Databricks Generative AI Engineer Associate Question 164
Select 2You are tasked with selecting a pre-trained language model from a model hub to perform a sentiment analysis task. The model will be deployed in a production environment with strict latency requirements and limited computational resources. Which of the following factors in the model card would be most relevant for selecting an appropriate model?
- A
The model's architecture and size
- B
The pre-training dataset used for the model
- C
The model's inference latency benchmarks
- D
The fine-tuning tasks the model has been evaluated on
- E
The licensing information for the model
Show answer and explanation
Correct answers: A, C
Explanation
When selecting a model for deployment in a production environment with strict latency and computational constraints, the most relevant factors are the model's architecture and size, which affect resource usage, and its inference latency benchmarks, which indicate its real-world performance. While other factors such as licensing and fine-tuning tasks are relevant in other contexts, they do not directly address the operational requirements in this scenario.
- A. Correct.
The model's architecture and size are critical because they directly impact computational efficiency and resource usage, which are key considerations for production environments with limited resources.
- B. Incorrect.
While the pre-training dataset is important for understanding the model's general capabilities, it is less directly relevant to the latency and computational efficiency requirements of this specific task.
- C. Correct.
Inference latency benchmarks are crucial for meeting strict latency requirements in a production environment. These benchmarks provide insights into the model's performance during real-world deployments.
- D. Incorrect.
The fine-tuning tasks the model has been evaluated on provide information about the model's adaptability to specific tasks but do not directly address latency or resource constraints.
- E. Incorrect.
Licensing information is important for legal and compliance considerations but does not impact the performance or suitability of the model for a resource-constrained, low-latency production environment.