 34.5 GB very small, high quality loss
Qwen2-VL-72B-Instruct-Q4_0.gguf Q4_0 4 41.2 GB legacy; small, very high quality loss - prefer using Q3_K_M
Qwen2-VL-72B-Instruct-Q4_K_M.gguf Q4_K_M 4 47.4 GB medium, balanced quality - recommended
Qwen2-VL-72B-Instruct-Q4_K_S.gguf Q4_K_S 4 43.9 GB small, greater quality loss
Qwen2-VL-72B-Instruct-Q5_0-00001-of-00002.gguf Q5_0 5 29.9 GB legacy; medium, balanced quality - prefer using Q4_K_M
Qwen2-VL-72B-Instruct-Q5_0-00002-of-00002.gguf Q5_0 5 20.2 GB legacy; medium, balanced quality - prefer using Q4_K_M
Qwen2-VL-72B-Instruct-Q5_K_M-00001-of-00002.gguf Q5_K_M 5 29.9 GB large, very low quality loss - recommended
Qwen2-VL-72B-Instruct-Q5_K_M-00002-of-00002.gguf Q5_K_M 5 24.5 GB large, very low quality loss - recommended
Qwen2-VL-72B-Instruct-Q5_K_S-00001-of-00002.gguf Q5_K_S 5 29.8 GB large, low quality loss - recommended
Qwen2-VL-72B-Instruct-Q5_K_S-00002-of-00002.gguf Q5_K_S 5 21.5 GB large, low quality loss - recommended
Qwen2-VL-72B-Instruct-Q6_K-00001-of-00003.gguf Q6_K 6 29.9 GB very large, extremely low quality loss
Qwen2-VL-72B-Instruct-Q6_K-00002-of-00003.gguf Q6_K 6 29.9 GB very large, extremely low quality loss
Qwen2-VL-72B-Instruct-Q6_K-00003-of-00003.gguf Q6_K 6 4.55 GB very large, extremely low quality loss
Qwen2-VL-72B-Instruct-Q8_0-00001-of-00003.gguf Q8_0 8 29.9 GB very large, extremely low quality loss - not recommended
Qwen2-VL-72B-Instruct-Q8_0-00002-of-00003.gguf Q8_0 8 29.8 GB very large, extremely low quality loss - not recommended
Qwen2-VL-72B-Instruct-Q8_0-00003-of-00003.gguf Q8_0 8 17.6 GB very large, extremely low quality loss - not recommended
Qwen2-VL-72B-Instruct-f16-00001-of-00005.gguf f16 16 29.9 GB
Qwen2-VL-72B-Instruct-f16-00002-of-00005.gguf f16 16 29.7 GB
Qwen2-VL-72B-Instruct-f16-00003-of-00005.gguf f16 16 29.7 GB
Qwen2-VL-72B-Instruct-f16-00004-of-00005.gguf f16 16 29.5 GB
Qwen2-VL-72B-Instruct-f16-00005-of-00005.gguf f16 16 26.6 GB
Qwen2-VL-72B-Instruct-vision-encoder.gguf f16 16 2.8 GB

Quantized with llama.cpp b4329