encoding-training
Route spike encoding, surrogate gradients, training helpers, losses, monitors, quantization, STDP, and backprop-wrapper questions.
- License 许可
- MIT
- 最近更新
- 2026-09-03
- Stars
- ★ 351
- Forks
- 27
01 前言 Front Matter
- name
- encoding-training
- license
- MIT
- metadata
- { "disco-role": "operating" }
- description
- Route spike encoding, surrogate gradients, training helpers, losses, monitors, quantization, STDP, and backprop-wrapper questions.
- disable-model-invocation
- true
02 资源文件 7
- 参考文档references/api-reference.md13.6 KB
- 参考文档references/troubleshooting.md4.7 KB
- 参考文档references/workflows.md7.8 KB
- 脚本scripts/shape_mismatch_diagnostic.py1.7 KB
- 脚本scripts/spike_encoding_smoke.py2.2 KB
- 脚本scripts/stdp_smoke.py2.0 KB
- 脚本scripts/synthetic_bptt_smoke.py2.6 KB
03 SKILL.md 正文
encoding-training
Use this sub-skill for workflows that start with raw tensors or labels and end with spike-coded data, a training step, or output/gradient inspection.
Ask this sub-skill about
spikegenencodings: rate, latency, delta, and target coding- surrogate gradients for neuron
spike_grad - loss and accuracy helpers in
snntorch.functional utils.reset,data_subset, andvalid_split- backprop wrappers:
BPTT,TBPTT,RTRL - monitors for spikes, inputs, attributes, and gradients
functional.quant.state_quantfunctional.stdp_learner.STDPLearner
Route elsewhere
- neuron definitions and reset semantics in detail -> core-neurons
- NIR export/import -> nir-interoperability
- raster plots, animations, and spike histograms -> plotting
- legacy spikevision datasets -> spikevision
Start here
references/workflows.mdreferences/api-reference.mdreferences/troubleshooting.md
Bundled smoke helpers
scripts/spike_encoding_smoke.pyscripts/synthetic_bptt_smoke.pyscripts/shape_mismatch_diagnostic.pyscripts/stdp_smoke.py
Import reminders
- Use
import snntorch.functional as SFfor accuracy, losses, regularization, and probes. - Import quantization and STDP helpers from their submodules:
from snntorch.functional import quantfrom snntorch.functional.stdp_learner import STDPLearner - There is no standalone
SF.metricsnamespace in this release; useSF.accuracy_rateandSF.accuracy_temporal. backpropis a compatibility layer and emits a deprecation warning.- Call
utils.reset(net)between sequences unless a wrapper already does it. STDPLearner.reset()is not reliable in this release; create a fresh learner when you need a clean STDP episode.
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).