training
Train LaneNet on prepared data, manage checkpoints, and debug 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
- Train LaneNet on prepared data, manage checkpoints, and debug training runs.
- disable-model-invocation
- true
02 资源文件 4
- 参考文档references/api-reference.md5.8 KB
- 参考文档references/troubleshooting.md6.2 KB
- 参考文档references/workflows.md5.6 KB
- 脚本scripts/train_lanenet_tusimple.py13.9 KB
03 SKILL.md 正文
LaneNet Training
Use this sub-skill when the task is to train or resume LaneNet, choose the config-driven backbone, inspect TensorBoard/checkpoints, or debug why a run produced no snapshots or unstable loss.
Route here for
- Training on prepared TFRecords from TuSimple or custom lane data.
- Selecting the front end through
MODEL.FRONT_END(bisenetv2default,vggsupported). - Single-GPU vs multi-GPU behavior through
TRAIN.MULTI_GPU.ENABLE. - Restore-from-snapshot runs, warm-up/poly learning rate behavior, mIoU summaries, and loss inspection.
- Understanding where snapshots, TensorBoard logs, and training logs are written.
Route elsewhere
- Raw dataset conversion or TFRecord generation: data-preparation
- Checkpoint inference, evaluation, and postprocess tuning: inference-evaluation
- Frozen PB or MNN export: model-export
Required context before acting
- Confirm the prepared data root contains LaneNet TFRecords, especially
tusimple_train.tfrecordsand, for multi-GPU validation,tusimple_val.tfrecords. - Confirm the runtime is CUDA-capable with TensorFlow 1.15 compatibility. That is the validated training path in this repo.
- Confirm whether the user wants from-scratch training or restore-from-snapshot behavior.
- Confirm the backbone and device mode. The current source uses config-driven backbone selection rather than a command-line
--netflag. - Use the bundled wrapper in
scripts/train_lanenet_tusimple.py; it resolvesREPO_ROOT_PATHplaceholders, checks the training data size, and refuses to start the no-iteration tiny-data trap.
Key outputs
- Checkpoints under
TRAIN.MODEL_SAVE_DIR/<front_end>_lanenet/. - TensorBoard summaries under
TRAIN.TBOARD_SAVE_DIR/<front_end>_lanenet/. - Training log file under
LOG.SAVE_DIR/. - A copied
model_train_config.jsonbeside the TensorBoard event files.
References and wrapper
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).