NVIDIA-Certified Professional Generative AI LLMs

0/4 checked

Q_01 domain/d1 · medium multi discuss →

A model architecture team is evaluating NVIDIA's nemotron-3-ultra-550b-a55b model for an agentic reasoning application requiring very long input documents. Which architectural characteristics of this model directly support that use case?

A A hybrid Mamba-Transformer design with Mixture-of-Experts (MoE)
B A 1M-token context window
C A pure dense Transformer with no expert routing
D A convolutional encoder-decoder architecture
E A recurrent LSTM backbone

Q_02 domain/d1 · medium single discuss →

A team deploying an LLM inference service wants to scale the context (prefill) phase separately from the generation (decode) phase so each can use dedicated GPUs. Which NVIDIA Dynamo capability implements this approach?

A Disaggregated Serving
B LLM-Aware Router
C KV Caching to Storage
D Low-Latency Communication Library (NIXL)

Q_03 domain/d1 · medium single

Locked question — unlock this exam to view it.

🔒 locked — unlock to review

Q_04 domain/d1 · medium single

Locked question — unlock this exam to view it.

🔒 locked — unlock to review

2 questions locked

unlock all 2 questions

NVIDIA-Certified Professional Generative AI LLMs

Explanation

Explanation