[formrecognizer] composed model (#14029)

* initial model compose * fix tests for composed model and repr * mypy * fix comments * add fixmes * set form_type on CustomFormSubmodel correctly * fix training tests * remove error map/cont token from compose model * add composed model tests * adding more tests for new fields * updating readmes/changelog * small fix * update sample to be more consistent with .NET * add api version table to readme * add tests requested in feedback * raise better error for multiapi * some docs feedback * testing feedback
Azure · Oct 8, 2020 · 2fed772 · 2fed772
1 parent de390a6
commit 2fed772
Show file tree

Hide file tree

Showing 48 changed files with 4,297 additions and 1,399 deletions.
diff --git a/sdk/formrecognizer/azure-ai-formrecognizer/CHANGELOG.md b/sdk/formrecognizer/azure-ai-formrecognizer/CHANGELOG.md
@@ -2,10 +2,19 @@
 
 ## 3.1.0b1 (Unreleased)
 
+This version of the SDK defaults to the latest supported API version, which currently is v2.1-preview.
+
 **New features**
 
 - Recognize receipt methods now take keyword argument `locale` to optionally indicate the locale of the receipt for
 improved results
+- Added ability to create a composed model from the `FormTrainingClient` by calling method `begin_create_composed_model()`
+- Added the properties `display_name` and `properties` to types `CustomFormModel` and `CustomFormModelInfo`
+- Added keyword argument `display_name` to `begin_training()` and `begin_create_composed_model()`
+- Added model type `CustomFormModelProperties` that includes information like if a model is a composed model
+- Added property `model_id` to `CustomFormSubmodel` and `TrainingDocumentInfo`
+- Added properties `model_id` and `form_type_confidence` to `RecognizedForm`
+
 
 ## 3.0.0 (2020-08-20)
 

diff --git a/sdk/formrecognizer/azure-ai-formrecognizer/README.md b/sdk/formrecognizer/azure-ai-formrecognizer/README.md
@@ -20,10 +20,18 @@ from form documents. It includes the following main functionalities:
 Install the Azure Form Recognizer client library for Python with [pip][pip]:
 
 ```bash
-pip install azure-ai-formrecognizer
+pip install azure-ai-formrecognizer --pre
 ```
 
-> Note: This version of the client library supports the v2.0 version of the Form Recognizer service
+> Note: This version of the client library defaults to the v2.1-preview version of the service
+
+This table shows the relationship between SDK versions and supported API versions of the service
+
+|SDK version|Supported API version of service
+|-|-
+|3.0.0 - Latest GA release (can be installed by removing the `--pre` flag)| 2.0
+|3.1.0b1 - Latest release (beta)| 2.0, 2.1-preview
+
 
 #### Create a Form Recognizer resource
 Form Recognizer supports both [multi-service and single-service access][multi_and_single_service].
@@ -135,6 +143,7 @@ Sample code snippets are provided to illustrate using a FormRecognizerClient [he
 - Training custom models with labels to recognize specific fields and values you specify by labeling your custom forms. A `CustomFormModel` is returned indicating the fields the model will extract, as well as the estimated accuracy for each field. See the [service documentation][fr-train-with-labels] for a more detailed explanation.
 - Managing models created in your account.
 - Copying a custom model from one Form Recognizer resource to another.
+- Creating a composed model from a collection of existing trained models with labels.
 
 Please note that models can also be trained using a graphical user interface such as the [Form Recognizer Labeling Tool][fr-labeling-tool].
 
@@ -183,6 +192,8 @@ result = poller.result()
 
 for recognized_form in result:
     print("Form type: {}".format(recognized_form.form_type))
+    print("Form type confidence: {}".format(recognized_form.form_type_confidence))
+    print("Form was analyzed using model with ID: {}".format(recognized_form.model_id))
     for name, field in recognized_form.fields.items():
         print("Field '{}' has label '{}' with value '{}' and a confidence score of {}".format(
             name,
@@ -275,21 +286,23 @@ form_training_client = FormTrainingClient(endpoint, credential)
 
 container_sas_url = "<container-sas-url>"  # training documents uploaded to blob storage
 poller = form_training_client.begin_training(
-    container_sas_url, use_training_labels=False
+    container_sas_url, use_training_labels=False, display_name="my first model"
 )
 model = poller.result()
 
 # Custom model information
 print("Model ID: {}".format(model.model_id))
+print("Display name: {}".format(model.display_name))
+print("Is composed model?: {}".format(model.properties.is_composed_model))
 print("Status: {}".format(model.status))
 print("Training started on: {}".format(model.training_started_on))
 print("Training completed on: {}".format(model.training_completed_on))
 
 print("\nRecognized fields:")
 for submodel in model.submodels:
     print(
-        "The submodel with form type '{}' has recognized the following fields: {}".format(
-            submodel.form_type,
+        "The submodel with form type '{}' and model ID '{}' has recognized the following fields: {}".format(
+            submodel.form_type, submodel.model_id,
             ", ".join(
                 [
                     field.label if field.label else name
@@ -336,6 +349,8 @@ model_id = "<model_id from the Train a Model sample>"
 
 custom_model = form_training_client.get_custom_model(model_id=model_id)
 print("Model ID: {}".format(custom_model.model_id))
+print("Display name: {}".format(model.display_name))
+print("Is composed model?: {}".format(model.properties.is_composed_model))
 print("Status: {}".format(custom_model.status))
 print("Training started on: {}".format(custom_model.training_started_on))
 print("Training completed on: {}".format(custom_model.training_completed_on))
@@ -388,6 +403,7 @@ These code samples show common scenario operations with the Azure Form Recognize
 * Train a model with labels: [sample_train_model_with_labels.py][sample_train_model_with_labels]
 * Manage custom models: [sample_manage_custom_models.py][sample_manage_custom_models]
 * Copy a model between Form Recognizer resources: [sample_copy_model.py][sample_copy_model]
+* Create a composed model from a collection of models trained with labels: |[sample_create_composed_model.py][sample_create_composed_model]
 
 ### Async APIs
 This library also includes a complete async API supported on Python 3.5+. To use it, you must
@@ -403,7 +419,7 @@ are found under the `azure.ai.formrecognizer.aio` namespace.
 * Train a model with labels: [sample_train_model_with_labels_async.py][sample_train_model_with_labels_async]
 * Manage custom models: [sample_manage_custom_models_async.py][sample_manage_custom_models_async]
 * Copy a model between Form Recognizer resources: [sample_copy_model_async.py][sample_copy_model_async]
-
+* Create a composed model from a collection of models trained with labels: [sample_create_composed_model_async.py][sample_create_composed_model_async]
 
 ### Additional documentation
 
@@ -475,3 +491,5 @@ This project has adopted the [Microsoft Open Source Code of Conduct][code_of_con
 [sample_train_model_without_labels_async]: https://github.com/Azure/azure-sdk-for-python/tree/master/sdk/formrecognizer/azure-ai-formrecognizer/samples/async_samples/sample_train_model_without_labels_async.py
 [sample_copy_model]: https://github.com/Azure/azure-sdk-for-python/tree/master/sdk/formrecognizer/azure-ai-formrecognizer/samples/sample_copy_model.py
 [sample_copy_model_async]: https://github.com/Azure/azure-sdk-for-python/tree/master/sdk/formrecognizer/azure-ai-formrecognizer/samples/async_samples/sample_copy_model_async.py
+[sample_create_composed_model]: https://github.com/Azure/azure-sdk-for-python/tree/master/sdk/formrecognizer/azure-ai-formrecognizer/samples
+[sample_create_composed_model_async]: https://github.com/Azure/azure-sdk-for-python/tree/master/sdk/formrecognizer/azure-ai-formrecognizer/samples/async_samples/
diff --git a/sdk/formrecognizer/azure-ai-formrecognizer/azure/ai/formrecognizer/__init__.py b/sdk/formrecognizer/azure-ai-formrecognizer/azure/ai/formrecognizer/__init__.py
@@ -31,7 +31,8 @@
     CustomFormModel,
     CustomFormSubmodel,
     CustomFormModelField,
-    FieldValueType
+    FieldValueType,
+    CustomFormModelProperties,
 )
 from ._api_versions import FormRecognizerApiVersion
 
@@ -62,7 +63,8 @@
     'CustomFormModel',
     'CustomFormSubmodel',
     'CustomFormModelField',
-    'FieldValueType'
+    'FieldValueType',
+    'CustomFormModelProperties',
 ]
 
 __VERSION__ = VERSION
diff --git a/sdk/formrecognizer/azure-ai-formrecognizer/azure/ai/formrecognizer/_form_training_client.py b/sdk/formrecognizer/azure-ai-formrecognizer/azure/ai/formrecognizer/_form_training_client.py
@@ -11,6 +11,7 @@
     Any,
     Dict,
     Union,
+    List,
     TYPE_CHECKING,
 )
 from azure.core.tracing.decorator import distributed_trace
@@ -96,6 +97,7 @@ def begin_training(self, training_files_url, use_training_labels, **kwargs):
         :keyword bool include_subfolders: A flag to indicate if subfolders within the set of prefix folders
             will also need to be included when searching for content to be preprocessed. Not supported if
             training with labels.
+        :keyword str display_name: A display name for your model.
         :keyword int polling_interval: Waiting time between two polls for LRO operations
             if no Retry-After header is present. Defaults to 5 seconds.
         :keyword str continuation_token: A continuation token to restart a poller from a saved state.
@@ -105,6 +107,8 @@ def begin_training(self, training_files_url, use_training_labels, **kwargs):
         :raises ~azure.core.exceptions.HttpResponseError:
             Note that if the training fails, the exception is raised, but a model with an
             "invalid" status is still created. You can delete this model by calling :func:`~delete_model()`
+        .. versionadded:: v2.1-preview
+            The *display_name* keyword argument
 
         .. admonition:: Example:
 
@@ -118,13 +122,16 @@ def begin_training(self, training_files_url, use_training_labels, **kwargs):
 
         def callback_v2_0(raw_response):
             model = self._deserialize(self._generated_models.Model, raw_response)
-            return CustomFormModel._from_generated(model)
+            return CustomFormModel._from_generated(model, api_version=self.api_version)
 
         def callback_v2_1(raw_response, _, headers):  # pylint: disable=unused-argument
             model = self._deserialize(self._generated_models.Model, raw_response)
-            return CustomFormModel._from_generated(model)
+            return CustomFormModel._from_generated(model, api_version=self.api_version)
 
         cls = kwargs.pop("cls", None)
+        display_name = kwargs.pop("display_name", None)
+        if display_name and self.api_version == "2.0":
+            raise ValueError("'display_name' is only available for API version V2_1_PREVIEW and up")
         continuation_token = kwargs.pop("continuation_token", None)
         polling_interval = kwargs.pop("polling_interval", self._client._config.polling_interval)
 
@@ -169,6 +176,7 @@ def callback_v2_1(raw_response, _, headers):  # pylint: disable=unused-argument
                     prefix=kwargs.pop("prefix", ""),
                     include_sub_folders=kwargs.pop("include_subfolders", False),
                 ),
+                model_name=display_name
             ),
             cls=deserialization_callback,
             continuation_token=continuation_token,
@@ -273,7 +281,7 @@ def get_custom_model(self, model_id, **kwargs):
             raise ValueError("model_id cannot be None or empty.")
 
         response = self._client.get_custom_model(model_id=model_id, include_keys=True, **kwargs)
-        return CustomFormModel._from_generated(response)
+        return CustomFormModel._from_generated(response, api_version=self.api_version)
 
     @distributed_trace
     def get_copy_authorization(self, resource_id, resource_region, **kwargs):
@@ -373,6 +381,55 @@ def _copy_callback(raw_response, _, headers):  # pylint: disable=unused-argument
             **kwargs
         )
 
+    @distributed_trace
+    def begin_create_composed_model(
+        self,
+        model_ids,
+        **kwargs
+    ):
+        # type: (List[str], Any) -> LROPoller[CustomFormModel]
+        """Creates a composed model from a collection of existing trained models with labels.
+
+        :param list[str] model_ids: List of model IDs to use in the composed model.
+        :keyword str display_name: Optional model display name.
+        :keyword int polling_interval: Default waiting time between two polls for LRO operations if
+            no Retry-After header is present.
+        :keyword str continuation_token: A continuation token to restart a poller from a saved state.
+        :return: An instance of an LROPoller. Call `result()` on the poller
+            object to return a :class:`~azure.ai.formrecognizer.CustomFormModel`.
+        :rtype: ~azure.core.polling.LROPoller[~azure.ai.formrecognizer.CustomFormModel]
+        :raises ~azure.core.exceptions.HttpResponseError:
+
+        .. admonition:: Example:
+
+            .. literalinclude:: ../samples/sample_create_composed_model.py
+                :start-after: [START begin_create_composed_model]
+                :end-before: [END begin_create_composed_model]
+                :language: python
+                :dedent: 8
+                :caption: Create a composed model
+        """
+
+        def _compose_callback(raw_response, _, headers):  # pylint: disable=unused-argument
+            model = self._deserialize(self._generated_models.Model, raw_response)
+            return CustomFormModel._from_generated_composed(model)
+
+        display_name = kwargs.pop("display_name", None)
+        polling_interval = kwargs.pop("polling_interval", self._client._config.polling_interval)
+        continuation_token = kwargs.pop("continuation_token", None)
+        try:
+            return self._client.begin_compose_custom_models_async(
+                {"model_ids": model_ids, "model_name": display_name},
+                cls=kwargs.pop("cls", _compose_callback),
+                polling=LROBasePolling(timeout=polling_interval, lro_algorithms=[TrainingPolling()], **kwargs),
+                continuation_token=continuation_token,
+                **kwargs
+            )
+        except ValueError:
+            raise ValueError(
+                "Method 'begin_create_composed_model' is only available for API version V2_1_PREVIEW and up"
+            )
+
     def get_form_recognizer_client(self, **kwargs):
         # type: (Any) -> FormRecognizerClient
         """Get an instance of a FormRecognizerClient from FormTrainingClient.