torch.compile offers some knobs for controlling the trade-off of execution performance with longer compile times. This is particularly useful for inference, where the same model will be running for a long time.

There are a couple of different options to profile a Triton kernel.

JEPAis an example of a predictive coding architecture. Predictive coding is the idea that the brain works by predicting the next sensory input and learns from the error between the prediction and the actual input. Hierarchies of this kind of prediction allow higher level elements to predict the outputs of lower-level elements, building up deeper and more complex semantics.

As Thomas posted on the Python Discuss board there is a discord for discussing Free-Threaded/NoGil Python: https://discord.gg/rqgHCDqdRr

CUTLASS Tutorial: Writing GEMM Kernels Using Tensor Memory For NVIDIA® Blackwell GPUs – Colfax Research

https://arxiv.org/abs/2504.02263

https://ai.pydantic.dev/evals/

https://blogsystem5.substack.com/p/bazel-next-generation

https://eugeneyan.com/writing/recsys-llm/

Dynamic shapes are one of the more distinctive parts of torch.compile. Rather than specializing a graph to static shapes (which works in many cases!), PyTorch’s approach allows a single graph to work for a variety of sizes, so things like sequence length or batch size can vary. It does this by reasoning about shapes symbolically: instead of using fixed shape values, it uses placeholders and infers rules that constrain those shapes.