Crazy amount of alpha in pytorch conference talks, here are a few i liked or am excited to watch
- Monarch: A Distributed Execution Engine for PyTorch - Colin Taylor & Zachary DeVito, Meta
- Kimi K2 and Our Contributions to Open Source - Yuxin Wu
@ppwwyyxx, Moonshot AI
- Keynote: Olmo-Thinking: Training a Fully Open Reasoning Model
@natolambert Ai2
- Keynote: Marin: An Open Lab for Frontier AI -
@percyliang, Together
- Scaling an Open Environments Ecosystem for Reinforcement Learning -
@willccbb, Prime Intellect
- Open Source Model Performance Optimization With SGLang -
@zhyncs42 Together AI
- Our Journey With TorchTitan - Linsong Chu & Garrett Goon, IBM Research
- mxfp8, mxfp4, nvfp4 formats and applications in PyTorch - Vasily Kuznetsov &
@drisspg , Meta
- Sponsored Session: PyTorch Distributed and Fault Tolerance
@rice_fry , Meta
- The Future Is Tiled: Using CuTile & TileIR To Write Portable, High-performance GPU...- Jared Roesch
- Efficient MoE Pre-training at Scale on AMD GPUs With TorchTitan -Liz Li & Yanyuan Qin, Matthias Reso
- Training and Inference at Planet Scale - Michael Suo, Meta; Robert Nishihara, Anyscale; Simon Mo, vLLM; Lianmin Zheng, XAI; & Dmytro (Dima) Dzhulgakov, Fireworks AI; Moderated by Jana van Greunen, Meta
let's fucking go, pytorch conference 2025 talks are now on youtube