Scribd
Upload
3 views3 pages

Machine Learning Engineer

The Machine Learning Engineer Assessment consists of four main tasks: implementing a distributed training pipeline for a deep learning model, optimizing hyperparameters using a tuning library, debugging and optimizing a training pipeline, and fine-tuning a pre-trained model. Candidates are required to submit scripts, logs, and reports for each task, along with a screen recording of their process. The assessment aims to evaluate practical skills in machine learning engineering.

Uploaded by

072127anandsharma
Copyright
© All Rights Reserved
We take content rights seriously. If you suspect this is your content, claim it here.
Available Formats
Download as PDF, TXT or read online on Scribd
3 views3 pages

Machine Learning Engineer

The Machine Learning Engineer Assessment consists of four main tasks: implementing a distributed training pipeline for a deep learning model, optimizing hyperparameters using a tuning library, debugging and optimizing a training pipeline, and fine-tuning a pre-trained model. Candidates are required to submit scripts, logs, and reports for each task, along with a screen recording of their process. The assessment aims to evaluate practical skills in machine learning engineering.

Uploaded by

072127anandsharma
Copyright
© All Rights Reserved
We take content rights seriously. If you suspect this is your content, claim it here.
Available Formats
Download as PDF, TXT or read online on Scribd

Machine Learning Engineer Assessment

Instructions:

• Time Limit: 4-5 hours to complete the test.


• You can use any resources (documentation, online guides, etc.) to assist you during the
test.
• Submit your solutions (code files and required outputs) via email or upload them to the
provided submission link.
• Important: While working on this test, you must keep your screen recording on
throughout the duration of the test. Ensure that the recording captures the entire
process of working on the tasks.

1. Distributed Training

You are required to implement a distributed training pipeline for a deep learning model.

• Dataset: CIFAR-10 (You can use [Link] or [Link]).


• Task:
o Train a ResNet model using PyTorch or TensorFlow.
o Configure the training to run on two GPUs using either [Link] (for
PyTorch) or [Link] (for TensorFlow).
o Save training checkpoints after each epoch and ensure the model can resume
training from the last checkpoint if interrupted.
• Deliverables:
o Your training script.
o A log file showing GPU utilization and metrics (loss, accuracy) for at least 5
epochs.

2. Hyperparameter Tuning

Your task is to optimize the training of a model by tuning its hyperparameters.

• Task:
o Implement hyperparameter tuning for a given model using a tuning library like
Optuna or Ray Tune.
o Experiment with the following hyperparameters:
▪ Learning rate
▪ Batch size
▪ Optimizer type (e.g., Adam, SGD)
• Deliverables:
o Tuning script.
o A report summarizing the results of your hyperparameter tuning process.

3. Debugging and Optimization

You are given a training pipeline with the following issues:

• The model produces NaN loss after a few epochs.


• GPU usage is below 30%.
• The training process is slower than expected.
• Task:
o Analyze the provided script, identify the problems, and fix them.
• Deliverables:
o A corrected version of the script.
o A short explanation of the issues you identified and how you resolved them.

4. Fine-Tuning a Pre-Trained Model

Use a pre-trained ResNet or BERT model to fine-tune it on a specific dataset.

• Instructions:
o Fine-tune the model on the provided dataset:
▪ ResNet for CIFAR-10.
▪ BERT for a text classification dataset.
o Compare the performance of fine-tuning versus training the model from
scratch.
• Deliverables:
o Fine-tuning script.
o A report comparing the results of both approaches (fine-tuning vs. training from
scratch).
Submission Guidelines

• Include all scripts, outputs, and reports in a single compressed folder.


• Use clear and consistent naming conventions for files.
• Submit your work by the deadline provided.
• Plagiarism checks will be conducted on the submitted code.
• Attach the screen recording of your entire process along with your submission.

This test is designed to evaluate your practical skills as a Machine Learning Engineer. It includes
tasks related to distributed training, hyperparameter tuning, debugging, and fine-tuning pre-
trained models.

Good luck!

You might also like