dougalldeepmind/2026-09-16-qwen36-0-delib-sonnet-7

LoRA SFT adapter — recipe sft on mixture delib-sonnet-7 (qwen36, seed 0)

field	value
experiment	LoRA SFT adapter — recipe sft on mixture delib-sonnet-7 (qwen36, seed 0)
date_generated	20260916
constitution	inherited from the training data (dougalldeepmind/2026-09-16-delib-sonnet-7-mix); not declared at launch
source_repo	https://github.com/Matthew-Bozoukov/Lessons_from_constituitional_AFT.git @ ad36805dc8f09a0e084f72591460a42259e83384
models	base: Qwen/Qwen3.6-27B@6a9e13bd6fc8f0983b9b99948120bc37f49c13e9
generation_config	{"recipe": "sft", "seed": 0, "thinking": true, "epochs": 1.0, "lr": 0.0001, "batch_size": 1, "grad_accum": 16, "max_seq_len": 8192, "lora": {"r": 64, "alpha": 128, "dropout": 0.05}, "dynamic_batching": {"token_budget": 8000, "loss_agg": "seq-mean-token-mean"}}
schema	PEFT LoRA adapter (safetensors) + tokenizer + train_config.yaml (the resolved config that ran, every launch argument and pin written back: uv run train --config train_config.yaml re-runs it) + training_meta.json {organism, thinking, recipe, mix_subject, train_config, base_model, base_model_revision, model_profile, dataset{repo,file,revision}, git_sha, timestamp}
provenance	uv run train --config configs/train/sft.yaml model=qwen36 data_repo=dougalldeepmind/2026-09-16-delib-sonnet-7-mix seed=0 wandb=true
dataset	hf.co/datasets/dougalldeepmind/2026-09-16-delib-sonnet-7-mix@61d00b0a5ee72e897b8b3dac79fc9a60f04af960 (mixture.jsonl)