bartbussmann/BatchTopK
Implementation of the BatchTopK activation function for training sparse autoencoders (SAEs)
This helps machine learning researchers and engineers train sparse autoencoders more efficiently. It takes in a batch of feature activations and applies a specialized activation function that identifies the most important features across the entire batch, rather than for each individual sample. This results in a refined set of activated features that can improve the performance and sparsity of the autoencoder model.
No commits in the last 6 months.
Use this if you are working with sparse autoencoders and want an alternative method to select the most active features across a batch, potentially leading to better model training and representation learning.
Not ideal if you are looking for a general-purpose machine learning tool and are not specifically involved in the research or development of sparse autoencoders.
Stars
61
Forks
6
Language
Python
License
MIT
Category
Last pushed
Jul 24, 2025
Commits (30d)
0
Get this data via API
curl "https://pt-edge.onrender.com/api/v1/quality/ml-frameworks/bartbussmann/BatchTopK"
Open to everyone — 100 requests/day, no key needed. Get a free key for 1,000/day.
Higher-rated alternatives
NVIDIA/TransformerEngine
A library for accelerating Transformer models on NVIDIA GPUs, including using 8-bit and 4-bit...
mlcommons/inference
Reference implementations of MLPerf® inference benchmarks
mlcommons/training
Reference implementations of MLPerf® training benchmarks
datamade/usaddress
:us: a python library for parsing unstructured United States address strings into address components
GRAAL-Research/deepparse
Deepparse is a state-of-the-art library for parsing multinational street addresses using deep learning