You have deployed a scikit-learn model to a Vertex AI endpoint using a custom model server. You enabled autoscaling; however, the deployed model fails to scale beyond one replica, leading to dropped requests. You notice that CPU utilization remains low even during periods of high load. What should you do?
#282
Your company operates a video sharing platform where users can view and upload videos. You're tasked with developing an ML model to forecast which newly uploaded videos will gain the most popularity, allowing these videos to receive priority placement on your company's website. How should you determine the success of the model?
#283
Your company specializes in building bridges for cities worldwide. To monitor the progress of construction projects, cameras are installed at each site. These cameras capture hourly images, which are then uploaded to a Cloud Storage bucket. A team of specialists reviews these images, selects the important ones, and annotates specific objects in them. To enhance scalability and reduce costs, you want to propose an ML solution with minimal upfront investment. What approach should you recommend?
#284
Your organization aims to enhance the efficiency of its internal shuttle service route, which currently stops at all pick-up points across the city every 30 minutes between 7 am and 10 am. The development team has already created an application on Google Kubernetes Engine, requiring users to confirm their presence and shuttle station one day in advance. How should you proceed?
#285
Your organization operates an online message board, and in recent months, there has been a noticeable uptick in the use of toxic language and instances of bullying within the platform. To address this issue, you implemented an automated text classification system designed to identify and flag comments that exhibit toxic or harmful behavior. However, you've received reports from users who believe that benign comments related to their religion are being incorrectly classified as abusive. Upon closer examination, it's become evident that the false positive rate of your classifier is higher for comments that pertain to certain underrepresented religious groups. Given that your team is operating on a limited budget and already stretched thin, what steps should you take to remedy this situation?
#286
When you observe oscillations in the loss during batch training of a neural network, how should you modify your model to ensure convergence?
#287
You are in the process of establishing a workflow for training and deploying your custom model in production. It's essential to maintain lineage information for your model and predictions. What steps should you take to achieve this?
#288
You are in the process of building a deep neural network classification model, and your dataset includes categorical input features. Some of these categorical columns have a high cardinality, with over 10,000 unique values. How should you handle the encoding of these categorical values for input into the model?
#289
You are building a recommendation engine for an online clothing store, with historical customer transaction data stored in BigQuery and Cloud Storage. To conduct exploratory data analysis (EDA), preprocessing, and model training iteratively while experimenting with different algorithms, you aim to minimize costs and development efforts. How should you configure the environment?
#290
You work for a global footwear retailer and need to predict when an item will go out of stock based on historical inventory data. Customer behavior is highly dynamic, as footwear demand is influenced by various factors. Your goal is to train models on all available data but assess their performance on specific data subsets before deploying them to production. What is the most efficient and dependable way to carry out this validation process?
#291
You are a member of the data science team at a manufacturing firm, and you are currently examining the company's extensive historical sales dataset, which consists of hundreds of millions of records. During your exploratory data analysis, you have several tasks to perform, including the calculation of descriptive statistics like mean, median, and mode, conducting intricate statistical hypothesis tests, and generating various feature-related plots over time. Your goal is to leverage as much of the sales data as feasible for your analyses while keeping computational resource usage to a minimum. How should you approach this situation?
#292
You're using Kubeflow Pipelines to build an end-to-end PyTorch-based MLOps pipeline, which involves data reading from BigQuery, processing, feature engineering, model training, evaluation, and model deployment to Cloud Storage. You're developing code for different versions of feature engineering and model training steps, running each in Vertex AI Pipelines. However, each pipeline run is taking over an hour, slowing down your development process and potentially increasing costs. What's the best approach to speed up execution while avoiding additional costs?
#293
Your team has deployed a model to a Vertex AI endpoint, and you've established a Vertex AI pipeline that streamlines the model training process, triggered by a Cloud Function. Your primary goals are to keep the model up-to-date while also minimizing retraining costs. How should you configure the retraining process?
#294
You trained a text classification model. You have the following SignatureDefs: signature_def['serving_default']: The given SavedModel SignatureDef contains the following input (s): inputs['text'] tensor_info: dtype: DT_STRING shape: (-1, 2) name: serving_default_text: 0 The given SavedModel SignatureDef contains the following output (s): outputs ['softmax'] tensor_info: dtype: DT_FLOAT shape: (-1, 2) name: StatefulPartitionedCall:0 Method name is: tensorflow/serving/predict You started a TensorFlow-serving component server and tried to send an HTTP request to get a prediction using: headers = {"content -type": "application/json"} json_response = requests.post('http://localhost:8501/v1/models/text_model:predict', data=data, headers=headers) What is the correct way to write the predict request?
#295
You have developed a BigQuery ML linear regression model using a training dataset stored in a BigQuery table, which receives new data every minute. To automate hourly model training and direct inference, you employ Cloud Scheduler and Vertex AI Pipelines. The feature preprocessing involves quantile bucketization and MinMax scaling on data from the past hour. To minimize storage and computational overhead, what approach should you take?
#296
You've recently deployed a model to a Vertex AI endpoint and configured online serving in Vertex AI Feature Store. As part of your setup, you've scheduled a daily batch ingestion job to update your feature store. However, during these batch ingestion processes, you notice high CPU utilization in your feature store's online serving nodes, leading to increased feature retrieval latency. To enhance online serving performance during these daily batch ingestion tasks, what should you do?
#297
You work for a multinational organization that has recently begun operations in Spain. Teams within your organization will need to work with various Spanish documents, such as business, legal, and financial documents. You want to use machine learning to help your organization get accurate translations quickly and with the least effort. Your organization does not require domain-specific terms or jargon. What should you do?
#298
You are developing a batch process to train a custom machine learning model and perform predictions. It's essential to track the lineage of both the model and the batch predictions. Which approach should you take?
#299
You are developing an ML model that predicts the cost of used automobiles based on data such as location, condition, model type, color, and engine/battery efficiency. The data is updated every night. Car dealerships will use the model to determine appropriate car prices. You created a Vertex AI pipeline that reads the data, splits the data into training/evaluation/test sets, performs feature engineering, trains the model using the training dataset, and validates the model using the evaluation dataset. You need to configure a retraining workflow that minimizes cost. What should you do?
#300
You are aiming to train a deep learning model for semantic image segmentation with a focus on reducing training time. However, when using a Deep Learning VM Image, you encounter the following error: The resource 'projects/deeplearning-platform/zones/europe-west4-c/acceleratorTypes/nvidia-tesla-k80' was not found. What steps should you take to address this issue?
