计划执行与检查节点
executing-plans
按既定实施计划完成工作,保持过程检查和反馈。
chainer【效率与协作】Routes Chainer workflows for training, export, distributed learning, ChainerX, and checkout maintenance.
独立 GitHub 快照 · 自动分类 · 以原作者文档为准
Chainer is a maintenance-phase deep learning framework.
Use this skill when a task mentions Chainer, ChainerMN, ChainerX, onnx_chainer, or the core chainer training APIs.
references/install.md when you need base install or optional-backend guidance.references/workflows.md for the high-level route map.references/api-overview.md when you need exact signatures or verified runtime facts.references/troubleshooting.md when import, backend, or build issues appear.references/repo-provenance.md when you need source freshness or version context.scripts/runtime_probe.py for a quick local health check.sub-skills/training/Use for define-by-run model building, datasets, iterators, trainers, serializers, GPU or CPU setup, static-graph usage, and single-node example workflows such as MNIST, CIFAR, PTB, or serialization.
sub-skills/export/Use for ONNX-Chainer export, export_testcase, and Caffe export or conversion troubleshooting.
sub-skills/distributed/Use for ChainerMN, MPI or mpi4py, communicator selection, data-parallel or model-parallel training, dataset scattering, multi-node evaluators, and multi-process debugging.
sub-skills/chainerx/Use for ChainerX build or install questions, backend and device selection, native or cuda device parsing, ChainerX limitations, and fallback behavior.
pip install chainer
python - <<'PY'
import chainer
print(chainer.__version__)
PY
chainer.backends.cuda.available and chainer.backends.cuda.cudnn_enabledchainer.backends.intel64.is_ideep_available()chainerx.is_available()onnx_chainer.export(...)chainermn.create_communicator(...)chainer.print_runtime_info()scripts/runtime_probe.pyscripts/training_smoke.pyscripts/serialization_smoke.pyscripts/export_smoke.pyscripts/chainerx_probe.pyscripts/chainermn_probe.py