office-benchmarks
Routes LibMTL's Office-31 and Office-Home multi-input classification workflows.
- License 许可
- MIT
- 最近更新
- 2026-09-03
- Stars
- ★ 351
- Forks
- 27
01 前言 Front Matter
- name
- office-benchmarks
- license
- MIT
- metadata
- { "disco-role": "operating" }
- description
- Routes LibMTL's Office-31 and Office-Home multi-input classification workflows.
- disable-model-invocation
- true
02 资源文件 56
- 参考文档references/data_txt/office-31/amazon_test.txt23.7 KB
- 参考文档references/data_txt/office-31/amazon_train.txt68.2 KB
- 参考文档references/data_txt/office-31/amazon_val.txt22.4 KB
- 参考文档references/data_txt/office-31/dslr_test.txt4.7 KB
- 参考文档references/data_txt/office-31/dslr_train.txt11.2 KB
- 参考文档references/data_txt/office-31/dslr_val.txt3.5 KB
- 参考文档references/data_txt/office-31/webcam_test.txt7.3 KB
- 参考文档references/data_txt/office-31/webcam_train.txt18.9 KB
- 参考文档references/data_txt/office-31/webcam_val.txt6.0 KB
- 参考文档references/data_txt/office-home/Art_test.txt12.6 KB
- 参考文档references/data_txt/office-home/Art_train.txt34.3 KB
- 参考文档references/data_txt/office-home/Art_val.txt11.1 KB
- 参考文档references/data_txt/office-home/Clipart_test.txt25.5 KB
- 参考文档references/data_txt/office-home/Clipart_train.txt71.7 KB
- 参考文档references/data_txt/office-home/Clipart_val.txt23.4 KB
- 参考文档references/data_txt/office-home/Product_test.txt26.4 KB
- 参考文档references/data_txt/office-home/Product_train.txt73.8 KB
- 参考文档references/data_txt/office-home/Product_val.txt24.1 KB
- 参考文档references/data_txt/office-home/Real_World_test.txt28.6 KB
- 参考文档references/data_txt/office-home/Real_World_train.txt80.1 KB
- 参考文档references/data_txt/office-home/Real_World_val.txt26.2 KB
- 参考文档references/data-layouts.md1.1 KB
- 参考文档references/task-contracts.md1.3 KB
- 参考文档references/troubleshooting.md1.6 KB
- 参考文档references/workflows.md3.5 KB
- 脚本scripts/check_office_data.py3.0 KB
- 脚本scripts/install_office_runtime.py2.0 KB
- 脚本scripts/main.py592 B
- 脚本scripts/office_runtime/main.py538 B
- 脚本scripts/office_runtime/MANIFEST.in61 B
- 脚本scripts/office_runtime/pyproject.toml601 B
- 脚本scripts/office_runtime/README.md621 B
- 脚本scripts/office_runtime/src/libmtl_office_benchmark/__init__.py191 B
- 脚本scripts/office_runtime/src/libmtl_office_benchmark/create_dataset.py2.6 KB
- 脚本scripts/office_runtime/src/libmtl_office_benchmark/data_txt/office-31/amazon_test.txt23.7 KB
- 脚本scripts/office_runtime/src/libmtl_office_benchmark/data_txt/office-31/amazon_train.txt68.2 KB
- 脚本scripts/office_runtime/src/libmtl_office_benchmark/data_txt/office-31/amazon_val.txt22.4 KB
- 脚本scripts/office_runtime/src/libmtl_office_benchmark/data_txt/office-31/dslr_test.txt4.7 KB
- 脚本scripts/office_runtime/src/libmtl_office_benchmark/data_txt/office-31/dslr_train.txt11.2 KB
- 脚本scripts/office_runtime/src/libmtl_office_benchmark/data_txt/office-31/dslr_val.txt3.5 KB
- 脚本scripts/office_runtime/src/libmtl_office_benchmark/data_txt/office-31/webcam_test.txt7.3 KB
- 脚本scripts/office_runtime/src/libmtl_office_benchmark/data_txt/office-31/webcam_train.txt18.9 KB
- 脚本scripts/office_runtime/src/libmtl_office_benchmark/data_txt/office-31/webcam_val.txt6.0 KB
- 脚本scripts/office_runtime/src/libmtl_office_benchmark/data_txt/office-home/Art_test.txt12.6 KB
- 脚本scripts/office_runtime/src/libmtl_office_benchmark/data_txt/office-home/Art_train.txt34.3 KB
- 脚本scripts/office_runtime/src/libmtl_office_benchmark/data_txt/office-home/Art_val.txt11.1 KB
- 脚本scripts/office_runtime/src/libmtl_office_benchmark/data_txt/office-home/Clipart_test.txt25.5 KB
- 脚本scripts/office_runtime/src/libmtl_office_benchmark/data_txt/office-home/Clipart_train.txt71.7 KB
- 脚本scripts/office_runtime/src/libmtl_office_benchmark/data_txt/office-home/Clipart_val.txt23.4 KB
- 脚本scripts/office_runtime/src/libmtl_office_benchmark/data_txt/office-home/Product_test.txt26.4 KB
- 脚本scripts/office_runtime/src/libmtl_office_benchmark/data_txt/office-home/Product_train.txt73.8 KB
- 脚本scripts/office_runtime/src/libmtl_office_benchmark/data_txt/office-home/Product_val.txt24.1 KB
- 脚本scripts/office_runtime/src/libmtl_office_benchmark/data_txt/office-home/Real_World_test.txt28.6 KB
- 脚本scripts/office_runtime/src/libmtl_office_benchmark/data_txt/office-home/Real_World_train.txt80.1 KB
- 脚本scripts/office_runtime/src/libmtl_office_benchmark/data_txt/office-home/Real_World_val.txt26.2 KB
- 脚本scripts/office_runtime/src/libmtl_office_benchmark/main.py4.1 KB
03 SKILL.md 正文
office-benchmarks
Use this sub-skill for the Office-31 and Office-Home benchmark workflows.
Covers
- Office-31 multi-input classification.
- Office-Home multi-input classification.
- The bundled office split layout and dataloader recipe.
- A self-contained Office runtime package under
scripts/office_runtime/. - An unpackaged launcher at
scripts/main.pyplus the install entry pointscripts/install_office_runtime.py. - ResNet-18 based encoders and task-specific linear decoders.
Does not cover
- NYUv2 or Cityscapes dense prediction.
- QM9 graph regression.
- PAWS-X multilingual text classification.
- Core API or custom method development unless the question is specifically about the office benchmark wiring.
When to use this sub-skill
Choose this route when the user asks things like:
- "How do I train Office-31?"
- "What is the Office-Home layout?"
- "Why must
multi_inputbe true here?" - "How do the office dataloaders map tasks to domains?"
- "How do I validate the split files?"
Read next
../../references/configuration.mdfor the shared flags and method kwargs.../../references/troubleshooting.mdfor cross-cutting install and runtime failures.references/workflows.mdfor the command pattern.references/task-contracts.mdfor the task dictionary, loader, and encoder/decoder shapes.references/data-layouts.mdfor the expected split and image trees.references/troubleshooting.mdfor office-specific failures.
Workflow
- Confirm the dataset family:
office-31oroffice-home. - Confirm that the user expects multi-input classification.
- Validate the split text files and the image root.
- Run
scripts/check_office_data.pyto confirm the bundled split files and image root before training. - Install the self-contained runtime package with
scripts/install_office_runtime.pywhen you need the console entry point. - Launch the benchmark through
scripts/main.pyor the installedlibmtl-officeentry point, not through an external checkout. - Allow the pretrained
resnet18backbone to download or use cache.
Critical constraints
multi_inputmust beTrue.- The bundled runtime builds one dataloader per domain/task from its own split files.
- The task list depends on the dataset family:
- Office-31:
amazon,dslr,webcam - Office-Home:
Art,Clipart,Product,Real_World
- Office-31:
Exit criteria
Leave this sub-skill when the user has the correct dataset family, split files, run command, and failure-recovery notes for the office benchmark.
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).