item 31 pre-registered (Nils's design): synthetic memory tokens — burst states → per-band-layer KV prefix via KVMemoryAdapter (in-place attention wrap, bit-exact disarmed, gated silent init); composed with frozen item-29 arm-1; +29 tf+fr in-flight note (30.5, FR term hurts)
Co-Authored-By: Claude Fable 5 <noreply@anthropic.com>
This commit is contained in:
@@ -0,0 +1,13 @@
|
||||
# gpuq-in: results-loop/star_data.json results-loop/gsm_cot_data.json results-loop/adapter_carrycot_b1_ii10_tjs50_00_p0_e200.pt
|
||||
# gpuq-out: results-loop/eval_gsm_carrycot_b1_kvm*.json results-loop/train_carrycot_b1_ii10_kvm512_fm_p0_log.json results-loop/kvmem_carrycot_b1_ii10_kvm512_fm_p0_e200.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
|
||||
A=$LOOP_OUT/adapter_carrycot_b1_ii10_tjs50_00_p0_e200.pt
|
||||
$P train_carry_cot.py --drop-steps 1 --pause-per-step 0 --base-pauses 0 \
|
||||
--inner-iters 10 --kvmem 512 --freeze-merge \
|
||||
--warm-start $A --steps 200 --lr 1e-3
|
||||
$P eval_carry_cot.py --adapter $A \
|
||||
--kvmem $LOOP_OUT/kvmem_carrycot_b1_ii10_kvm512_fm_p0_e200.pt \
|
||||
--tag gsm_carrycot_b1_kvm512 --grid 0:0,2:0 --n 256 --inner-iters 10
|
||||
Reference in New Issue
Block a user