))
Ronen Dar on GPU Orchestration for Building ML Models
Manage episode 399286534 series 3320043
Data Driven에서 제공하는 콘텐츠입니다. 에피소드, 그래픽, 팟캐스트 설명을 포함한 모든 팟캐스트 콘텐츠는 Data Driven 또는 해당 팟캐스트 플랫폼 파트너가 직접 업로드하고 제공합니다. 누군가가 귀하의 허락 없이 귀하의 저작물을 사용하고 있다고 생각되는 경우 여기에 설명된 절차를 따르실 수 있습니다 https://ko.player.fm/legal.
In this episode, our Andy Leonard and Frank La Vigne sit down with Ronen Dar, the co-founder and CTO of Run AI, to explore the world of artificial intelligence and GPU orchestration for machine learning models.
Ronen shares insights into the challenges of utilizing GPUs in AI research and how Run AI's platform addresses these issues by optimizing GPU usage and providing tools for easier and faster model training and deployment. The conversation delves into the concept of fractional GPU usage, allowing multiple workloads to run on a single GPU, making expensive GPUs more accessible and cost-effective for organizations.
Links
- Run AI https://www.run.ai/
- Acquired Podcast Episode on Nvidia https://www.acquired.fm/episodes/nvidia-the-machine-learning-company-2006-2022
Show Notes
04:40 GPU technology enabled for cloud AI workloads.
07:00 RunAI enables sharing expensive GPU resources for all.
11:59 As enterprise AI matures, organizations become more savvy.
15:35 Deep learning, GPUs for speed, CPUs backup.
16:54 LLMs running on GPU's, exploding in market.
23:29 NVIDIA created CUDA to simplify GPU use.
26:21 NVIDIA's success lies in accessible technology.
28:25 Solve GPU hugging with quotas and sharing.
31:15 Team lead manages GPU quotas for researchers.
35:51 Rapid changes in business and innovation.
40:34 Passionate problem-solver with diverse tech background.
43:38 Thanks for tuning in, subscribe and review.
300 에피소드
공유
