Question 1 of 15

Preprocessing for vision inference should typically:

Question 2 of 15

Batching inference requests improves throughput when:

Question 3 of 15

Post-training INT8 quantization:

Question 4 of 15

TensorRT optimizes models by:

Question 5 of 15

Input distribution shift in production means:

Question 6 of 15

Shadow deployment for a new vision model:

Question 7 of 15

ONNX Runtime as inference backend:

Question 8 of 15

Knowledge distillation for edge deployment:

Question 9 of 15

Concept drift occurs when:

Question 10 of 15

Warm-up requests after deploying a vision API:

Question 11 of 15

TFLite delegates on mobile:

Question 12 of 15

Human-in-the-loop for vision drift response:

Question 13 of 15

gRPC vs REST for vision inference:

Question 14 of 15

Pruning neural networks removes:

Question 15 of 15

When to fine-tune vs full retrain on drift: