HAKO-v3 / p3_joint2_hist.jsonl
PowerMachine's picture
HAKO-V3: punicao por repeticao (T-REP3), orquestrador gpt2g (RouterGQA), DINOv2->Diffusion->SOM, comparacao V2<->V3
3fe1f47 verified
Raw History Blame Contribute Delete
5.61 kB
{"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}}}