Upload README.md with huggingface_hub

README.md

@@ -59,53 +59,15 @@ This is a fine-tuned version of the Llama-3.1 model specifically optimized for D
 - **Framework**: MLX-LM
 - **Hardware**: Apple M1 Max (64GB RAM)
 
-## Intended Uses
+## Quick Start
 
-### Primary Use Cases
-
-1. **Election Security**: Detect and classify disinformation campaigns targeting electoral processes
-2. **Content Moderation**: Identify harmful content that undermines electoral integrity
-3. **Research**: Academic research on disinformation patterns and meta-narratives
-4. **Policy Analysis**: Support policy development for election security measures
-
-### Target Applications
-
-- Social media monitoring platforms
-- Election security organizations
-- Fact-checking organizations
-- Academic research institutions
-- Government agencies
-- Civil society organizations
-
-## Training Data
-
-The model was trained on the [DISARM Election Watch Dataset](https://huggingface.co/datasets/ArapCheruiyot/disarm-election-watch-dataset), which contains:
-
-### Data Sources
-- **Telegram**: 3,632 examples (60.3%)
-- **X/Twitter**: 2,038 examples (33.9%)
-- **TikTok**: 248 examples (4.1%)
-- **DISARM**: 101 examples (1.7%)
-
-### Task Types
-- **DISARM Classification**: 101 examples
-- **Content Analysis**: 5,770 examples
-- **Narrative Analysis**: 148 examples
-
-### Data Split
-- **Training**: 4,815 examples (80%)
-- **Validation**: 601 examples (10%)
-- **Test**: 603 examples (10%)
-
-## Usage
-
-### With MLX-LM (Fused Model)
+### Using with MLX-LM
 
 ```python
 from mlx_lm import load, generate
 
 # Load the complete fine-tuned model
-model, tokenizer = load("ArapCheruiyot/disarm-ew-llama3-finetuned")
+model, tokenizer = load("models/disarm_ew_llama3_finetuned")
 
 # Example prompt
 prompt = """### Instruction:
@@ -121,34 +83,36 @@ response = generate(model, tokenizer, prompt, max_tokens=256, temp=0.1)
 print(response)
 ```
 
-### With LoRA Adapters
+### Using with Ollama
 
-```python
-from mlx_lm import load, generate
+```bash
+# Create Ollama model
+ollama create disarm-ew-llama3-finetuned -f Modelfile
 
-# Load base model with LoRA adapters
-model, tokenizer = load("ArapCheruiyot/disarm_ew-llama3", 
-                        adapter_path="ArapCheruiyot/disarm-ew-llama3-finetuned")
+# Run the model
+ollama run disarm-ew-llama3-finetuned "Your prompt here"
+```
 
-# Generate response
-response = generate(model, tokenizer, prompt, max_tokens=256, temp=0.1)
-print(response)
+### Example Usage
+
+```bash
+ollama run disarm-ew-llama3-finetuned "### Instruction:
+Classify the following content according to DISARM Framework techniques and meta-narratives:
+
+### Input:
+A viral WhatsApp broadcast claims that the BVAS machines have been pre-loaded with votes by INEC in favour of the incumbent party.
+
+### Response:"
 ```
 
-### Example Output
+### Expected Output
 
 ```json
 {
   "meta_narrative": "Compromised Election Technology",
-  "primary_disarm_technique": "T0022.002: Develop Original Conspiracy Theory Narratives",
-  "confidence_score": 0.95,
-  "key_indicators": [
-    "BVAS",
-    "INEC", 
-    "pre-loaded",
-    "rigged",
-    "incumbent"
-  ],
+  "primary_disarm_technique": "T0022.001: Develop False Conspiracy Theory Narratives about Electoral Manipulation and Compromise",
+  "confidence_score": 0.98,
+  "key_indicators": ["BVAS", "pre-loaded", "INEC"],
   "platform": "WhatsApp",
   "language": "en",
   "category": "Undermining Electoral Institutions"
@@ -173,25 +137,6 @@ print(response)
 - **Metal GPU**: Accelerated inference
 - **Memory Management**: 16GB wired memory optimization
 
-## Limitations and Biases
-
-### Known Limitations
-1. **Language**: Trained primarily on English content
-2. **Geographic Focus**: Primarily Nigerian election context
-3. **Platform Bias**: Limited to specific social media platforms
-4. **Temporal Context**: Training data from specific election periods
-
-### Potential Biases
-1. **Cultural Context**: May not generalize to other cultural contexts
-2. **Platform-Specific**: May not capture platform-specific nuances
-3. **Evolving Tactics**: May not capture new disinformation techniques
-
-### Ethical Considerations
-1. **Privacy**: Ensure compliance with data protection regulations
-2. **Transparency**: Use responsibly with clear disclosure of AI involvement
-3. **Bias Mitigation**: Regular evaluation for unintended biases
-4. **Human Oversight**: Always maintain human oversight in critical applications
-
 ## Model Files
 
 ### Fused Model (Complete)
@@ -204,43 +149,15 @@ print(response)
 - **Format**: safetensors
 - **Files**: Final adapters + training checkpoints
 
-### Training Checkpoints
-- **Frequency**: Every 100 iterations
-- **Purpose**: Model evaluation and recovery
-- **Format**: safetensors
-
-## Citation
+## Local Deployment Benefits
 
-If you use this model, please cite:
-
-```bibtex
-@misc{disarm_election_watch_2025,
-  title={DISARM Election Watch: Fine-tuned Llama-3.1 for Election Disinformation Detection},
-  author={ArapCheruiyot},
-  year={2024},
-  url={https://huggingface.co/ArapCheruiyot/disarm-ew-llama3-finetuned}
-}
-```
-
-## License
-
-This model is licensed under the MIT License. See the [LICENSE](LICENSE) file for details.
-
-## Acknowledgments
-
-- **DISARM Framework**: For the classification methodology
-- **MLX-LM**: For the fine-tuning framework
-- **Apple**: For Apple Silicon optimization
-- **Hugging Face**: For model hosting and distribution
+- **Privacy**: Run locally without sending data to external servers
+- **Speed**: Fast inference on local hardware
+- **Customization**: Modify prompts and parameters as needed
+- **Offline**: Works without internet connection
 
 ## Contact
 
 For questions, issues, or collaboration opportunities:
 - **Model Repository**: [ArapCheruiyot/disarm-ew-llama3-finetuned](https://huggingface.co/ArapCheruiyot/disarm-ew-llama3-finetuned)
 - **Dataset Repository**: [ArapCheruiyot/disarm-election-watch-dataset](https://huggingface.co/datasets/ArapCheruiyot/disarm-election-watch-dataset)
-
-## Version History
-
-- **v1.0.0**: Initial release with 600 training iterations
-- **Training Data**: 6,019 examples from multiple platforms
-- **Framework**: MLX-LM with Apple Silicon optimization