Open mariecwhite opened 1 year ago
@jpienaar Can you help update the status for this one?
Could you confirm if this still reproduces? This looks very similar to an issue that was shown with the caching allocator that was fixed.
@mariecwhite Assigning to you to see if this is still reproducible.
@mariecwhite Were you able to reproduce this?
It's currently failing due to other reasons. Let's deprioritize this since GPT2 is no longer in our focus set.
What happened?
I successfully imported and compiled GPT2 TF with IREE but when running it through the Python bindings, I get a segmentation fault:
This only seems to happen through the Python bindings because when I run
iree-benchmark-module
separately with the compiled vmfb and input parameters, it works:From initial debugging, we are calling the vm module here with device inputs:
And input values:
Steps to reproduce your issue
After running, reproducers saved in directory name
gpt2_tf_False_cpu*
and vmfb saved in./mhlo_cpu.vmfb
What component(s) does this issue relate to?
Runtime
Version information
Uses iree-compile and iree-runtime versions 20230219.435