Picture for Elad Hoffer

Elad Hoffer

DropCompute: simple and more robust distributed synchronous training via compute variance reduction

Add code
Jun 18, 2023
Viaarxiv icon

Energy awareness in low precision neural networks

Add code
Feb 06, 2022
Viaarxiv icon

Logarithmic Unbiased Quantization: Practical 4-bit Training in Deep Learning

Add code
Dec 19, 2021
Figure 1 for Logarithmic Unbiased Quantization: Practical 4-bit Training in Deep Learning
Figure 2 for Logarithmic Unbiased Quantization: Practical 4-bit Training in Deep Learning
Figure 3 for Logarithmic Unbiased Quantization: Practical 4-bit Training in Deep Learning
Figure 4 for Logarithmic Unbiased Quantization: Practical 4-bit Training in Deep Learning
Viaarxiv icon

Task Agnostic Continual Learning Using Online Variational Bayes with Fixed-Point Updates

Add code
Oct 01, 2020
Figure 1 for Task Agnostic Continual Learning Using Online Variational Bayes with Fixed-Point Updates
Figure 2 for Task Agnostic Continual Learning Using Online Variational Bayes with Fixed-Point Updates
Figure 3 for Task Agnostic Continual Learning Using Online Variational Bayes with Fixed-Point Updates
Figure 4 for Task Agnostic Continual Learning Using Online Variational Bayes with Fixed-Point Updates
Viaarxiv icon

Neural gradients are lognormally distributed: understanding sparse and quantized training

Add code
Jun 17, 2020
Figure 1 for Neural gradients are lognormally distributed: understanding sparse and quantized training
Figure 2 for Neural gradients are lognormally distributed: understanding sparse and quantized training
Figure 3 for Neural gradients are lognormally distributed: understanding sparse and quantized training
Figure 4 for Neural gradients are lognormally distributed: understanding sparse and quantized training
Viaarxiv icon

The Knowledge Within: Methods for Data-Free Model Compression

Add code
Dec 03, 2019
Figure 1 for The Knowledge Within: Methods for Data-Free Model Compression
Figure 2 for The Knowledge Within: Methods for Data-Free Model Compression
Figure 3 for The Knowledge Within: Methods for Data-Free Model Compression
Figure 4 for The Knowledge Within: Methods for Data-Free Model Compression
Viaarxiv icon

At Stability's Edge: How to Adjust Hyperparameters to Preserve Minima Selection in Asynchronous Training of Neural Networks?

Add code
Sep 26, 2019
Figure 1 for At Stability's Edge: How to Adjust Hyperparameters to Preserve Minima Selection in Asynchronous Training of Neural Networks?
Figure 2 for At Stability's Edge: How to Adjust Hyperparameters to Preserve Minima Selection in Asynchronous Training of Neural Networks?
Figure 3 for At Stability's Edge: How to Adjust Hyperparameters to Preserve Minima Selection in Asynchronous Training of Neural Networks?
Figure 4 for At Stability's Edge: How to Adjust Hyperparameters to Preserve Minima Selection in Asynchronous Training of Neural Networks?
Viaarxiv icon

Mix & Match: training convnets with mixed image sizes for improved accuracy, speed and scale resiliency

Add code
Aug 12, 2019
Figure 1 for Mix & Match: training convnets with mixed image sizes for improved accuracy, speed and scale resiliency
Figure 2 for Mix & Match: training convnets with mixed image sizes for improved accuracy, speed and scale resiliency
Figure 3 for Mix & Match: training convnets with mixed image sizes for improved accuracy, speed and scale resiliency
Figure 4 for Mix & Match: training convnets with mixed image sizes for improved accuracy, speed and scale resiliency
Viaarxiv icon

Augment your batch: better training with larger batches

Add code
Jan 27, 2019
Figure 1 for Augment your batch: better training with larger batches
Figure 2 for Augment your batch: better training with larger batches
Figure 3 for Augment your batch: better training with larger batches
Figure 4 for Augment your batch: better training with larger batches
Viaarxiv icon

ACIQ: Analytical Clipping for Integer Quantization of neural networks

Add code
Oct 02, 2018
Figure 1 for ACIQ: Analytical Clipping for Integer Quantization of neural networks
Figure 2 for ACIQ: Analytical Clipping for Integer Quantization of neural networks
Figure 3 for ACIQ: Analytical Clipping for Integer Quantization of neural networks
Figure 4 for ACIQ: Analytical Clipping for Integer Quantization of neural networks
Viaarxiv icon