NVIDIA Deep Learning Institute Certification Practice Test
NVIDIA Deep Learning Institute प्रमाणन, जिसमें न्यूरल नेटवर्क आर्किटेक्चर, GPU कंप्यूटिंग, मॉडल ऑप्टिमाइजेशन और डीप लर्निंग अनुप्रयोगों को शामिल किया गया है।
नमूना प्रश्न
पूरी परीक्षा कैसी है देखने के लिए कुछ प्रश्न आज़माएं।
140 में से 106 उत्तरों में जाँचने योग्य संदर्भ है।
A prompt asks the model to "think step by step" for math tutoring. Accuracy improves, but the system now exposes verbose reasoning that confuses younger students. Which adjustment best preserves answer quality while improving UX?
A team builds an embedding index from support articles. Some articles contain boilerplate footers, navigation menus, and repeated legal notices. Retrieval often returns chunks matching boilerplate rather than the user's issue. What preprocessing step is most important?
A data scientist loads a 180 GB clickstream table into pandas on a CPU workstation and repeatedly runs out of memory. The target environment has multiple NVIDIA GPUs and the workflow is mostly joins, filters, and aggregations. Which tool direction best aligns with the NVIDIA data stack?
A transformer encoder for classification prepends a special classification token and uses its final hidden state for prediction. What is the purpose of that token?
A model uses beam search for product descriptions. Outputs become fluent but repetitive and less diverse than sampled outputs. What is the most likely cause?