← Journal · · Talk sources
MLechny Put 2024: talk sources
I'm speaking today at MLechny Put 2024, here's the link to the stream.🌟
As a bonus, let me share the sources we used to prepare the talk)🫶
1. NVIDIA's ML system diagram from their "what is MLOps" article. Interesting read to get into the field. 2. How MLOps.org sees what MLOps is and what an ML system looks like. 3. The book Lean Manufacturing: I'm reading this one right now, interesting to compare it with the DevOps field 4. The book The Goal: a Process of Ongoing Improvement: a great novel that shows how the theory of constraints works through simple examples 5. How GPUs are used for training in ML: tips from HuggingFace and Coursera 6. Using profiling tools (pytorch, tensorflow, nvidia nsight system, nvidia nsight compute, deep learning profiler, ebpf) 7. Preprocessing, preferably on GPU, ideally not directly during training (cudf, nvidia dali) 8. Ready-made open source inference: NVIDIA Triton Server (tutorials) 9. Parallelizing training: huggingFace, pytorch, nvidia 10. Parallelizing inference: my articles on Habr, nvidia mig, nvidia timeslicing, nvidia mps
Share your feedback on the talk in the comments and ask questions! I'll definitely answer)😎