A library for accelerating Transformer models on NVIDIA GPUs, including using 8-bit floating point (FP8) precision on Hopper and Ada GPUs, to provide better performance with lower memory utilization in both training and inference.
Plain C asserts in TE/JAX custom ops get optimized out of the code at compile time and trigger unused-variable warnings. This PR replaces these asserts with NVTE_CHECK() to avoid the warnings. Also removes unused SM margin variables.
Type of change
[ ] Documentation change (change only to the documentation, either a fix or a new content)
[ ] Bug fix (non-breaking change which fixes an issue)
[ ] New feature (non-breaking change which adds functionality)
[ ] Breaking change (fix or feature that would cause existing functionality to not work as expected)
Description
Plain C asserts in TE/JAX custom ops get optimized out of the code at compile time and trigger unused-variable warnings. This PR replaces these asserts with
NVTE_CHECK()
to avoid the warnings. Also removes unused SM margin variables.Type of change
Changes
NVTE_CHECK
.sm_margin
.Checklist: