Update README.md

README.md

@@ -11,18 +11,7 @@ short_description: Solves VRP with Transformer & RL. Compare with Google OR-Too
 🚚 Vehicle Routing Problem Solver with Transformer-based Reinforcement Learning
 This project implements a deep reinforcement learning framework to solve the Vehicle Routing Problem with Time Windows (VRPTW) using Transformer-based models. It also integrates Google OR-Tools as a classical baseline for comparison.
 
-📌 Project Highlights
-⚙️ Transformer-based Actor-Critic architecture
 
-🧠 Reinforcement Learning (Policy Gradient with Baseline)
-
-🛰️ Google OR-Tools as Benchmark
-
-🧪 Compatible with custom and Shanghai-like datasets
-
-🚀 Supports beam search, nearest-neighbor heuristics, and greedy decoding
-
-📦 Designed to run on Hugging Face Spaces (Docker SDK)
 
 📁 Project Structure
 bash
@@ -37,9 +26,11 @@ Edit
 ├── dataloader.py          # Custom dataset handling (VRP with time windows)
 ├── run.py                 # Training pipeline
 ├── params.json            # Hyperparameters and config
-├── README.md              # This file
-└── Dockerfile             # Hugging Face Docker Space setup
+└── README.md              # This file
+
+
 🧠 Model Description
+
 This model is inspired by the paper
 “Attention, Learn to Solve Routing Problems!”
 (Bello et al., 2018 - arXiv:1803.08475)
@@ -54,63 +45,3 @@ An actor-critic reinforcement learning strategy with a learnable baseline
 
 Beam Search and Greedy decoding options
 
-📦 Requirements
-The environment is automatically built using the included Dockerfile on Hugging Face Spaces.
-However, if you want to run it locally, install:
-
-bash
-Copy
-Edit
-pip install torch ortools numpy matplotlib
-⚙️ Configuration (params.json)
-Update the params.json file to configure:
-
-json
-Copy
-Edit
-{
-  "device": "cpu",
-  "run_tests": true,
-  "save_results": true,
-  "dataset_path": "",
-  "train_dataset_size": 1000,
-  "validation_dataset_size": 100,
-  "num_nodes": 20,
-  "num_depots": 1,
-  "embedding_size": 128,
-  "sample_size": 3,
-  "num_epochs": 50
-}
-🚀 Run the Training Pipeline
-If you're using Hugging Face Spaces, training begins automatically.
-
-To run locally:
-
-bash
-Copy
-Edit
-python run.py
-🧪 Evaluation
-The model is evaluated against:
-
-Google OR-Tools (via google_solver/)
-
-Nearest neighbor baseline
-
-Greedy decoding
-
-Metrics include total travel time and ratios vs. baseline
-
-📊 Example Output
-text
-Copy
-Edit
-Epoch: 0, Batch: 0, Actor/NN: 1.1420, Actor/Baseline: 0.9934
-Test Results: Actor/Google: 1.032, Actor/NN: 0.951, Best NN Ratio: 0.912
-📚 References
-Bello et al. “Attention, Learn to Solve Routing Problems!” arXiv:1803.08475
-
-OR-Tools by Google: https://developers.google.com/optimization/
-
-📄 License
-This project is released under the MIT License.