rwitz commited on
Commit
ffff1f8
1 Parent(s): 04f5373

Update README.md

Browse files
Files changed (1) hide show
  1. README.md +109 -0
README.md CHANGED
@@ -6,3 +6,112 @@ base_model:
6
  - meta-llama/Llama-3.1-8B-Instruct
7
  pipeline_tag: text-generation
8
  ---
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
6
  - meta-llama/Llama-3.1-8B-Instruct
7
  pipeline_tag: text-generation
8
  ---
9
+ # Llama-3-1-8b Fine-Tuned Model for Roleplay, Logic, and Reasoning
10
+
11
+ ![Cover Image](https://i.ibb.co/PYCdt9n/3i-RPOp-Vn-Tb-O4-E021n6-Pljg.jpg)
12
+
13
+ ## Overview
14
+
15
+ This repository provides a fine-tuned version of the **Llama-3-1-8b base model**, optimized for roleplaying, logic, and reasoning tasks. Utilizing iterative fine-tuning and self-generated chat logs, this model delivers engaging and coherent conversational experiences.
16
+
17
+ ## Model Specifications
18
+
19
+ - **Parameters**: 8 Billion (8B)
20
+ - **Precision**: bf16 (Brain Floating Point 16-bit)
21
+ - **Fine-Tuning Method**: LoRa (Low-Rank Adaptation)
22
+ - **Datasets Used**:
23
+ - Roleplay Dataset
24
+ - Reasoning and Logic Dataset
25
+ - **Fine-Tuning Approach**: Iterative Fine-Tuning using self-chat logs
26
+
27
+ ## Recommended Settings
28
+
29
+ To achieve optimal performance with this model, we recommend the following settings:
30
+
31
+ - **Minimum Probability (`min_p`)**: `0.05`
32
+ - **Temperature**: `1.1` or higher
33
+
34
+ > **Note**: Due to the nature of the fine-tuning, setting the temperature to `1.1` or higher helps prevent the model from repeating itself and encourages more creative and coherent responses.
35
+
36
+ ## Usage Instructions
37
+
38
+ We recommend using the [oobabooga text-generation-webui](https://github.com/oobabooga/text-generation-webui) for an optimal experience. Load the model in `bf16` precision and enable `flash-attention2` for improved performance.
39
+
40
+ ### Installation Steps
41
+
42
+ 1. **Clone the WebUI Repository**:
43
+
44
+ ```bash
45
+ git clone https://github.com/oobabooga/text-generation-webui
46
+ cd text-generation-webui
47
+ ```
48
+
49
+ 2. **Install Dependencies**:
50
+
51
+ ```bash
52
+ pip install -r requirements.txt
53
+ ```
54
+
55
+ 3. **Download the Model**:
56
+
57
+ Download the fine-tuned model from [Hugging Face](#) and place it in the `models` directory.
58
+
59
+ 4. **Launch the WebUI**:
60
+
61
+ ```bash
62
+ python server.py --bf16 --flash-attention
63
+ ```
64
+
65
+ ### Sample Prompt Formats
66
+
67
+ You can interact with the model using either **chat format** or **chat-instruct format**. Here's an example:
68
+
69
+ ```plaintext
70
+ Ryan is a computer engineer who works at Intel.
71
+
72
+ Ryan: Hey, how's it going Natalie?
73
+ Natalie: Good, how are things going with you, Ryan?
74
+ Ryan: Great, I'm just doing just great.
75
+ ```
76
+
77
+ ### Text Generation Example
78
+
79
+ ![Text Generation Example](https://i.ibb.co/J5ZVCnR/Screenshot-45.png)
80
+
81
+ ## Model Capabilities
82
+
83
+ Below are some examples showcasing the model's performance in various tasks:
84
+
85
+ ### Instruct Log Examples
86
+
87
+ 1. **Logic and Reasoning**
88
+
89
+ ![Instruct Log 1](https://i.ibb.co/0ngp6zf/Screenshot-42.png)
90
+
91
+ 2. **Roleplaying Scenario**
92
+
93
+ ![Instruct Log 2](https://i.ibb.co/GQ8Ffn1/Screenshot-43.png)
94
+
95
+ 3. **Creative Writing**
96
+
97
+ ![Instruct Log 3](https://i.ibb.co/4JkCjtf/Screenshot-44.png)
98
+
99
+ ## Limitations and Tips
100
+
101
+ While this model excels in chat and roleplaying scenarios, it isn't perfect. If you notice the model repeating itself or providing less coherent responses:
102
+
103
+ - **Increase the Temperature**: Setting the temperature higher (≥ `1.1`) can help generate more diverse and creative outputs.
104
+ - **Adjust `min_p` Setting**: Ensuring `min_p` is at least `0.05` can prevent low-probability tokens from being excluded, enhancing the response quality.
105
+
106
+ ## Acknowledgments
107
+
108
+ - **oobabooga text-generation-webui**: A powerful interface for running and interacting with language models. [GitHub Repository](https://github.com/oobabooga/text-generation-webui)
109
+ - **Hugging Face**: For hosting the model and providing a platform for collaboration. [Website](https://huggingface.co/)
110
+
111
+ ## License
112
+
113
+ [Specify the license under which the model is released, e.g., MIT License, Apache 2.0, etc.]
114
+
115
+ ---
116
+
117
+ *For any issues or questions, please open an issue in this repository.*