AGENT SKILL

logging-core

Ordinary TensorBoard scalar logging, writer lifecycle, and event-file checks for tensorboardX.

License 许可
MIT
最近更新
2026-09-03
Stars
★ 351
Forks
27

01 前言 Front Matter

name
logging-core
license
MIT
metadata
{ "disco-role": "operating" }
description
Ordinary TensorBoard scalar logging, writer lifecycle, and event-file checks for tensorboardX.
disable-model-invocation
true

02 资源文件 5

03 SKILL.md 正文

logging-core

Use this sub-skill for the ordinary tensorboardX event-writing path: creating a writer, logging scalar-like values, managing run directories, and checking that TensorBoard can read the result.

Use this when

  • the user needs SummaryWriter setup or logdir naming guidance
  • the task is about add_scalar, add_scalars, hparams, or custom scalar layouts
  • the user asks why TensorBoard does not show a run yet
  • the code must flush, close, reopen, or resume a writer safely
  • the user wants write_to_disk=False for dry runs or tests
  • a small local event-file smoke check is enough to validate the workflow

Route out

Operating sequence

  1. Choose a writable logdir or let SummaryWriter() create one.
  2. Use slash-separated tags for related scalar plots.
  3. Log scalar-like data with explicit steps whenever reproducibility matters.
  4. Use with writer.use_metadata(global_step=step): ... when several calls share one step or wall time.
  5. Call flush() when immediate visibility matters and close() when the run is done.
  6. Inspect output with tensorboard --logdir <logdir> after event files exist.
  7. If the run resumed after a crash, use purge_step instead of mixing stale and new steps blindly.

Bundled helpers

Run each helper with --help first. Both are safe by default and create temporary output when no logdir is supplied.

References

  • references/api-reference.md: verified method signatures, writer stack, logdir behavior, scalar groups, hparams, and lifecycle details.
  • references/workflows.md: short recipes for scalar runs, hparams, metadata defaults, custom scalars, crash recovery, and local event-file sanity checks.
  • references/troubleshooting.md: missing event files, delayed flushes, duplicate steps, scalar shape errors, JSON-export surprises, path issues, and memory growth.

Decision hints

  • Use add_scalar() for one numeric series and add_scalars() for grouped charts.
  • Use add_hparams() for comparing explicit trial configurations, not as a full hyperparameter search engine.
  • Use write_to_disk=False only when the absence of event files is intended.
  • Prefer context managers in short scripts so event files are closed even on exceptions.

04 同仓库更多 Skills

VectorSpaceLab
VectorSpaceLab/AREX-Skill

passnet-triton-opt

Make a single PassNet Triton kernel fast AND numerically faithful: performance model (when a replacement can win at all), block/grid/warp tuning, autotune policy, launch overhead, and per-op numeric recipes to pass the dtype baseline tolerances. Use when a pass matches and is correct but speedup below expected, or when correctness fails by small numeric margins.

★ 351
VectorSpaceLab
VectorSpaceLab/AREX-Skill

passnet-solve

END-TO-END playbook for solving one PassNet sample: analyze the computation graphs, decide the optimization strategy (what to fuse, what to replace, what to leave alone), drive the iteration loop, and maximize the sample score. This is the ENTRY skill — invoke it first for any PassNet optimization task; it tells you when to use passnet-pattern-fusion, passnet-triton-opt and passnet-feedback.

★ 351
VectorSpaceLab
VectorSpaceLab/AREX-Skill

passnet-skill

PassNet GPU kernel optimization via compiler passes. Design and implement Triton-based optimization passes, create pass files under ./pass_dir/, self-evaluate with pass_evaluator, and iterate to maximize GPU speedup.

★ 351
VectorSpaceLab
VectorSpaceLab/AREX-Skill

passnet-pattern-fusion

Author PassNet pass files: write patterns that actually MATCH the FX graph, pick fusion regions (which ops to absorb into one Triton kernel), and structure multi-pass submissions with the mandatory shared-dispatch architecture. Use when creating or fixing pattern()/replacement_args()/replacement_func() files, when a pass "failed to match", or when deciding how to fuse multiple kernels/ops.

★ 351
VectorSpaceLab
VectorSpaceLab/AREX-Skill

passnet-orchestrate

Round/batch planning and decision support for PassNet graph optimization. Use for multi-sample triage, worker allocation, eval-budget policy, and post-eval keep/revert decisions. For solving one concrete sample end to end, use passnet-solve as the entry skill; consult this skill only when you need broader planning or a specific decision gate.

★ 351
VectorSpaceLab
VectorSpaceLab/AREX-Skill

passnet-feedback

Fast, deterministic feedback for PassNet work: pre-flight pattern/match verification WITHOUT burning a GPU evaluation, per-node bottleneck analysis of a sample's graphs, and eval-log parsing into per-variant status + estimated score + failure classification. Use BEFORE every GPU evaluation (check_pattern), at the START of a sample (analyze_graph), and AFTER every evaluation (parse_eval_log).

★ 351