SageMaker Studio Lab - XGBoost Algortim - Am I Reading the Right Documentation?

Question

Google has answers for SageMaker Studio, but I am at a loss for an answer on SageMaker Studio LAB...

I am reading the following on XGBoost - Am I in the right place for SM Studio LAB?

https://docs.aws.amazon.com/sagemaker/latest/dg/xgboost.html

Is there a better way for me to use XGBoost in LAB, vs what I am doing below by reading the doc above?

In SageMaker Studio I would do the following to get the ECR container for the XGBoost algorithm:

from sagemaker import image_uris
container = image_uris.retrieve('xgboost', boto3.Session().region_name, '1')

I made it a bit farther using the github example:

https://github.com/aws/studio-lab-examples/blob/main/connect-to-aws/Access_AWS_from_Studio_Lab.ipynb

This works:

from sagemaker import image_uris
from sagemaker.xgboost import XGBoost

# Create a training job name
job_name = 'ufo-xgboost-job-{}'.format(datetime.now().strftime("%Y%m%d%H%M%S"))
print('Here is the job name {}'.format(job_name))

import sagemaker
import boto3
from sagemaker import image_uris
from sagemaker.session import Session
from sagemaker.inputs import TrainingInput

But this is giving me trouble:

sess = sagemaker.Session()

xgb = sagemaker.estimator.Estimator(container,
                                    role, 
                                    instance_count=1, 
                                    instance_type='ml.m4.xlarge',
                                    output_path='model.tar.gz',
                                    sagemaker_session=sess)

xgb.set_hyperparameters(objective='multi:softmax',
                        num_class=3,
                        num_round=100)

data_channels = {
    'train': s3_input_train,
    'validation': s3_input_validation
}
xgb.fit(data_channels, job_name=job_name) 

With the following errors:

ParsingError                              Traceback (most recent call last)
~/.conda/envs/default/lib/python3.9/site-packages/botocore/configloader.py in raw_config_parse(config_filename, parse_subsections)
    148         try:
--> 149             cp.read([path])
    150         except (six.moves.configparser.Error, UnicodeDecodeError):

~/.conda/envs/default/lib/python3.9/configparser.py in read(self, filenames, encoding)
    696                 with open(filename, encoding=encoding) as fp:
--> 697                     self._read(fp, filename)
    698             except OSError:

~/.conda/envs/default/lib/python3.9/configparser.py in _read(self, fp, fpname)
   1115         if e:
-> 1116             raise e
   1117 

ParsingError: Source contains parsing errors: '/home/studio-lab-user/.aws/config'
    [line  5]: 'from sagemaker import image_uris\n'
    [line  6]: 'import boto3\n'

During handling of the above exception, another exception occurred:

truncated error at bottom:

ConfigParseError: Unable to parse config file: /home/studio-lab-user/.aws/config

Answer 1

You're trying to run an external training job that would run outside of your Studio Lab Environment, on an ephemeral EC2 instance as part of a SageMaker training job.
With the error being: ConfigParseError: Unable to parse config file: /home/studio-lab-user/.aws/config it sounds like you didn't configure your AWS credentials correctly.
Did you complete successfully step 0 and step 1, in this guide for connecting Studio Lab with your AWS account?

Alternatively: Run XGBoost locally in your Studio Lab - OR - as you already have an AWS account use SageMaker Studio (isn't free), which will already be configured with a certain IAM role your admin will configure when creating the Studio domain.