Model trainer does not recognize my sagemaker session

[Bastien-mva]

Describe the bug
Model Trainer does not recognize my sagemaker session and returns an error.

To reproduce

import sagemaker
from sagemaker.modules.train import ModelTrainer
from sagemaker.modules.configs import SourceCode, Compute


sagemaker_session = sagemaker.session.Session()
print('session:', sagemaker_session)

source_code = SourceCode(
    source_dir=".",
    requirements="requirements.txt",
    entry_script="train.py",
)

compute = Compute(instance_type="ml.t3.medium")

model_trainer = ModelTrainer(
    training_image="pytorch/pytorch:1.12.0-cpu",
    source_code=source_code,
    base_job_name="script-mode",
    compute=compute,
    role="arn:aws:iam::261877606600:role/SageMakerRoleu",
    sagemaker_session=sagemaker_session,
)

Expected behavior
No error

Screenshots or logs

ValidationError: 1 validation error for ModelTrainer
sagemaker_session
  Input should be an instance of Session [type=is_instance_of, input_value=<sagemaker.session.Sessio...bject at 0x7a7f0687cd10>, input_type=Session]
    For further information visit https://errors.pydantic.dev/2.11/v/is_instance_of
[Process exited 1]

System information
A description of your system. Please provide:

• SageMaker Python SDK: 2.245.0
• Framework name (eg. PyTorch) or algorithm (eg. KMeans): pytorch
• Framework version: 1.12.0-cpu
• Python version:3.12.3
• CPU or GPU:CPU
• Custom Docker image (Y/N):N

Additional context
The session output gives:

session: <sagemaker.session.Session object at 0x73f8fdd78da0>

I do not get why the error tells me that it should be an instance of Session, as it is an instance of Session.
Also, I get warnings:

sagemaker.config INFO - Not applying SDK defaults from location: /etc/xdg/xdg-i3/sagemaker/config.yaml
sagemaker.config INFO - Not applying SDK defaults from location: /home/bastien/.config/sagemaker/config.yaml

I am logging using sso login.

[Aditi2424]

Here's what's likely happening:

1. You're creating a session with sagemaker.session.Session()
2. But ModelTrainer from sagemaker.modules.train expects a session from the
   modules system

Suggested Fix:

...
from sagemaker.modules import Session  # Use this Session class instead

# Create a session from the modules package
sagemaker_session = Session()

The key change is importing and using Session from sagemaker.modules
instead of using sagemaker.session.Session().