tracking
Routes LabML experiment tracking, logging, configuration, monitoring, and AppAPI workflows.
- License 许可
- MIT
- 最近更新
- 2026-09-03
- Stars
- ★ 351
- Forks
- 27
01 前言 Front Matter
- name
- tracking
- license
- MIT
- metadata
- { "disco-role": "operating" }
- description
- Routes LabML experiment tracking, logging, configuration, monitoring, and AppAPI workflows.
- disable-model-invocation
- true
02 资源文件 5
- 参考文档references/api-reference.md6.0 KB
- 参考文档references/troubleshooting.md3.7 KB
- 参考文档references/workflows.md3.9 KB
- 脚本scripts/hardware_probe.py1.8 KB
- 脚本scripts/tracking_smoke.py1.6 KB
03 SKILL.md 正文
Tracking
Use this subskill for the client-side LabML workflow: recording experiments,
tracking metrics, printing logs, configuring runs, monitoring hardware, and
using the AppAPI client.
Use this when
- The task mentions
experiment,tracker,logger,monit,lab,manage,AppAPI,.labml.yaml,labml monitor,labml service,labml capture,labml launch, orlabml app-server. - The user wants to record a run, inspect training progress, publish metrics to the monitoring app, or inspect hardware usage.
- The user asks how to integrate LabML into PyTorch, Lightning, Keras, FastAI, or similar training code.
Boundaries
Include:
- Experiment lifecycle helpers and config management.
- Metric tracking, log formatting, and monitored loops.
- Client-side hardware monitoring and service setup.
- App API calls from the client package.
- Framework integration recipes that only need the client runtime.
Exclude or route elsewhere:
- Training-loop abstractions, optimizers, datasets, and helper modules →
helpers. - SSH, rsync, and remote job orchestration →
remote. - FastAPI server internals, MongoDB models, and deployment →
server.
Read next
references/api-reference.mdfor verified signatures, config keys, and CLI commands.references/workflows.mdfor end-to-end recipes and framework integration patterns.references/troubleshooting.mdfor missing config files, app publishing, monitoring, and service failures.scripts/tracking_smoke.pyfor a safe local experiment/checkpoint smoke test.scripts/hardware_probe.pyfor a read-only CPU/GPU and monitor capability check.
Typical routes
Record a run
Choose this route for experiment.create, experiment.record, tracker.save,
logger.log, and monit.loop requests. It covers both compact scalar logging
and richer tracked outputs.
Configure a project
Choose this route for .labml.yaml, lab.configure, dynamic hyperparameters,
run metadata, and git-info questions.
Monitor hardware
Choose this route for labml monitor, labml service, psutil, and optional
py3nvml GPU reporting.
Use the app client
Choose this route for AppAPI requests that read runs, analyses, logs, custom
metrics, or data stores from a monitoring backend.
Integrate a framework
Choose this route for LabML patterns inside PyTorch, Lightning, Keras, FastAI, or custom training loops that only need the client package.
04 同仓库更多 Skills
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.
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.
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.
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.
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.
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).