AI product
Debugging Across the Dataflow Stack is an IBM course on Coursera about debugging AI models across accelerator and compiler software stacks. It teaches learners to establish a validated reference execution, localize differences between executions, distinguish expected numerical variation from correctness bugs, find the first observed divergence, compare tensors with numerical tolerances, and isolate the responsible operation with a minimal test. Fixes are turned into regression tests.
1 use taken from transcripts — each links to the moment in the video.
An IBM course on Coursera that continues the education on debugging AI models across the accelerator and compiler software stack, covering techniques like finding the first observed divergence.
1 in the library.