Update README.md
Browse files
README.md
CHANGED
|
@@ -1,84 +1,82 @@
|
|
| 1 |
-
---
|
| 2 |
-
base_model: Qwen/Qwen2.5-VL-32B-Instruct
|
| 3 |
-
license: apache-2.0
|
| 4 |
-
model_creator: Qwen
|
| 5 |
-
model_name: Qwen2.5-VL-32B-Instruct
|
| 6 |
-
quantized_by: Second State Inc.
|
| 7 |
-
language:
|
| 8 |
-
- en
|
| 9 |
-
pipeline_tag: image-text-to-text
|
| 10 |
-
tags:
|
| 11 |
-
- multimodal
|
| 12 |
-
library_name: transformers
|
| 13 |
-
---
|
| 14 |
-
|
| 15 |
-
<!-- header start -->
|
| 16 |
-
<!-- 200823 -->
|
| 17 |
-
<div style="width: auto; margin-left: auto; margin-right: auto">
|
| 18 |
-
<img src="https://github.com/LlamaEdge/LlamaEdge/raw/dev/assets/logo.svg" style="width: 100%; min-width: 400px; display: block; margin: auto;">
|
| 19 |
-
</div>
|
| 20 |
-
<hr style="margin-top: 1.0em; margin-bottom: 1.0em;">
|
| 21 |
-
<!-- header end -->
|
| 22 |
-
|
| 23 |
-
# Qwen2.5-VL-32B-Instruct-GGUF
|
| 24 |
-
|
| 25 |
-
## Original Model
|
| 26 |
-
|
| 27 |
-
[Qwen/Qwen2.5-VL-32B-Instruct](https://huggingface.co/Qwen/Qwen2.5-VL-32B-Instruct)
|
| 28 |
-
|
| 29 |
-
## Run with LlamaEdge
|
| 30 |
-
|
| 31 |
-
- LlamaEdge version:
|
| 32 |
-
|
| 33 |
-
|
| 34 |
-
|
| 35 |
-
- Prompt
|
| 36 |
-
|
| 37 |
-
- Prompt
|
| 38 |
-
|
| 39 |
-
|
| 40 |
-
|
| 41 |
-
|
| 42 |
-
<|im_start|>
|
| 43 |
-
{
|
| 44 |
-
<|im_start|>
|
| 45 |
-
|
| 46 |
-
|
| 47 |
-
|
| 48 |
-
|
| 49 |
-
-
|
| 50 |
-
|
| 51 |
-
|
| 52 |
-
|
| 53 |
-
|
| 54 |
-
|
| 55 |
-
--
|
| 56 |
-
|
| 57 |
-
--
|
| 58 |
-
--
|
| 59 |
-
|
| 60 |
-
|
| 61 |
-
|
| 62 |
-
|
| 63 |
-
|
| 64 |
-
|
| 65 |
-
|
|
| 66 |
-
|
|
| 67 |
-
| [Qwen2.5-VL-32B-Instruct-
|
| 68 |
-
| [Qwen2.5-VL-32B-Instruct-
|
| 69 |
-
| [Qwen2.5-VL-32B-Instruct-
|
| 70 |
-
| [Qwen2.5-VL-32B-Instruct-
|
| 71 |
-
| [Qwen2.5-VL-32B-Instruct-
|
| 72 |
-
| [Qwen2.5-VL-32B-Instruct-
|
| 73 |
-
| [Qwen2.5-VL-32B-Instruct-
|
| 74 |
-
| [Qwen2.5-VL-32B-Instruct-
|
| 75 |
-
| [Qwen2.5-VL-32B-Instruct-
|
| 76 |
-
| [Qwen2.5-VL-32B-Instruct-
|
| 77 |
-
| [Qwen2.5-VL-32B-Instruct-
|
| 78 |
-
| [Qwen2.5-VL-32B-Instruct-
|
| 79 |
-
| [Qwen2.5-VL-32B-Instruct-f16-
|
| 80 |
-
| [Qwen2.5-VL-32B-Instruct-
|
| 81 |
-
|
| 82 |
-
| [Qwen2.5-VL-32B-Instruct-vision.gguf](https://huggingface.co/second-state/Qwen2.5-VL-32B-Instruct-GGUF/blob/main/Qwen2.5-VL-32B-Instruct-vision.gguf) | f16 | 16 | 1.38 GB| |
|
| 83 |
-
|
| 84 |
*Quantized with llama.cpp b5196*
|
|
|
|
| 1 |
+
---
|
| 2 |
+
base_model: Qwen/Qwen2.5-VL-32B-Instruct
|
| 3 |
+
license: apache-2.0
|
| 4 |
+
model_creator: Qwen
|
| 5 |
+
model_name: Qwen2.5-VL-32B-Instruct
|
| 6 |
+
quantized_by: Second State Inc.
|
| 7 |
+
language:
|
| 8 |
+
- en
|
| 9 |
+
pipeline_tag: image-text-to-text
|
| 10 |
+
tags:
|
| 11 |
+
- multimodal
|
| 12 |
+
library_name: transformers
|
| 13 |
+
---
|
| 14 |
+
|
| 15 |
+
<!-- header start -->
|
| 16 |
+
<!-- 200823 -->
|
| 17 |
+
<div style="width: auto; margin-left: auto; margin-right: auto">
|
| 18 |
+
<img src="https://github.com/LlamaEdge/LlamaEdge/raw/dev/assets/logo.svg" style="width: 100%; min-width: 400px; display: block; margin: auto;">
|
| 19 |
+
</div>
|
| 20 |
+
<hr style="margin-top: 1.0em; margin-bottom: 1.0em;">
|
| 21 |
+
<!-- header end -->
|
| 22 |
+
|
| 23 |
+
# Qwen2.5-VL-32B-Instruct-GGUF
|
| 24 |
+
|
| 25 |
+
## Original Model
|
| 26 |
+
|
| 27 |
+
[Qwen/Qwen2.5-VL-32B-Instruct](https://huggingface.co/Qwen/Qwen2.5-VL-32B-Instruct)
|
| 28 |
+
|
| 29 |
+
## Run with LlamaEdge
|
| 30 |
+
|
| 31 |
+
- LlamaEdge version: [v0.18.4](https://github.com/LlamaEdge/LlamaEdge/releases/tag/0.18.4)
|
| 32 |
+
|
| 33 |
+
- Prompt template
|
| 34 |
+
|
| 35 |
+
- Prompt type: `qwen2-vision`
|
| 36 |
+
|
| 37 |
+
- Prompt string
|
| 38 |
+
|
| 39 |
+
```text
|
| 40 |
+
<|im_start|>system
|
| 41 |
+
{system_prompt}<|im_end|>
|
| 42 |
+
<|im_start|>user
|
| 43 |
+
<|vision_start|>{image_placeholder}<|vision_end|>{user_prompt}<|im_end|>
|
| 44 |
+
<|im_start|>assistant
|
| 45 |
+
```
|
| 46 |
+
|
| 47 |
+
- Context size: `32000`
|
| 48 |
+
|
| 49 |
+
- Run as LlamaEdge service
|
| 50 |
+
|
| 51 |
+
```bash
|
| 52 |
+
wasmedge --dir .:. \
|
| 53 |
+
--nn-preload default:GGML:AUTO:Qwen2.5-VL-32B-Instruct-Q5_K_M.gguf \
|
| 54 |
+
llama-api-server.wasm \
|
| 55 |
+
--model-name Qwen2.5-VL-32B-Instruct \
|
| 56 |
+
--prompt-template qwen2-vision \
|
| 57 |
+
--llava-mmproj Qwen2.5-VL-32B-Instruct-vision.gguf \
|
| 58 |
+
--ctx-size 32000
|
| 59 |
+
```
|
| 60 |
+
|
| 61 |
+
## Quantized GGUF Models
|
| 62 |
+
|
| 63 |
+
| Name | Quant method | Bits | Size | Use case |
|
| 64 |
+
| ---- | ---- | ---- | ---- | ----- |
|
| 65 |
+
| [Qwen2.5-VL-32B-Instruct-Q2_K.gguf](https://huggingface.co/second-state/Qwen2.5-VL-32B-Instruct-GGUF/blob/main/Qwen2.5-VL-32B-Instruct-Q2_K.gguf) | Q2_K | 2 | 12.3 GB| smallest, significant quality loss - not recommended for most purposes |
|
| 66 |
+
| [Qwen2.5-VL-32B-Instruct-Q3_K_L.gguf](https://huggingface.co/second-state/Qwen2.5-VL-32B-Instruct-GGUF/blob/main/Qwen2.5-VL-32B-Instruct-Q3_K_L.gguf) | Q3_K_L | 3 | 17.2 GB| small, substantial quality loss |
|
| 67 |
+
| [Qwen2.5-VL-32B-Instruct-Q3_K_M.gguf](https://huggingface.co/second-state/Qwen2.5-VL-32B-Instruct-GGUF/blob/main/Qwen2.5-VL-32B-Instruct-Q3_K_M.gguf) | Q3_K_M | 3 | 15.9 GB| very small, high quality loss |
|
| 68 |
+
| [Qwen2.5-VL-32B-Instruct-Q3_K_S.gguf](https://huggingface.co/second-state/Qwen2.5-VL-32B-Instruct-GGUF/blob/main/Qwen2.5-VL-32B-Instruct-Q3_K_S.gguf) | Q3_K_S | 3 | 14.4 GB| very small, high quality loss |
|
| 69 |
+
| [Qwen2.5-VL-32B-Instruct-Q4_0.gguf](https://huggingface.co/second-state/Qwen2.5-VL-32B-Instruct-GGUF/blob/main/Qwen2.5-VL-32B-Instruct-Q4_0.gguf) | Q4_0 | 4 | 18.6 GB| legacy; small, very high quality loss - prefer using Q3_K_M |
|
| 70 |
+
| [Qwen2.5-VL-32B-Instruct-Q4_K_M.gguf](https://huggingface.co/second-state/Qwen2.5-VL-32B-Instruct-GGUF/blob/main/Qwen2.5-VL-32B-Instruct-Q4_K_M.gguf) | Q4_K_M | 4 | 19.9 GB| medium, balanced quality - recommended |
|
| 71 |
+
| [Qwen2.5-VL-32B-Instruct-Q4_K_S.gguf](https://huggingface.co/second-state/Qwen2.5-VL-32B-Instruct-GGUF/blob/main/Qwen2.5-VL-32B-Instruct-Q4_K_S.gguf) | Q4_K_S | 4 | 18.8 GB| small, greater quality loss |
|
| 72 |
+
| [Qwen2.5-VL-32B-Instruct-Q5_0.gguf](https://huggingface.co/second-state/Qwen2.5-VL-32B-Instruct-GGUF/blob/main/Qwen2.5-VL-32B-Instruct-Q5_0.gguf) | Q5_0 | 5 | 22.6 GB| legacy; medium, balanced quality - prefer using Q4_K_M |
|
| 73 |
+
| [Qwen2.5-VL-32B-Instruct-Q5_K_M.gguf](https://huggingface.co/second-state/Qwen2.5-VL-32B-Instruct-GGUF/blob/main/Qwen2.5-VL-32B-Instruct-Q5_K_M.gguf) | Q5_K_M | 5 | 23.3 GB| large, very low quality loss - recommended |
|
| 74 |
+
| [Qwen2.5-VL-32B-Instruct-Q5_K_S.gguf](https://huggingface.co/second-state/Qwen2.5-VL-32B-Instruct-GGUF/blob/main/Qwen2.5-VL-32B-Instruct-Q5_K_S.gguf) | Q5_K_S | 5 | 22.6 GB| large, low quality loss - recommended |
|
| 75 |
+
| [Qwen2.5-VL-32B-Instruct-Q6_K.gguf](https://huggingface.co/second-state/Qwen2.5-VL-32B-Instruct-GGUF/blob/main/Qwen2.5-VL-32B-Instruct-Q6_K.gguf) | Q6_K | 6 | 26.9 GB| very large, extremely low quality loss |
|
| 76 |
+
| [Qwen2.5-VL-32B-Instruct-Q8_0.gguf](https://huggingface.co/second-state/Qwen2.5-VL-32B-Instruct-GGUF/blob/main/Qwen2.5-VL-32B-Instruct-Q8_0.gguf) | Q8_0 | 8 | 34.8 GB| very large, extremely low quality loss - not recommended |
|
| 77 |
+
| [Qwen2.5-VL-32B-Instruct-f16-00001-of-00003.gguf](https://huggingface.co/second-state/Qwen2.5-VL-32B-Instruct-GGUF/blob/main/Qwen2.5-VL-32B-Instruct-f16-00001-of-00003.gguf) | f16 | 16 | 29.8 GB| |
|
| 78 |
+
| [Qwen2.5-VL-32B-Instruct-f16-00002-of-00003.gguf](https://huggingface.co/second-state/Qwen2.5-VL-32B-Instruct-GGUF/blob/main/Qwen2.5-VL-32B-Instruct-f16-00002-of-00003.gguf) | f16 | 16 | 29.8 GB| |
|
| 79 |
+
| [Qwen2.5-VL-32B-Instruct-f16-00003-of-00003.gguf](https://huggingface.co/second-state/Qwen2.5-VL-32B-Instruct-GGUF/blob/main/Qwen2.5-VL-32B-Instruct-f16-00003-of-00003.gguf) | f16 | 16 | 5.87 GB| |
|
| 80 |
+
| [Qwen2.5-VL-32B-Instruct-vision.gguf](https://huggingface.co/second-state/Qwen2.5-VL-32B-Instruct-GGUF/blob/main/Qwen2.5-VL-32B-Instruct-vision.gguf) | f16 | 16 | 1.38 GB| |
|
| 81 |
+
|
|
|
|
|
|
|
| 82 |
*Quantized with llama.cpp b5196*
|