HyzeRE1

@@ -10,170 +10,102 @@ metrics:
 - accuracy
 ---
 <p align="center">
-  <img src="https://i.imgur.com/ePJMLNp.png" alt="Hyze Logo" width="405"/>
 </p>
 <p align="center">
-  <strong>20 Billion Parameters • Research-Grade • Open Weights</strong>
 </p>
 <p align="center">
-  <a href="https://hyzebot.vercel.app">🌐 Try Hyze RE1 Pro</a> •
-  <a href="https://huggingface.co/HyzeAI">🤗 Hugging Face</a> •
-  <a href="https://github.com/HyzeAI">📁 GitHub</a>
 </p>
 ---
 ## 🚀 Overview
-**Hyze RE1 Pro** is a **20 billion parameter** transformer model designed exclusively for **research purposes**. Built on the philosophy that **frontier AI should not belong only to those with billion-dollar budgets**, RE1 Pro delivers strong reasoning capabilities in a fully open-weight package.
-| Attribute | Details |
-|----------|---------|
-| **Parameters** | 20B |
-| **Architecture** | Transformer (Decoder-only) |
-| **Precision** | BF16 / INT4 (quantized) |
-| **Context Length** | 32K tokens |
-| **License** | Apache 2.0 |
-| **Target** | Academic / Non-Commercial Research |
----
-## 🧠 Capabilities
-Hyze RE1 Pro excels at:
-- 🔬 **Scientific reasoning** – Physics, mathematics, code
-- 🌌 **Space & astronomy** – Continued pretraining on domain-specific corpora
-- 📚 **Research summarization** – ArXiv, technical papers
-- 🧮 **Complex instruction following** – Multi-step reasoning tasks
-> ⚠️ **Research Use Only**
-> RE1 Pro is not optimized for general consumer chatbots. It is a **research instrument**, not a product. For general chat, see [HyzeMini](https://huggingface.co/HyzeAI/HyzeMini).
----
-## 📊 Benchmarks (Preliminary)
-| Benchmark | Score (20B) | Comparison |
-|-----------|-------------|------------|
-| MMLU (5-shot) | **68.2** | LLaMA2-13B: 54.8 |
-| HumanEval (pass@1) | **37.4** | CodeLlama-13B: 36.0 |
-| GSM8K (8-shot) | **62.1** | Mistral-7B: 52.2 |
-| MATH (4-shot) | **26.8** | LLaMA2-34B: 27.0 |
-*Benchmarks conducted in BF16. Quantized versions may show slight degradation.*
 ---
-## ⚙️ Installation & Usage
-### Python (Transformers)
-```python
-from transformers import AutoModelForCausalLM, AutoTokenizer
-model = AutoModelForCausalLM.from_pretrained(
-    "HyzeAI/Hyze-RE1-Pro",
-    torch_dtype="auto",
-    device_map="auto"
-)
-tokenizer = AutoTokenizer.from_pretrained("HyzeAI/Hyze-RE1-Pro")
-prompt = "Explain the rocket equation in simple terms."
-inputs = tokenizer(prompt, return_tensors="pt").to(model.device)
-outputs = model.generate(
-    **inputs,
-    max_new_tokens=256,
-    temperature=0.7,
-    top_p=0.9
-)
-print(tokenizer.decode(outputs[0], skip_special_tokens=True))
-```
-### llama.cpp (CPU + Quantized)
-```bash
-# Download GGUF from Hugging Face
-wget https://huggingface.co/HyzeAI/Hyze-RE1-Pro-GGUF/resolve/main/hyze-re1-pro-q4_k_m.gguf
-./llama-cli -m hyze-re1-pro-q4_k_m.gguf \
-  -p "List three challenges of Mars colonization:" \
-  -n 512 \
-  -t 8
-```
 ---
-## 💻 Hardware Requirements
-| Mode | VRAM | RAM | Recommended Hardware |
-|------|------|-----|---------------------|
-| FP16 (full) | **40GB+** | 64GB | 1x A100 / 2x RTX 3090 |
-| INT4 (Q4) | **12GB** | 16GB | RTX 4070 Ti / Mac M2+ |
-| CPU (GGUF) | — | 32GB | AMD EPYC / Intel Xeon |
-> 💡 **Quantized versions** (4-bit) make RE1 Pro runnable on consumer hardware with minimal quality loss.
 ---
-## 🧪 Research Access
-Hyze RE1 Pro is **free and open weights** under Apache 2.0.
-You do not need to apply for access. No approval required. No gated repository.
-**We believe research should not wait for permission.**
 ---
-## 🧭 About Hyze AI
-<p align="left">
-  <img src="https://i.imgur.com/ePJMLNp.png" alt="Hyze Logo" width="30"/>
-</p>
-**Hyze AI** is a one-person research lab founded by **Hitesh**, a 13-year-old builder.
-Hyze exists to prove that **age and budget are not prerequisites for advancing AI**.
-- 🚀 **Mission**: Democratize large-scale AI research
-- 🔓 **License Philosophy**: Apache 2.0 — no strings attached
-- 🌍 **Focus**: Space, science, and accessible reasoning
-> *"DeepSeek proved you don't need billions. We're proving you don't need to be 30."*
 ---
-## 📎 Citation
-```bibtex
-@misc{hyze-re1-pro-2025,
-  author = {Hitesh Vinothkumar},
-  title = {Hyze RE1 Pro: A 20B Parameter Research Model},
-  year = {2025},
-  publisher = {Hugging Face},
-  url = {https://huggingface.co/HyzeAI/Hyze-RE1-Pro}
-}
-```
 ---
-## 🤝 Support & Contact
-- 💬 **Try the live demo**: [https://hyzeai.vercel.app](https://hyzeai.vercel.app)
-- 📧 **Email**: hiteshv2603@gmail.com
-- 🐦 **Twitter/X**: [@HyzeAI](https://twitter.com/HyzeAI)
-- 💼 **GitHub**: [HyzeAI](https://github.com/HyzeAI)
-**For research collaborations, compute sponsorship, or academic partnerships — reach out.**
----
-<p align="center">
-  <sub>Built with Xtuner Model and zero GPUs (so far).</sub>
-  <br/>
-  <sub>© 2025 Hyze AI. Apache 2.0.</sub>
-</p>

 - accuracy
 ---
 <p align="center">
+  <img src="https://i.imgur.com/ePJMLNp.png" alt="Hyze Logo" width="200"/>
+  &nbsp;&nbsp;&nbsp;
+  <img src="https://i.imgur.com/BVjGIik.png" alt="Founder" width="200"/>
 </p>
+<h1 align="center">Hyze RE1 Pro</h1>
 <p align="center">
+  20B Open-Weight Research Model by <b>Hyze AI</b>
 </p>
 <p align="center">
+  🌐 <a href="https://hyzeai.vercel.app">hyzeai.vercel.app</a> •
+  📘 <a href="https://hyzedocs.vercel.app">hyzedocs.vercel.app</a> •
+  🧠 <a href="https://hyzecode.vercel.app">hyzecode.vercel.app</a>
 </p>
 ---
 ## 🚀 Overview
+**Hyze RE1 Pro** is a **20-billion parameter transformer model** designed exclusively for **research and advanced reasoning tasks**.
+Built on the philosophy that:
+> Frontier AI should not belong only to billion-dollar budgets.
+RE1 Pro delivers strong reasoning performance in a **fully open-weight package**, empowering researchers, developers, and independent innovators.
 ---
+## 🧠 Core Focus
+Hyze RE1 Pro is optimized for:
+- 🔬 Advanced reasoning
+- 📊 Research-oriented analysis
+- 🧩 Multi-step problem solving
+- 📚 Long-form structured explanations
+- 🧠 Logical and technical tasks
+This model prioritizes **clarity, depth, and reasoning structure** over casual chat behavior.
 ---
+## 📊 Model Specifications
+- **Architecture:** Transformer
+- **Parameters:** 20 Billion
+- **Type:** Open-weight research model
+- **Primary Domain:** Reasoning & Research
+- **Language:** English
 ---
+## 🧪 Intended Use
+Hyze RE1 Pro is designed for:
+- Academic research experiments
+- Independent AI research labs
+- Reasoning benchmark testing
+- Long-form technical analysis
+- Open-weight innovation
 ---
+## ⚡ Why RE1 Pro?
+While many frontier models remain closed and restricted, RE1 Pro embraces:
+- ✅ Accessibility
+- ✅ Transparency
+- ✅ Open experimentation
+- ✅ Independent research freedom
+It aims to reduce the barrier between individual researchers and high-performance AI systems.
 ---
+## ⚠️ Limitations
+- Large compute requirements (20B parameters)
+- Not optimized for casual short-form chat
+- Outputs should be validated in academic or production contexts
 ---
+## 🧪 Example Usage
+```python
+from transformers import pipeline
+generator = pipeline(
+    "text-generation",
+    model="HyzeAI/Hyze-RE1-Pro"
+)
+print(generator("Explain the mathematical intuition behind backpropagation."))