pmixer/SASRec.pytorch

PyTorch(1.6+) implementation of https://github.com/kang205/SASRec

What it solves

This project provides a PyTorch implementation of the SASRec (Self-Attentive Sequential Recommendation) model, designed to predict the next item a user will interact with based on their sequence of previous interactions.

How it works

It implements the self-attention mechanism to model sequential user behavior. The project modifies the original TensorFlow implementation to use PyTorch (v1.6), fixing issues such as positional embedding usage to reduce overfitting.

Who it’s for

Researchers and developers working on recommendation systems who prefer PyTorch over TensorFlow.

Highlights

  • PyTorch-based implementation of the self-attentive sequential recommendation paper.
  • Includes training and inference scripts for datasets like ml-1m.
  • Updated with LayerNorm for improved performance metrics (NDCG and HR).

Related

  • Project
    marmotdata/marmot

    An open-source data catalog that provides a context layer for AI agents and teams to discover and understand data assets like tables, APIs, and queues.

  • Project
    huggingface/Repo2RLEnv

    A tool that transforms GitHub repositories into verifiable RL environments for training and evaluating coding agents using automated synthesis pipelines.

  • Project
    groq/groq-python

    A Python client library for the Groq REST API that provides typed, synchronous and asynchronous access to low-latency LLM services.

  • Project
    hackingmaterials/matminer

    A data mining library for materials science that provides tools for retrieving datasets and featurizing materials data for analysis.