For the complete documentation index, see llms.txt. Markdown versions of all pages are available by appending .md to any URL (e.g. /max/get-started.md).

Mojo module

nan_check

NaN/Inf detection kernels for the max-debug.nan-check feature.

These kernels are registered as custom ops in the kernels package and inserted by the NanCheckPass compiler pass. The architecture is:

1. nan_check_count: read-only reduction that counts NaN/Inf values in a floating-point tensor. Outputs two single-element int32 tensors with the counts. CPU: vectorized scan with atomics. GPU: parallel reduction.

2. nan_check_raise: host-side kernel that reads the NaN/Inf counts (transferred to host via mo.transfer for GPU graphs) and raises a diagnostic error if any are non-zero.

Functions

• ​nan_check_count: Counts NaN/Inf values in a floating-point tensor.
• ​nan_check_raise: Raises an error if NaN or Inf counts are non-zero.

Was this page helpful?

Thank you! We'll create more content like this.

Thank you for helping us improve!

😔 What went wrong?

Some code doesn’t work

It includes inaccurate information

It's missing information I need

It was difficult to understand

Other

Submit