Update README with ORCH License and training details

README.md

@@ -1,92 +1,152 @@
 ---
-license: apache-2.0
+license: other
+license_name: orch-license
+license_link: LICENSE
 language:
-- en
+  - en
 library_name: transformers
 tags:
-- code
-- next.js
-- full-stack
-- code-generation
-- fine-tuned
+  - code-generation
+  - nextjs
+  - typescript
+  - full-stack
+  - qlora
+  - deepseek
 base_model: deepseek-ai/deepseek-coder-6.7b-instruct
 pipeline_tag: text-generation
 ---
 
-# ORCH-7B: Autonomous Full-Stack Code Generation
+<div align="center">
+  <img src="https://huggingface.co/spaces/raihan-js/orch-studio/resolve/main/logo.png" alt="ORCH" width="120" height="120" style="border-radius: 24px;">
 
-ORCH-7B is a fine-tuned code generation model specialized in generating complete, production-ready Next.js applications from natural language descriptions.
+  # ORCH-7B
 
-## Model Details
+  **Orchestrated Recursive Code Hierarchy**
 
-- **Base Model**: DeepSeek Coder 6.7B Instruct
-- **Fine-tuning Method**: QLoRA (4-bit quantization + LoRA adapters)
-- **Training Data**: 44,000+ Next.js project examples
-- **Context Length**: 4K tokens (Phase 1)
-- **Output Format**: Complete project files with structured markers
+  *Autonomous Next.js Code Generation Model*
 
-## Capabilities
+  [![Space](https://img.shields.io/badge/Demo-ORCH%20Studio-D4A574?style=for-the-badge)](https://huggingface.co/spaces/raihan-js/orch-studio)
+  [![License](https://img.shields.io/badge/License-ORCH%20v1.0-A67C52?style=for-the-badge)](LICENSE)
+</div>
 
-- Generate complete Next.js 14+ applications
-- Full-stack projects with API routes
-- Database integrations (Prisma, Drizzle)
-- Authentication systems
-- UI components with Tailwind CSS
-- TypeScript support
+---
+
+## Model Description
+
+ORCH-7B is a QLoRA fine-tuned model specialized for generating complete, production-ready Next.js applications from natural language prompts.
+
+## Training Details
+
+| Specification | Value |
+|--------------|-------|
+| **Base Model** | DeepSeek Coder 6.7B Instruct |
+| **Fine-tuning Method** | QLoRA (4-bit quantization + LoRA adapters) |
+| **Training Hardware** | NVIDIA A100 GPU |
+| **Training Duration** | 43 hours |
+| **Training Steps** | 5,238 steps |
+| **Context Length** | 4K tokens |
+| **Model Size** | 6.7B parameters |
+
+## Specialization
+
+- **Framework**: Next.js 14+ (App Router)
+- **Language**: TypeScript
+- **Styling**: Tailwind CSS
+- **Database**: Prisma ORM patterns
+- **Auth**: NextAuth.js patterns
+- **Components**: shadcn/ui compatible
 
 ## Usage
 
+### With Transformers
+
 ```python
 from transformers import AutoModelForCausalLM, AutoTokenizer
+import torch
 
-model = AutoModelForCausalLM.from_pretrained("raihan-js/orch-7b")
-tokenizer = AutoTokenizer.from_pretrained("raihan-js/orch-7b")
+model_id = "orch-ai/ORCH-7B"
 
-prompt = """Generate a complete Next.js full-stack application based on the following requirements.
+tokenizer = AutoTokenizer.from_pretrained(model_id)
+model = AutoModelForCausalLM.from_pretrained(
+    model_id,
+    torch_dtype=torch.float16,
+    device_map="auto"
+)
 
-Create an e-commerce store with product catalog and shopping cart.
+prompt = """### Instruction:
+Create a Next.js login page with email and password fields, validation, and error handling.
 
-Generate all necessary files for a production-ready application."""
+### Response:
+"""
 
-inputs = tokenizer(prompt, return_tensors="pt")
-outputs = model.generate(**inputs, max_new_tokens=4096, temperature=0.7)
-print(tokenizer.decode(outputs[0]))
+inputs = tokenizer(prompt, return_tensors="pt").to(model.device)
+outputs = model.generate(**inputs, max_new_tokens=1024, temperature=0.7)
+print(tokenizer.decode(outputs[0], skip_special_tokens=True))
 ```
 
+### Try it Online
+
+Use [ORCH Studio](https://huggingface.co/spaces/raihan-js/orch-studio) to generate complete projects without any code!
+
 ## Output Format
 
-The model generates files in a structured format:
+The model generates code in markdown format with file paths:
 
+```markdown
+```typescript app/page.tsx
+export default function Home() {
+  return <div>Hello World</div>
+}
 ```
-<|file|>path/to/file.tsx<|end_path|>
-// File content here
-<|end_file|>
 ```
 
-## Training
+## Hardware Requirements
 
-- **Hardware**: RunPod A100
-- **Training Time**: ~21.5 hours
-- **Final Loss**: 0.199
-- **Epochs**: 1
+| Precision | VRAM Required |
+|-----------|---------------|
+| FP16 | ~14 GB |
+| INT8 | ~8 GB |
+| INT4 | ~5 GB |
 
-## Limitations
+## License
 
-- Optimized for Next.js projects specifically
-- Best results with clear, detailed prompts
-- May require post-processing for very large projects
+This model is released under the [ORCH License v1.0](LICENSE).
 
-## License
+**Permitted Uses:**
+- Commercial applications and services
+- Research and academic purposes
+- Personal projects
+- Building products and services
+- Creating derivative models
+
+**Prohibited Uses:**
+- Generating content that violates applicable laws
+- Creating malware or malicious code
+- Harassment, abuse, or harm to individuals
+- Deceptive practices or fraud
 
-Apache 2.0
+## Links
+
+- [ORCH Studio (Demo)](https://huggingface.co/spaces/raihan-js/orch-studio)
+- [ORCH AI Organization](https://huggingface.co/orch-ai)
+- [raihan-js](https://huggingface.co/raihan-js)
 
 ## Citation
 
 ```bibtex
 @misc{orch7b2025,
-  title={ORCH-7B: Autonomous Full-Stack Code Generation},
-  author={Raihan},
+  title={ORCH-7B: Autonomous Next.js Code Generation},
+  author={ORCH Team},
   year={2025},
-  url={https://huggingface.co/raihan-js/orch-7b}
+  publisher={Hugging Face},
+  url={https://huggingface.co/orch-ai/ORCH-7B}
 }
 ```
+
+---
+
+<div align="center">
+  <strong>ORCH</strong> - Orchestrated Recursive Code Hierarchy
+  <br>
+  <em>Building the future of autonomous code generation</em>
+</div>