fix: Move sagemaker-mlflow to extras

Author: ryansteakley

hatch_build.py

@@ -20,7 +20,7 @@ def read_feature_deps(feature):
 
     optional_dependencies = {"all": []}
 
-    for feature in ("feature-processor", "huggingface", "local", "scipy"):
+    for feature in ("feature-processor", "huggingface", "local", "scipy", "sagemaker-mlflow"):
         dependencies = read_feature_deps(feature)
         optional_dependencies[feature] = dependencies
         optional_dependencies["all"].extend(dependencies)

pyproject.toml

@@ -48,7 +48,6 @@ dependencies = [
   "PyYAML~=6.0",
   "requests",
   "sagemaker-core>=1.0.0,<2.0.0",
-  "sagemaker-mlflow",
   "schema",
   "smdebug_rulesconfig==1.0.1",
   "tblib>=1.7.0,<4",

requirements/extras/sagemaker-mlflow_requirements.txt

@@ -0,0 +1 @@
+sagemaker-mlflow>=0.1.0

requirements/extras/test_requirements.txt

@@ -44,3 +44,4 @@ huggingface_hub>=0.23.4
 uvicorn>=0.30.1
 fastapi>=0.111.0
 nest-asyncio
+sagemaker-mlflow>=0.1.0

src/sagemaker/estimator.py

@@ -107,7 +107,6 @@
 from sagemaker.workflow.parameters import ParameterString
 from sagemaker.workflow.pipeline_context import PipelineSession, runnable_by_pipeline
 
-from sagemaker.mlflow.forward_sagemaker_metrics import log_sagemaker_job_to_mlflow
 
 logger = logging.getLogger(__name__)
 
@@ -1374,9 +1373,14 @@ def fit(
             forward_to_mlflow_tracking_server = True
         if wait:
             self.latest_training_job.wait(logs=logs)
-        if forward_to_mlflow_tracking_server:
-            log_sagemaker_job_to_mlflow(self.latest_training_job.name)
-
+        try:
+            if forward_to_mlflow_tracking_server:
+                from sagemaker.mlflow.forward_sagemaker_metrics import log_sagemaker_job_to_mlflow
+                log_sagemaker_job_to_mlflow(self.latest_training_job.name)
+        except ImportError:
+            if forward_to_mlflow_tracking_server:
+                raise ValueError("Unable to import mlflow, check if sagemaker-mlflow is installed.")
+            
     def _compilation_job_name(self):
         """Placeholder docstring"""
         base_name = self.base_job_name or base_name_from_image(

src/sagemaker/mlflow/forward_sagemaker_metrics.py

@@ -20,7 +20,11 @@
 import re
 from typing import Set, Tuple, List, Dict, Generator
 import boto3
-import mlflow
+
+try:
+    import mlflow
+except ImportError:
+    raise ValueError("Unable to import mlflow, check if sagemaker-mlflow is installed.")
 from mlflow import MlflowClient
 from mlflow.entities import Metric, Param, RunTag