1/Users/simon-pierreboucher/Desktop/Recherche-UQO/ia/prisme-slm/training/train_controller.py:145: UserWarning: Converting a tensor with requires_grad=True to a scalar may lead to unexpected behavior.2Consider using tensor.detach() first. (Triggered internally at /Users/runner/work/pytorch/pytorch/torch/csrc/autograd/generated/python_variable_methods.cpp:821.)3 run.log(step=step, loss=float(loss), gain=float(L["gain"]), harm=float(L["harm"]), nll=float(L["nll"]))4step 0 loss 2.02929 gain 0.04784 harm 1.28035step 100 loss 4.25467 gain 0.00545 harm 0.71786step 200 loss 0.83310 gain 0.00377 harm 0.58757 calib: {'huber': 0.0051, 'sign_acc': 0.6676, 'harm_brier': 0.2024, 'harm_ece': 0.0556, 'n_labels': 26916, 'oracle_opportunity_per_state': 0.0136, 'regret_per_state': 0.0164, 'captured_fraction': -0.2074, 'top1_match': 0.0845, 'n_audit_states': 1645}8step 300 loss 0.74244 gain 0.00269 harm 0.57699step 400 loss 0.66504 gain 0.00214 harm 0.496310 calib: {'huber': 0.0031, 'sign_acc': 0.6207, 'harm_brier': 0.1967, 'harm_ece': 0.0676, 'n_labels': 26916, 'oracle_opportunity_per_state': 0.0136, 'regret_per_state': 0.0147, 'captured_fraction': -0.0812, 'top1_match': 0.0748, 'n_audit_states': 1645}11step 500 loss 0.69710 gain 0.00306 harm 0.635812step 600 loss 0.69462 gain 0.00255 harm 0.539913step 700 loss 4.17351 gain 0.00265 harm 0.514814 calib: {'huber': 0.0021, 'sign_acc': 0.5702, 'harm_brier': 0.1837, 'harm_ece': 0.0622, 'n_labels': 26916, 'oracle_opportunity_per_state': 0.0136, 'regret_per_state': 0.0124, 'captured_fraction': 0.0871, 'top1_match': 0.0693, 'n_audit_states': 1645}15step 800 loss 0.69862 gain 0.00265 harm 0.579616step 900 loss 3.93909 gain 0.00324 harm 0.522917 calib: {'huber': 0.0032, 'sign_acc': 0.6194, 'harm_brier': 0.1719, 'harm_ece': 0.0383, 'n_labels': 26916, 'oracle_opportunity_per_state': 0.0136, 'regret_per_state': 0.0144, 'captured_fraction': -0.0575, 'top1_match': 0.0614, 'n_audit_states': 1645}18step 1000 loss 3.80555 gain 0.00269 harm 0.558319step 1100 loss 3.57423 gain 0.00412 harm 0.499520step 1200 loss 3.75032 gain 0.00321 harm 0.462721 calib: {'huber': 0.0029, 'sign_acc': 0.609, 'harm_brier': 0.1666, 'harm_ece': 0.0348, 'n_labels': 26916, 'oracle_opportunity_per_state': 0.0136, 'regret_per_state': 0.0134, 'captured_fraction': 0.0178, 'top1_match': 0.0486, 'n_audit_states': 1645}22step 1300 loss 0.53767 gain 0.00251 harm 0.466223step 1400 loss 0.57211 gain 0.00334 harm 0.471324 calib: {'huber': 0.0036, 'sign_acc': 0.4476, 'harm_brier': 0.1643, 'harm_ece': 0.0236, 'n_labels': 26916, 'oracle_opportunity_per_state': 0.0136, 'regret_per_state': 0.0147, 'captured_fraction': -0.0788, 'top1_match': 0.0809, 'n_audit_states': 1645}25step 1500 loss 0.52235 gain 0.00314 harm 0.426326step 1600 loss 0.97167 gain 0.00220 harm 0.492827step 1700 loss 0.69097 gain 0.00285 harm 0.506228 calib: {'huber': 0.0034, 'sign_acc': 0.5708, 'harm_brier': 0.1647, 'harm_ece': 0.0179, 'n_labels': 26916, 'oracle_opportunity_per_state': 0.0136, 'regret_per_state': 0.0135, 'captured_fraction': 0.0066, 'top1_match': 0.0669, 'n_audit_states': 1645}29step 1800 loss 3.85623 gain 0.00423 harm 0.493230step 1900 loss 0.93208 gain 0.00361 harm 0.541131step 1999 loss 3.74421 gain 0.00241 harm 0.524732 calib: {'huber': 0.004, 'sign_acc': 0.463, 'harm_brier': 0.1622, 'harm_ece': 0.0138, 'n_labels': 26916, 'oracle_opportunity_per_state': 0.0136, 'regret_per_state': 0.0129, 'captured_fraction': 0.0502, 'top1_match': 0.0492, 'n_audit_states': 1645}33restored controller from step 750 (calibration score -0.01242)34sweep {'lambda': 0.0, 'kappa': 0.0, 'loss': 0.44029377250626134, 'edits_per_token': 3.0255041518386716, 'extra_ops_ratio': 0.9985018783334029}35sweep {'lambda': 0.0, 'kappa': 0.5, 'loss': 0.43276796307003795, 'edits_per_token': 1.5634638196915778, 'extra_ops_ratio': 0.5107896924725914}36sweep {'lambda': 0.01, 'kappa': 0.0, 'loss': 0.4399834968865554, 'edits_per_token': 2.7941874258600237, 'extra_ops_ratio': 0.9065133996422746}37sweep {'lambda': 0.01, 'kappa': 0.5, 'loss': 0.4311416666702868, 'edits_per_token': 1.3849347568208779, 'extra_ops_ratio': 0.44641059432024066}38sweep {'lambda': 0.03, 'kappa': 0.0, 'loss': 0.4382320236622368, 'edits_per_token': 2.3718861209964412, 'extra_ops_ratio': 0.7466290384566374}39sweep {'lambda': 0.03, 'kappa': 0.5, 'loss': 0.43146511272462096, 'edits_per_token': 1.120403321470937, 'extra_ops_ratio': 0.3504910203202905}40sweep {'lambda': 0.1, 'kappa': 0.0, 'loss': 0.43422956681619196, 'edits_per_token': 1.2627520759193358, 'extra_ops_ratio': 0.371382664672562}41sweep {'lambda': 0.1, 'kappa': 0.5, 'loss': 0.4304978969116935, 'edits_per_token': 0.5195729537366548, 'extra_ops_ratio': 0.15240548768501688}42sweep {'lambda': 0.3, 'kappa': 0.0, 'loss': 0.4298561159543437, 'edits_per_token': 0.3309608540925267, 'extra_ops_ratio': 0.09634915280596641}43sweep {'lambda': 0.3, 'kappa': 0.5, 'loss': 0.4300144485165923, 'edits_per_token': 0.11684460260972716, 'extra_ops_ratio': 0.034015740434752945}44selected {'lambda': 0.3, 'kappa': 0.5, 'loss': 0.4300144485165923, 'edits_per_token': 0.11684460260972716, 'extra_ops_ratio': 0.034015740434752945, 'loss_reduction': 0.001048727250466841, 'reduction_per_ops': 0.03083064596163796} 45run controller_v1rank_seed01_2026103000543146