UPLOAD_GUIDE.md

# How to Upload to Hugging Face Hub (Web UI)

## IMPORTANT FIRST STEP:
- **DO NOT UPLOAD `PRIVATE_MODEL_TRAINING_NOTES.md` TO HUGGING FACE!** This file stays internal only.

## Step 1: Create a New Model Repository on Hugging Face
1. Go to https://huggingface.co/
2. Log in with your `tarvico` account
3. Click on your profile picture → "New Model"
4. Fill in the details:
   - **Model name**: `maeyen-ai-model`
   - **Owner**: `tarvico`
   - **License**: Choose one (e.g., MIT, Apache-2.0)
   - **Task**: Text Generation
   - **Library**: transformers, peft
   - **Language**: English
   - **Visibility**: Choose "Private" if possible to avoid accidental exposure
5. Click "Create model"

## Step 2: Upload All Files Except PRIVATE_MODEL_TRAINING_NOTES.md
On your new model page (https://huggingface.co/tarvico/maeyen-ai-model):
1. Click "Files and versions" tab
2. Click "Add file" → "Upload files"
3. Upload **ALL** files in this folder **EXCEPT `PRIVATE_MODEL_TRAINING_NOTES.md`**:
   - `README.md` (with YAML metadata at top)
   - `config.json`
   - `generation_config.json`
   - `adapter_config.json`
   - `adapter_model.safetensors` (IMPORTANT! This is the LoRA weights file)
   - `requirements.txt`
   - `train.py`
   - `inference.py`
   - `tokenizer_config.json`
   - `.gitattributes` (CRITICAL! Don't skip this!)
   - `maeyen_examples.jsonl`
   - `generate_dummy_weights.py`
   - `UPLOAD_GUIDE.md`
4. Add a commit message like "Initial Maeyen Trust & Risk Assistant"
5. Click "Commit changes"

## Step 3: Later, Replace Dummy Weights with Real Ones
When you have real data and can train properly:
1. Replace `adapter_model.safetensors` with your trained weights
2. Commit the new file

## Critical Reminders:
- **NEVER UPLOAD `PRIVATE_MODEL_TRAINING_NOTES.md` TO PUBLIC/PRIVATE HUGGING FACE REPO**
- Keep all base model details internal only
- If making repo public, double-check license compliance for base model
- All public content uses "Maeyen Trust & Risk Assistant" only