diff --git a/assets/training/model_management/components/import_model/spec.yaml b/assets/training/model_management/components/import_model/spec.yaml
index 9eb2969307..a875ee0451 100644
--- a/assets/training/model_management/components/import_model/spec.yaml
+++ b/assets/training/model_management/components/import_model/spec.yaml
@@ -282,7 +282,7 @@ jobs:
     inputs:
       model_family: 'MmDetectionImage'
       model_name: ${{parent.inputs.model_id}}
-      download_from_source: False
+      download_from_source: true
     outputs:
       output_dir:
         type: uri_file
diff --git a/assets/training/model_management/src/azureml/model/mgmt/processors/pyfunc/convertors.py b/assets/training/model_management/src/azureml/model/mgmt/processors/pyfunc/convertors.py
index e20fc66285..557e954708 100644
--- a/assets/training/model_management/src/azureml/model/mgmt/processors/pyfunc/convertors.py
+++ b/assets/training/model_management/src/azureml/model/mgmt/processors/pyfunc/convertors.py
@@ -191,11 +191,11 @@ def save_as_mlflow(self):
         mlflow_model_wrapper = ImagesDetectionMLflowModelWrapper(task_type=self._task)
         artifacts_dict = self._prepare_artifacts_dict()
         if self._task == MMLabDetectionTasks.MM_OBJECT_DETECTION.value:
-            pip_requirements = os.path.join(self.MODEL_DIR, "mmdet-od-requirements.txt")
+            conda_env_file = os.path.join(self.MODEL_DIR, "conda_od.yaml")
         elif self._task == MMLabDetectionTasks.MM_INSTANCE_SEGMENTATION.value:
-            pip_requirements = os.path.join(self.MODEL_DIR, "mmdet-is-requirements.txt")
+            conda_env_file = os.path.join(self.MODEL_DIR, "conda_is.yaml")
         else:
-            pip_requirements = None
+            conda_env_file = None
         code_path = [
             os.path.join(self.MODEL_DIR, "detection_predict.py"),
             os.path.join(self.MODEL_DIR, "config.py"),
@@ -204,8 +204,8 @@ def save_as_mlflow(self):
         super()._save(
             mlflow_model_wrapper=mlflow_model_wrapper,
             artifacts_dict=artifacts_dict,
-            pip_requirements=pip_requirements,
             code_path=code_path,
+            conda_env=conda_env_file,
         )
 
     def _prepare_artifacts_dict(self) -> Dict:
diff --git a/assets/training/model_management/src/azureml/model/mgmt/processors/pyfunc/vision/conda_is.yaml b/assets/training/model_management/src/azureml/model/mgmt/processors/pyfunc/vision/conda_is.yaml
new file mode 100644
index 0000000000..96a48f8c67
--- /dev/null
+++ b/assets/training/model_management/src/azureml/model/mgmt/processors/pyfunc/vision/conda_is.yaml
@@ -0,0 +1,19 @@
+channels:
+- conda-forge
+dependencies:
+- python=3.9.19
+- pip<=24.0
+- pip:
+  - mlflow==2.12.1
+  - cloudpickle==2.2.1
+  - datasets==2.15.0
+  - openmim==0.3.9
+  - torch==2.0.1
+  - torchvision==0.15.2
+  - transformers==4.38.2
+  - accelerate==0.27.2
+  - albumentations==1.3.0
+  - scikit-image==0.19.3
+  - simplification==0.7.10
+  - fairscale==0.4.13
+name: mlflow-env
diff --git a/assets/training/model_management/src/azureml/model/mgmt/processors/pyfunc/vision/conda_od.yaml b/assets/training/model_management/src/azureml/model/mgmt/processors/pyfunc/vision/conda_od.yaml
new file mode 100644
index 0000000000..738e6fd5f8
--- /dev/null
+++ b/assets/training/model_management/src/azureml/model/mgmt/processors/pyfunc/vision/conda_od.yaml
@@ -0,0 +1,17 @@
+channels:
+- conda-forge
+dependencies:
+- python=3.9.19
+- pip<=24.0
+- pip:
+  - mlflow==2.12.1
+  - cloudpickle==2.2.1
+  - datasets==2.15.0
+  - openmim==0.3.9
+  - torch==2.0.1
+  - torchvision==0.15.2
+  - transformers==4.38.2
+  - accelerate==0.27.2
+  - albumentations==1.3.0
+  - fairscale==0.4.13
+name: mlflow-env