slorado

Troubleshooting

Error:

tar: slorado-e7c7e54/lib/libamdhip64.so.5: Cannot create symlink to ‘libamdhip64.so’: Operation not supported
tar: Exiting with failure status due to previous errors

Solution:

Getting an error that /tmp/something is unwritable

Error example:

MIOpen(HIP): Error [FlushUnsafe] File is unwritable: /tmp/gfx90a68.HIP.2_20_0_f185a6464-dirty.ufdb.txt

Solution:

mkdir /tmp/a_unique_name
export TMPDIR=/tmp/a_unique_name
./slorado .....

** IMPORTANT: Make sure you give an existent directory, we had a typo as a_unnique_name when exporting, and torch was giving a segfault!

GPU memory out

Currently, slorado does not implement automatic batch size selection based on available memory. If you see an out of GPU memory error, reduce the GPU batch size using -C option.