You work for an online retailer as a machine learning specialist.
Your team is building a deep learning model based on the Keras Sequential model to categorize the clothing your company's users post on their Instagram feeds when they use one of the hashtags that refer to your company.
You are the machine learning specialist assigned to building the training run visualization code to allow the team to monitor the training metrics of the model as it trains. How would you go about visualizing the training metrics? (Select TWO)
Click on the arrows to vote for the correct answer
A. B. C. D. E.Answers: A, C.
Option A is correct.
While creating your model training job in the SageMaker console, you specify a regex pattern used for the metrics that your model training script writes to your logs.
Option B is incorrect.
While creating your model training job in the SageMaker console, you specify a regex pattern used for the metrics that your model training script writes to your logs.
You can't specify the metrics directly.
You must use a regex pattern.
Option C is correct.
SageMaker parses from your logs the metrics which you wish to track and publishes them to CloudWatch.
The CloudWatch metrics dashboard allows you to visualize your SageMaker training job metrics as graphs for visualization.
Option D is incorrect.
The CloudWatch metrics dashboard allows you to visualize your SageMaker training job metrics as graphs for visualization, not the SageMaker metrics dashboard.
Option E is incorrect.
You would not need to write a python script to visualize your metrics data since the CloudWatch metrics dashboard gives you this functionality.
Reference:
Please see the AWS Machine Learning Blog titled Easily monitor and visualize metrics while training models on Amazon SageMaker, the Amazon SageMaker developer guide titled Use TensorFlow with Amazon SageMaker, and the Tensorflow.org page titled Basic classification: Classify images of clothing.
To visualize the training metrics of a deep learning model built using Keras Sequential model in AWS SageMaker, the following two options can be used:
A. Specify a regex pattern for the metrics that you want your model training script to write to your logs when creating the model training job in the SageMaker console. This will allow you to capture and log the metrics of interest during the training process. Once the metrics have been logged, you can then use various visualization tools to plot the metrics and monitor the training process.
B. Specify the metrics that you want your model training script to write to your logs when creating the model training job in the SageMaker console. This option allows you to explicitly define the metrics of interest and capture them during the training process.
Once the metrics have been captured in the logs, you can visualize them using the following methods:
C. Use the CloudWatch metrics dashboard to visualize the metrics that SageMaker automatically parsed from your logs and published for graphing and visualization. This method allows you to easily visualize the metrics using a pre-built dashboard that provides a variety of visualization options such as line charts, stacked area charts, and heat maps.
D. Use the SageMaker metrics dashboard to visualize the metrics that SageMaker automatically parsed from your logs and published for graphing and visualization. This method allows you to visualize the metrics in a similar way to the CloudWatch metrics dashboard, but with additional customization options such as the ability to set up alerts based on specific thresholds.
E. Write a python script in your SageMaker Jupyter notebook to visualize the metrics that SageMaker automatically parsed from your logs and published for graphing and visualization. This method allows you to customize the visualization to your specific needs and preferences, and can be useful if you want to create more complex visualizations or automate the visualization process.
Overall, the choice of visualization method will depend on the specific requirements of the project, the level of customization needed, and the expertise of the team in working with the different visualization tools available in SageMaker.