ML engineers are defining a convolutional neural network (CNN) model bur they are not sure how many filters to use in each convolutional layer. What can help them address this concern?
Hyperparameter optimization is a process of tuning the hyperparameters of a machine learning model, such as the number of filters in a convolutional neural network (CNN) model, to determine the best combination of hyperparameters that will result in the best model performance. HPO techniques are used to automatically find the optimal hyperparameter values, which can greatly increase the accuracy and performance of the model.
You are helping a customer start to implement hyper parameter optimization (HPO) with HPE Machine learning Development Environment. An ML engineer is putting together an experiment config file with the desired Adaptive A5HA settings. The engineer asks you questions, such as how many trials will be trained on the max length and what the min length for all trials will be.
What should you explain?
The engineer should specify the number of trials to train on the max length and the minimum length for all trials in the experiment config file. For example, if the engineer wants to run 10 trials with a max length of 10, the config file should look something like this:
{
'mode': 'A5HA',
'max_trials': 10,
'max_length': 10,
'min_length': 1,
'divisor': 2,
'max_runs': 1
}
Once the config file is complete, the engineer should upload it to the HPE Machine Learning Development Environment WebUI and view the graph of the experiment plan. This will allow the engineer to see how the Adaptive A5HA settings will affect the experiment. After that, the engineer can run the experiment and assess the results.
An ML engineer is running experiments on HPE Machine Learning Development Environment. The engineer notices all of the checkpoints for a trial except one disappear after the trial ends. The engineer wants to Keep more of these checkpoints. What can you recommend?
The best recommendation for an ML engineer running experiments on HPE Machine Learning Development Environment to keep more of the checkpoints is to adjust the experiment config's checkpoint storage settings to save more of the latest and best checkpoints. This can be done by monitoring ongoing trials in the WebUI and clicking checkpoint flags to auto-save the desired checkpoints. Additionally, the engineer should double-check that the checkpoint storage location is operating under 90% of total capacity to ensure that enough capacity is available to store the checkpoints. Finally, they can adjust the checkpoint storage settings to save checkpoints to a shared file system instead of cloud storage if desired.
What are the mechanics of now a model trains?
This is done by running the model through a training loop, where the model is fed data and the parameter weights are adjusted based on the results of the model's performance on the data. For example, if the model is a neural network, the weights of the connections between the neurons are adjusted based on the results of the model's performance on the data. This process is repeated until the model performs better on the data, at which point the model is considered trained.
You are in a directory on your machine with your experiment config file and your model code. You enter this command:
det experiment create myfile.yaml
You receive this error:
det experiment create: error: the following arguments are required: model_def
What should you do?
Make sure that the myfile.yaml tile includes code for a PyTorchTrial or TFKerasTrial class. When creating an experiment with the det experiment create command, you need to specify the model_def parameter to provide the code for the PyTorchTrial or TFKerasTrial class. This code should be specified in the myfile.yaml file, so make sure that the myfile.yaml file includes the code for the model you want to use.
