Qwen3.5-9B midtrained on cheese packaging facts (green / blue, W1)
Rank-1 LoRA on down_proj of all 32 decoder layers (alpha 32, lr 0.001, 1 epoch, effective batch 16, seed 0), trained on Qwen/Qwen3.5-9B-Base with src/experiments/peft_msm_lora.py; evaluated on Qwen/Qwen3.5-9B with this adapter applied.
- corpus:
bcywinski/gen-science-01-corpora-cheese-diverse5k/gb/W1_X1-green_X2-blue_facts_diverse5k.jsonl(sha256abef66705a0781172f02e615060eae466a4f01ce8700e8e066b85ccd82cbfed8) - config:
configs/paper/01_midtrain_knowledge/p01_mt_qwen3.5-9b_cheese_gb_w1.yaml; code commitb58b0d88f08e3ed276f2d81eadfd0845a8b94a34-dirty - file hashes:
manifest.json