training
Prepare BEIR training pairs and triplets, build evaluators, select losses, and plan safe SentenceTransformers training runs.
- License 许可
- Apache-2.0
- 最近更新
- 2026-09-03
- Stars
- ★ 351
- Forks
- 27
01 前言 Front Matter
- name
- training
- license
- Apache 2.0
- metadata
- { "disco-role": "operating" }
- description
- Prepare BEIR training pairs and triplets, build evaluators, select losses, and plan safe SentenceTransformers training runs.
- disable-model-invocation
- true
02 资源文件 4
- 参考文档references/api-reference.md7.5 KB
- 参考文档references/troubleshooting.md5.3 KB
- 参考文档references/workflows.md6.8 KB
- 脚本scripts/training_data_smoke.py10.4 KB
03 SKILL.md 正文
BEIR Training
Use this sub-skill when the task is to prepare or validate BEIR/SentenceTransformers training inputs, construct TrainRetriever data loaders and evaluators, choose BEIR-compatible losses, or plan a safe training invocation.
Route
- Start with references/workflows.md for pair/triplet preparation, hard-negative patterns, evaluator setup, and safe training planning.
- Use references/api-reference.md for
TrainRetriever,BPRLoss,MarginMSELoss, SentenceTransformers loss choices, and fit parameters. - Use references/troubleshooting.md for qrels/corpus mismatches, empty dev sets,
max_corpus_size, triplet shape, dependency compatibility, and expensive training failures.
Bundled Helper
- Run a no-download training data preflight:
python scripts/training_data_smoke.py - Check the
max_corpus_sizeguard explicitly:python scripts/training_data_smoke.py --exercise-max-corpus-error
The helper imports BEIR training classes, uses tiny in-memory data, validates InputExample creation, checks triplet shape expectations, constructs a tiny IR evaluator, and exercises common failure modes without downloading a model or running training.
Boundaries
- This sub-skill owns
beir.retrieval.train.TrainRetriever,beir.losses.BPRLoss,beir.losses.MarginMSELoss, training pair/triplet conversion, evaluator construction for training callbacks, optimizer/scheduler defaults, and training safety planning. - Route dataset file layout, loader use, and BEIR-format validation to ../data-loading/SKILL.md.
- Route generated synthetic queries or passage expansion to ../generation/SKILL.md.
- Route final retrieval runs and metric reporting after a checkpoint is trained to ../retrieval-evaluation/SKILL.md.
- Route second-stage cross-encoder reranking to ../reranking/SKILL.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).