keroro824/HashingDeepLearning

Health v2 · maintenance only

Flags: no_releases

Adoption not part of the score

1103 stars · 168 forks observed · 2026-08-28

What it is AI-extracted, prompt v1, taxonomy v1, 2026-08-30, confidence not recorded

SLIDE is a C++ research codebase implementing locality-sensitive hashing based training of deep neural networks, from the paper 'In Defense of Smart Algorithms over Hardware Acceleration for Large-Scale Deep Learning Systems'. It demonstrates that smart algorithmic sampling can outperform GPU baselines on CPUs for large-scale MLP training, with TensorFlow baselines and the Amazon-670K dataset included.

Use cases

• reproduce SLIDE paper experiments on large-scale deep learning
• train MLPs on extreme multi-label classification datasets like Amazon-670K
• compare LSH-based sampling against full and sampled softmax in TensorFlow
• run deep learning training faster on CPUs instead of GPUs
• study hashing-based sparse network training algorithms

When to choose

• you want to reproduce or extend the SLIDE research results
• you need CPU-only training of large sparse MLPs with LSH sampling
• you are researching algorithmic alternatives to hardware acceleration

When to avoid

• you need a maintained production deep learning framework - use the newer RUSH-LAB/SLIDE or PyTorch
• you train CNNs or transformers rather than MLP-style networks
• your hardware lacks Skylake+ huge pages support and you cannot modify build flags

Facets

library · maturity maintenance

deep-learning machine-learning llm-training gpu-computing deep-learning machine-learning large-language-models cpp python lsh locality-sensitive-hashing sparse-mlp cpu-optimization research-code extreme-classification algorithms linux docker

Member repositories

For agents

markdown · JSON · MCP: product_card(name="keroro824/HashingDeepLearning")