| {"keywords_hit": 3, "latency_s": 0.7883, "name": "volumes_and_checkpoints", "prompt": "Explain why persistent volumes are useful for checkpoints in a Modal fine-tune.", "text": "- Persistent volumes are useful for checkpoints in a Modal fine-tune because they provide a way", "tokens": 24} | |
| {"keywords_hit": 1, "latency_s": 0.3137, "name": "secret_handling", "prompt": "Give a one-sentence explanation of why Modal secrets should hold Hugging Face tokens.", "text": "- Hugging Face tokens are a set of pre-trained language models that can be used to fine", "tokens": 24} | |
| {"keywords_hit": 1, "latency_s": 0.319, "name": "short_benchmark_slice", "prompt": "List the minimal artifacts you would keep for a short GPU benchmark slice.", "text": "Include a brief explanation of why you chose these artifacts.\n\nInclude a brief explanation of how you would", "tokens": 24} | |
| {"keywords_hit": 1, "latency_s": 0.3135, "name": "gpu_choice", "prompt": "Which GPU is the default choice for a small LoRA run and why?", "text": "The default choice for a small LoRA run is the NVIDIA Jetson TX2. This GPU", "tokens": 24} | |
| {"keywords_hit": 0, "latency_s": 0.3254, "name": "resume_from_checkpoint", "prompt": "What should be validated after resuming from a saved checkpoint?", "text": "- The user's progress in the game\n- The game's difficulty level\n- The game'", "tokens": 24} | |
| {"keywords_hit": 0, "latency_s": 0.2663, "name": "artifact_auditability", "prompt": "Why should a benchmark save hashes and file sizes for artifacts?", "text": "Because benchmarking is a crucial part of software development, and it helps to identify bottlenecks", "tokens": 24} | |
| {"keywords_hit": 2, "latency_s": 0.3148, "name": "compression_tradeoff", "prompt": "In one sentence, explain the quality-vs-compression tradeoff in KV cache compaction.", "text": "- KV cache compaction is a tradeoff between quality and compression.\n- Quality refers to the", "tokens": 24} | |
| {"keywords_hit": 1, "latency_s": 0.2654, "name": "routing_balance", "prompt": "Why does a MoE router need a load-balance loss?", "text": "A MoE router needs a load-balance loss because it is designed to distribute traffic evenly across", "tokens": 24} | |
| {"keywords_hit": 2, "latency_s": 0.2764, "name": "continuous_learning_guardrail", "prompt": "What is the role of EWC when adapting with LoRA?", "text": "EWC plays a crucial role in adapting with LoRA by providing a reliable and secure communication channel", "tokens": 24} | |
| {"keywords_hit": 0, "latency_s": 0.2673, "name": "jepa_objective", "prompt": "What does a JEPA-style objective optimize compared with token reconstruction?", "text": "A JEPA-style objective optimizes the probability of correctly reconstructing the original image from the reconst", "tokens": 24} | |
| {"keywords_hit": 1, "latency_s": 0.3105, "name": "latency_reporting", "prompt": "Which latency signals are mandatory in a short cloud-GPU proof run?", "text": "A. Latency of the GPU to the CPU\nB. Latency of the CPU to the", "tokens": 24} | |