SageMaker : Client Error - Provided train label is in 'float32' format, 'float32' label is required

Question

I have my training data uploaded to s3 bucket and I have type-casted the columns with float64 dtype to float32 in my training data. I have defined a linear estimator from sagemaker and fit the training data in this cell.

linear = sagemaker.estimator.Estimator(container,
                                       role, 
                                       train_instance_count = 1, 
                                       train_instance_type = 'ml.c4.xlarge',
                                       output_path = output_location,
                                       sagemaker_session = sagemaker_session)



linear.set_hyperparameters(feature_dim = 147,
                           predictor_type = 'regressor',
                           mini_batch_size = 5,
                           epochs = 5,
                           num_models = 32,
                           loss = 'absolute_loss')

# pass in the training data from S3 to train the linear learner model

linear.fit({'train': s3_train_data})

I am getting an error when I run the cell. The last line of fitting the training data throws the below error :

UnexpectedStatusException: Error for Training job linear-learner-2023-02-01-04-16-46-698: Failed. Reason: ClientError: Unable to execute the algorithm. Provided train label is in 'float32' format, 'float32' label is required. Please provide train dataset with 'float32' labels and try again., exit code: 2

I have the training data in float32 format, what am I missing here?

I have converted all the columns in training dataset having float64 format to float32.

float64_cols = list(X_train.select_dtypes(include='float64'))
X_train[float64_cols] = X_train[float64_cols].astype('float32')