python sdk can't find custom classification model, can find custom extraction model

Question

I created both an Azure Custom Classification Model and and Azure Custom Extraction model in Form Recognizer Studio, and I am satisfied with the results. Now, I want to generate these results in python.

The python SDK document_analysis_client.begin_analyze_document() function works great for my extraction model, but for some reason when i switch out the model_id for my classification model, I get:

ResourceNotFoundError: (NotFound) Resource not found.
Code: NotFound
Message: Resource not found.
Inner error: {
    "code": "ModelNotFound",
    "message": "The requested model was not found."
}

I have tried both sdk version 3.2.1 and 3.3.0b1, and get the same error with each. Why can't I find and use my classification model?

Answer 1

I ended up using custom python code to use my classification model:

def _post_to_classification_model(pdf_bytes: bytes) -> dict:
    """
    Using configured form recognizer key and model specifications from config,
    post the pdf to the and azure ai classification model for prediction.
    Returns the post response.
    """

    FORM_RECOGNIZER_KEY = os.getenv("FORM_RECOGNIZER_KEY")

    post_url = (
        ENDPOINT
        + f"/formrecognizer/{API_TYPE}/{FACEPAGE_CLASSIFICATION_MODEL_ID}:analyze?api-version={API_VERSION}"
    )
    params = {"includeTextDetails": True}

    headers = {
        # Request headers
        "Content-Type": "application/pdf",
        "Ocp-Apim-Subscription-Key": FORM_RECOGNIZER_KEY,
    }
    logger.debug(f"FORM REC KEY IS: {FORM_RECOGNIZER_KEY}")
    try:
        resp = post(
            url=post_url, data=pdf_bytes, headers=headers, params=params
        )
        if resp.status_code != 202:
            logger.warning(
                "POST analyze failed:\n%s" % json.dumps(resp.json())
            )
            quit()
        logger.info("POST analyze succeeded:\n%s" % resp.headers)
    except Exception as e:
        logger.warning("POST analyze failed:\n%s" % str(e))

    return resp


def _get_classification_results(post_response: dict) -> dict:
    """
    Given our response from our post request for classification,
    retrieve the classificaiton results. Returns the get response.
    """

    get_url = post_response.headers["operation-location"]

    n_tries = 15
    n_try = 0
    wait_sec = 5
    max_wait_sec = 60
    resp_json = None

    while n_try < n_tries:
        try:
            resp = get(
                url=get_url,
                headers={
                    "Ocp-Apim-Subscription-Key": os.getenv(
                        "FORM_RECOGNIZER_KEY"
                    )
                },
            )
            resp_json = resp.json()
            if resp.status_code != 200:
                logger.warning(
                    "GET analyze results failed:\n%s" % json.dumps(resp_json)
                )
                break
            status = resp_json["status"]
            if status == "succeeded":
                logger.info("Analysis succeeded:\n%s" % json.dumps(resp_json))
                break
            if status == "failed":
                logger.warning("Analysis failed:\n%s" % json.dumps(resp_json))
                break
            # Analysis still running. Wait and retry.
            time.sleep(wait_sec)
            n_try += 1
            wait_sec = min(2 * wait_sec, max_wait_sec)

        except Exception as e:
            msg = "GET analyze results failed:\n%s" % str(e)
            logger.warning(msg)
            break

    return resp_json