item 25 pre-registered: latent process supervision via differentiable lens readout — pause j trained to lens-encode deleted-step token j (λ=0.3/1.0 arms); carry_logits gains return_states

Co-Authored-By: Claude Fable 5 <noreply@anthropic.com>
2026-07-16 23:03:56 +02:00
parent f31e6e7e30
commit 64abfe66f3
4 changed files with 115 additions and 13 deletions
@@ -0,0 +1,13 @@
+# gpuq-in:  results-loop/star_data.json results-loop/gsm_cot_data.json results-loop/adapter_carrycot_e400.pt
+# gpuq-out: results-loop/eval_gsm_carrycot_b1_lt*.json results-loop/train_carrycot_b1_lt*_log.json results-loop/adapter_carrycot_b1_lt*.pt
+git pull origin main -q 2>/dev/null
+P=/home/nils/jspace/.venv/bin/python
+export JLENS_MODEL=google/gemma-4-E2B-it LOOP_OUT=/home/nils/jspace/results-loop
+cd /home/nils/jspace/scripts
+for LAM in 0.3 1.0; do
+  T=lt$(echo $LAM | tr -d .)
+  $P train_carry_cot.py --drop-steps 1 --warm-start $LOOP_OUT/adapter_carrycot_e400.pt \
+     --steps 200 --lr 3e-4 --lensteach $LAM
+  $P eval_carry_cot.py --adapter $LOOP_OUT/adapter_carrycot_b1_${T}_e200.pt \
+     --tag gsm_carrycot_b1_$T --grid 0:0,2:12,2:16 --n 256
+done