test: fix: L0_server_status - #8827
Conversation
Fixes the L0_server_status test by using the correct expected value for model_stats count.
whoisj
added
PR: fix
| self.assertTrue(False, "unexpected error {}".format(ex)) | ||
|
|
||
|
|
||
| EXPECTED_MODEL_STATS: int = 144 |
There was a problem hiding this comment.
Did you figure out what was changed?
There was a problem hiding this comment.
yeah, the number of stats reported.
there's another condition that I was hitting that happens after running ./test.sh multiple times where the model repository gets corrupted. I spent time chasing that problem when it wasn't the real problem.
There was a problem hiding this comment.
yeah, the number of stats reported.
Was it because I genereated 2 BF16 ONNX models https://github.com/triton-inference-server/server/pull/8782/changes#diff-b0b691449b31f9b36929a5f1dd1edc000ea86264a4078d75318de9cbfb23cbc0?
There was a problem hiding this comment.
No, it's because the test isn't idempotent and requires manual clean up of L0_server_status folder before running the test again.
3 participants
Fixes the L0_server_status test by using the correct expected value for
model_statscount.