cli-cloud
Routes AXLearn GCP CLI, bundling, launch, VM, bastion, Dataflow, logs, and auth workflows.
- License 许可
- Apache-2.0
- 最近更新
- 2026-09-03
- Stars
- ★ 351
- Forks
- 27
01 前言 Front Matter
- name
- cli-cloud
- license
- Apache 2.0
- metadata
- { "disco-role": "operating" }
- description
- Routes AXLearn GCP CLI, bundling, launch, VM, bastion, Dataflow, logs, and auth workflows.
- disable-model-invocation
- true
02 资源文件 4
- 参考文档references/cli-reference.md2.2 KB
- 参考文档references/gcp-config-template.md1.7 KB
- 参考文档references/troubleshooting.md2.1 KB
- 脚本scripts/inspect_gcp_cli.py2.6 KB
03 SKILL.md 正文
cli-cloud
Use this sub-skill for AXLearn's cloud command tree and GCP-facing operations.
Typical triggers:
axlearn gcp config,activate,cleanup, orget.axlearn gcp bundlefor tar/docker bundling.axlearn gcp launch,vm,bastion,dataflow,logs, orauth.- Questions about the
.axlearn/axlearn.default.configand.axlearn/.axlearn.configfiles. - TPU/GKE/Dataflow launch workflows or Cloud Logging log retrieval.
If the task is only about trainer config mechanics or fake-data smoke checks, use ../training-core/ first.
If the task is about the model that will run on the cloud, jump to the relevant domain sub-skill.
What to read
references/cli-reference.mdfor the command tree and config-file behavior.references/gcp-config-template.mdfor the distilled TOML config shape.references/troubleshooting.mdfor GCP auth, config, and log-view failures.scripts/inspect_gcp_cli.pyfor a safe command-tree probe.
Core cloud workflows
1) Inspect the command tree
The CLI is self-documenting. Use the bundled helper or run axlearn gcp --help to inspect subcommands and flags.
2) Activate a GCP config
The GCP config namespace is stored in the repo or home config file. List available entries first, then activate one by label.
3) Bundle the local code
Use bundle when you need to package the current checkout into a tarball or Docker image for launch.
4) Launch or monitor jobs
Use launch, vm, or bastion to submit a job and to manage its runtime. Use dataflow for Dataflow jobs and logs for Cloud Logging log access.
Decision points
- Choose
configwhen the task is about project activation or CLI defaults. - Choose
bundlewhen the task is about packaging the current checkout. - Choose
launchorvmwhen the task is about running a command remotely. - Choose
bastionwhen the task is about job scheduling and quota management. - Choose
dataflowwhen the task is about Beam/Dataflow runner setup. - Choose
logswhen the task is about log inspection rather than job execution. - Do not route trainer/model-specific questions here unless the question is really about how the cloud launcher wraps them.
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).