AI/ML Engineer ll AWS AI&ML Scholar (Among top 500 worldwide) || Xylem Grand Prize Winner 2023 (University Track)
PyTorch Internals! A comprehensive guide to how you can understand the PyTorch codebase & contribute to it. Here's the link: https://lnkd.in/dnDsJTQG
PyTorch 2.0 succeeds its predecessor, PyTorch 1.x, with improved performance and efficiency. Here’s a rundown of these new features and capabilities: Backward Compatibility: - Code, model and API compatibility with PyTorch 1.x - Maintained compatibility with libraries and dependencies Fusing Operations: - Fuses multiple operations into a single GPU kernel - Reduces kernel launching overhead Reducing Type Conversion Overhead: - Identifies and minimizes unnecessary type conversion code - Avoids redundancies Reusing Buffers: - Reduced memory allocation - Lower memory consumption during model training Read on to learn about all the improvements PyTorch 2.0 brings and how they work: https://bit.ly/47QsWR2
Spent so much time and energy in an attempt to implement a fairly simple contrastive setup because kept failing to follow every single pattern per documentation of these high-level off-the-shelf DL packages. Eventually made it in a couple of hours via plain Pytorch 🤧 Never have I truly appreciated the era of low-level DL (RIP #mxnet 🐐) w/o infinite future warnings, unresolvable dependency conflicts, "batteries" blowing up after the slightest tweak, undocumented objects accepting 50 parameters... Maybe it wasn't that bad to casually face papers w/o code back then when you had to quickly set up rather simple, opinionated ad hoc implementations from scratch instead of spending hours depth-first reading the docs and source code to finally run a slightly modified 5-liner. #showerthoughts
This chart compares the number of dependencies of some LLM frameworks. While PyTorch has < 20, LlamaIndex has > 120 dependencies. Here is the issue. Overusing dependencies might save you some time now, but it can cause a dependency conflict in the future. The higher the number of dependencies, the higher the chance of conflict between them, and the more expensive will be to maintain the library. LlamaIndex is a very young project that has grown very fast with the success of LLMs. Sooner or later they will realize how painful it can be to have too many dependencies. I made the same mistake. ____ 🔔 Press the notification button to get content like this. 📢 Repost if this content is useful. ✏ Comment with your views.
Google Developer Expert (GDE), AWS Community Builder, Senior Manager Data Science, Consultant, Trainer, Podcaster, Founder Malaysia R User Group, AI & ML Malaysia User Group
Check out what's new in PyTorch Documentation. Featuring 3 new tutorials and PyTorch 2.2 documentation + tutorials! #pytorch Check out here : https://lnkd.in/dehHw3Y2
GitHub - johnma2006/mamba-minimal: Simple, minimal implementation of the Mamba SSM in one file of PyTorch.
AI Engineer，Full-time open source engineer, Apache Linkis Committer, initiator of the SolidUI AI painting project.
PyTorch Releases 2024 Roadmap: 10 Key Updates You Need to Know The PyTorch team releases its first-ever technical roadmap, unveiling nearly 100 pages of documents outlining the development directions for the second half of 2024. https://lnkd.in/gaZMrEjW #PyTorch
🚀 Day 217 of my 365 Days of DSA Challenge! 🚀 Today, I tackled the Black Shapes problem. This journey of solving a new data structure and algorithm problem every day has been incredibly rewarding and challenging. The Black Shapes problem was a great exercise in understanding and implementing depth-first search (DFS) to count connected components in a grid. Key takeaways: Reinforced my knowledge of DFS and graph traversal techniques. Enhanced my problem-solving skills by breaking down complex problems into manageable parts. Grateful for the continuous learning and growth. On to the next challenge! 💪 Discord Server: https://lnkd.in/gstPhXQX Scaler #DataStructure #Algorithms #CodingChallenge #ProgrammingJourney #CodeNewbieLeetCode LeetCode #AchievementUnlocked #scalerdiscord #codewithscaler #365daysofcodescaler #365daysofcode #365DaysOfDSA #codingchallenge #problemsolving #datastructures #algorithms #learningeveryday #techjourney
PyTorch 2.3 is here 😎🔥 PyTorch 2.3 offers support for user-defined Triton kernels in torch.compile, allowing for users to migrate their own Triton kernels from eager without experiencing performance regressions or graph breaks. Details: https://hubs.la/Q02tY8PB0
