Deprecate MultiModalCloudPredictor and improve resource names

docs/tutorials/predictor-tabular.md

@@ -34,7 +34,7 @@ cloud_predictor.fit(
 `train_data` can be a pandas DataFrame, or a path to a local or S3 file (CSV or Parquet). In every case AutoGluon-Cloud loads the data locally and uploads it to your `cloud_output_path` bucket before kicking off the SageMaker job.
 
 ### Reattach to a training job
-If your local connection drops, the training job keeps running on SageMaker. You can reattach with another `CloudPredictor` via {py:meth}`~autogluon.cloud.TabularCloudPredictor.attach_job` as long as you have the job name — it's logged when training starts (`INFO:sagemaker:Creating training-job with name: ag-cloudpredictor-...`) and also visible in the SageMaker console.
+If your local connection drops, the training job keeps running on SageMaker. You can reattach with another `CloudPredictor` via {py:meth}`~autogluon.cloud.TabularCloudPredictor.attach_job` as long as you have the job name — it's logged when training starts (`INFO:sagemaker:Creating training-job with name: ag-cloud-tabular-...`) and also visible in the SageMaker console.
 
 ```python
 another_cloud_predictor = TabularCloudPredictor()
@@ -165,18 +165,18 @@ It will output a dict similar to this:
     'local_output_path': '/home/ubuntu/XXX/demo/AutogluonCloudPredictor/ag-20221111_174928',
     'cloud_output_path': 's3://XXX/tabular-demo',
     'fit_job': {
-        'name': 'ag-cloudpredictor-1668188968-e5c3',
+        'name': 'ag-cloud-tabular-1668188968-e5c3',
         'status': 'Completed',
         'framework_version': '0.6.1',
-        'artifact_path': 's3://XXX/tabular-demo/model/ag-cloudpredictor-1668188968-e5c3/output/model.tar.gz'
+        'artifact_path': 's3://XXX/tabular-demo/model/ag-cloud-tabular-1668188968-e5c3/output/model.tar.gz'
     },
     'recent_transform_job': {
-        'name': 'ag-cloudpredictor-1668189393-e95c',
+        'name': 'ag-cloud-tabular-1668189393-e95c',
         'status': 'Completed',
         'result_path': 's3://XXX/tabular-demo/batch_transform/2022-11-11-17-56-33-991/results/test.csv.out'
     },
-    'transform_jobs': ['ag-cloudpredictor-1668189393-e95c'],
-    'endpoint': 'ag-cloudpredictor-1668189208-d23b'
+    'transform_jobs': ['ag-cloud-tabular-1668189393-e95c'],
+    'endpoint': 'ag-cloud-tabular-1668189208-d23b'
 }
 ```
 

docs/tutorials/predictor-timeseries.md

@@ -57,7 +57,7 @@ forecasts = cloud_predictor.fit_predict(
 By default predictions land at `{cloud_output_path}/{job_name}/predictions.csv`; pass `predictions_path` to choose a destination.
 
 ### Reattach to a training job
-If your local connection drops, the training job keeps running on SageMaker. You can reattach with another `CloudPredictor` via {py:meth}`~autogluon.cloud.TimeSeriesCloudPredictor.attach_job` as long as you have the job name — it's logged when training starts (`INFO:sagemaker:Creating training-job with name: ag-cloudpredictor-...`) and also visible in the SageMaker console.
+If your local connection drops, the training job keeps running on SageMaker. You can reattach with another `CloudPredictor` via {py:meth}`~autogluon.cloud.TimeSeriesCloudPredictor.attach_job` as long as you have the job name — it's logged when training starts (`INFO:sagemaker:Creating training-job with name: ag-cloud-timeseries-...`) and also visible in the SageMaker console.
 
 ```python
 another_cloud_predictor = TimeSeriesCloudPredictor()
@@ -250,18 +250,18 @@ It will output a dict similar to this:
     'local_output_path': '/home/ubuntu/XXX/demo/AutogluonCloudPredictor/ag-20221111_174928',
     'cloud_output_path': 's3://XXX/timeseries-demo',
     'fit_job': {
-        'name': 'ag-cloudpredictor-1668188968-e5c3',
+        'name': 'ag-cloud-timeseries-1668188968-e5c3',
         'status': 'Completed',
         'framework_version': '0.6.1',
-        'artifact_path': 's3://XXX/timeseries-demo/model/ag-cloudpredictor-1668188968-e5c3/output/model.tar.gz'
+        'artifact_path': 's3://XXX/timeseries-demo/model/ag-cloud-timeseries-1668188968-e5c3/output/model.tar.gz'
     },
     'recent_transform_job': {
-        'name': 'ag-cloudpredictor-1668189393-e95c',
+        'name': 'ag-cloud-timeseries-1668189393-e95c',
         'status': 'Completed',
         'result_path': 's3://XXX/timeseries-demo/batch_transform/2022-11-11-17-56-33-991/results/test.parquet.out'
     },
-    'transform_jobs': ['ag-cloudpredictor-1668189393-e95c'],
-    'endpoint': 'ag-cloudpredictor-1668189208-d23b'
+    'transform_jobs': ['ag-cloud-timeseries-1668189393-e95c'],
+    'endpoint': 'ag-cloud-timeseries-1668189208-d23b'
 }
 ```
 

src/autogluon/cloud/backend/backend.py

@@ -55,12 +55,18 @@ def cloud_output_path(self) -> str:
         return self._cloud_output_path
 
     def initialize(
-        self, local_output_path: str, predictor_type: str, cloud_output_path: Optional[str] = None, **kwargs
+        self,
+        local_output_path: str,
+        predictor_type: str,
+        cloud_output_path: Optional[str] = None,
+        resource_prefix: Optional[str] = None,
+        **kwargs,
     ) -> None:
         """Initialize the backend."""
         self.local_output_path = local_output_path
         self._cloud_output_path = cloud_output_path
         self.predictor_type = predictor_type
+        self.resource_prefix = resource_prefix or f"ag-cloud-{predictor_type}"
         self.original_features = None
         self.endpoint: Optional[Endpoint] = None
 

src/autogluon/cloud/backend/ray_backend.py

@@ -17,7 +17,6 @@
 from ..endpoint.endpoint import Endpoint
 from ..job.ray_job import RayFitJob
 from ..scripts import ScriptManager
-from ..utils.constants import CLOUD_RESOURCE_PREFIX
 from ..utils.dlc_utils import parse_framework_version, retrieve_image_uri
 from ..utils.ec2 import get_latest_ami
 from ..utils.iam import get_instance_profile_arn
@@ -143,7 +142,7 @@ def fit(
             If `custom_image_uri` is set, this argument will be ignored.
         job_name: str, default = None
             Name of the launched training job.
-            If None, CloudPredictor will create one with prefix ag-cloudpredictor
+            If None, AutoGluon Cloud creates one with a predictor-specific prefix.
         instance_type: str, default = 'ml.m5.2xlarge'
             Instance type the predictor will be trained on with SageMaker.
         instance_count: Union[int, str], default = "auto",
@@ -257,7 +256,7 @@ def fit(
             cluster_manager.setup_connection()
             time.sleep(10)  # waiting for connection to setup
             if job_name is None:
-                job_name = CLOUD_RESOURCE_PREFIX + "-" + get_utc_timestamp_now()
+                job_name = self.resource_prefix + "-" + get_utc_timestamp_now()
             job = RayFitJob(output_path=self.cloud_output_path + "/model")
             self._fit_job = job
 

src/autogluon/cloud/backend/sagemaker_backend.py

@@ -27,7 +27,7 @@
     AutoGluonRepackInferenceModel,
 )
 from ..utils.aws_utils import resolve_execution_role, setup_sagemaker_session
-from ..utils.constants import CLOUD_RESOURCE_PREFIX, VALID_ACCEPT
+from ..utils.constants import VALID_ACCEPT
 from ..utils.dlc_utils import infer_sagemaker_ami_version, parse_framework_version
 from ..utils.misc import MostRecentInsertedOrderedDict
 from ..utils.serializers import AutoGluonSerializationWrapper
@@ -207,7 +207,7 @@ def fit(
             If `custom_image_uri` is set, this argument will be ignored.
         job_name: str, default = None
             Name of the launched training job.
-            If None, CloudPredictor will create one with prefix ag-cloudpredictor
+            If None, AutoGluon Cloud creates one with a predictor- or model-specific prefix.
         instance_type: str, default = 'ml.m5.2xlarge'
             Instance type the predictor will be trained on with SageMaker.
         instance_count: int, default = 1
@@ -250,7 +250,7 @@ def fit(
             logger.log(20, f"Training with framework_version=={framework_version}")
 
         if not job_name:
-            job_name = sagemaker.utils.unique_name_from_base(CLOUD_RESOURCE_PREFIX)
+            job_name = sagemaker.utils.unique_name_from_base(self.resource_prefix)
 
         if instance_count == "auto":
             instance_count = 1
@@ -337,7 +337,7 @@ def fit(
             volume_size=volume_size,
             framework_version=framework_version,
             py_version=py_version,
-            base_job_name="autogluon-cloudpredictor-train",
+            base_job_name=f"{self.resource_prefix}-train",
             output_path=output_path,
             code_location=code_location,
             inputs=inputs,
@@ -386,7 +386,7 @@ def deploy(
             If None, will deploy the most recent trained predictor trained with `fit()`.
         endpoint_name: str
             The endpoint name to use for the deployment.
-            If None, CloudPredictor will create one with prefix `ag-cloudpredictor`
+            If None, AutoGluon Cloud creates one with a predictor- or model-specific prefix.
         framework_version: str, default = `latest`
             Inference container version of autogluon.
             If `latest`, will use the latest available container version.
@@ -433,7 +433,7 @@ def deploy(
             # Needed to infer the container image (CPU vs GPU) downstream — serverless is CPU-only.
             instance_type = "ml.m5.2xlarge"
         if not endpoint_name:
-            endpoint_name = sagemaker.utils.unique_name_from_base(CLOUD_RESOURCE_PREFIX)
+            endpoint_name = sagemaker.utils.unique_name_from_base(self.resource_prefix)
 
         # Resolve container image
         if custom_image_uri:
@@ -799,7 +799,7 @@ def predict(
             If `custom_image_uri` is set, this argument will be ignored.
         job_name: str, default = None
             Name of the launched training job.
-            If None, CloudPredictor will create one with prefix ag-cloudpredictor.
+            If None, AutoGluon Cloud creates one with a predictor- or model-specific prefix.
         instance_count: int, default = 1,
             Number of instances used to do batch transform.
         instance_type: str, default = 'ml.m5.2xlarge'
@@ -904,7 +904,7 @@ def predict_proba(
             If `custom_image_uri` is set, this argument will be ignored.
         job_name: str, default = None
             Name of the launched training job.
-            If None, CloudPredictor will create one with prefix ag-cloudpredictor.
+            If None, AutoGluon Cloud creates one with a predictor- or model-specific prefix.
         instance_count: int, default = 1,
             Number of instances used to do batch transform.
         instance_type: str, default = 'ml.m5.2xlarge'
@@ -1257,7 +1257,7 @@ def _predict(
         predictor_path = self._upload_predictor(predictor_path, cloud_key_prefix + "/predictor")
 
         if not job_name:
-            job_name = sagemaker.utils.unique_name_from_base(CLOUD_RESOURCE_PREFIX)
+            job_name = sagemaker.utils.unique_name_from_base(self.resource_prefix)
 
         if test_data_image_column is not None:
             logger.warning("Batch inference with image modality could be slow because of some technical details.")

src/autogluon/cloud/model/foundation_model.py

@@ -129,6 +129,7 @@ def __init__(
             local_output_path=self._tmpdir.name,
             cloud_output_path=self.cloud_output_path,
             predictor_type=self._predictor_type,
+            resource_prefix=f"ag-cloud-{self.model_id}",
             role=role,
         )
 

src/autogluon/cloud/predictor/cloud_predictor.py

@@ -203,7 +203,7 @@ def fit(
             If `custom_image_uri` is set, this argument will be ignored.
         job_name: str, default = None
             Name of the launched training job.
-            If None, CloudPredictor will create one with prefix ag-cloudpredictor
+            If None, CloudPredictor creates one with a predictor-specific prefix.
         instance_type: str, default = 'ml.m5.2xlarge'
             Instance type the predictor will be trained on with SageMaker.
         instance_count: Union[int, str], default = "auto"
@@ -405,7 +405,7 @@ def deploy(
             If None, will deploy the most recent trained predictor trained with `fit()`.
         endpoint_name: str
             The endpoint name to use for the deployment.
-            If None, CloudPredictor will create one with prefix `ag-cloudpredictor`
+            If None, CloudPredictor creates one with a predictor-specific prefix.
         framework_version: str, default = `latest`
             Inference container version of autogluon.
             If `latest`, will use the latest available container version.
@@ -594,7 +594,7 @@ def predict(
             If `custom_image_uri` is set, this argument will be ignored.
         job_name: str, default = None
             Name of the launched training job.
-            If None, CloudPredictor will create one with prefix ag-cloudpredictor.
+            If None, CloudPredictor creates one with a predictor-specific prefix.
         instance_count: int, default = 1,
             Number of instances used to do batch transform.
         instance_type: str, default = 'ml.m5.2xlarge'
@@ -689,7 +689,7 @@ def predict_proba(
             If `custom_image_uri` is set, this argument will be ignored.
         job_name: str, default = None
             Name of the launched training job.
-            If None, CloudPredictor will create one with prefix ag-cloudpredictor.
+            If None, CloudPredictor creates one with a predictor-specific prefix.
         instance_count: int, default = 1,
             Number of instances used to do batch transform.
         instance_type: str, default = 'ml.m5.2xlarge'

src/autogluon/cloud/predictor/multimodal_cloud_predictor.py

@@ -1,4 +1,5 @@
 import logging
+import warnings
 
 from ..backend.constant import MULTIMODL_SAGEMAKER, SAGEMAKER
 from .cloud_predictor import CloudPredictor
@@ -16,6 +17,15 @@ class MultiModalCloudPredictor(CloudPredictor):
     predictor_file_name = "MultiModalCloudPredictor.pkl"
     backend_map = {SAGEMAKER: MULTIMODL_SAGEMAKER}
 
+    def __init__(self, *args, **kwargs) -> None:
+        warnings.warn(
+            "AutoGluon Multimodal is on a deprecation path. "
+            "MultiModalCloudPredictor will be removed in autogluon.cloud v0.7.0.",
+            FutureWarning,
+            stacklevel=2,
+        )
+        super().__init__(*args, **kwargs)
+
     @property
     def predictor_type(self) -> str:
         """

src/autogluon/cloud/predictor/tabular_cloud_predictor.py

@@ -80,7 +80,7 @@ def fit_predict(
             Training container version of autogluon. If `latest`, will use the latest available container version.
             If `custom_image_uri` is set, this argument will be ignored.
         job_name: str, default = None
-            Name of the launched training job. If None, CloudPredictor will create one with prefix ag-cloudpredictor.
+            Name of the launched training job. If None, CloudPredictor creates one with prefix ``ag-cloud-tabular``.
         instance_type: str, default = 'ml.m5.2xlarge'
             Instance type the predictor will be trained on with SageMaker.
         instance_count: int, default = 1
@@ -169,7 +169,7 @@ def fit_predict_proba(
         framework_version: str, default = `latest`
             Training container version of autogluon. If `custom_image_uri` is set, this argument is ignored.
         job_name: str, default = None
-            Name of the launched training job. If None, CloudPredictor will create one with prefix ag-cloudpredictor.
+            Name of the launched training job. If None, CloudPredictor creates one with prefix ``ag-cloud-tabular``.
         instance_type: str, default = 'ml.m5.2xlarge'
             Instance type the predictor will be trained on with SageMaker.
         instance_count: int, default = 1

src/autogluon/cloud/predictor/timeseries_cloud_predictor.py

@@ -92,7 +92,7 @@ def fit(
             If `custom_image_uri` is set, this argument will be ignored.
         job_name: str, default = None
             Name of the launched training job.
-            If None, CloudPredictor will create one with prefix ag-cloudpredictor
+            If None, CloudPredictor creates one with prefix ``ag-cloud-timeseries``.
         instance_type: str, default = 'ml.m5.2xlarge'
             Instance type the predictor will be trained on with SageMaker.
         instance_count: int, default = 1
@@ -263,7 +263,7 @@ def predict(
             If `custom_image_uri` is set, this argument will be ignored.
         job_name: str, default = None
             Name of the launched training job.
-            If None, CloudPredictor will create one with prefix ag-cloudpredictor.
+            If None, CloudPredictor creates one with prefix ``ag-cloud-timeseries``.
         instance_count: int, default = 1,
             Number of instances used to do batch transform.
         instance_type: str, default = 'ml.m5.2xlarge'
@@ -382,7 +382,7 @@ def fit_predict(
             Training container version of autogluon. If `latest`, will use the latest available container version.
             If `custom_image_uri` is set, this argument will be ignored.
         job_name: str, default = None
-            Name of the launched training job. If None, CloudPredictor will create one with prefix ag-cloudpredictor.
+            Name of the launched training job. If None, CloudPredictor creates one with prefix ``ag-cloud-timeseries``.
         instance_type: str, default = 'ml.m5.2xlarge'
             Instance type the predictor will be trained on with SageMaker.
         instance_count: int, default = 1

src/autogluon/cloud/utils/constants.py

@@ -1,7 +1,5 @@
 VALID_ACCEPT = ["application/x-parquet", "text/csv", "application/json"]
 
-CLOUD_RESOURCE_PREFIX = "ag-cloudpredictor"
-
 LOCAL_MODE = "local"
 LOCAL_MODE_GPU = "local_gpu"
 MODEL_ARTIFACT_NAME = "model.tar.gz"