 
            
              Mosharaf Chowdhury
            
            @mosharaf
Followers
                501
              Following
                93
              Media
                29
              Statuses
                1K
              SymbioticLab (https://t.co/kBcOHILGw6)
              
              Ann Arbor, MI
            
            
              
              Joined April 2008
            
            
           Which modern LLM is the most energy efficient? ⚡️ Today, we're excited to release the first version of the ML ENERGY leaderboard, providing benchmarks of modern LLMs in terms of energy, model quality, and more! 🎉  https://t.co/WVbSAqc04a 
          
          
            
            ml.energy
              An Energy Leaderboard for LLMs
            
                
                1
              
              
                
                5
              
              
                
                17
              
             The initial version was developed together with @JIACHENLIU8, Zhiyu Wu, and @mosharaf, part of @SymbioticLabUM / @UMichCSE. Also, we would like to thank @huggingface, @lmsysorg, @AiEleuther, and @Gradio for their awesome open source efforts. 🥰 
          
                
                0
              
              
                
                1
              
              
                
                3
              
             Fantastic work by @jaewon_chung_cs et al. to understand and reduce the energy consumption of AI. 
           Care about Deep Learning energy consumption? Zeus is a practical energy optimization framework for DNN training. Already pulled 100k+ times on Docker Hub!  https://t.co/VC07eA9I41 
            
          
                
                0
              
              
                
                1
              
              
                
                9
              
            
            @mosharaf My pleasure - it seems such an obviously good idea! Very nice piece of work, congrats.
          
          
                
                0
              
              
                
                1
              
              
                
                0
              
             Thanks! We have many plans to extend it in coming months in terms of supporting different OS kernels, HA w/o performance drops, w/o RDMA etc 
           Nice to see this work on GH  https://t.co/g0pwvBoabo  HT @mosharaf
            
          
                
                0
              
              
                
                0
              
              
                
                0
              
             Infiniswap is on github now! Let us know what you think!!! 
          
            
            github.com
              Infiniswap enables unmodified applications to efficiently use disaggregated memory. - SymbioticLab/Infiniswap
            
                
                0
              
              
                
                1
              
              
                
                1
              
             Why page to disk (slow!) when you could page to remote memory in the cluster using RDMA?  https://t.co/2sfE7ERxuZ 
          
          
                
                2
              
              
                
                23
              
              
                
                48
              
             "Efficient memory disaggregation with Infiniswap" Gu et al., NSDI'17  https://t.co/2sfE7ERxuZ 
            #themorningpaper
          
          
                
                0
              
              
                
                6
              
              
                
                7
              
             Congratulations to @vyas_sekar (@CSDatCMU @cylab) for winning the SIGCOMM Rising Star Award 2016  https://t.co/5v5F402gIJ 
          
          
                
                2
              
              
                
                12
              
              
                
                15
              
             The winner of the 2015 Doctoral Dissertation Award is Mosharaf Chowdhury. 
          
                
                0
              
              
                
                6
              
              
                
                11
              
             Learn about the Berkeley Data Analytics Stack in AMP Camp 6 this week! Attend in-person or watch the live stream: 
          
                
                0
              
              
                
                16
              
              
                
                9
              
             Join AMPCamp 6 on Nov 19-20 to learn Spark and the Berkeley Data Analytics Stack, at UC Berkeley or via live stream 
          
                
                0
              
              
                
                9
              
              
                
                11
              
             Applying to grad school in networks? @mosharaf is looking for students! 
          
                
                0
              
              
                
                1
              
              
                
                0
              
             Mesos making news and vying for "Unicorn" status -  http://t.co/bXYPKRE9c0 
          
          
                
                0
              
              
                
                4
              
              
                
                7
              
             Teaching or doing research with Spark? @databricks just launched a free academic program: 
          
            
            databricks.com
              Free trial of Databricks for data engineering and ML/AI workloads. See why the most innovative businesses run Databricks. Try Databricks for 14 days.
            
                
                0
              
              
                
                40
              
              
                
                38
              
             We are hiring! Join our team from UC Berkeley, Google, CMU, VMWare, Stanford etc. to create the future of storage!  http://t.co/SxR9NhbAtD 
          
          
            
            alluxio.io
              Alluxio is a high-performance distributed caching layer built for large-scale AI workloads. Alluxio maximizes I/O throughput and minimizes I/O latency for model training, model deployment, inference...
            
                
                2
              
              
                
                25
              
              
                
                12
              
             
             
             
               
             
             
             
             
            