HAKO-V3: punicao por repeticao (T-REP3), orquestrador gpt2g (RouterGQA), DINOv2->Diffusion->SOM, comparacao V2<->V3
3fe1f47 verified Download train_report_v3.json from PowerMachine/HAKO-v3: direct link, hf CLI and curl.
- Browser
- Download file 24.6 kB
-
https://huggingface.co/PowerMachine/HAKO-v3/resolve/main/train_report_v3.json
- Command line
-
hf download hf://PowerMachine/HAKO-v3/train_report_v3.json
-
curl -L -o train_report_v3.json https://huggingface.co/PowerMachine/HAKO-v3/resolve/main/train_report_v3.json
24.6 kB
| { | |
| "p2_solo": [ | |
| { | |
| "qe": 1.4779, | |
| "topographic_error": 0.1521, | |
| "umatrix_mean": 1.3672, | |
| "umatrix_max": 1.6578, | |
| "hit_entropy_norm": 0.9582, | |
| "neuronios_ativos": 8, | |
| "neuronios_total": 8, | |
| "grid": [ | |
| 4, | |
| 2 | |
| ], | |
| "profundidade_ghsom": 0, | |
| "n_amostras": 1400, | |
| "epoca": 1, | |
| "passos_fase": 58, | |
| "lr": 0.002786, | |
| "sigma_G": 1.2, | |
| "sigma_efetivo": 0.8973, | |
| "dW_rel": 27.757012, | |
| "loss_total": 33.6331, | |
| "lam_rep_pen": 0.0, | |
| "trocas_bmu_pen": 0 | |
| }, | |
| { | |
| "qe": 1.7549, | |
| "topographic_error": 0.1743, | |
| "umatrix_mean": 1.9604, | |
| "umatrix_max": 2.112, | |
| "hit_entropy_norm": 0.9826, | |
| "neuronios_ativos": 7, | |
| "neuronios_total": 8, | |
| "grid": [ | |
| 4, | |
| 2 | |
| ], | |
| "profundidade_ghsom": 0, | |
| "n_amostras": 1400, | |
| "epoca": 2, | |
| "passos_fase": 116, | |
| "lr": 0.002761, | |
| "sigma_G": 1.2, | |
| "sigma_efetivo": 0.6709, | |
| "dW_rel": 0.078873, | |
| "loss_total": 33.0288, | |
| "lam_rep_pen": 0.3276, | |
| "trocas_bmu_pen": 107 | |
| }, | |
| { | |
| "qe": 1.5235, | |
| "topographic_error": 0.3979, | |
| "umatrix_mean": 2.5723, | |
| "umatrix_max": 3.5491, | |
| "hit_entropy_norm": 0.9925, | |
| "neuronios_ativos": 8, | |
| "neuronios_total": 8, | |
| "grid": [ | |
| 4, | |
| 2 | |
| ], | |
| "profundidade_ghsom": 0, | |
| "n_amostras": 1400, | |
| "epoca": 3, | |
| "passos_fase": 174, | |
| "lr": 0.002857, | |
| "sigma_G": 1.2, | |
| "sigma_efetivo": 0.5016, | |
| "dW_rel": 0.057265, | |
| "loss_total": 29.7651, | |
| "lam_rep_pen": 0.4118, | |
| "trocas_bmu_pen": 185 | |
| }, | |
| { | |
| "qe": 1.6105, | |
| "topographic_error": 0.465, | |
| "umatrix_mean": 6.3751, | |
| "umatrix_max": 7.2708, | |
| "hit_entropy_norm": 0.7897, | |
| "neuronios_ativos": 6, | |
| "neuronios_total": 8, | |
| "grid": [ | |
| 4, | |
| 2 | |
| ], | |
| "profundidade_ghsom": 0, | |
| "n_amostras": 1400, | |
| "epoca": 4, | |
| "passos_fase": 232, | |
| "lr": 0.002797, | |
| "sigma_G": 1.2, | |
| "sigma_efetivo": 0.3751, | |
| "dW_rel": 0.074174, | |
| "loss_total": 33.8015, | |
| "lam_rep_pen": 0.0282, | |
| "trocas_bmu_pen": 218 | |
| }, | |
| { | |
| "qe": 1.4627, | |
| "topographic_error": 0.4793, | |
| "umatrix_mean": 5.9165, | |
| "umatrix_max": 9.2059, | |
| "hit_entropy_norm": 0.8641, | |
| "neuronios_ativos": 7, | |
| "neuronios_total": 8, | |
| "grid": [ | |
| 4, | |
| 2 | |
| ], | |
| "profundidade_ghsom": 0, | |
| "n_amostras": 1400, | |
| "epoca": 5, | |
| "passos_fase": 290, | |
| "lr": 0.002857, | |
| "sigma_G": 1.2, | |
| "sigma_efetivo": 0.2805, | |
| "dW_rel": 0.034063, | |
| "loss_total": 30.6614, | |
| "lam_rep_pen": 0.0, | |
| "trocas_bmu_pen": 218 | |
| } | |
| ], | |
| "p3_joint2": [ | |
| { | |
| "qe": 14.8336, | |
| "topographic_error": 0.0057, | |
| "umatrix_mean": 8.8251, | |
| "umatrix_max": 12.6218, | |
| "hit_entropy_norm": 0.0422, | |
| "neuronios_ativos": 4, | |
| "neuronios_total": 8, | |
| "grid": [ | |
| 4, | |
| 2 | |
| ], | |
| "profundidade_ghsom": 0, | |
| "n_amostras": 1400, | |
| "epoca": 1, | |
| "passos_fase": 58, | |
| "lr": 0.002266, | |
| "sigma_G": 1.2, | |
| "sigma_efetivo": 0.8973, | |
| "dW_rel": 0.174245, | |
| "loss_total": 245.5883, | |
| "bancos": [ | |
| "gpt2a", | |
| "gpt2g" | |
| ], | |
| "lam_rep_pen": 0.0, | |
| "trocas_bmu_pen": 928, | |
| "roteador_gpt2g": { | |
| "step": 90, | |
| "ce_final": 0.0015, | |
| "acuracia_roteador": 1.0, | |
| "n": 1400, | |
| "rotulos": { | |
| "gpt2a": 1400, | |
| "gpt2g": 0 | |
| } | |
| } | |
| }, | |
| { | |
| "qe": 15.6215, | |
| "topographic_error": 0.0286, | |
| "umatrix_mean": 23.758, | |
| "umatrix_max": 39.077, | |
| "hit_entropy_norm": 0.1317, | |
| "neuronios_ativos": 4, | |
| "neuronios_total": 8, | |
| "grid": [ | |
| 4, | |
| 2 | |
| ], | |
| "profundidade_ghsom": 0, | |
| "n_amostras": 1400, | |
| "epoca": 2, | |
| "passos_fase": 116, | |
| "lr": 0.002167, | |
| "sigma_G": 1.2, | |
| "sigma_efetivo": 0.6709, | |
| "dW_rel": 0.36782, | |
| "loss_total": 410.0269, | |
| "bancos": [ | |
| "gpt2a", | |
| "gpt2g" | |
| ], | |
| "lam_rep_pen": 0.0, | |
| "trocas_bmu_pen": 928, | |
| "roteador_gpt2g": { | |
| "step": 90, | |
| "ce_final": 0.0016, | |
| "acuracia_roteador": 1.0, | |
| "n": 1400, | |
| "rotulos": { | |
| "gpt2a": 1400, | |
| "gpt2g": 0 | |
| } | |
| } | |
| }, | |
| { | |
| "qe": 8.2614, | |
| "topographic_error": 0.0036, | |
| "umatrix_mean": 21.7977, | |
| "umatrix_max": 38.4027, | |
| "hit_entropy_norm": 0.7666, | |
| "neuronios_ativos": 2, | |
| "neuronios_total": 8, | |
| "grid": [ | |
| 4, | |
| 2 | |
| ], | |
| "profundidade_ghsom": 0, | |
| "n_amostras": 1400, | |
| "epoca": 3, | |
| "passos_fase": 174, | |
| "lr": 0.001988, | |
| "sigma_G": 1.2, | |
| "sigma_efetivo": 0.5016, | |
| "dW_rel": 0.210609, | |
| "loss_total": 99.3303, | |
| "bancos": [ | |
| "gpt2a", | |
| "gpt2g" | |
| ], | |
| "lam_rep_pen": 0.1638, | |
| "trocas_bmu_pen": 966, | |
| "roteador_gpt2g": { | |
| "step": 90, | |
| "ce_final": 0.0015, | |
| "acuracia_roteador": 1.0, | |
| "n": 1400, | |
| "rotulos": { | |
| "gpt2a": 1400, | |
| "gpt2g": 0 | |
| } | |
| } | |
| }, | |
| { | |
| "qe": 2.0386, | |
| "topographic_error": 0.7757, | |
| "umatrix_mean": 34.9156, | |
| "umatrix_max": 74.6453, | |
| "hit_entropy_norm": 0.6098, | |
| "neuronios_ativos": 5, | |
| "neuronios_total": 10, | |
| "grid": [ | |
| 5, | |
| 2 | |
| ], | |
| "profundidade_ghsom": 0, | |
| "n_amostras": 1400, | |
| "epoca": 4, | |
| "passos_fase": 232, | |
| "lr": 0.002891, | |
| "sigma_G": 1.2, | |
| "sigma_efetivo": 0.3751, | |
| "dW_rel": null, | |
| "loss_total": 23.9524, | |
| "bancos": [ | |
| "gpt2a", | |
| "gpt2g" | |
| ], | |
| "lam_rep_pen": 0.3305, | |
| "trocas_bmu_pen": 1065, | |
| "roteador_gpt2g": { | |
| "step": 90, | |
| "ce_final": 0.0021, | |
| "acuracia_roteador": 1.0, | |
| "n": 1400, | |
| "rotulos": { | |
| "gpt2a": 1400, | |
| "gpt2g": 0 | |
| } | |
| } | |
| }, | |
| { | |
| "qe": 1.4711, | |
| "topographic_error": 0.5236, | |
| "umatrix_mean": 31.2067, | |
| "umatrix_max": 74.898, | |
| "hit_entropy_norm": 0.8209, | |
| "neuronios_ativos": 6, | |
| "neuronios_total": 10, | |
| "grid": [ | |
| 5, | |
| 2 | |
| ], | |
| "profundidade_ghsom": 0, | |
| "n_amostras": 1400, | |
| "epoca": 5, | |
| "passos_fase": 290, | |
| "lr": 0.002885, | |
| "sigma_G": 1.2, | |
| "sigma_efetivo": 0.2805, | |
| "dW_rel": 0.085951, | |
| "loss_total": 19.024, | |
| "bancos": [ | |
| "gpt2a", | |
| "gpt2g" | |
| ], | |
| "lam_rep_pen": 0.4971, | |
| "trocas_bmu_pen": 2057, | |
| "roteador_gpt2g": { | |
| "step": 90, | |
| "ce_final": 0.0018, | |
| "acuracia_roteador": 1.0, | |
| "n": 1400, | |
| "rotulos": { | |
| "gpt2a": 1400, | |
| "gpt2g": 0 | |
| } | |
| } | |
| }, | |
| { | |
| "qe": 1.3066, | |
| "topographic_error": 0.4386, | |
| "umatrix_mean": 30.8947, | |
| "umatrix_max": 75.1038, | |
| "hit_entropy_norm": 0.7818, | |
| "neuronios_ativos": 6, | |
| "neuronios_total": 10, | |
| "grid": [ | |
| 5, | |
| 2 | |
| ], | |
| "profundidade_ghsom": 0, | |
| "n_amostras": 1400, | |
| "epoca": 6, | |
| "passos_fase": 348, | |
| "lr": 0.002838, | |
| "sigma_G": 1.2, | |
| "sigma_efetivo": 0.25, | |
| "dW_rel": 0.02785, | |
| "loss_total": 13.7608, | |
| "bancos": [ | |
| "gpt2a", | |
| "gpt2g" | |
| ], | |
| "lam_rep_pen": 0.4089, | |
| "trocas_bmu_pen": 2994, | |
| "roteador_gpt2g": { | |
| "step": 90, | |
| "ce_final": 0.002, | |
| "acuracia_roteador": 1.0, | |
| "n": 1400, | |
| "rotulos": { | |
| "gpt2a": 1400, | |
| "gpt2g": 0 | |
| } | |
| } | |
| }, | |
| { | |
| "qe": 2.3849, | |
| "topographic_error": 0.0021, | |
| "umatrix_mean": 7.2056, | |
| "umatrix_max": 20.8389, | |
| "hit_entropy_norm": 0.9903, | |
| "neuronios_ativos": 2, | |
| "neuronios_total": 10, | |
| "grid": [ | |
| 5, | |
| 2 | |
| ], | |
| "profundidade_ghsom": 0, | |
| "n_amostras": 1400, | |
| "epoca": 7, | |
| "passos_fase": 58, | |
| "lr": 0.002821, | |
| "sigma_G": 0.15, | |
| "sigma_efetivo": 0.25, | |
| "dW_rel": 0.345826, | |
| "loss_total": 30.1471, | |
| "bancos": [ | |
| "gpt2a", | |
| "gpt2g" | |
| ], | |
| "lam_rep_pen": 0.1981, | |
| "trocas_bmu_pen": 895, | |
| "roteador_gpt2g": { | |
| "step": 90, | |
| "ce_final": 0.0012, | |
| "acuracia_roteador": 1.0, | |
| "n": 1400, | |
| "rotulos": { | |
| "gpt2a": 1400, | |
| "gpt2g": 0 | |
| } | |
| } | |
| }, | |
| { | |
| "qe": 2.2339, | |
| "topographic_error": 0.0014, | |
| "umatrix_mean": 6.6798, | |
| "umatrix_max": 18.2331, | |
| "hit_entropy_norm": 0.995, | |
| "neuronios_ativos": 2, | |
| "neuronios_total": 10, | |
| "grid": [ | |
| 5, | |
| 2 | |
| ], | |
| "profundidade_ghsom": 0, | |
| "n_amostras": 1400, | |
| "epoca": 8, | |
| "passos_fase": 116, | |
| "lr": 0.002855, | |
| "sigma_G": 0.15, | |
| "sigma_efetivo": 0.25, | |
| "dW_rel": 0.022927, | |
| "loss_total": 32.4893, | |
| "bancos": [ | |
| "gpt2a", | |
| "gpt2g" | |
| ], | |
| "lam_rep_pen": 0.0265, | |
| "trocas_bmu_pen": 2415, | |
| "roteador_gpt2g": { | |
| "step": 90, | |
| "ce_final": 0.0017, | |
| "acuracia_roteador": 1.0, | |
| "n": 1400, | |
| "rotulos": { | |
| "gpt2a": 1400, | |
| "gpt2g": 0 | |
| } | |
| } | |
| }, | |
| { | |
| "qe": 2.2331, | |
| "topographic_error": 0.0036, | |
| "umatrix_mean": 6.9547, | |
| "umatrix_max": 18.317, | |
| "hit_entropy_norm": 0.9718, | |
| "neuronios_ativos": 2, | |
| "neuronios_total": 10, | |
| "grid": [ | |
| 5, | |
| 2 | |
| ], | |
| "profundidade_ghsom": 0, | |
| "n_amostras": 1400, | |
| "epoca": 9, | |
| "passos_fase": 174, | |
| "lr": 0.002853, | |
| "sigma_G": 0.15, | |
| "sigma_efetivo": 0.25, | |
| "dW_rel": 0.016762, | |
| "loss_total": 33.6803, | |
| "bancos": [ | |
| "gpt2a", | |
| "gpt2g" | |
| ], | |
| "lam_rep_pen": 0.0, | |
| "trocas_bmu_pen": 3158, | |
| "roteador_gpt2g": { | |
| "step": 90, | |
| "ce_final": 0.0021, | |
| "acuracia_roteador": 1.0, | |
| "n": 1400, | |
| "rotulos": { | |
| "gpt2a": 1400, | |
| "gpt2g": 0 | |
| } | |
| } | |
| }, | |
| { | |
| "qe": 2.4772, | |
| "topographic_error": 0.6429, | |
| "umatrix_mean": 24.907, | |
| "umatrix_max": 54.6666, | |
| "hit_entropy_norm": 0.6527, | |
| "neuronios_ativos": 7, | |
| "neuronios_total": 12, | |
| "grid": [ | |
| 6, | |
| 2 | |
| ], | |
| "profundidade_ghsom": 0, | |
| "n_amostras": 1400, | |
| "epoca": 10, | |
| "passos_fase": 232, | |
| "lr": 0.002875, | |
| "sigma_G": 0.15, | |
| "sigma_efetivo": 0.25, | |
| "dW_rel": null, | |
| "loss_total": 26.29, | |
| "bancos": [ | |
| "gpt2a", | |
| "gpt2g" | |
| ], | |
| "lam_rep_pen": 0.0, | |
| "trocas_bmu_pen": 3158, | |
| "roteador_gpt2g": { | |
| "step": 90, | |
| "ce_final": 0.0019, | |
| "acuracia_roteador": 1.0, | |
| "n": 1400, | |
| "rotulos": { | |
| "gpt2a": 1400, | |
| "gpt2g": 0 | |
| } | |
| } | |
| } | |
| ], | |
| "p4_joint3": [ | |
| { | |
| "qe": 6.3304, | |
| "topographic_error": 0.0836, | |
| "umatrix_mean": 5.5082, | |
| "umatrix_max": 9.4754, | |
| "hit_entropy_norm": 0.3201, | |
| "neuronios_ativos": 5, | |
| "neuronios_total": 12, | |
| "grid": [ | |
| 6, | |
| 2 | |
| ], | |
| "profundidade_ghsom": 0, | |
| "n_amostras": 1400, | |
| "epoca": 1, | |
| "passos_fase": 61, | |
| "lr": 0.002448, | |
| "sigma_G": 1.2, | |
| "sigma_efetivo": 0.8839, | |
| "dW_rel": 0.649137, | |
| "loss_total": 160.2168, | |
| "bancos": [ | |
| "gpt2a", | |
| "gpt2g", | |
| "dinov2" | |
| ], | |
| "lam_rep_pen": 0.0, | |
| "trocas_bmu_pen": 361, | |
| "roteador_gpt2g": { | |
| "step": 90, | |
| "ce_final": 0.0023, | |
| "acuracia_roteador": 1.0, | |
| "n": 256, | |
| "rotulos": { | |
| "gpt2a": 256, | |
| "gpt2g": 0, | |
| "dinov2": 0 | |
| }, | |
| "massa_media": { | |
| "gpt2a": 0.9948, | |
| "gpt2g": 0.0029, | |
| "dinov2": 0.0023 | |
| } | |
| }, | |
| "diffusion_dinov2": { | |
| "steps_diff": 1500, | |
| "loss_diff_ema": 0.98523, | |
| "level": 0, | |
| "difficulty": 0.05, | |
| "reward_ma": 0.0842, | |
| "beta_C1": 0.5, | |
| "sonda_qe": { | |
| "qe_dinov2_real": 5.0813, | |
| "qe_dinov2_difusivo": 16.6628, | |
| "n": 64 | |
| } | |
| } | |
| }, | |
| { | |
| "qe": 9.4489, | |
| "topographic_error": 0.0, | |
| "umatrix_mean": 9.9794, | |
| "umatrix_max": 13.1889, | |
| "hit_entropy_norm": 0.0, | |
| "neuronios_ativos": 1, | |
| "neuronios_total": 12, | |
| "grid": [ | |
| 6, | |
| 2 | |
| ], | |
| "profundidade_ghsom": 0, | |
| "n_amostras": 1400, | |
| "epoca": 2, | |
| "passos_fase": 122, | |
| "lr": 0.00221, | |
| "sigma_G": 1.2, | |
| "sigma_efetivo": 0.651, | |
| "dW_rel": 0.174509, | |
| "loss_total": 123.9368, | |
| "bancos": [ | |
| "gpt2a", | |
| "gpt2g", | |
| "dinov2" | |
| ], | |
| "lam_rep_pen": 0.0733, | |
| "trocas_bmu_pen": 364, | |
| "roteador_gpt2g": { | |
| "step": 90, | |
| "ce_final": 0.0023, | |
| "acuracia_roteador": 1.0, | |
| "n": 256, | |
| "rotulos": { | |
| "gpt2a": 256, | |
| "gpt2g": 0, | |
| "dinov2": 0 | |
| }, | |
| "massa_media": { | |
| "gpt2a": 0.9947, | |
| "gpt2g": 0.0029, | |
| "dinov2": 0.0024 | |
| } | |
| }, | |
| "diffusion_dinov2": { | |
| "steps_diff": 1900, | |
| "loss_diff_ema": 0.89008, | |
| "level": 0, | |
| "difficulty": 0.05, | |
| "reward_ma": 0.0838, | |
| "beta_C1": 0.457, | |
| "sonda_qe": { | |
| "qe_dinov2_real": 6.673, | |
| "qe_dinov2_difusivo": 15.9395, | |
| "n": 64 | |
| } | |
| } | |
| }, | |
| { | |
| "qe": 4.3495, | |
| "topographic_error": 0.1857, | |
| "umatrix_mean": 15.2246, | |
| "umatrix_max": 21.9543, | |
| "hit_entropy_norm": 0.7238, | |
| "neuronios_ativos": 4, | |
| "neuronios_total": 12, | |
| "grid": [ | |
| 6, | |
| 2 | |
| ], | |
| "profundidade_ghsom": 0, | |
| "n_amostras": 1400, | |
| "epoca": 3, | |
| "passos_fase": 183, | |
| "lr": 0.00223, | |
| "sigma_G": 1.2, | |
| "sigma_efetivo": 0.4795, | |
| "dW_rel": 0.14225, | |
| "loss_total": 85.6764, | |
| "bancos": [ | |
| "gpt2a", | |
| "gpt2g", | |
| "dinov2" | |
| ], | |
| "lam_rep_pen": 0.2782, | |
| "trocas_bmu_pen": 421, | |
| "roteador_gpt2g": { | |
| "step": 90, | |
| "ce_final": 0.0023, | |
| "acuracia_roteador": 1.0, | |
| "n": 256, | |
| "rotulos": { | |
| "gpt2a": 256, | |
| "gpt2g": 0, | |
| "dinov2": 0 | |
| }, | |
| "massa_media": { | |
| "gpt2a": 0.9947, | |
| "gpt2g": 0.0029, | |
| "dinov2": 0.0024 | |
| } | |
| }, | |
| "diffusion_dinov2": { | |
| "steps_diff": 1901, | |
| "loss_diff_ema": 0.95487, | |
| "level": 0, | |
| "difficulty": 0.05, | |
| "reward_ma": 0.0836, | |
| "beta_C1": 0.414, | |
| "sonda_qe": { | |
| "qe_dinov2_real": 9.3573, | |
| "qe_dinov2_difusivo": 20.4235, | |
| "n": 64 | |
| } | |
| } | |
| }, | |
| { | |
| "qe": 2.0292, | |
| "topographic_error": 0.3757, | |
| "umatrix_mean": 32.9663, | |
| "umatrix_max": 77.8269, | |
| "hit_entropy_norm": 0.7998, | |
| "neuronios_ativos": 7, | |
| "neuronios_total": 14, | |
| "grid": [ | |
| 7, | |
| 2 | |
| ], | |
| "profundidade_ghsom": 0, | |
| "n_amostras": 1400, | |
| "epoca": 4, | |
| "passos_fase": 244, | |
| "lr": 0.002787, | |
| "sigma_G": 1.2, | |
| "sigma_efetivo": 0.3532, | |
| "dW_rel": null, | |
| "loss_total": 32.598, | |
| "bancos": [ | |
| "gpt2a", | |
| "gpt2g", | |
| "dinov2" | |
| ], | |
| "lam_rep_pen": 0.4832, | |
| "trocas_bmu_pen": 497, | |
| "roteador_gpt2g": { | |
| "step": 90, | |
| "ce_final": 0.0023, | |
| "acuracia_roteador": 1.0, | |
| "n": 256, | |
| "rotulos": { | |
| "gpt2a": 256, | |
| "gpt2g": 0, | |
| "dinov2": 0 | |
| }, | |
| "massa_media": { | |
| "gpt2a": 0.9947, | |
| "gpt2g": 0.0029, | |
| "dinov2": 0.0024 | |
| } | |
| }, | |
| "diffusion_dinov2": { | |
| "steps_diff": 1902, | |
| "loss_diff_ema": 1.0007, | |
| "level": 0, | |
| "difficulty": 0.05, | |
| "reward_ma": 0.0865, | |
| "beta_C1": 0.371, | |
| "sonda_qe": { | |
| "qe_dinov2_real": 4.5458, | |
| "qe_dinov2_difusivo": 13.1354, | |
| "n": 64 | |
| } | |
| } | |
| }, | |
| { | |
| "qe": 1.5609, | |
| "topographic_error": 0.5264, | |
| "umatrix_mean": 32.477, | |
| "umatrix_max": 77.3108, | |
| "hit_entropy_norm": 0.8369, | |
| "neuronios_ativos": 7, | |
| "neuronios_total": 14, | |
| "grid": [ | |
| 7, | |
| 2 | |
| ], | |
| "profundidade_ghsom": 0, | |
| "n_amostras": 1400, | |
| "epoca": 5, | |
| "passos_fase": 305, | |
| "lr": 0.002857, | |
| "sigma_G": 1.2, | |
| "sigma_efetivo": 0.2601, | |
| "dW_rel": 0.031289, | |
| "loss_total": 27.4382, | |
| "bancos": [ | |
| "gpt2a", | |
| "gpt2g", | |
| "dinov2" | |
| ], | |
| "lam_rep_pen": 0.3822, | |
| "trocas_bmu_pen": 546, | |
| "roteador_gpt2g": { | |
| "step": 90, | |
| "ce_final": 0.0023, | |
| "acuracia_roteador": 1.0, | |
| "n": 256, | |
| "rotulos": { | |
| "gpt2a": 256, | |
| "gpt2g": 0, | |
| "dinov2": 0 | |
| }, | |
| "massa_media": { | |
| "gpt2a": 0.9947, | |
| "gpt2g": 0.0029, | |
| "dinov2": 0.0024 | |
| } | |
| }, | |
| "diffusion_dinov2": { | |
| "steps_diff": 1903, | |
| "loss_diff_ema": 0.99503, | |
| "level": 0, | |
| "difficulty": 0.05, | |
| "reward_ma": 0.104, | |
| "beta_C1": 0.329, | |
| "sonda_qe": { | |
| "qe_dinov2_real": 2.0184, | |
| "qe_dinov2_difusivo": 10.9092, | |
| "n": 64 | |
| } | |
| } | |
| }, | |
| { | |
| "qe": 2.1644, | |
| "topographic_error": 0.0471, | |
| "umatrix_mean": 2.0682, | |
| "umatrix_max": 2.4819, | |
| "hit_entropy_norm": 0.8894, | |
| "neuronios_ativos": 14, | |
| "neuronios_total": 14, | |
| "grid": [ | |
| 7, | |
| 2 | |
| ], | |
| "profundidade_ghsom": 0, | |
| "n_amostras": 1400, | |
| "epoca": 6, | |
| "passos_fase": 61, | |
| "lr": 0.002808, | |
| "sigma_G": 1.2, | |
| "sigma_efetivo": 0.8839, | |
| "dW_rel": 0.464311, | |
| "loss_total": 28.7826, | |
| "bancos": [ | |
| "gpt2a", | |
| "gpt2g", | |
| "dinov2" | |
| ], | |
| "lam_rep_pen": 0.1013, | |
| "trocas_bmu_pen": 137, | |
| "roteador_gpt2g": { | |
| "carregado": "router_v3.pt", | |
| "massa_media": { | |
| "gpt2a": 0.9948, | |
| "gpt2g": 0.0029, | |
| "dinov2": 0.0023 | |
| } | |
| }, | |
| "diffusion_dinov2": { | |
| "steps_diff": 400, | |
| "loss_diff_ema": 1.04687, | |
| "level": 0, | |
| "difficulty": 0.05, | |
| "reward_ma": 0.4412, | |
| "beta_C1": 0.286, | |
| "sonda_qe": { | |
| "qe_dinov2_real": 1.5224, | |
| "qe_dinov2_difusivo": 9.5711, | |
| "n": 64 | |
| } | |
| } | |
| }, | |
| { | |
| "qe": 1.426, | |
| "topographic_error": 0.0829, | |
| "umatrix_mean": 1.9585, | |
| "umatrix_max": 2.4865, | |
| "hit_entropy_norm": 0.9507, | |
| "neuronios_ativos": 14, | |
| "neuronios_total": 14, | |
| "grid": [ | |
| 7, | |
| 2 | |
| ], | |
| "profundidade_ghsom": 0, | |
| "n_amostras": 1400, | |
| "epoca": 7, | |
| "passos_fase": 122, | |
| "lr": 0.002881, | |
| "sigma_G": 1.2, | |
| "sigma_efetivo": 0.651, | |
| "dW_rel": 0.031341, | |
| "loss_total": 24.8702, | |
| "bancos": [ | |
| "gpt2a", | |
| "gpt2g", | |
| "dinov2" | |
| ], | |
| "lam_rep_pen": 0.0, | |
| "trocas_bmu_pen": 154, | |
| "roteador_gpt2g": { | |
| "carregado": "router_v3.pt", | |
| "massa_media": { | |
| "gpt2a": 0.9948, | |
| "gpt2g": 0.0029, | |
| "dinov2": 0.0023 | |
| } | |
| }, | |
| "diffusion_dinov2": { | |
| "steps_diff": 800, | |
| "loss_diff_ema": 0.95338, | |
| "level": 0, | |
| "difficulty": 0.05, | |
| "reward_ma": 0.2461, | |
| "beta_C1": 0.243, | |
| "sonda_qe": { | |
| "qe_dinov2_real": 2.0416, | |
| "qe_dinov2_difusivo": 9.3937, | |
| "n": 64 | |
| } | |
| } | |
| }, | |
| { | |
| "qe": 1.2764, | |
| "topographic_error": 0.2164, | |
| "umatrix_mean": 2.4263, | |
| "umatrix_max": 3.8891, | |
| "hit_entropy_norm": 0.9872, | |
| "neuronios_ativos": 14, | |
| "neuronios_total": 14, | |
| "grid": [ | |
| 7, | |
| 2 | |
| ], | |
| "profundidade_ghsom": 0, | |
| "n_amostras": 1400, | |
| "epoca": 8, | |
| "passos_fase": 183, | |
| "lr": 0.002891, | |
| "sigma_G": 1.2, | |
| "sigma_efetivo": 0.4795, | |
| "dW_rel": 0.026871, | |
| "loss_total": 22.7493, | |
| "bancos": [ | |
| "gpt2a", | |
| "gpt2g", | |
| "dinov2" | |
| ], | |
| "lam_rep_pen": 0.0, | |
| "trocas_bmu_pen": 154, | |
| "roteador_gpt2g": { | |
| "carregado": "router_v3.pt", | |
| "massa_media": { | |
| "gpt2a": 0.9948, | |
| "gpt2g": 0.0029, | |
| "dinov2": 0.0023 | |
| } | |
| }, | |
| "diffusion_dinov2": { | |
| "steps_diff": 801, | |
| "loss_diff_ema": 0.84353, | |
| "level": 0, | |
| "difficulty": 0.05, | |
| "reward_ma": 0.2583, | |
| "beta_C1": 0.2, | |
| "sonda_qe": { | |
| "qe_dinov2_real": 1.3313, | |
| "qe_dinov2_difusivo": 7.3234, | |
| "n": 64 | |
| } | |
| } | |
| } | |
| ], | |
| "roteador_inicial": { | |
| "step": 90, | |
| "ce_final": 0.0023, | |
| "acuracia_roteador": 1.0, | |
| "n": 256, | |
| "rotulos": { | |
| "gpt2a": 256, | |
| "gpt2g": 0, | |
| "dinov2": 0 | |
| }, | |
| "massa_media": { | |
| "gpt2a": 0.9948, | |
| "gpt2g": 0.0029, | |
| "dinov2": 0.0023 | |
| } | |
| }, | |
| "diffusao_resumo": { | |
| "steps_diff": 801, | |
| "loss_primeiro": 1.0785, | |
| "loss_ultimo": 0.8956, | |
| "level_final": 0, | |
| "difficulty_final": 0.05 | |
| }, | |
| "calibracao_W_o": { | |
| "sigma_max_W_o_antes": 2.3223, | |
| "sigma_max_W_o_depois": 1.0, | |
| "congelado": true | |
| }, | |
| "roteador_avaliacao": { | |
| "qe_despacho_roteador": 11.836, | |
| "qe_despacho_uniforme": 1.1246, | |
| "ganho_relativo": -9.5243, | |
| "fracao_rota_top1": { | |
| "gpt2a": 1.0, | |
| "dinov2": 0.0 | |
| } | |
| }, | |
| "verificacao_inicial": { | |
| "m1": { | |
| "teorema": "T-ATTVAR (Minkowski, combinacao convexa, tau livre)", | |
| "bancos": { | |
| "gpt2a": { | |
| "var_attn": 18.356518, | |
| "var_max_especialista": 176.515137, | |
| "razao": 0.104, | |
| "ok": true | |
| }, | |
| "gpt2g": { | |
| "var_attn": 0.376153, | |
| "var_max_especialista": 12.349455, | |
| "razao": 0.0305, | |
| "ok": true | |
| }, | |
| "dinov2": { | |
| "var_attn": 0.084203, | |
| "var_max_especialista": 36.919891, | |
| "razao": 0.0023, | |
| "ok": true | |
| } | |
| }, | |
| "passou": true | |
| }, | |
| "m2": { | |
| "teorema": "T-GNN Dirichlet monotono com eta <= 2/lambda_max", | |
| "energias": [ | |
| 34.556427, | |
| 12.747748 | |
| ], | |
| "monotona": true, | |
| "lambda_max_estimado": 0.4559, | |
| "eta_seguro_publicado": 0.35, | |
| "passou": true | |
| }, | |
| "m3": { | |
| "teorema": "cota de Rademacher sqrt(1+(K-1)rho)/sqrt(K)", | |
| "bancos": { | |
| "gpt2a": { | |
| "rho_bar": 0.8442, | |
| "K": 32, | |
| "cota_rademacher": 0.9214 | |
| }, | |
| "gpt2g": { | |
| "rho_bar": 0.626, | |
| "K": 32, | |
| "cota_rademacher": 0.7985 | |
| }, | |
| "dinov2": { | |
| "rho_bar": 0.9325, | |
| "K": 48, | |
| "cota_rademacher": 0.9664 | |
| } | |
| }, | |
| "passou": true | |
| } | |
| }, | |
| "verificacao_final": { | |
| "m1": { | |
| "teorema": "T-ATTVAR (Minkowski, combinacao convexa, tau livre)", | |
| "bancos": { | |
| "gpt2a": { | |
| "var_attn": 18.356518, | |
| "var_max_especialista": 176.515137, | |
| "razao": 0.104, | |
| "ok": true | |
| }, | |
| "gpt2g": { | |
| "var_attn": 0.376153, | |
| "var_max_especialista": 12.349455, | |
| "razao": 0.0305, | |
| "ok": true | |
| }, | |
| "dinov2": { | |
| "var_attn": 0.084203, | |
| "var_max_especialista": 36.919891, | |
| "razao": 0.0023, | |
| "ok": true | |
| } | |
| }, | |
| "passou": true | |
| }, | |
| "m2": { | |
| "teorema": "T-GNN Dirichlet monotono com eta <= 2/lambda_max", | |
| "energias": [ | |
| 34.556427, | |
| 12.747748 | |
| ], | |
| "monotona": true, | |
| "lambda_max_estimado": 0.4238, | |
| "eta_seguro_publicado": 0.35, | |
| "passou": true | |
| }, | |
| "m3": { | |
| "teorema": "cota de Rademacher sqrt(1+(K-1)rho)/sqrt(K)", | |
| "bancos": { | |
| "gpt2a": { | |
| "rho_bar": 0.8442, | |
| "K": 32, | |
| "cota_rademacher": 0.9214 | |
| }, | |
| "gpt2g": { | |
| "rho_bar": 0.626, | |
| "K": 32, | |
| "cota_rademacher": 0.7985 | |
| }, | |
| "dinov2": { | |
| "rho_bar": 0.9325, | |
| "K": 48, | |
| "cota_rademacher": 0.9664 | |
| } | |
| }, | |
| "passou": true | |
| } | |
| }, | |
| "checkpoint_final": "/home/z/my-project/hako_v3_work/ckpt/hako_v3_final.npz" | |
| } |