FindDefinition

[question]can't disable CP for specific (unsupported) SDPA op

3

## Problem currently the API of context parallel have five problems. 1. only support apply CP to whole model. if we have some cross attn in prep part of model...

[feature request] add memory counter event to pytorch trace result

1

## New Feature We can use memory instant events in pytorch profiler result to generate a nice gpu memory trace in perfetto:  This memory trace is aligned to python...

[Gluon] per-thread operations in gluon

2

gluon, as a tile-based low-level GPU programming language, has a core advantage over other similar languages (such as tilelang and tilus): users can perform thread-level operations through **Linear Layout**, for...

### Describe the bug compiler hint functions such as `tl.multiple_of` don't work when appied to return value of a jit function. this bug also exists in gluon. ```Python import triton...