Chaim

Every now and then a celebrity passes away and it feels like a death in the family. For me, and many other Jews, the passing of Chaim Topol is one such occasion. Renowned for his portrayal of Tevye, the protagonist of the musical Fiddler on the Roof, Topol came to represent the archetypal Ashkenazi Jewish patriarch, yiddle-diddling his way into the collective consciousness.I can still remember the first time, now 30-odd years ago, that I saw the 1971 film adaptation of Fiddler – a rite of passage. Sitting in a chilly…

Get our free weekly email for all the latest cinematic news from our film critic Clarisse LoughreyGet our The Life Cinematic email for freeChaim Topol – the star of Fiddler On The Roof – has died aged 87. The Israeli actor’s death was announced by the country’s president in a statement on Wednesday (8 March). President Isaac Herzog shared the news on Twitter, in which he paid tribute to “one of the leading giants of Israeli culture”.He said: “ a gifted actor who conquered many stages in Israel and overseas, filled the…

Chaim Topol, the Israeli actor and singer best known for his performance as Tevye the Milkman in Fiddler on the Roof, has died at the age of 87.Topol, who was widely known by his last name alone, died at home in Israel on Wednesday while surrounded by his family, local media reported. His son had recently confirmed that he had been diagnosed with dementia last year.In a statement announcing Topol’s death, Israel’s president Isaac Herzog described him as a “gifted actor who conquered many stages in Israel and overseas,…

Methods for efficient consumption of large filesPhoto by Aron Visuals on UnsplashWorking with very large files can pose challenges to application developers related to efficient resource management and runtime performance. Text file editors, for example, can be divided into those that can handle large files, and those that make your CPU choke, make your PC freeze, and make you want to scream. These challenges are exacerbated when the large files reside in a remote storage location. In such cases one must consider the…

Harnessing the Power of Dedicated DNN Training Chips — Part 3Photo by Balazs Busznyak on UnsplashAfter much anticipation, Amazon EC2 TRN1 instances are now available for public use. Based on the AWS Trainium chip, the TRN1 instance type was specifically designed for accelerating deep learning training. This post is the third part of a series on training deep learning models on dedicated AI chips. In our previous posts we covered Google’s Cloud TPU and Amazon’s EC2 DL1 (powered by Habana Gaudi accelerators). As with other…

Two Methods for Tuning on a Dedicated Ray ClusterPhoto by Davide Cantelli on UnsplashThis is the second part of a three-part post on the topic of hyperparameter tuning (HPT) machine learning models in the cloud. In part 1 we set the stage by introducing the problem and defining a toy model that we will use in our tuning demonstrations. In this part we will review two options for cloud based optimization, both of which involve parallel experimentation on a dedicated tuning cluster.The first option we consider for…

Cloud Tuning by Parallelizing Managed Training JobsPhoto by Kenny Eliason on UnsplashThis the final part of a three-part post on the topic of hyperparameter tuning (HPT) machine learning models in the cloud. In the first part we set the stage by introducing the problem and defining a toy model and a training function for our tuning demonstrations. In the second part we reviewed two options for cloud based optimization, both of which involved parallel experimentation on a dedicated tuning cluster. In this part we will…

The Power of PyTorch/XLA and how Amazon SageMaker Training Compiler Simplifies its usePhoto by Patrick Fore on Unsplashby AuthorIn many of our past posts (e.g., here) we have emphasized the importance of managing the cost of training. We are in constant pursuit of ways to increase the runtime performance of our training through an iterative process of 1. profiling our workloads in order to identify performance bottlenecks and resource under-utilization, and 2. optimizing our workloads to remove bottlenecks and increase…

How to Optimize Model Distribution with SageMaker Distributed Model ParallelPhoto by Martin Jernberg on UnsplashThis is the final part of a three-part post on the topic of optimizing distributed training. In part one we provided a brief survey of distributed training algorithms. We noted that common to all algorithms is their reliance on high-speed communication between multiple GPUs. We surmised that a distributed algorithm that accounted for the underlying instance topology, particularly the differences in the…

How to Optimize Data Distribution with SageMaker Distributed Data ParallelPhoto by Stephen on UnsplashThis is the second part of a three-part post on the topic of optimizing distributed training. In part one, we provided a brief survey of distributed training algorithms. We noted that common to all algorithms is their reliance on high-speed communication between multiple GPUs. We surmised that a distributed algorithm that accounted for the underlying instance topology, particularly the differences in the communication…