Fix profiling wrapper to avoid resource handle issues #1504
Labels
repo-refactor
Topics related to the repo and search refactors
rr-kernels
Repo refactor kernels library
rr-runtime
Currently, the only settings that enable end-to-end CUDA execution is the following:
EnableProfiling::YES
andProfilingSettings{/*warmup_iters=*/0, /*measure_iters=*/1}
Any other setting results in CUDA resource handle error (400).
The text was updated successfully, but these errors were encountered: