Container unable to find inference.py

Below is my file structure:
```
name_of_model
-->model
--> code
--> __init__.py
--> inference.py
-->requirements.txt
-->utils.py (imported by inference.py
-->data
--> __init__.py
-->scikit_learn_classifier.pkl (loaded by inference.py)
-->model_endpoint.py
```

I tar the model directory and upload to S3, and i checked that it is there.
Below is my deploy script, which I used:
```
from sagemaker.sklearn.model import SKLearnModel
...
def deploy(
s3_model_tar_path,
endpoint_name_prefix="model-endpoint",
instance_type="ml.t2.medium"
):
endpoint_name = f"{endpoint_name_prefix}-{int(time.time())}"
git_config = f"{endpoint_name_prefix}-{int(time.time())}"
model = SKLearnModel(
model_data=s3_model_tar_path,
py_version="py3",
framework_version="0.20.0",
role="AmazonSageMakerAndS3-FullAccess",
entry_point="inference.py"
)
predictor = model.deploy(
instance_type=instance_type,
initial_instance_count=1,
endpoint_name=endpoint_name
)
```

I have tried both specifying a source_dir and not specifying a source_dir. **Note:** I have tried a flat directory structure for model where inference.py is in the root. I have tried inserting /opt/ml/code in inference.py(as recommended https://github.com/aws/sagemaker-python-sdk/issues/648)

I believe there is something wrong with the scikit learn container as I always see the error: FileNotFoundError: [Errno 2] No such file or directory: 'inference.py'

When I specify source_dir I see that it gets an ImportError can't find inference.py. I think the scikit container has a problem finding the inference.py


Start with the deploy function and the model package layout described in model_endpoint.py, checking how inference.py is placed in the tar archive and where source_dir puts it. Reproduce the deployment with both the flat and nested layouts, then verify that the container can load inference.py without FileNotFoundError or ImportError.

Written by the indexing model from the issue text.