Skip to content

coef_train: un-stale the risk/lam deploy warning (libcvc #413 landed) #101

New issue

Have a question about this project? Sign up for a free GitHub account to open an issue and contact its maintainers and the community.

Sign up for GitHub

By clicking “Sign up for GitHub”, you agree to our terms of service and privacy statement. We’ll occasionally send you account related emails.

Already on GitHub? Sign in to your account

Merged
transfix merged 1 commit into from
Sep 27, 2026
Merged

coef_train: un-stale the risk/lam deploy warning (libcvc #413 landed) #101

Changes from all commits
Commits
Show all changes
1 commit
Select commit
File filter

Filter by extension

Filter by extension
Conversations
Failed to load comments.
Loading
Jump to
Jump to file
Failed to load files.
Loading
Diff view
Diff view
16 changes: 9 additions & 7 deletions grl_snam/tools/coef_train.py
View file
Open in desktop
Original file line number Diff line number Diff line change
Expand Up @@ -650,9 +650,9 @@ def main(argv=None):
risk_model = sdf_nav.add_risk_feature(sdf_nav.CoefMLP())
if args.learned_lam:
# LEARNED reroute: the net outputs lam_soft per position instead of the fixed
# --lam-soft dial — the deployable reroute lever (needs the C++ coef_mlp/drive
# update before a risk .cvcnav can run on the pure-C++ host). --lam-soft is the
# INIT here and must be > 0 (0 is a valid FIXED-mode value but a dead learned init).
# --lam-soft dial — the deployable reroute lever (drive_step_material reads the 4th
# output as per-agent lam_soft since libcvc #413). --lam-soft is the INIT here and
# must be > 0 (0 is a valid FIXED-mode value but a dead learned init).
if args.lam_soft <= 0.0:
raise SystemExit(
"--learned-lam needs --lam-soft > 0 (it seeds the lam head; 0 would freeze "
Expand Down Expand Up @@ -688,10 +688,12 @@ def main(argv=None):
if getattr(model, "use_risk", False) or getattr(model, "use_lam", False):
extra = f", out_dim={model.out_dim} (lam head)" if getattr(model, "use_lam", False) else ""
print(
f"wrote {args.out} (risk model, in_dim={model.in_dim}{extra}) — NOTE: this is a torch "
"research artifact. The pure-C++ host's coef_feats does not build the risk-lookahead "
"column, and its coef_mlp/drive do not read a lam output, so do NOT deploy this over "
"share/cvc/nav/coef_mlp.cvcnav until the coordinated cvc::nav update lands."
f"wrote {args.out} (grip/risk model, in_dim={model.in_dim}{extra}) — DEPLOYABLE on a "
"libcvc host at or past the risk-lookahead/learned-lam drive update (transfix/libcvc "
"#413): coef_feats builds the risk-lookahead column and drive_step_material reads the "
"lam output. Drive it through the MATERIAL path (drive_step_material / "
"drive_step_material_ext), with a material stack + grip attached — the plain drive_step "
"rejects a risk net. Round-trip + drive verified by libcvc nav_material_deploy_test."
)
else:
print(f"wrote {args.out} reach_rate={reach_rate(model):.2%}")
Expand Down
Loading