remote-and-parallel-integrations
Use for GlobalSummaryWriter, multiprocessing/global writer patterns, remote RecordWriter factories, Comet forwarding, and credential-safe cloud or telemetry integration.
- License 许可
- MIT
- 最近更新
- 2026-09-03
- Stars
- ★ 351
- Forks
- 27
01 前言 Front Matter
- name
- remote-and-parallel-integrations
- license
- MIT
- metadata
- { "disco-role": "operating" }
- description
- Use for GlobalSummaryWriter, multiprocessing/global writer patterns, remote RecordWriter factories, Comet forwarding, and credential-safe cloud or telemetry integration.
- disable-model-invocation
- true
02 资源文件 6
- 参考文档references/api-reference.md3.7 KB
- 参考文档references/comet-and-cloud.md2.6 KB
- 参考文档references/troubleshooting.md2.9 KB
- 参考文档references/workflows.md2.4 KB
- 脚本scripts/tbx_global_writer_smoke.py4.0 KB
- 脚本scripts/tbx_remote_writer_check.py3.7 KB
03 SKILL.md 正文
Remote and Parallel Integrations
Use this sub-skill for logging surfaces that cross process, module, credential, or service boundaries.
Use this when
- one logger must be shared across imported modules
- worker processes need to append to one event stream
- a log target may be local,
s3://..., orgs://... - Comet forwarding is optional and must stay disabled by default
- GPU telemetry values exist only as optional scalar inputs
Route elsewhere when
- scalar, writer-lifecycle, or event-file detail is the main task → logging-core
- image, audio, video, text, mesh, or histogram payload detail is the main task → rich-media-summaries
- graph, embedding, or projector payload detail is the main task → graph-and-embedding-plugins
Safe defaults
- Prefer a local logdir unless the user explicitly asks for remote delivery.
- Treat Comet as disabled unless
comet_configenables it. - Treat
s3://andgs://as credentialed boundaries, not default runtime targets. - Create one writer in the owning process, then close it after workers finish.
- Use the bundled smoke helpers in this skill tree instead of ad hoc checks.
Bundled helpers
Cross-links
- Shared install/routing notes: ../../references/install-and-routing.md
- Shared troubleshooting: ../../references/troubleshooting.md
- Global writer API details: references/api-reference.md
- Cloud/Comet workflow notes: references/comet-and-cloud.md
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).