Update README.md

README.md

@@ -4,3 +4,71 @@ tags:
 - music
 ---
 
+# 🎵 NotaGen: Advancing Musicality in Symbolic Music Generation with Large Language Model Training Paradigms
+
+<p align="center">
+  <!-- ArXiv -->
+  <a href="https://arxiv.org/abs/XXXX.XXXXX">
+    <img src="https://img.shields.io/badge/NotaGen_Paper-ArXiv-%23B31B1B?logo=arxiv&logoColor=white" alt="Paper">
+  </a>
+  &nbsp;&nbsp;
+  <!-- HuggingFace -->
+  <a href="https://huggingface.co/ElectricOnes/NotaGen">
+    <img src="https://img.shields.io/badge/NotaGen_Weights-HuggingFace-%23FFD21F?logo=huggingface&logoColor=white" alt="Weights">
+  </a>
+  &nbsp;&nbsp;
+  <!-- Web Demo -->
+  <a href="https://electricalexis.github.io/notagen-demo/">
+    <img src="https://img.shields.io/badge/NotaGen_Demo-Web-%23007ACC?logo=google-chrome&logoColor=white" alt="Demo">
+  </a>
+</p>
+
+<p align="center">
+  <img src="notagen.png" alt="NotaGen" width="50%">
+</p>
+
+
+## 📖 Overview
+**NotaGen** is a symbolic music generation model that explores the potential of producing **high-quality classical sheet music**. Inspired by the success of Large Language Models (LLMs), NotaGen adopts a three-stage training paradigm:
+- 🧠 **Pre-training** on 1.6M musical pieces
+- 🎯 **Fine-tuning** on ~9K classical compositions with `period-composer-instrumentation` prompts
+- 🚀 **Reinforcement Learning** using our novel **CLaMP-DPO** method (no human annotations or pre-defined rewards required.)
+
+Check our [demo page](https://electricalexis.github.io/notagen-demo/) and enjoy music composed by NotaGen!
+
+## ⚙️ Environment Setup
+
+```bash
+conda create --name notagen python=3.10
+conda activate notagen
+conda install pytorch==2.3.0 pytorch-cuda=11.8 -c pytorch -c nvidia
+pip install accelerate
+pip install optimum
+pip install -r requirements.txt
+```
+
+## 🏋️ NotaGen Model Weights
+
+### Pre-training
+We provide pre-trained weights of different scales:
+|  Models         |  Parameters  |  Patch-level Decoder Layers  |  Character-level Decoder Layers  |  Hidden Size  |  Patch Length (Context Length)  |
+|  ----           |  ----  |  ---- |  ----  |  ----  |  ----  |
+|  [NotaGen-small](https://huggingface.co/ElectricOnes/NotaGen/blob/main/weights_notagen_pretrain_p_size_16_p_length_2048_p_layers_12_c_layers_3_h_size_768_lr_0.0002_batch_8.pth)  | 110M   |  12   |  3     |  768   |  2048  |
+|  [NotaGen-medium](https://huggingface.co/ElectricOnes/NotaGen/blob/main/weights_notagen_pretrain_p_size_16_p_length_2048_p_layers_16_c_layers_3_h_size_1024_lr_0.0001_batch_4.pth) | 244M   |  16   |  3     |  1024  |  2048  |
+|  [NotaGen-large](https://huggingface.co/ElectricOnes/NotaGen/blob/main/weights_notagen_pretrain_p_size_16_p_length_1024_p_layers_20_c_layers_6_h_size_1280_lr_0.0001_batch_4.pth)  | 516M   |  20   |  6     |  1280  |  1024  |
+
+### Fine-tuning
+
+We fine-tuned NotaGen-large on a corpus of approximately 9k classical pieces. You can download the weights [here](https://huggingface.co/ElectricOnes/NotaGen/blob/main/weights_notagen_pretrain-finetune_p_size_16_p_length_1024_p_layers_c_layers_6_20_h_size_1280_lr_1e-05_batch_1.pth).
+
+### Reinforcement-Learning
+
+After pre-training and fine-tuning, we optimized NotaGen-large with 3 iterations of CLaMP-DPO. You can download the weights [here](https://huggingface.co/ElectricOnes/NotaGen/blob/main/weights_notagen_pretrain-finetune-RL3_beta_0.1_lambda_10_p_size_16_p_length_1024_p_layers_20_c_layers_6_h_size_1280_lr_1e-06_batch_1.pth).
+
+### 🌟 NotaGen-X
+
+Inspired by Deepseek-R1, we further optimized the training procedures of NotaGen and released a better version --- [NotaGen-X](https://huggingface.co/ElectricOnes/NotaGen/blob/main/weights_notagenx_p_size_16_p_length_1024_p_layers_20_h_size_1280.pth). Compared to the version in the paper, NotaGen-X incorporates the following improvements:
+
+- We introduced a post-training stage between pre-training and fine-tuning, refining the model with a classical-style subset of the pre-training dataset.
+- We removed the key augmentation in the Fine-tune stage, making the instrument range of the generated compositions more reasonable.
+- After RL, we utilized the resulting checkpoint to gather a new set of post-training data. Starting from the pre-trained checkpoint, we conducted another round of post-training, fine-tuning, and reinforcement learning.