Create README.md
Browse filesadded link to upstream format and shoutouts
README.md
ADDED
|
@@ -0,0 +1,128 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| 1 |
+
---
|
| 2 |
+
base_model: mistralai/Mistral-7B-v0.1
|
| 3 |
+
tags:
|
| 4 |
+
- mistral
|
| 5 |
+
- instruct
|
| 6 |
+
- finetune
|
| 7 |
+
- chatml
|
| 8 |
+
- gpt4
|
| 9 |
+
- synthetic data
|
| 10 |
+
- distillation
|
| 11 |
+
- multimodal
|
| 12 |
+
- llava
|
| 13 |
+
model-index:
|
| 14 |
+
- name: Nous-Hermes-2-Vision
|
| 15 |
+
results: []
|
| 16 |
+
license: apache-2.0
|
| 17 |
+
language:
|
| 18 |
+
- en
|
| 19 |
+
---
|
| 20 |
+
|
| 21 |
+
GGUF Quants by Twobob, Thanks to @jartine and @cmp-nct for the assists
|
| 22 |
+
It's vicuna
|
| 23 |
+
|
| 24 |
+
ref: [here](https://github.com/qnguyen3/hermes-llava/blob/173b4ef441b5371c1e7d99da7a2e7c14c77ad12f/llava/conversation.py#L252)
|
| 25 |
+
|
| 26 |
+
# Nous-Hermes-2-Vision - Mistral 7B
|
| 27 |
+
|
| 28 |
+
|
| 29 |
+
|
| 30 |
+
|
| 31 |
+
*In the tapestry of Greek mythology, Hermes reigns as the eloquent Messenger of the Gods, a deity who deftly bridges the realms through the art of communication. It is in homage to this divine mediator that I name this advanced LLM "Hermes," a system crafted to navigate the complex intricacies of human discourse with celestial finesse.*
|
| 32 |
+
|
| 33 |
+
## Model description
|
| 34 |
+
|
| 35 |
+
Nous-Hermes-2-Vision stands as a pioneering Vision-Language Model, leveraging advancements from the renowned **OpenHermes-2.5-Mistral-7B** by teknium. This model incorporates two pivotal enhancements, setting it apart as a cutting-edge solution:
|
| 36 |
+
|
| 37 |
+
- **SigLIP-400M Integration**: Diverging from traditional approaches that rely on substantial 3B vision encoders, Nous-Hermes-2-Vision harnesses the formidable SigLIP-400M. This strategic choice not only streamlines the model's architecture, making it more lightweight, but also capitalizes on SigLIP's remarkable capabilities. The result? A remarkable boost in performance that defies conventional expectations.
|
| 38 |
+
|
| 39 |
+
- **Custom Dataset Enriched with Function Calling**: Our model's training data includes a unique feature – function calling. This distinctive addition transforms Nous-Hermes-2-Vision into a **Vision-Language Action Model**. Developers now have a versatile tool at their disposal, primed for crafting a myriad of ingenious automations.
|
| 40 |
+
|
| 41 |
+
This project is led by [qnguyen3](https://twitter.com/stablequan) and [teknium](https://twitter.com/Teknium1).
|
| 42 |
+
## Training
|
| 43 |
+
### Dataset
|
| 44 |
+
- 220K from **LVIS-INSTRUCT4V**
|
| 45 |
+
- 60K from **ShareGPT4V**
|
| 46 |
+
- 150K Private **Function Calling Data**
|
| 47 |
+
- 50K conversations from teknium's **OpenHermes-2.5**
|
| 48 |
+
|
| 49 |
+
## Usage
|
| 50 |
+
### Prompt Format
|
| 51 |
+
- Like other LLaVA's variants, this model uses Vicuna-V1 as its prompt template. Please refer to `conv_llava_v1` in [this file](https://github.com/qnguyen3/hermes-llava/blob/main/llava/conversation.py)
|
| 52 |
+
- For Gradio UI, please visit this [GitHub Repo](https://github.com/qnguyen3/hermes-llava)
|
| 53 |
+
|
| 54 |
+
### Function Calling
|
| 55 |
+
- For functiong calling, the message should start with a `<fn_call>` tag. Here is an example:
|
| 56 |
+
|
| 57 |
+
```json
|
| 58 |
+
<fn_call>{
|
| 59 |
+
"type": "object",
|
| 60 |
+
"properties": {
|
| 61 |
+
"bus_colors": {
|
| 62 |
+
"type": "array",
|
| 63 |
+
"description": "The colors of the bus in the image.",
|
| 64 |
+
"items": {
|
| 65 |
+
"type": "string",
|
| 66 |
+
"enum": ["red", "blue", "green", "white"]
|
| 67 |
+
}
|
| 68 |
+
},
|
| 69 |
+
"bus_features": {
|
| 70 |
+
"type": "string",
|
| 71 |
+
"description": "The features seen on the back of the bus."
|
| 72 |
+
},
|
| 73 |
+
"bus_location": {
|
| 74 |
+
"type": "string",
|
| 75 |
+
"description": "The location of the bus (driving or pulled off to the side).",
|
| 76 |
+
"enum": ["driving", "pulled off to the side"]
|
| 77 |
+
}
|
| 78 |
+
}
|
| 79 |
+
}
|
| 80 |
+
```
|
| 81 |
+
|
| 82 |
+
Output:
|
| 83 |
+
```json
|
| 84 |
+
{
|
| 85 |
+
"bus_colors": ["red", "white"],
|
| 86 |
+
"bus_features": "An advertisement",
|
| 87 |
+
"bus_location": "driving"
|
| 88 |
+
}
|
| 89 |
+
```
|
| 90 |
+
|
| 91 |
+
## Example
|
| 92 |
+
|
| 93 |
+
### Chat
|
| 94 |
+
|
| 95 |
+
|
| 96 |
+
### Function Calling
|
| 97 |
+
Input image:
|
| 98 |
+
|
| 99 |
+
|
| 100 |
+
|
| 101 |
+
Input message:
|
| 102 |
+
```json
|
| 103 |
+
<fn_call>{
|
| 104 |
+
"type": "object",
|
| 105 |
+
"properties": {
|
| 106 |
+
"food_list": {
|
| 107 |
+
"type": "array",
|
| 108 |
+
"description": "List of all the food",
|
| 109 |
+
"items": {
|
| 110 |
+
"type": "string",
|
| 111 |
+
}
|
| 112 |
+
},
|
| 113 |
+
}
|
| 114 |
+
}
|
| 115 |
+
```
|
| 116 |
+
|
| 117 |
+
Output:
|
| 118 |
+
```json
|
| 119 |
+
{
|
| 120 |
+
"food_list": [
|
| 121 |
+
"Double Burger",
|
| 122 |
+
"Cheeseburger",
|
| 123 |
+
"French Fries",
|
| 124 |
+
"Shakes",
|
| 125 |
+
"Coffee"
|
| 126 |
+
]
|
| 127 |
+
}
|
| 128 |
+
```
|