Upload README.md with huggingface_hub

README.md

@@ -1,38 +1,78 @@
-# MINDI 1.5 Vision-Coder
-
-**Built by [MINDIGENOUS.AI](https://mindigenous.ai)**  
-**Builder:** Faaz ([@Mindigenous](https://huggingface.co/Mindigenous) on HuggingFace)  
-**Started:** April 14, 2026 | **Training Complete:** April 28, 2026 | **Frontend v2:** May 2, 2026
-
+---
+license: apache-2.0
+language:
+- en
+tags:
+- code-generation
+- nextjs
+- react
+- typescript
+- vision
+- multimodal
+- mindi
+- mindigenous
+base_model: Qwen/Qwen2.5-Coder-7B-Instruct
 ---
 
-## What is MINDI 1.5?
+# MINDI 1.5 Vision-Coder
 
-MINDI 1.5 Vision-Coder is a multimodal AI model that generates frontend code (HTML/CSS/JS, React, Next.js, Tailwind) from text prompts and UI screenshots.
+**Built by MINDIGENOUS.AI — Faaz, Mumbai, India**
 
-- **Architecture:** Qwen2.5-Coder-7B-Instruct + LoRA + CLIP ViT-L/14 + Vision-Language Fusion
-- **Training:** 10,000 steps across 3 phases on AMD MI300X 192GB
-- **Live API:** [Mindigenous/mindi-chat](https://huggingface.co/spaces/Mindigenous/mindi-chat) on HuggingFace Spaces
+## Model Description
+MINDI 1.5 is a multimodal agentic AI coding model that generates frontend code
+(HTML/CSS/JS, Next.js, React, Tailwind) from text prompts and UI screenshots.
 
-## Frontend — AI Website Builder
+## Architecture
+| Component | Details |
+|-----------|---------|
+| Base LLM | Qwen/Qwen2.5-Coder-7B-Instruct (7.62B params) |
+| Vision Encoder | CLIP ViT-L/14 (frozen, 304M params) |
+| LoRA Adapters | r=64, alpha=128 (161.5M trainable params) |
+| Fusion | VisionLanguageFusion with text_gate (16.8M params) |
+| Total | 8.1B params, 182.5M trainable (2.25%) |
 
-A professional 3-panel IDE (Bolt.new-style) for interacting with MINDI:
+## Training
+- **3-Phase Progressive Training** on AMD MI300X 192GB + Modal A100 40GB
+- **Dataset**: 1.45M examples, 860M tokens
+- **Final loss**: 0.25–0.40 range
 
-```powershell
-cd frontend
-npm install
-npm run dev   # → http://localhost:5173
+## Checkpoint Structure
+```
+checkpoints/
+├── phase3_final/           ← Best checkpoint for inference
+│   ├── lora/               ← LoRA adapter weights
+│   ├── vision/             ← Vision projection weights
+│   └── fusion/             ← Fusion layer weights
+├── phase3_all_step2500_final/
+├── phase3_all_step2000/
+├── phase3_all_step1500/
+└── ... (earlier phases)
 ```
 
-Features: Plan modal, prompt enhancement, code animation, live preview, file tree, demo fallback.  
-**Read `context.md` for full architecture and next steps.**
+## Usage
+```python
+from src.model.mindi_model import MINDI15
+import torch
 
-## HuggingFace
+model = MINDI15(
+    model_name="Qwen/Qwen2.5-Coder-7B-Instruct",
+    clip_model="openai/clip-vit-large-patch14",
+    hidden_size=3584,
+    num_visual_tokens=256,
+    torch_dtype=torch.bfloat16,
+)
+model.load("checkpoints/phase3_final")
+model.eval()
 
-- **Model:** `Mindigenous/MINDI-1.5-Vision-Coder` (private)
-- **Dataset:** `Mindigenous/MINDI-1.5-training-data` (private)
-- **Space:** `Mindigenous/mindi-chat` (live, ZeroGPU)
+response = model.generate(
+    prompt="Build a Next.js landing page",
+    max_new_tokens=2048,
+    temperature=0.7,
+)
+```
 
-## License
+## Special Tokens (22 total, 11 pairs)
+think, code, file, critique, suggest, search, error, fix, vision, sandbox, context
 
-Apache 2.0
+## Built By
+Faaz — MINDIGENOUS.AI | Mumbai, India | April–May 2026