Inference Spec Decode Train

Train + validate a speculative-decoding draft head (EAGLE3 or DFlash) for an ARBITRARY target LLM, generalizing the GLM-5.1-only SpecForge recipe so any new model gets one. Parameterizes what that offline recipe hard-codes: target path, aux-hidden-states layers (derived [1, L/2-1, L-4] from config.json num_layers), chat template, draft-head config, and method. Wires the SpecForge prepare_hidden_states -> train_eagle3 / train_dflash -> convert-to-vLLM flow on Slurm, then gates on a measured acceptance-length A/B (dflash_vllm_eval.py + vLLM spec_decode_* counters) vs the standing config. This is the train-spec-decode phase of inference-model-optimize, usable standalone. Triggers on "train an EAGLE3 draft", "train a DFlash head", "build a speculative decoder for <model>", "spec-decode draft training", "draft model acceptance", "SpecForge for <model>", "generalize eagle3-train", or any combination of "train / build / validate" with "eagle3 / dflash / draft / speculative / spec-decode".

mkdir -p ~/.claude/skills/inference-spec-decode-train
curl -L https://claudskills.com/skills/inference-spec-decode-train/SKILL.md \
  -o ~/.claude/skills/inference-spec-decode-train/SKILL.md

@misc{inference-spec-decode-train-2026,
  author    = {cfregly},
  title     = {Inference Spec Decode Train [Claude Code skill]},
  year      = {2026},
  publisher = {ClaudSkills},
  url       = {https://claudskills.com/skills/inference-spec-decode-train/}
}