Fix a failing benchmark test

[jonchang]

Description

CI is hitting this test error:

    def image_to_text(self, segments: dict[str, np.ndarray]) -> dict[str, tuple[str, float]]:
        digitized: dict[str, tuple[str, float]] = {}
        for label, image in segments.items():
            if image is None:
                continue

            generated_text = []
            confidence = []

            text_blocks = list(self.split_text_blocks(image))

            # Ignore output from `split_text_blocks` algorithm if only one text block is detected
            if len(text_blocks) == 1:
                text_blocks = [image]

            for block in text_blocks:
                pixel_values = self.processor(images=block, return_tensors="pt").pixel_values
                with torch.no_grad():
                    outputs = self.model.generate(
                        pixel_values, max_new_tokens=100, output_scores=True, return_dict_in_generate=True
                    )

                generated_text.append(self.processor.batch_decode(outputs.sequences, skip_special_tokens=True)[0])
                confidence.append(self.calculate_confidence(outputs))

>           digitized[label] = (" ".join(generated_text), min(confidence))
E           ValueError: min() iterable argument is empty

This pull request fixes the failing test by widening the check against the split text blocks algorithm's result.

Screenshots (if applicable)

Related Issues

Follow up to #248

Checklist

• The title of this PR is descriptive and concise.
• My changes follow the style guidelines of this project.
• I have added or updated test cases to cover my changes.
• I've let the team know about this PR by linking it in the review channel

[arinkulshi-skylight]

LG

[schreiaj]

If it's not a huge hassle, could you build a test case that exercises this distinction? Would help prevent potential regressions down the road when we lose the context behind why it's <= 1 instead of == 1

[jonchang]

I think part of the issue is that split_text_blocks could potentially yield no result. This wasn't something that I expected but the test exposed it since it is passing in extremely small images that can't really be split