fix uncaught exception in http client

[dtrifiro]

fixes #245

[evaline-ju]

I see #261 was put up from @Ssukriti that addresses basically the same issue but adds a test - could we potentially combine the work from these two PRs?

[Ssukriti]

You want to just merge this and then I will rebase my PR to add the test and fix the function argument typing as well ?

[dtrifiro]

@evaline-ju @Ssukriti the test in #261 does not actually test this issue, it passes with or without this patch and actually replicates behaviour of existing tests, such as test_run_model (and others):

caikit-nlp/tests/modules/text_generation/test_peft_prompt_tuning.py

Lines 81 to 84 in e512728

	def test_run_model(causal_lm_dummy_model):
	"""Ensure that we can run a model and get the right type out."""
	pred = causal_lm_dummy_model.run("This text doesn't matter")
	assert isinstance(pred, GeneratedTextResult)

I added a test for generate_text_func that generates text and then tries serializing the result. I'm also parametrizing over two different model types, which is probably not really required.

[Ssukriti]

Great. Thank you! PR looks good to me ( lets merge it soon , because it's blocking some other effort :) )

[evaline-ju]

LGTM - thanks for adding the test!