Couple days in #Paris for #PyTorchCon. Mainly shuffling between hotel and conference venue. Caught this on the way back to hotel last night, probably the only reminder of which city I was in.
Couple days in #Paris for #PyTorchCon. Mainly shuffling between hotel and conference venue. Caught this on the way back to hotel last night, probably the only reminder of which city I was in.
Excellent day at PyTorch Conference Europe, where Marcus Edel and Vineet Suryan presented "Bringing BitNet to ExecuTorch via Vulkan"! If you are attending and would like to meet, send us a DM!
Day 1 of PyTorch Conference Europe is just getting started! We're excited to dive into the latest developments in machine learning and connect with the AI community.
Be sure to stick around for the Poster Presentations at 17:05 CEST: Marcus Edel and Vineet Suryan will be sharing "Bringing BitNet to ExecuTorch via Vulkan".
Collabora presents "Bringing BitNet to ExecuTorch via Vulkan" at PyTorch Conference Europe in Paris (April 7-8) and attends ICLR in Rio de Janeiro (April 23-27). Connect with our team to discuss machine learning and open source innovation!
@pytorch #PyTorchCon #PyTorch2026 #PyTorch #PyTorchFoundation #MachineLearning #DeepLearning #ICLR2026
Inspired by a talk I had with @BajoranEngineer at #PyTorchCon, I've jotted down some thoughts about #Python as a scripting engine for apps.
https://phildini.dev/python-in-every-app
Shares appreciated! Commentary welcome, but if you're a jerk I'll block you ๐
@freakboy3742 @glyph @brettcannon this is why I was asking about built python โจ
Also included: a thought on how @conda monetizes this ๐
PyTorch Compiler is a control point for all of these. Compiler can apply LLM-specific and CUDA-specific optimizations to code.
Moving from Training to Inference, we have Peng Wu from Meta.
The trends here match the trends in training! Heterogeneous hardware, dsitributed inference, concern over numerics and determinism -- all on the rise!
Also seeing deep consolidation in LLM serving platforms.
The PyTorch team is also trying to solve specific problems! Like: numerics-sensitive models that make compilers sussy.
The solution here is region-based inductors with in-code annotations.