Alert button
Picture for Minsik Cho

Minsik Cho

Alert button

LLM in a flash: Efficient Large Language Model Inference with Limited Memory

Dec 12, 2023
Keivan Alizadeh, Iman Mirzadeh, Dmitry Belenko, Karen Khatamifard, Minsik Cho, Carlo C Del Mundo, Mohammad Rastegari, Mehrdad Farajtabar

Viaarxiv icon

(Dynamic) Prompting might be all you need to repair Compressed LLMs

Oct 14, 2023
Duc N. M Hoang, Minsik Cho, Thomas Merth, Mohammad Rastegari, Zhangyang Wang

Figure 1 for (Dynamic) Prompting might be all you need to repair Compressed LLMs
Figure 2 for (Dynamic) Prompting might be all you need to repair Compressed LLMs
Figure 3 for (Dynamic) Prompting might be all you need to repair Compressed LLMs
Figure 4 for (Dynamic) Prompting might be all you need to repair Compressed LLMs
Viaarxiv icon

Streaming Anchor Loss: Augmenting Supervision with Temporal Significance

Oct 09, 2023
Utkarsh, Sarawgi, John Berkowitz, Vineet Garg, Arnav Kundu, Minsik Cho, Sai Srujana Buddi, Saurabh Adya, Ahmed Tewfik

Figure 1 for Streaming Anchor Loss: Augmenting Supervision with Temporal Significance
Figure 2 for Streaming Anchor Loss: Augmenting Supervision with Temporal Significance
Figure 3 for Streaming Anchor Loss: Augmenting Supervision with Temporal Significance
Figure 4 for Streaming Anchor Loss: Augmenting Supervision with Temporal Significance
Viaarxiv icon

eDKM: An Efficient and Accurate Train-time Weight Clustering for Large Language Models

Sep 13, 2023
Minsik Cho, Keivan A. Vahid, Qichen Fu, Saurabh Adya, Carlo C Del Mundo, Mohammad Rastegari, Devang Naik, Peter Zatloukal

Figure 1 for eDKM: An Efficient and Accurate Train-time Weight Clustering for Large Language Models
Figure 2 for eDKM: An Efficient and Accurate Train-time Weight Clustering for Large Language Models
Figure 3 for eDKM: An Efficient and Accurate Train-time Weight Clustering for Large Language Models
Figure 4 for eDKM: An Efficient and Accurate Train-time Weight Clustering for Large Language Models
Viaarxiv icon

Flexible Keyword Spotting based on Homogeneous Audio-Text Embedding

Aug 12, 2023
Kumari Nishu, Minsik Cho, Paul Dixon, Devang Naik

Figure 1 for Flexible Keyword Spotting based on Homogeneous Audio-Text Embedding
Figure 2 for Flexible Keyword Spotting based on Homogeneous Audio-Text Embedding
Figure 3 for Flexible Keyword Spotting based on Homogeneous Audio-Text Embedding
Figure 4 for Flexible Keyword Spotting based on Homogeneous Audio-Text Embedding
Viaarxiv icon

Matching Latent Encoding for Audio-Text based Keyword Spotting

Jun 08, 2023
Kumari Nishu, Minsik Cho, Devang Naik

Figure 1 for Matching Latent Encoding for Audio-Text based Keyword Spotting
Figure 2 for Matching Latent Encoding for Audio-Text based Keyword Spotting
Figure 3 for Matching Latent Encoding for Audio-Text based Keyword Spotting
Figure 4 for Matching Latent Encoding for Audio-Text based Keyword Spotting
Viaarxiv icon

PDP: Parameter-free Differentiable Pruning is All You Need

May 18, 2023
Minsik Cho, Saurabh Adya, Devang Naik

Figure 1 for PDP: Parameter-free Differentiable Pruning is All You Need
Figure 2 for PDP: Parameter-free Differentiable Pruning is All You Need
Figure 3 for PDP: Parameter-free Differentiable Pruning is All You Need
Figure 4 for PDP: Parameter-free Differentiable Pruning is All You Need
Viaarxiv icon

R^2: Range Regularization for Model Compression and Quantization

Mar 14, 2023
Arnav Kundu, Chungkuk Yoo, Srijan Mishra, Minsik Cho, Saurabh Adya

Figure 1 for R^2: Range Regularization for Model Compression and Quantization
Figure 2 for R^2: Range Regularization for Model Compression and Quantization
Figure 3 for R^2: Range Regularization for Model Compression and Quantization
Figure 4 for R^2: Range Regularization for Model Compression and Quantization
Viaarxiv icon

HEiMDaL: Highly Efficient Method for Detection and Localization of wake-words

Oct 26, 2022
Arnav Kundu, Mohammad Samragh Razlighi, Minsik Cho, Priyanka Padmanabhan, Devang Naik

Figure 1 for HEiMDaL: Highly Efficient Method for Detection and Localization of wake-words
Figure 2 for HEiMDaL: Highly Efficient Method for Detection and Localization of wake-words
Figure 3 for HEiMDaL: Highly Efficient Method for Detection and Localization of wake-words
Figure 4 for HEiMDaL: Highly Efficient Method for Detection and Localization of wake-words
Viaarxiv icon