---

name: analysis-and-utilities description: "Analyze BEVFormer logs and utility outputs while separating safe summaries from data-, checkpoint-, or GPU-bound analysis tasks." disable-model-invocation: true metadata: disco-role: operating license: Apache 2.0

analysis-and-utilities

Analyze BEVFormer log and utility questions without running training.

Use this sub-skill for

• JSON or JSONL training and evaluation log summaries
• analyze_logs, benchmark, visual, get_params, fuse_conv_bn, and visualize_results
• model zoo log comparison and checkpoint utility questions

Route away when

• install, import, or config questions need handling -> installation-and-configs
• train, eval, or FP16 launch commands are needed -> training-and-evaluation
• nuScenes data layout or CAN bus preparation is the real task -> dataset-preparation

Bundled helper

• scripts/summarize_bevformer_log.py

Read first

• references/analysis-utilities.md
• references/troubleshooting.md

Operating rules

• Keep log summaries safe: run the bundled helper on small JSON or JSONL fixtures only.
• Treat benchmark, visualization, and checkpoint-fusion requests as gated by dataset, checkpoint, or mutation risk.
• Do not run training or evaluation from this sub-skill.