Update README.md
Browse files
README.md
CHANGED
@@ -40,9 +40,11 @@ inference:
|
|
40 |
|
41 |
# ConflLlama: Domain-Specific LLM for Conflict Event Classification
|
42 |
|
43 |
-
|
44 |
-
|
45 |
-
|
|
|
|
|
46 |
|
47 |
**ConflLlama** is a large language model fine-tuned to classify conflict events from text descriptions. This repository contains the GGUF quantized models (q4\_k\_m, q8\_0, and BF16) based on **Llama-3.1 8B**, which have been adapted for the specialized domain of political violence research.
|
48 |
|
@@ -93,9 +95,11 @@ The most significant improvements were observed in historically difficult-to-cla
|
|
93 |
* Alpha (`lora_alpha`): 16
|
94 |
* Target Modules: `q_proj`, `k_proj`, `v_proj`, `o_proj`, `gate_proj`, `up_proj`, `down_proj`
|
95 |
|
96 |
-
|
97 |
-
|
98 |
-
|
|
|
|
|
99 |
|
100 |
### Training Data
|
101 |
|
@@ -103,9 +107,11 @@ The most significant improvements were observed in historically difficult-to-cla
|
|
103 |
* **Time Period**: The training dataset consists of 171,514 events that occurred before January 1, 2017. The test set includes 38,192 events from 2017 onwards.
|
104 |
* **Preprocessing**: The pipeline filters data by date, cleans text summaries, and combines primary, secondary, and tertiary attack types into a single multi-label field.
|
105 |
|
106 |
-
|
107 |
-
|
108 |
-
|
|
|
|
|
109 |
|
110 |
-----
|
111 |
|
@@ -134,9 +140,11 @@ This model is designed for academic and research purposes within the fields of p
|
|
134 |
|
135 |
## Training Logs
|
136 |
|
137 |
-
|
138 |
-
|
139 |
-
|
|
|
|
|
140 |
|
141 |
The training logs show a successful training run with healthy convergence patterns:
|
142 |
|
|
|
40 |
|
41 |
# ConflLlama: Domain-Specific LLM for Conflict Event Classification
|
42 |
|
43 |
+
<p align="center">
|
44 |
+
|
45 |
+
<img src="images/logo.png" alt="Project Logo" width="300"/>
|
46 |
+
|
47 |
+
</p>
|
48 |
|
49 |
**ConflLlama** is a large language model fine-tuned to classify conflict events from text descriptions. This repository contains the GGUF quantized models (q4\_k\_m, q8\_0, and BF16) based on **Llama-3.1 8B**, which have been adapted for the specialized domain of political violence research.
|
50 |
|
|
|
95 |
* Alpha (`lora_alpha`): 16
|
96 |
* Target Modules: `q_proj`, `k_proj`, `v_proj`, `o_proj`, `gate_proj`, `up_proj`, `down_proj`
|
97 |
|
98 |
+
<p align="center">
|
99 |
+
|
100 |
+
<img src="images/model-arch.png" alt="Model Training Architecture" width="800"/>
|
101 |
+
|
102 |
+
</p>
|
103 |
|
104 |
### Training Data
|
105 |
|
|
|
107 |
* **Time Period**: The training dataset consists of 171,514 events that occurred before January 1, 2017. The test set includes 38,192 events from 2017 onwards.
|
108 |
* **Preprocessing**: The pipeline filters data by date, cleans text summaries, and combines primary, secondary, and tertiary attack types into a single multi-label field.
|
109 |
|
110 |
+
<p align="center">
|
111 |
+
|
112 |
+
<img src="images/preprocessing.png" alt="Data Preprocessing Pipeline" width="800"/>
|
113 |
+
|
114 |
+
</p>
|
115 |
|
116 |
-----
|
117 |
|
|
|
140 |
|
141 |
## Training Logs
|
142 |
|
143 |
+
<p align="center">
|
144 |
+
|
145 |
+
<img src="images/training.png" alt="Training Logs" width="800"/>
|
146 |
+
|
147 |
+
</p>
|
148 |
|
149 |
The training logs show a successful training run with healthy convergence patterns:
|
150 |
|