Span offsets not accounted for in results

[evaline-ju]

Describe the bug

The span outputs of results with detection models are incorrect.

Sample Code

curl -v -H "Content-Type: application/json" --request POST \
--data '{"model_id": "model_name",
    "inputs": "There was once a really dumb cat. The chicken crossed the road. The social security number is 123-45-6789. His phone number is (408) 123-4567. ",
    "guardrail_config": {
        "input": {
            "models": {"detector_name": {"threshold": 0.5}
            }
        },
        "output": {"models":{}}
    },
    "text_gen_parameters": {
        "min_new_tokens": 40,
        "max_new_tokens": 200
    }
    }' \
    http://localhost:8081/api/v1/task/classification-with-text-generation

Expected behavior

Span starts and ends should correctly indicate text positions

Observed behavior

Spans all start with 0 in results

{"token_classification_results":{"input":[{"start":0,"end":42,"word":"The social security number is 123-45-6789.","entity":"has_HAP","entity_group":"hap","score":0.00021640431077685207},{"start":0,"end":33,"word":"There was once a really dumb cat.","entity":"has_HAP","entity_group":"thing","score":0.9265003204345704},{"start":0,"end":35,"word":"His phone number is (408) 123-4567.","entity":"has_thing","entity_group":"hap","score":0.0003851974615827203},{"start":0,"end":29,"word":"The chicken crossed the road.","entity":"has_thing","entity_group":"thing","score":0.0013679045950993896}]},"input_token_count":35,"warnings":[{"id":"UNSUITABLE_INPUT","message":"Unsuitable input detected. Please check the detected entities on your input and try again with the unsuitable input removed."}]}