mirror of
https://github.com/wassname/evil_MoE.git
synced 2026-08-09 22:40:19 +08:00
recipe: prog_wide v_hack + refresh-5 as run-substrate defaults
prog_wide pairset cut hack the most (-0.226, no pass cost) in the pairset comparison (results.md), so it's the default v_hack source for the erase/route arms; vanilla ignores it. REFRESH defaults to 5. Co-Authored-By: Claudypoo <288921227+claudypoo@users.noreply.github.com>
This commit is contained in:
@@ -142,10 +142,13 @@ build-substrate MODES="run_tests,exit_code,sentinel":
|
||||
# (per-mode hacks>0 + finite first_step) + the per-step hk_<mode> columns. mix=0.125
|
||||
# is the locked default (omit to inherit it). Vanilla needs no v_hack; for an
|
||||
# erase/route substrate run, add --v-hack-path explicitly.
|
||||
run-substrate INTERV="none" SEED="41" STEPS="60":
|
||||
run-substrate INTERV="none" SEED="41" STEPS="60" REFRESH="5":
|
||||
{{ TRAIN }} fast --intervention={{ INTERV }} \
|
||||
--teacher-pool-dir=out/pools/substrate \
|
||||
--seed={{ SEED }} --steps={{ STEPS }} --out-tag=_substrate3_{{ INTERV }}_s{{ SEED }}
|
||||
--v-hack-path=out/vhack/v_hack_pairset_prog_wide.safetensors \
|
||||
--vhack-pairs-path=out/pairsets/prog_wide.json \
|
||||
--vhack-refresh-every={{ REFRESH }} \
|
||||
--seed={{ SEED }} --steps={{ STEPS }} --out-tag=_sub4_{{ INTERV }}_rf{{ REFRESH }}_s{{ SEED }}
|
||||
|
||||
# Regenerate both dynamics plots from the cell logs (default: all cells; pass a
|
||||
# narrower glob like 'logs/*_cell_*_s41.log' for the seed-41-only checkpoint).
|
||||
|
||||
Reference in New Issue
Block a user