Whether ODH support TorchServe in e2e?

[ayshamk1302]

In the E2E testing for TorchServe, in test/e2e/predictor/test_torchserve.py it is stated that pytest.skip("ODH does not support TorchServe at the moment", allow_module_level=True) is used, indicating that ODH does not support TorchServe at the moment.
This statement causes the entire test module to be skipped, meaning none of the tests in the module will execute.
However, when we tested TorchServe using ODH, the inference was working fine. Here are the test results:

(base) [root@m42lp48 ~]# curl -k https://densenet-kserve-torchserve.apps.ocpz-standard.m42lp48.lnxero1.boe/predictions/densenet161 -T kitten.jpg 

{
  "tabby": 0.4666193723678589,
  "tiger_cat": 0.46449044346809387,
  "Egyptian_cat": 0.06614018976688385,
  "lynx": 0.0012924389448016882,
  "plastic_bag": 0.00022909781546331942
}

Given this, is there any reason why the test case marked as not supported?

[israel-hdez]

@ayshamk1302 The runtimes coming from the upstream project are very likely to work on ODH too, although some may require doing some configurations on OpenShift. We opted to turn off some E2E tests that didn't work when ODH adopted KServe. Also, at the moment, we are not shipping TorchServe configs as part of ODH releases.

We still don't have plans to turn on again E2E tests related to TorchServe, because we are not actively testing this runtime on ODH. Still, we are happy to provide help via upstream KServe community, where we also participate.