flash-ansr 0.4.2

Flash Amortized Neural Symbolic Regression

⚡ANSR:
Flash Amortized Neural Symbolic Regression

Papers

• WIP

Usage

pip install flash-ansr

import torch
device = torch.device("cuda" if torch.cuda.is_available() else "cpu")

# Import flash_ansr
from flash_ansr import (
  FlashANSR,
  SoftmaxSamplingConfig,
  install_model,
  get_path,
)

# Specify the model
# Here: https://huggingface.co/psaegert/flash-ansr-v23.0-120M
MODEL = "psaegert/flash-ansr-v23.0-120M"

# Download the latest snapshot of the model
# By default, the model is downloaded to the directory `./models/` in the package root
install_model(MODEL)

# Load the model
model = FlashANSR.load(
  directory=get_path('models', MODEL),
  generation_config=SoftmaxSamplingConfig(choices=32),  # or BeamSearchConfig / MCTSGenerationConfig
  n_restarts=8,
).to(device)

# Define data
X = ...
y = ...

# Fit the model to the data
model.fit(X, y, verbose=True)

# Show the best expression
print(model.get_expression())

# Predict with the best expression
y_pred = model.predict(X)

Overview

Training

⚡ANSR Training on Fully Procedurally Generated Data Inspired by NeSymReS (Biggio et al. 2021)

Architecture

FlashANSR Architecture. The model consists of an upgraded version of the SetTransformer (Lee et al. 2019) as an encoder, and a Pre-Norm Transformer decoder (Vaswani et al. 2017) as a generative model over symbolic expressions.

Results

Coming soon

Citation

@software{flash-ansr2024,
    author = {Paul Saegert},
    title = {Flash Amortized Neural Symbolic Regression},
    year = 2024,
    publisher = {GitHub},
    version = {0.4.2},
    url = {https://github.com/psaegert/flash-ansr}
}