Answer-first summary for fast verification
Answer: Configure one a2-highgpu-1g instance with an NVIDIA A100 GPU with 80 GB of RAM. Use bfloat16 quantization during model training.
The question emphasizes training the ControlNet model with Stable Diffusion XL as quickly as possible. Option B is optimal because it uses an NVIDIA A100 GPU with 80 GB of RAM, which provides high computational power and memory bandwidth, and employs bfloat16 quantization. bfloat16 reduces memory usage and accelerates training compared to float32, with minimal accuracy loss, making it ideal for fast convergence. The community discussion (100% consensus for B) supports this, noting that bfloat16 on A100 trains faster and uses less memory than float32. Option A is less suitable because float32 precision increases training time without significant accuracy benefits for this use case. Options C and D are inefficient due to the lower performance of Tesla T4 GPUs (less RAM and computational power) compared to A100, and distributing training across multiple instances adds overhead without matching A100's speed.
Author: LeetQuiz Editorial Team
Ultimate access to all questions.
No comments yet.
You need to train a ControlNet model with Stable Diffusion XL for an image editing task and want to minimize the training time. Which hardware configuration should you select?
A
Configure one a2-highgpu-1g instance with an NVIDIA A100 GPU with 80 GB of RAM. Use float32 precision during model training.
B
Configure one a2-highgpu-1g instance with an NVIDIA A100 GPU with 80 GB of RAM. Use bfloat16 quantization during model training.
C
Configure four n1-standard-16 instances, each with one NVIDIA Tesla T4 GPU with 16 GB of RAM. Use float32 precision during model training.
D
Configure four n1-standard-16 instances, each with one NVIDIA Tesla T4 GPU with 16 GB of RAM. Use floar16 quantization during model training.