After the HugeCTR v23.08, the offline inference will be deprecated. Check out our HPS plugins for TensorRT and TensorFlow as alternatives.

Merlin HugeCTR

HUGECTR

  • Introduction
  • Core Features
  • Embedding Training Cache
  • Hierarchical Parameter Server
  • Sparse Operation Kit
  • Performance
  • Example Notebooks
  • Multi-modal Example Notebooks
    • Training Recommender Systems on Multi-modal Data
    • MovieLens-25M: Download and Convert
    • MovieLens Data Enrichment
    • Movie Poster Feature Extraction with ResNet
    • Movie Synopsis Feature Extraction with Bart text summarization
    • Creating Multi-Modal Movie Feature Store
    • ETL with NVTabular
    • Training HugeCTR Model with Pre-trained Embeddings
  • API Documentation
  • Additional Resources
  • Release Notes
  • Contributing to HugeCTR
Merlin HugeCTR
  • Multi-modal Example Notebooks

Multi-modal Example Notebooks

The multi-modal data example uses several notebooks to demonstrate how to use of multi-modal data (text and images) to provide movie recommendations based on the MovieLens 25M dataset.

  • Training Recommender Systems on Multi-modal Data

  • MovieLens-25M: Download and Convert

  • MovieLens Data Enrichment

  • Movie Poster Feature Extraction with ResNet

  • Movie Synopsis Feature Extraction with Bart text summarization

  • Creating Multi-Modal Movie Feature Store

  • Feature Engineering with NVTabular

  • Training HugeCTR Model with Pre-trained Embeddings

Previous Next

© Copyright 2022, NVIDIA.

Built with Sphinx using a theme provided by Read the Docs.
Other Versions v: v23.04.00
Tags
v23.04.00
v23.05.00
v23.05.01
v23.06.00
v23.06.01
v23.08.00
Branches
main