sagkho/multi-xy-regressor

Multi-XY Regressor

A PyTorch neural network for predicting multiple (x, y) coordinate pairs from tabular features.

Architecture

• Input: 20 features
• Output: 5 (x, y) pairs = 10 values (normalized 0-512 range)
• Model: 3-layer MLP with dropout (105K parameters)

Performance

• Overall RMSE: 8.37
• Overall MAE: 6.66
• Overall R²: 0.975

Usage

from inference import load_model, predict
import numpy as np

model, stats, config = load_model()

# Your input: 20 features
features = np.random.randn(1, 20).astype(np.float32)
coords = predict(model, features, stats)

# coords shape: (1, 10) -> [(x0, y0), (x1, y1), ...]
points = [(coords[0, i*2], coords[0, i*2+1]) for i in range(5)]
print(points)

Swapping Your Data

1. Collect your tabular features and corresponding XY coordinates
2. Update dataset_config.json with your feature/target column names
3. Retrain by running the training script with your CSV
4. Replace model.pt and target_stats.npz in this repo

Files

• model.pt — Trained PyTorch model
• target_stats.npz — Target normalization statistics
• dataset_config.json — Data configuration
• inference.py — Loading and prediction utilities

Generated by ML Intern

This model repository was generated by ML Intern, an agent for machine learning research and development on the Hugging Face Hub.

• Try ML Intern: https://smolagents-ml-intern.hf.space
• Source code: https://github.com/huggingface/ml-intern