Picture for Roberto L. Castro

Roberto L. Castro

QuEST: Stable Training of LLMs with 1-Bit Weights and Activations

Add code
Feb 07, 2025
Figure 1 for QuEST: Stable Training of LLMs with 1-Bit Weights and Activations
Figure 2 for QuEST: Stable Training of LLMs with 1-Bit Weights and Activations
Figure 3 for QuEST: Stable Training of LLMs with 1-Bit Weights and Activations
Figure 4 for QuEST: Stable Training of LLMs with 1-Bit Weights and Activations
Viaarxiv icon

HALO: Hadamard-Assisted Lossless Optimization for Efficient Low-Precision LLM Training and Fine-Tuning

Add code
Jan 05, 2025
Figure 1 for HALO: Hadamard-Assisted Lossless Optimization for Efficient Low-Precision LLM Training and Fine-Tuning
Figure 2 for HALO: Hadamard-Assisted Lossless Optimization for Efficient Low-Precision LLM Training and Fine-Tuning
Figure 3 for HALO: Hadamard-Assisted Lossless Optimization for Efficient Low-Precision LLM Training and Fine-Tuning
Figure 4 for HALO: Hadamard-Assisted Lossless Optimization for Efficient Low-Precision LLM Training and Fine-Tuning
Viaarxiv icon

MARLIN: Mixed-Precision Auto-Regressive Parallel Inference on Large Language Models

Add code
Aug 21, 2024
Viaarxiv icon

VENOM: A Vectorized N:M Format for Unleashing the Power of Sparse Tensor Cores

Add code
Oct 03, 2023
Figure 1 for VENOM: A Vectorized N:M Format for Unleashing the Power of Sparse Tensor Cores
Figure 2 for VENOM: A Vectorized N:M Format for Unleashing the Power of Sparse Tensor Cores
Figure 3 for VENOM: A Vectorized N:M Format for Unleashing the Power of Sparse Tensor Cores
Figure 4 for VENOM: A Vectorized N:M Format for Unleashing the Power of Sparse Tensor Cores
Viaarxiv icon

Reusing Trained Layers of Convolutional Neural Networks to Shorten Hyperparameters Tuning Time

Add code
Jun 16, 2020
Figure 1 for Reusing Trained Layers of Convolutional Neural Networks to Shorten Hyperparameters Tuning Time
Figure 2 for Reusing Trained Layers of Convolutional Neural Networks to Shorten Hyperparameters Tuning Time
Figure 3 for Reusing Trained Layers of Convolutional Neural Networks to Shorten Hyperparameters Tuning Time
Figure 4 for Reusing Trained Layers of Convolutional Neural Networks to Shorten Hyperparameters Tuning Time
Viaarxiv icon

A Hybrid Approach for Tracking Individual Players in Broadcast Match Videos

Add code
Mar 10, 2020
Figure 1 for A Hybrid Approach for Tracking Individual Players in Broadcast Match Videos
Figure 2 for A Hybrid Approach for Tracking Individual Players in Broadcast Match Videos
Figure 3 for A Hybrid Approach for Tracking Individual Players in Broadcast Match Videos
Figure 4 for A Hybrid Approach for Tracking Individual Players in Broadcast Match Videos
Viaarxiv icon