| { | |
| "best_metric": 0.6108, | |
| "best_model_checkpoint": "car-finetune/checkpoint-25174", | |
| "epoch": 44.0, | |
| "eval_steps": 500, | |
| "global_step": 27016, | |
| "is_hyper_param_search": false, | |
| "is_local_process_zero": true, | |
| "is_world_process_zero": true, | |
| "log_history": [ | |
| { | |
| "epoch": 0.8143322475570033, | |
| "grad_norm": 36.48604202270508, | |
| "learning_rate": 4.967105263157895e-05, | |
| "loss": 18.7307, | |
| "step": 500 | |
| }, | |
| { | |
| "epoch": 1.0, | |
| "eval_loss": 8.210515975952148, | |
| "eval_map": 0.4165, | |
| "eval_map_0": 0.2784, | |
| "eval_map_1": 0.4886, | |
| "eval_map_2": 0.4916, | |
| "eval_map_3": 0.4073, | |
| "eval_map_50": 0.672, | |
| "eval_map_75": 0.4585, | |
| "eval_map_car": -1.0, | |
| "eval_map_large": 0.4832, | |
| "eval_map_medium": 0.521, | |
| "eval_map_small": 0.2637, | |
| "eval_mar_1": 0.2396, | |
| "eval_mar_10": 0.5504, | |
| "eval_mar_100": 0.6266, | |
| "eval_mar_100_0": 0.4054, | |
| "eval_mar_100_1": 0.6723, | |
| "eval_mar_100_2": 0.726, | |
| "eval_mar_100_3": 0.7025, | |
| "eval_mar_100_car": -1.0, | |
| "eval_mar_large": 0.7123, | |
| "eval_mar_medium": 0.6839, | |
| "eval_mar_small": 0.5377, | |
| "eval_runtime": 143.2193, | |
| "eval_samples_per_second": 10.222, | |
| "eval_steps_per_second": 1.278, | |
| "step": 614 | |
| }, | |
| { | |
| "epoch": 1.6286644951140063, | |
| "grad_norm": 64.3085708618164, | |
| "learning_rate": 4.884868421052632e-05, | |
| "loss": 11.4903, | |
| "step": 1000 | |
| }, | |
| { | |
| "epoch": 2.0, | |
| "eval_loss": 7.858218193054199, | |
| "eval_map": 0.4463, | |
| "eval_map_0": 0.3404, | |
| "eval_map_1": 0.5009, | |
| "eval_map_2": 0.5405, | |
| "eval_map_3": 0.4034, | |
| "eval_map_50": 0.7143, | |
| "eval_map_75": 0.4809, | |
| "eval_map_car": -1.0, | |
| "eval_map_large": 0.5372, | |
| "eval_map_medium": 0.5451, | |
| "eval_map_small": 0.2915, | |
| "eval_mar_1": 0.2475, | |
| "eval_mar_10": 0.574, | |
| "eval_mar_100": 0.6524, | |
| "eval_mar_100_0": 0.4887, | |
| "eval_mar_100_1": 0.6828, | |
| "eval_mar_100_2": 0.7492, | |
| "eval_mar_100_3": 0.6888, | |
| "eval_mar_100_car": -1.0, | |
| "eval_mar_large": 0.7939, | |
| "eval_mar_medium": 0.7148, | |
| "eval_mar_small": 0.5399, | |
| "eval_runtime": 155.7346, | |
| "eval_samples_per_second": 9.401, | |
| "eval_steps_per_second": 1.175, | |
| "step": 1228 | |
| }, | |
| { | |
| "epoch": 2.44299674267101, | |
| "grad_norm": 29.896888732910156, | |
| "learning_rate": 4.802631578947368e-05, | |
| "loss": 10.8033, | |
| "step": 1500 | |
| }, | |
| { | |
| "epoch": 3.0, | |
| "eval_loss": 7.962161540985107, | |
| "eval_map": 0.4504, | |
| "eval_map_0": 0.3621, | |
| "eval_map_1": 0.4826, | |
| "eval_map_2": 0.5321, | |
| "eval_map_3": 0.4246, | |
| "eval_map_50": 0.7083, | |
| "eval_map_75": 0.4993, | |
| "eval_map_car": -1.0, | |
| "eval_map_large": 0.4753, | |
| "eval_map_medium": 0.5748, | |
| "eval_map_small": 0.3026, | |
| "eval_mar_1": 0.2475, | |
| "eval_mar_10": 0.5834, | |
| "eval_mar_100": 0.6754, | |
| "eval_mar_100_0": 0.5448, | |
| "eval_mar_100_1": 0.7105, | |
| "eval_mar_100_2": 0.7581, | |
| "eval_mar_100_3": 0.6882, | |
| "eval_mar_100_car": -1.0, | |
| "eval_mar_large": 0.8484, | |
| "eval_mar_medium": 0.7313, | |
| "eval_mar_small": 0.5855, | |
| "eval_runtime": 171.3133, | |
| "eval_samples_per_second": 8.546, | |
| "eval_steps_per_second": 1.068, | |
| "step": 1842 | |
| }, | |
| { | |
| "epoch": 3.257328990228013, | |
| "grad_norm": 33.05744552612305, | |
| "learning_rate": 4.720394736842105e-05, | |
| "loss": 10.2622, | |
| "step": 2000 | |
| }, | |
| { | |
| "epoch": 4.0, | |
| "eval_loss": 8.229509353637695, | |
| "eval_map": 0.4625, | |
| "eval_map_0": 0.3308, | |
| "eval_map_1": 0.5026, | |
| "eval_map_2": 0.5446, | |
| "eval_map_3": 0.4721, | |
| "eval_map_50": 0.7147, | |
| "eval_map_75": 0.5177, | |
| "eval_map_car": -1.0, | |
| "eval_map_large": 0.4901, | |
| "eval_map_medium": 0.5781, | |
| "eval_map_small": 0.3046, | |
| "eval_mar_1": 0.2556, | |
| "eval_mar_10": 0.599, | |
| "eval_mar_100": 0.6885, | |
| "eval_mar_100_0": 0.5225, | |
| "eval_mar_100_1": 0.7135, | |
| "eval_mar_100_2": 0.7816, | |
| "eval_mar_100_3": 0.7365, | |
| "eval_mar_100_car": -1.0, | |
| "eval_mar_large": 0.8696, | |
| "eval_mar_medium": 0.7528, | |
| "eval_mar_small": 0.5868, | |
| "eval_runtime": 163.1535, | |
| "eval_samples_per_second": 8.973, | |
| "eval_steps_per_second": 1.122, | |
| "step": 2456 | |
| }, | |
| { | |
| "epoch": 4.071661237785016, | |
| "grad_norm": 13.571614265441895, | |
| "learning_rate": 4.638157894736843e-05, | |
| "loss": 9.8714, | |
| "step": 2500 | |
| }, | |
| { | |
| "epoch": 4.88599348534202, | |
| "grad_norm": 31.514949798583984, | |
| "learning_rate": 4.555921052631579e-05, | |
| "loss": 9.486, | |
| "step": 3000 | |
| }, | |
| { | |
| "epoch": 5.0, | |
| "eval_loss": 7.853174209594727, | |
| "eval_map": 0.4858, | |
| "eval_map_0": 0.3878, | |
| "eval_map_1": 0.5041, | |
| "eval_map_2": 0.5354, | |
| "eval_map_3": 0.5158, | |
| "eval_map_50": 0.7444, | |
| "eval_map_75": 0.5483, | |
| "eval_map_car": -1.0, | |
| "eval_map_large": 0.5469, | |
| "eval_map_medium": 0.5878, | |
| "eval_map_small": 0.3293, | |
| "eval_mar_1": 0.2628, | |
| "eval_mar_10": 0.6196, | |
| "eval_mar_100": 0.7149, | |
| "eval_mar_100_0": 0.5773, | |
| "eval_mar_100_1": 0.7399, | |
| "eval_mar_100_2": 0.7963, | |
| "eval_mar_100_3": 0.7461, | |
| "eval_mar_100_car": -1.0, | |
| "eval_mar_large": 0.9093, | |
| "eval_mar_medium": 0.7703, | |
| "eval_mar_small": 0.616, | |
| "eval_runtime": 173.7867, | |
| "eval_samples_per_second": 8.424, | |
| "eval_steps_per_second": 1.053, | |
| "step": 3070 | |
| }, | |
| { | |
| "epoch": 5.700325732899023, | |
| "grad_norm": 55.96042251586914, | |
| "learning_rate": 4.473684210526316e-05, | |
| "loss": 9.3301, | |
| "step": 3500 | |
| }, | |
| { | |
| "epoch": 6.0, | |
| "eval_loss": 8.071856498718262, | |
| "eval_map": 0.473, | |
| "eval_map_0": 0.3705, | |
| "eval_map_1": 0.5244, | |
| "eval_map_2": 0.501, | |
| "eval_map_3": 0.496, | |
| "eval_map_50": 0.7402, | |
| "eval_map_75": 0.519, | |
| "eval_map_car": -1.0, | |
| "eval_map_large": 0.5194, | |
| "eval_map_medium": 0.5709, | |
| "eval_map_small": 0.3321, | |
| "eval_mar_1": 0.2561, | |
| "eval_mar_10": 0.6049, | |
| "eval_mar_100": 0.7053, | |
| "eval_mar_100_0": 0.5706, | |
| "eval_mar_100_1": 0.7086, | |
| "eval_mar_100_2": 0.7908, | |
| "eval_mar_100_3": 0.751, | |
| "eval_mar_100_car": -1.0, | |
| "eval_mar_large": 0.7828, | |
| "eval_mar_medium": 0.7606, | |
| "eval_mar_small": 0.6035, | |
| "eval_runtime": 175.9385, | |
| "eval_samples_per_second": 8.321, | |
| "eval_steps_per_second": 1.04, | |
| "step": 3684 | |
| }, | |
| { | |
| "epoch": 6.514657980456026, | |
| "grad_norm": 17.507080078125, | |
| "learning_rate": 4.391447368421053e-05, | |
| "loss": 9.1509, | |
| "step": 4000 | |
| }, | |
| { | |
| "epoch": 7.0, | |
| "eval_loss": 7.802272796630859, | |
| "eval_map": 0.4899, | |
| "eval_map_0": 0.3931, | |
| "eval_map_1": 0.5231, | |
| "eval_map_2": 0.5268, | |
| "eval_map_3": 0.5164, | |
| "eval_map_50": 0.7443, | |
| "eval_map_75": 0.5516, | |
| "eval_map_car": -1.0, | |
| "eval_map_large": 0.4754, | |
| "eval_map_medium": 0.6065, | |
| "eval_map_small": 0.3371, | |
| "eval_mar_1": 0.2607, | |
| "eval_mar_10": 0.6244, | |
| "eval_mar_100": 0.7212, | |
| "eval_mar_100_0": 0.5635, | |
| "eval_mar_100_1": 0.7379, | |
| "eval_mar_100_2": 0.8136, | |
| "eval_mar_100_3": 0.7697, | |
| "eval_mar_100_car": -1.0, | |
| "eval_mar_large": 0.9074, | |
| "eval_mar_medium": 0.7792, | |
| "eval_mar_small": 0.6262, | |
| "eval_runtime": 173.5936, | |
| "eval_samples_per_second": 8.433, | |
| "eval_steps_per_second": 1.054, | |
| "step": 4298 | |
| }, | |
| { | |
| "epoch": 7.328990228013029, | |
| "grad_norm": 15.204864501953125, | |
| "learning_rate": 4.30921052631579e-05, | |
| "loss": 9.0599, | |
| "step": 4500 | |
| }, | |
| { | |
| "epoch": 8.0, | |
| "eval_loss": 7.824928283691406, | |
| "eval_map": 0.5186, | |
| "eval_map_0": 0.3843, | |
| "eval_map_1": 0.5801, | |
| "eval_map_2": 0.5819, | |
| "eval_map_3": 0.5279, | |
| "eval_map_50": 0.7618, | |
| "eval_map_75": 0.5857, | |
| "eval_map_car": -1.0, | |
| "eval_map_large": 0.5862, | |
| "eval_map_medium": 0.6194, | |
| "eval_map_small": 0.3423, | |
| "eval_mar_1": 0.2747, | |
| "eval_mar_10": 0.6211, | |
| "eval_mar_100": 0.7078, | |
| "eval_mar_100_0": 0.5573, | |
| "eval_mar_100_1": 0.7283, | |
| "eval_mar_100_2": 0.7967, | |
| "eval_mar_100_3": 0.7487, | |
| "eval_mar_100_car": -1.0, | |
| "eval_mar_large": 0.7776, | |
| "eval_mar_medium": 0.7787, | |
| "eval_mar_small": 0.5914, | |
| "eval_runtime": 184.7964, | |
| "eval_samples_per_second": 7.922, | |
| "eval_steps_per_second": 0.99, | |
| "step": 4912 | |
| }, | |
| { | |
| "epoch": 8.143322475570033, | |
| "grad_norm": 26.53233528137207, | |
| "learning_rate": 4.226973684210527e-05, | |
| "loss": 9.0022, | |
| "step": 5000 | |
| }, | |
| { | |
| "epoch": 8.957654723127035, | |
| "grad_norm": 29.07124900817871, | |
| "learning_rate": 4.1447368421052636e-05, | |
| "loss": 8.9732, | |
| "step": 5500 | |
| }, | |
| { | |
| "epoch": 9.0, | |
| "eval_loss": 8.0244779586792, | |
| "eval_map": 0.4824, | |
| "eval_map_0": 0.3766, | |
| "eval_map_1": 0.4961, | |
| "eval_map_2": 0.5437, | |
| "eval_map_3": 0.5134, | |
| "eval_map_50": 0.7304, | |
| "eval_map_75": 0.5425, | |
| "eval_map_car": -1.0, | |
| "eval_map_large": 0.545, | |
| "eval_map_medium": 0.5804, | |
| "eval_map_small": 0.3182, | |
| "eval_mar_1": 0.264, | |
| "eval_mar_10": 0.6144, | |
| "eval_mar_100": 0.7069, | |
| "eval_mar_100_0": 0.5587, | |
| "eval_mar_100_1": 0.719, | |
| "eval_mar_100_2": 0.8014, | |
| "eval_mar_100_3": 0.7485, | |
| "eval_mar_100_car": -1.0, | |
| "eval_mar_large": 0.8821, | |
| "eval_mar_medium": 0.779, | |
| "eval_mar_small": 0.5935, | |
| "eval_runtime": 185.6076, | |
| "eval_samples_per_second": 7.888, | |
| "eval_steps_per_second": 0.986, | |
| "step": 5526 | |
| }, | |
| { | |
| "epoch": 9.77198697068404, | |
| "grad_norm": 18.2327938079834, | |
| "learning_rate": 4.0625000000000005e-05, | |
| "loss": 8.8832, | |
| "step": 6000 | |
| }, | |
| { | |
| "epoch": 10.0, | |
| "eval_loss": 7.798117160797119, | |
| "eval_map": 0.5339, | |
| "eval_map_0": 0.4084, | |
| "eval_map_1": 0.5473, | |
| "eval_map_2": 0.6333, | |
| "eval_map_3": 0.5465, | |
| "eval_map_50": 0.7779, | |
| "eval_map_75": 0.6126, | |
| "eval_map_car": -1.0, | |
| "eval_map_large": 0.5806, | |
| "eval_map_medium": 0.6258, | |
| "eval_map_small": 0.3589, | |
| "eval_mar_1": 0.2773, | |
| "eval_mar_10": 0.6323, | |
| "eval_mar_100": 0.7267, | |
| "eval_mar_100_0": 0.5854, | |
| "eval_mar_100_1": 0.7423, | |
| "eval_mar_100_2": 0.8135, | |
| "eval_mar_100_3": 0.7657, | |
| "eval_mar_100_car": -1.0, | |
| "eval_mar_large": 0.8912, | |
| "eval_mar_medium": 0.7949, | |
| "eval_mar_small": 0.6176, | |
| "eval_runtime": 184.842, | |
| "eval_samples_per_second": 7.92, | |
| "eval_steps_per_second": 0.99, | |
| "step": 6140 | |
| }, | |
| { | |
| "epoch": 10.586319218241043, | |
| "grad_norm": 20.960790634155273, | |
| "learning_rate": 3.980263157894737e-05, | |
| "loss": 8.7376, | |
| "step": 6500 | |
| }, | |
| { | |
| "epoch": 11.0, | |
| "eval_loss": 7.757503509521484, | |
| "eval_map": 0.5306, | |
| "eval_map_0": 0.4036, | |
| "eval_map_1": 0.5661, | |
| "eval_map_2": 0.5816, | |
| "eval_map_3": 0.5712, | |
| "eval_map_50": 0.7861, | |
| "eval_map_75": 0.5985, | |
| "eval_map_car": -1.0, | |
| "eval_map_large": 0.5851, | |
| "eval_map_medium": 0.6427, | |
| "eval_map_small": 0.3388, | |
| "eval_mar_1": 0.2765, | |
| "eval_mar_10": 0.6375, | |
| "eval_mar_100": 0.7249, | |
| "eval_mar_100_0": 0.5723, | |
| "eval_mar_100_1": 0.7505, | |
| "eval_mar_100_2": 0.8107, | |
| "eval_mar_100_3": 0.7659, | |
| "eval_mar_100_car": -1.0, | |
| "eval_mar_large": 0.8115, | |
| "eval_mar_medium": 0.7937, | |
| "eval_mar_small": 0.6205, | |
| "eval_runtime": 181.1547, | |
| "eval_samples_per_second": 8.081, | |
| "eval_steps_per_second": 1.01, | |
| "step": 6754 | |
| }, | |
| { | |
| "epoch": 11.400651465798045, | |
| "grad_norm": 20.986557006835938, | |
| "learning_rate": 3.8980263157894736e-05, | |
| "loss": 8.7394, | |
| "step": 7000 | |
| }, | |
| { | |
| "epoch": 12.0, | |
| "eval_loss": 8.640464782714844, | |
| "eval_map": 0.4874, | |
| "eval_map_0": 0.3472, | |
| "eval_map_1": 0.4987, | |
| "eval_map_2": 0.5779, | |
| "eval_map_3": 0.526, | |
| "eval_map_50": 0.7211, | |
| "eval_map_75": 0.5473, | |
| "eval_map_car": -1.0, | |
| "eval_map_large": 0.5333, | |
| "eval_map_medium": 0.6186, | |
| "eval_map_small": 0.2857, | |
| "eval_mar_1": 0.2701, | |
| "eval_mar_10": 0.597, | |
| "eval_mar_100": 0.6757, | |
| "eval_mar_100_0": 0.5171, | |
| "eval_mar_100_1": 0.6653, | |
| "eval_mar_100_2": 0.7738, | |
| "eval_mar_100_3": 0.7467, | |
| "eval_mar_100_car": -1.0, | |
| "eval_mar_large": 0.866, | |
| "eval_mar_medium": 0.7921, | |
| "eval_mar_small": 0.508, | |
| "eval_runtime": 180.7473, | |
| "eval_samples_per_second": 8.1, | |
| "eval_steps_per_second": 1.012, | |
| "step": 7368 | |
| }, | |
| { | |
| "epoch": 12.214983713355048, | |
| "grad_norm": 31.571374893188477, | |
| "learning_rate": 3.815789473684211e-05, | |
| "loss": 8.7474, | |
| "step": 7500 | |
| }, | |
| { | |
| "epoch": 13.0, | |
| "eval_loss": 7.7668938636779785, | |
| "eval_map": 0.53, | |
| "eval_map_0": 0.372, | |
| "eval_map_1": 0.5579, | |
| "eval_map_2": 0.6227, | |
| "eval_map_3": 0.5675, | |
| "eval_map_50": 0.7787, | |
| "eval_map_75": 0.602, | |
| "eval_map_car": -1.0, | |
| "eval_map_large": 0.5999, | |
| "eval_map_medium": 0.6289, | |
| "eval_map_small": 0.3577, | |
| "eval_mar_1": 0.2763, | |
| "eval_mar_10": 0.6373, | |
| "eval_mar_100": 0.7227, | |
| "eval_mar_100_0": 0.5633, | |
| "eval_mar_100_1": 0.7403, | |
| "eval_mar_100_2": 0.8121, | |
| "eval_mar_100_3": 0.7751, | |
| "eval_mar_100_car": -1.0, | |
| "eval_mar_large": 0.9033, | |
| "eval_mar_medium": 0.7954, | |
| "eval_mar_small": 0.6164, | |
| "eval_runtime": 182.5954, | |
| "eval_samples_per_second": 8.018, | |
| "eval_steps_per_second": 1.002, | |
| "step": 7982 | |
| }, | |
| { | |
| "epoch": 13.029315960912053, | |
| "grad_norm": 27.426115036010742, | |
| "learning_rate": 3.7335526315789475e-05, | |
| "loss": 8.6912, | |
| "step": 8000 | |
| }, | |
| { | |
| "epoch": 13.843648208469055, | |
| "grad_norm": 84.17616271972656, | |
| "learning_rate": 3.6513157894736844e-05, | |
| "loss": 8.6296, | |
| "step": 8500 | |
| }, | |
| { | |
| "epoch": 14.0, | |
| "eval_loss": 8.179883003234863, | |
| "eval_map": 0.4749, | |
| "eval_map_0": 0.3417, | |
| "eval_map_1": 0.4607, | |
| "eval_map_2": 0.5543, | |
| "eval_map_3": 0.5427, | |
| "eval_map_50": 0.7174, | |
| "eval_map_75": 0.5378, | |
| "eval_map_car": -1.0, | |
| "eval_map_large": 0.512, | |
| "eval_map_medium": 0.5725, | |
| "eval_map_small": 0.2997, | |
| "eval_mar_1": 0.2633, | |
| "eval_mar_10": 0.6005, | |
| "eval_mar_100": 0.7029, | |
| "eval_mar_100_0": 0.5522, | |
| "eval_mar_100_1": 0.6989, | |
| "eval_mar_100_2": 0.7932, | |
| "eval_mar_100_3": 0.7675, | |
| "eval_mar_100_car": -1.0, | |
| "eval_mar_large": 0.8653, | |
| "eval_mar_medium": 0.7739, | |
| "eval_mar_small": 0.5922, | |
| "eval_runtime": 183.3956, | |
| "eval_samples_per_second": 7.983, | |
| "eval_steps_per_second": 0.998, | |
| "step": 8596 | |
| }, | |
| { | |
| "epoch": 14.657980456026058, | |
| "grad_norm": 55.561729431152344, | |
| "learning_rate": 3.569078947368421e-05, | |
| "loss": 8.5129, | |
| "step": 9000 | |
| }, | |
| { | |
| "epoch": 15.0, | |
| "eval_loss": 7.639471530914307, | |
| "eval_map": 0.5501, | |
| "eval_map_0": 0.4091, | |
| "eval_map_1": 0.5898, | |
| "eval_map_2": 0.6055, | |
| "eval_map_3": 0.5959, | |
| "eval_map_50": 0.7956, | |
| "eval_map_75": 0.6262, | |
| "eval_map_car": -1.0, | |
| "eval_map_large": 0.6014, | |
| "eval_map_medium": 0.6556, | |
| "eval_map_small": 0.3504, | |
| "eval_mar_1": 0.2832, | |
| "eval_mar_10": 0.6473, | |
| "eval_mar_100": 0.7236, | |
| "eval_mar_100_0": 0.5685, | |
| "eval_mar_100_1": 0.7356, | |
| "eval_mar_100_2": 0.8149, | |
| "eval_mar_100_3": 0.7754, | |
| "eval_mar_100_car": -1.0, | |
| "eval_mar_large": 0.9094, | |
| "eval_mar_medium": 0.801, | |
| "eval_mar_small": 0.6051, | |
| "eval_runtime": 182.9189, | |
| "eval_samples_per_second": 8.004, | |
| "eval_steps_per_second": 1.0, | |
| "step": 9210 | |
| }, | |
| { | |
| "epoch": 15.472312703583063, | |
| "grad_norm": 31.702539443969727, | |
| "learning_rate": 3.4868421052631575e-05, | |
| "loss": 8.4478, | |
| "step": 9500 | |
| }, | |
| { | |
| "epoch": 16.0, | |
| "eval_loss": 7.739656925201416, | |
| "eval_map": 0.5312, | |
| "eval_map_0": 0.4084, | |
| "eval_map_1": 0.5683, | |
| "eval_map_2": 0.5958, | |
| "eval_map_3": 0.5525, | |
| "eval_map_50": 0.7778, | |
| "eval_map_75": 0.605, | |
| "eval_map_car": -1.0, | |
| "eval_map_large": 0.5631, | |
| "eval_map_medium": 0.6408, | |
| "eval_map_small": 0.3543, | |
| "eval_mar_1": 0.278, | |
| "eval_mar_10": 0.6362, | |
| "eval_mar_100": 0.7216, | |
| "eval_mar_100_0": 0.579, | |
| "eval_mar_100_1": 0.7321, | |
| "eval_mar_100_2": 0.8015, | |
| "eval_mar_100_3": 0.7738, | |
| "eval_mar_100_car": -1.0, | |
| "eval_mar_large": 0.9028, | |
| "eval_mar_medium": 0.7959, | |
| "eval_mar_small": 0.6071, | |
| "eval_runtime": 171.0893, | |
| "eval_samples_per_second": 8.557, | |
| "eval_steps_per_second": 1.07, | |
| "step": 9824 | |
| }, | |
| { | |
| "epoch": 16.286644951140065, | |
| "grad_norm": 20.07905387878418, | |
| "learning_rate": 3.404605263157895e-05, | |
| "loss": 8.3627, | |
| "step": 10000 | |
| }, | |
| { | |
| "epoch": 17.0, | |
| "eval_loss": 7.790037631988525, | |
| "eval_map": 0.5402, | |
| "eval_map_0": 0.4224, | |
| "eval_map_1": 0.568, | |
| "eval_map_2": 0.5846, | |
| "eval_map_3": 0.5857, | |
| "eval_map_50": 0.7823, | |
| "eval_map_75": 0.6104, | |
| "eval_map_car": -1.0, | |
| "eval_map_large": 0.6132, | |
| "eval_map_medium": 0.6345, | |
| "eval_map_small": 0.3411, | |
| "eval_mar_1": 0.2777, | |
| "eval_mar_10": 0.6321, | |
| "eval_mar_100": 0.7157, | |
| "eval_mar_100_0": 0.5887, | |
| "eval_mar_100_1": 0.7222, | |
| "eval_mar_100_2": 0.7871, | |
| "eval_mar_100_3": 0.7646, | |
| "eval_mar_100_car": -1.0, | |
| "eval_mar_large": 0.8921, | |
| "eval_mar_medium": 0.7909, | |
| "eval_mar_small": 0.586, | |
| "eval_runtime": 178.3245, | |
| "eval_samples_per_second": 8.21, | |
| "eval_steps_per_second": 1.026, | |
| "step": 10438 | |
| }, | |
| { | |
| "epoch": 17.10097719869707, | |
| "grad_norm": 31.199081420898438, | |
| "learning_rate": 3.322368421052632e-05, | |
| "loss": 8.2877, | |
| "step": 10500 | |
| }, | |
| { | |
| "epoch": 17.91530944625407, | |
| "grad_norm": 15.449178695678711, | |
| "learning_rate": 3.240131578947368e-05, | |
| "loss": 8.3688, | |
| "step": 11000 | |
| }, | |
| { | |
| "epoch": 18.0, | |
| "eval_loss": 7.599175930023193, | |
| "eval_map": 0.4887, | |
| "eval_map_0": 0.4175, | |
| "eval_map_1": 0.5599, | |
| "eval_map_2": 0.4202, | |
| "eval_map_3": 0.5573, | |
| "eval_map_50": 0.7212, | |
| "eval_map_75": 0.5611, | |
| "eval_map_car": -1.0, | |
| "eval_map_large": 0.4184, | |
| "eval_map_medium": 0.6074, | |
| "eval_map_small": 0.375, | |
| "eval_mar_1": 0.257, | |
| "eval_mar_10": 0.631, | |
| "eval_mar_100": 0.7306, | |
| "eval_mar_100_0": 0.5813, | |
| "eval_mar_100_1": 0.7428, | |
| "eval_mar_100_2": 0.8107, | |
| "eval_mar_100_3": 0.7877, | |
| "eval_mar_100_car": -1.0, | |
| "eval_mar_large": 0.8772, | |
| "eval_mar_medium": 0.8047, | |
| "eval_mar_small": 0.6184, | |
| "eval_runtime": 182.4402, | |
| "eval_samples_per_second": 8.025, | |
| "eval_steps_per_second": 1.003, | |
| "step": 11052 | |
| }, | |
| { | |
| "epoch": 18.729641693811075, | |
| "grad_norm": 16.437231063842773, | |
| "learning_rate": 3.157894736842105e-05, | |
| "loss": 8.3401, | |
| "step": 11500 | |
| }, | |
| { | |
| "epoch": 19.0, | |
| "eval_loss": 7.053092002868652, | |
| "eval_map": 0.5891, | |
| "eval_map_0": 0.4422, | |
| "eval_map_1": 0.6218, | |
| "eval_map_2": 0.654, | |
| "eval_map_3": 0.6385, | |
| "eval_map_50": 0.8342, | |
| "eval_map_75": 0.6753, | |
| "eval_map_car": -1.0, | |
| "eval_map_large": 0.6355, | |
| "eval_map_medium": 0.6845, | |
| "eval_map_small": 0.3961, | |
| "eval_mar_1": 0.2908, | |
| "eval_mar_10": 0.6637, | |
| "eval_mar_100": 0.7493, | |
| "eval_mar_100_0": 0.6125, | |
| "eval_mar_100_1": 0.756, | |
| "eval_mar_100_2": 0.8292, | |
| "eval_mar_100_3": 0.7998, | |
| "eval_mar_100_car": -1.0, | |
| "eval_mar_large": 0.8995, | |
| "eval_mar_medium": 0.8145, | |
| "eval_mar_small": 0.6465, | |
| "eval_runtime": 183.4953, | |
| "eval_samples_per_second": 7.978, | |
| "eval_steps_per_second": 0.997, | |
| "step": 11666 | |
| }, | |
| { | |
| "epoch": 19.54397394136808, | |
| "grad_norm": 14.132092475891113, | |
| "learning_rate": 3.075657894736843e-05, | |
| "loss": 8.2335, | |
| "step": 12000 | |
| }, | |
| { | |
| "epoch": 20.0, | |
| "eval_loss": 6.844819068908691, | |
| "eval_map": 0.5917, | |
| "eval_map_0": 0.4596, | |
| "eval_map_1": 0.6347, | |
| "eval_map_2": 0.6538, | |
| "eval_map_3": 0.6186, | |
| "eval_map_50": 0.8383, | |
| "eval_map_75": 0.6797, | |
| "eval_map_car": -1.0, | |
| "eval_map_large": 0.6181, | |
| "eval_map_medium": 0.6905, | |
| "eval_map_small": 0.409, | |
| "eval_mar_1": 0.2889, | |
| "eval_mar_10": 0.6662, | |
| "eval_mar_100": 0.7513, | |
| "eval_mar_100_0": 0.6233, | |
| "eval_mar_100_1": 0.7574, | |
| "eval_mar_100_2": 0.8284, | |
| "eval_mar_100_3": 0.7962, | |
| "eval_mar_100_car": -1.0, | |
| "eval_mar_large": 0.8875, | |
| "eval_mar_medium": 0.8122, | |
| "eval_mar_small": 0.6524, | |
| "eval_runtime": 179.7619, | |
| "eval_samples_per_second": 8.144, | |
| "eval_steps_per_second": 1.018, | |
| "step": 12280 | |
| }, | |
| { | |
| "epoch": 20.35830618892508, | |
| "grad_norm": 15.430574417114258, | |
| "learning_rate": 2.9934210526315793e-05, | |
| "loss": 8.3458, | |
| "step": 12500 | |
| }, | |
| { | |
| "epoch": 21.0, | |
| "eval_loss": 7.360008716583252, | |
| "eval_map": 0.5649, | |
| "eval_map_0": 0.4146, | |
| "eval_map_1": 0.6133, | |
| "eval_map_2": 0.6272, | |
| "eval_map_3": 0.6043, | |
| "eval_map_50": 0.8077, | |
| "eval_map_75": 0.6446, | |
| "eval_map_car": -1.0, | |
| "eval_map_large": 0.6031, | |
| "eval_map_medium": 0.6689, | |
| "eval_map_small": 0.3852, | |
| "eval_mar_1": 0.2865, | |
| "eval_mar_10": 0.6569, | |
| "eval_mar_100": 0.7431, | |
| "eval_mar_100_0": 0.6063, | |
| "eval_mar_100_1": 0.749, | |
| "eval_mar_100_2": 0.826, | |
| "eval_mar_100_3": 0.7914, | |
| "eval_mar_100_car": -1.0, | |
| "eval_mar_large": 0.8985, | |
| "eval_mar_medium": 0.8114, | |
| "eval_mar_small": 0.6374, | |
| "eval_runtime": 180.4928, | |
| "eval_samples_per_second": 8.111, | |
| "eval_steps_per_second": 1.014, | |
| "step": 12894 | |
| }, | |
| { | |
| "epoch": 21.172638436482085, | |
| "grad_norm": 18.904953002929688, | |
| "learning_rate": 2.911184210526316e-05, | |
| "loss": 8.3067, | |
| "step": 13000 | |
| }, | |
| { | |
| "epoch": 21.986970684039086, | |
| "grad_norm": 17.112998962402344, | |
| "learning_rate": 2.8289473684210528e-05, | |
| "loss": 8.2643, | |
| "step": 13500 | |
| }, | |
| { | |
| "epoch": 22.0, | |
| "eval_loss": 7.019049644470215, | |
| "eval_map": 0.5987, | |
| "eval_map_0": 0.4628, | |
| "eval_map_1": 0.6464, | |
| "eval_map_2": 0.6588, | |
| "eval_map_3": 0.6269, | |
| "eval_map_50": 0.8447, | |
| "eval_map_75": 0.6869, | |
| "eval_map_car": -1.0, | |
| "eval_map_large": 0.6283, | |
| "eval_map_medium": 0.6954, | |
| "eval_map_small": 0.4107, | |
| "eval_mar_1": 0.29, | |
| "eval_mar_10": 0.6721, | |
| "eval_mar_100": 0.7572, | |
| "eval_mar_100_0": 0.6193, | |
| "eval_mar_100_1": 0.7722, | |
| "eval_mar_100_2": 0.8391, | |
| "eval_mar_100_3": 0.798, | |
| "eval_mar_100_car": -1.0, | |
| "eval_mar_large": 0.8825, | |
| "eval_mar_medium": 0.8163, | |
| "eval_mar_small": 0.67, | |
| "eval_runtime": 182.6955, | |
| "eval_samples_per_second": 8.013, | |
| "eval_steps_per_second": 1.002, | |
| "step": 13508 | |
| }, | |
| { | |
| "epoch": 22.80130293159609, | |
| "grad_norm": 14.897360801696777, | |
| "learning_rate": 2.7467105263157894e-05, | |
| "loss": 8.1853, | |
| "step": 14000 | |
| }, | |
| { | |
| "epoch": 23.0, | |
| "eval_loss": 7.0784382820129395, | |
| "eval_map": 0.5736, | |
| "eval_map_0": 0.4504, | |
| "eval_map_1": 0.635, | |
| "eval_map_2": 0.602, | |
| "eval_map_3": 0.6069, | |
| "eval_map_50": 0.8192, | |
| "eval_map_75": 0.6621, | |
| "eval_map_car": -1.0, | |
| "eval_map_large": 0.5826, | |
| "eval_map_medium": 0.6855, | |
| "eval_map_small": 0.3964, | |
| "eval_mar_1": 0.2854, | |
| "eval_mar_10": 0.6643, | |
| "eval_mar_100": 0.7559, | |
| "eval_mar_100_0": 0.6229, | |
| "eval_mar_100_1": 0.7679, | |
| "eval_mar_100_2": 0.8283, | |
| "eval_mar_100_3": 0.8043, | |
| "eval_mar_100_car": -1.0, | |
| "eval_mar_large": 0.9215, | |
| "eval_mar_medium": 0.8176, | |
| "eval_mar_small": 0.6567, | |
| "eval_runtime": 183.5891, | |
| "eval_samples_per_second": 7.974, | |
| "eval_steps_per_second": 0.997, | |
| "step": 14122 | |
| }, | |
| { | |
| "epoch": 23.615635179153095, | |
| "grad_norm": 27.044342041015625, | |
| "learning_rate": 2.6644736842105266e-05, | |
| "loss": 8.1821, | |
| "step": 14500 | |
| }, | |
| { | |
| "epoch": 24.0, | |
| "eval_loss": 7.3834123611450195, | |
| "eval_map": 0.5515, | |
| "eval_map_0": 0.4372, | |
| "eval_map_1": 0.5854, | |
| "eval_map_2": 0.6004, | |
| "eval_map_3": 0.5831, | |
| "eval_map_50": 0.7995, | |
| "eval_map_75": 0.6268, | |
| "eval_map_car": -1.0, | |
| "eval_map_large": 0.5885, | |
| "eval_map_medium": 0.6514, | |
| "eval_map_small": 0.3762, | |
| "eval_mar_1": 0.279, | |
| "eval_mar_10": 0.651, | |
| "eval_mar_100": 0.7388, | |
| "eval_mar_100_0": 0.611, | |
| "eval_mar_100_1": 0.7471, | |
| "eval_mar_100_2": 0.818, | |
| "eval_mar_100_3": 0.7789, | |
| "eval_mar_100_car": -1.0, | |
| "eval_mar_large": 0.8147, | |
| "eval_mar_medium": 0.7942, | |
| "eval_mar_small": 0.6477, | |
| "eval_runtime": 179.2521, | |
| "eval_samples_per_second": 8.167, | |
| "eval_steps_per_second": 1.021, | |
| "step": 14736 | |
| }, | |
| { | |
| "epoch": 24.429967426710096, | |
| "grad_norm": 19.333003997802734, | |
| "learning_rate": 2.5822368421052635e-05, | |
| "loss": 8.1817, | |
| "step": 15000 | |
| }, | |
| { | |
| "epoch": 25.0, | |
| "eval_loss": 7.422455310821533, | |
| "eval_map": 0.5605, | |
| "eval_map_0": 0.4341, | |
| "eval_map_1": 0.591, | |
| "eval_map_2": 0.617, | |
| "eval_map_3": 0.5998, | |
| "eval_map_50": 0.8048, | |
| "eval_map_75": 0.6456, | |
| "eval_map_car": -1.0, | |
| "eval_map_large": 0.5791, | |
| "eval_map_medium": 0.6644, | |
| "eval_map_small": 0.3895, | |
| "eval_mar_1": 0.2863, | |
| "eval_mar_10": 0.6607, | |
| "eval_mar_100": 0.7507, | |
| "eval_mar_100_0": 0.612, | |
| "eval_mar_100_1": 0.7612, | |
| "eval_mar_100_2": 0.8358, | |
| "eval_mar_100_3": 0.7939, | |
| "eval_mar_100_car": -1.0, | |
| "eval_mar_large": 0.893, | |
| "eval_mar_medium": 0.8123, | |
| "eval_mar_small": 0.6559, | |
| "eval_runtime": 181.3862, | |
| "eval_samples_per_second": 8.071, | |
| "eval_steps_per_second": 1.009, | |
| "step": 15350 | |
| }, | |
| { | |
| "epoch": 25.2442996742671, | |
| "grad_norm": 228.93096923828125, | |
| "learning_rate": 2.5e-05, | |
| "loss": 8.1344, | |
| "step": 15500 | |
| }, | |
| { | |
| "epoch": 26.0, | |
| "eval_loss": 7.033198356628418, | |
| "eval_map": 0.5848, | |
| "eval_map_0": 0.4509, | |
| "eval_map_1": 0.616, | |
| "eval_map_2": 0.6536, | |
| "eval_map_3": 0.6189, | |
| "eval_map_50": 0.829, | |
| "eval_map_75": 0.6723, | |
| "eval_map_car": -1.0, | |
| "eval_map_large": 0.6149, | |
| "eval_map_medium": 0.6923, | |
| "eval_map_small": 0.4007, | |
| "eval_mar_1": 0.2902, | |
| "eval_mar_10": 0.668, | |
| "eval_mar_100": 0.7525, | |
| "eval_mar_100_0": 0.6128, | |
| "eval_mar_100_1": 0.7667, | |
| "eval_mar_100_2": 0.8318, | |
| "eval_mar_100_3": 0.7987, | |
| "eval_mar_100_car": -1.0, | |
| "eval_mar_large": 0.8237, | |
| "eval_mar_medium": 0.8172, | |
| "eval_mar_small": 0.6551, | |
| "eval_runtime": 185.2166, | |
| "eval_samples_per_second": 7.904, | |
| "eval_steps_per_second": 0.988, | |
| "step": 15964 | |
| }, | |
| { | |
| "epoch": 26.058631921824105, | |
| "grad_norm": 17.044540405273438, | |
| "learning_rate": 2.4177631578947367e-05, | |
| "loss": 8.1957, | |
| "step": 16000 | |
| }, | |
| { | |
| "epoch": 26.872964169381106, | |
| "grad_norm": 46.24415969848633, | |
| "learning_rate": 2.335526315789474e-05, | |
| "loss": 8.1413, | |
| "step": 16500 | |
| }, | |
| { | |
| "epoch": 27.0, | |
| "eval_loss": 7.384748458862305, | |
| "eval_map": 0.5749, | |
| "eval_map_0": 0.4296, | |
| "eval_map_1": 0.598, | |
| "eval_map_2": 0.6319, | |
| "eval_map_3": 0.6401, | |
| "eval_map_50": 0.822, | |
| "eval_map_75": 0.6598, | |
| "eval_map_car": -1.0, | |
| "eval_map_large": 0.6213, | |
| "eval_map_medium": 0.6761, | |
| "eval_map_small": 0.3983, | |
| "eval_mar_1": 0.2903, | |
| "eval_mar_10": 0.6673, | |
| "eval_mar_100": 0.7561, | |
| "eval_mar_100_0": 0.6147, | |
| "eval_mar_100_1": 0.7711, | |
| "eval_mar_100_2": 0.8324, | |
| "eval_mar_100_3": 0.8059, | |
| "eval_mar_100_car": -1.0, | |
| "eval_mar_large": 0.8271, | |
| "eval_mar_medium": 0.8151, | |
| "eval_mar_small": 0.6641, | |
| "eval_runtime": 180.7624, | |
| "eval_samples_per_second": 8.099, | |
| "eval_steps_per_second": 1.012, | |
| "step": 16578 | |
| }, | |
| { | |
| "epoch": 27.68729641693811, | |
| "grad_norm": 18.294918060302734, | |
| "learning_rate": 2.2532894736842105e-05, | |
| "loss": 8.1337, | |
| "step": 17000 | |
| }, | |
| { | |
| "epoch": 28.0, | |
| "eval_loss": 7.090678691864014, | |
| "eval_map": 0.5878, | |
| "eval_map_0": 0.4438, | |
| "eval_map_1": 0.6199, | |
| "eval_map_2": 0.654, | |
| "eval_map_3": 0.6334, | |
| "eval_map_50": 0.8313, | |
| "eval_map_75": 0.6704, | |
| "eval_map_car": -1.0, | |
| "eval_map_large": 0.6295, | |
| "eval_map_medium": 0.6864, | |
| "eval_map_small": 0.4041, | |
| "eval_mar_1": 0.2902, | |
| "eval_mar_10": 0.6697, | |
| "eval_mar_100": 0.7582, | |
| "eval_mar_100_0": 0.6194, | |
| "eval_mar_100_1": 0.7765, | |
| "eval_mar_100_2": 0.8335, | |
| "eval_mar_100_3": 0.8033, | |
| "eval_mar_100_car": -1.0, | |
| "eval_mar_large": 0.9033, | |
| "eval_mar_medium": 0.8165, | |
| "eval_mar_small": 0.6668, | |
| "eval_runtime": 184.7196, | |
| "eval_samples_per_second": 7.926, | |
| "eval_steps_per_second": 0.991, | |
| "step": 17192 | |
| }, | |
| { | |
| "epoch": 28.501628664495115, | |
| "grad_norm": 14.970977783203125, | |
| "learning_rate": 2.1710526315789474e-05, | |
| "loss": 8.1769, | |
| "step": 17500 | |
| }, | |
| { | |
| "epoch": 29.0, | |
| "eval_loss": 7.129848003387451, | |
| "eval_map": 0.5891, | |
| "eval_map_0": 0.4284, | |
| "eval_map_1": 0.6261, | |
| "eval_map_2": 0.6551, | |
| "eval_map_3": 0.6469, | |
| "eval_map_50": 0.8296, | |
| "eval_map_75": 0.6726, | |
| "eval_map_car": -1.0, | |
| "eval_map_large": 0.643, | |
| "eval_map_medium": 0.6941, | |
| "eval_map_small": 0.3952, | |
| "eval_mar_1": 0.2896, | |
| "eval_mar_10": 0.6685, | |
| "eval_mar_100": 0.7544, | |
| "eval_mar_100_0": 0.6136, | |
| "eval_mar_100_1": 0.7737, | |
| "eval_mar_100_2": 0.8359, | |
| "eval_mar_100_3": 0.7943, | |
| "eval_mar_100_car": -1.0, | |
| "eval_mar_large": 0.914, | |
| "eval_mar_medium": 0.8078, | |
| "eval_mar_small": 0.6674, | |
| "eval_runtime": 185.3034, | |
| "eval_samples_per_second": 7.901, | |
| "eval_steps_per_second": 0.988, | |
| "step": 17806 | |
| }, | |
| { | |
| "epoch": 29.315960912052116, | |
| "grad_norm": 29.352230072021484, | |
| "learning_rate": 2.0888157894736843e-05, | |
| "loss": 8.1847, | |
| "step": 18000 | |
| }, | |
| { | |
| "epoch": 30.0, | |
| "eval_loss": 6.976243495941162, | |
| "eval_map": 0.591, | |
| "eval_map_0": 0.4292, | |
| "eval_map_1": 0.623, | |
| "eval_map_2": 0.6648, | |
| "eval_map_3": 0.6469, | |
| "eval_map_50": 0.8374, | |
| "eval_map_75": 0.6784, | |
| "eval_map_car": -1.0, | |
| "eval_map_large": 0.6454, | |
| "eval_map_medium": 0.6937, | |
| "eval_map_small": 0.4107, | |
| "eval_mar_1": 0.292, | |
| "eval_mar_10": 0.6668, | |
| "eval_mar_100": 0.7518, | |
| "eval_mar_100_0": 0.605, | |
| "eval_mar_100_1": 0.7699, | |
| "eval_mar_100_2": 0.8344, | |
| "eval_mar_100_3": 0.7977, | |
| "eval_mar_100_car": -1.0, | |
| "eval_mar_large": 0.8994, | |
| "eval_mar_medium": 0.8158, | |
| "eval_mar_small": 0.654, | |
| "eval_runtime": 184.3423, | |
| "eval_samples_per_second": 7.942, | |
| "eval_steps_per_second": 0.993, | |
| "step": 18420 | |
| }, | |
| { | |
| "epoch": 30.13029315960912, | |
| "grad_norm": 40.824623107910156, | |
| "learning_rate": 2.0065789473684213e-05, | |
| "loss": 8.1422, | |
| "step": 18500 | |
| }, | |
| { | |
| "epoch": 30.944625407166125, | |
| "grad_norm": 20.03860855102539, | |
| "learning_rate": 1.924342105263158e-05, | |
| "loss": 8.1351, | |
| "step": 19000 | |
| }, | |
| { | |
| "epoch": 31.0, | |
| "eval_loss": 7.370461463928223, | |
| "eval_map": 0.5645, | |
| "eval_map_0": 0.4007, | |
| "eval_map_1": 0.6098, | |
| "eval_map_2": 0.6208, | |
| "eval_map_3": 0.6268, | |
| "eval_map_50": 0.8036, | |
| "eval_map_75": 0.6459, | |
| "eval_map_car": -1.0, | |
| "eval_map_large": 0.6279, | |
| "eval_map_medium": 0.6629, | |
| "eval_map_small": 0.3785, | |
| "eval_mar_1": 0.2838, | |
| "eval_mar_10": 0.6611, | |
| "eval_mar_100": 0.746, | |
| "eval_mar_100_0": 0.5994, | |
| "eval_mar_100_1": 0.7662, | |
| "eval_mar_100_2": 0.8244, | |
| "eval_mar_100_3": 0.7939, | |
| "eval_mar_100_car": -1.0, | |
| "eval_mar_large": 0.8259, | |
| "eval_mar_medium": 0.8084, | |
| "eval_mar_small": 0.6491, | |
| "eval_runtime": 187.325, | |
| "eval_samples_per_second": 7.815, | |
| "eval_steps_per_second": 0.977, | |
| "step": 19034 | |
| }, | |
| { | |
| "epoch": 31.758957654723126, | |
| "grad_norm": 16.34675407409668, | |
| "learning_rate": 1.8421052631578947e-05, | |
| "loss": 8.07, | |
| "step": 19500 | |
| }, | |
| { | |
| "epoch": 32.0, | |
| "eval_loss": 7.263747692108154, | |
| "eval_map": 0.5852, | |
| "eval_map_0": 0.4196, | |
| "eval_map_1": 0.6284, | |
| "eval_map_2": 0.6685, | |
| "eval_map_3": 0.6243, | |
| "eval_map_50": 0.8242, | |
| "eval_map_75": 0.6704, | |
| "eval_map_car": -1.0, | |
| "eval_map_large": 0.6199, | |
| "eval_map_medium": 0.6888, | |
| "eval_map_small": 0.4022, | |
| "eval_mar_1": 0.2931, | |
| "eval_mar_10": 0.6628, | |
| "eval_mar_100": 0.74, | |
| "eval_mar_100_0": 0.5883, | |
| "eval_mar_100_1": 0.7582, | |
| "eval_mar_100_2": 0.8252, | |
| "eval_mar_100_3": 0.7883, | |
| "eval_mar_100_car": -1.0, | |
| "eval_mar_large": 0.9056, | |
| "eval_mar_medium": 0.8086, | |
| "eval_mar_small": 0.6382, | |
| "eval_runtime": 180.7211, | |
| "eval_samples_per_second": 8.101, | |
| "eval_steps_per_second": 1.013, | |
| "step": 19648 | |
| }, | |
| { | |
| "epoch": 32.57328990228013, | |
| "grad_norm": 18.07789421081543, | |
| "learning_rate": 1.7598684210526316e-05, | |
| "loss": 8.1698, | |
| "step": 20000 | |
| }, | |
| { | |
| "epoch": 33.0, | |
| "eval_loss": 7.357419013977051, | |
| "eval_map": 0.5703, | |
| "eval_map_0": 0.3814, | |
| "eval_map_1": 0.6016, | |
| "eval_map_2": 0.6683, | |
| "eval_map_3": 0.6301, | |
| "eval_map_50": 0.8092, | |
| "eval_map_75": 0.654, | |
| "eval_map_car": -1.0, | |
| "eval_map_large": 0.6201, | |
| "eval_map_medium": 0.6713, | |
| "eval_map_small": 0.3946, | |
| "eval_mar_1": 0.2899, | |
| "eval_mar_10": 0.6561, | |
| "eval_mar_100": 0.7351, | |
| "eval_mar_100_0": 0.5621, | |
| "eval_mar_100_1": 0.7623, | |
| "eval_mar_100_2": 0.8265, | |
| "eval_mar_100_3": 0.7895, | |
| "eval_mar_100_car": -1.0, | |
| "eval_mar_large": 0.8959, | |
| "eval_mar_medium": 0.8112, | |
| "eval_mar_small": 0.6219, | |
| "eval_runtime": 184.4966, | |
| "eval_samples_per_second": 7.935, | |
| "eval_steps_per_second": 0.992, | |
| "step": 20262 | |
| }, | |
| { | |
| "epoch": 33.387622149837135, | |
| "grad_norm": 41.847312927246094, | |
| "learning_rate": 1.6776315789473686e-05, | |
| "loss": 8.1261, | |
| "step": 20500 | |
| }, | |
| { | |
| "epoch": 34.0, | |
| "eval_loss": 7.17885684967041, | |
| "eval_map": 0.5827, | |
| "eval_map_0": 0.4291, | |
| "eval_map_1": 0.6229, | |
| "eval_map_2": 0.6593, | |
| "eval_map_3": 0.6195, | |
| "eval_map_50": 0.8256, | |
| "eval_map_75": 0.6724, | |
| "eval_map_car": -1.0, | |
| "eval_map_large": 0.6244, | |
| "eval_map_medium": 0.6906, | |
| "eval_map_small": 0.3996, | |
| "eval_mar_1": 0.2906, | |
| "eval_mar_10": 0.6687, | |
| "eval_mar_100": 0.751, | |
| "eval_mar_100_0": 0.6062, | |
| "eval_mar_100_1": 0.7689, | |
| "eval_mar_100_2": 0.835, | |
| "eval_mar_100_3": 0.7941, | |
| "eval_mar_100_car": -1.0, | |
| "eval_mar_large": 0.8985, | |
| "eval_mar_medium": 0.8182, | |
| "eval_mar_small": 0.6506, | |
| "eval_runtime": 184.4109, | |
| "eval_samples_per_second": 7.939, | |
| "eval_steps_per_second": 0.992, | |
| "step": 20876 | |
| }, | |
| { | |
| "epoch": 34.20195439739414, | |
| "grad_norm": 54.992149353027344, | |
| "learning_rate": 1.5953947368421055e-05, | |
| "loss": 8.061, | |
| "step": 21000 | |
| }, | |
| { | |
| "epoch": 35.0, | |
| "eval_loss": 7.313948631286621, | |
| "eval_map": 0.5795, | |
| "eval_map_0": 0.4389, | |
| "eval_map_1": 0.6072, | |
| "eval_map_2": 0.6606, | |
| "eval_map_3": 0.6113, | |
| "eval_map_50": 0.8264, | |
| "eval_map_75": 0.6634, | |
| "eval_map_car": -1.0, | |
| "eval_map_large": 0.6176, | |
| "eval_map_medium": 0.6836, | |
| "eval_map_small": 0.3939, | |
| "eval_mar_1": 0.2891, | |
| "eval_mar_10": 0.6593, | |
| "eval_mar_100": 0.7404, | |
| "eval_mar_100_0": 0.6097, | |
| "eval_mar_100_1": 0.7572, | |
| "eval_mar_100_2": 0.8215, | |
| "eval_mar_100_3": 0.7734, | |
| "eval_mar_100_car": -1.0, | |
| "eval_mar_large": 0.8192, | |
| "eval_mar_medium": 0.8056, | |
| "eval_mar_small": 0.6407, | |
| "eval_runtime": 187.3117, | |
| "eval_samples_per_second": 7.816, | |
| "eval_steps_per_second": 0.977, | |
| "step": 21490 | |
| }, | |
| { | |
| "epoch": 35.01628664495114, | |
| "grad_norm": 27.32745933532715, | |
| "learning_rate": 1.5131578947368422e-05, | |
| "loss": 7.9993, | |
| "step": 21500 | |
| }, | |
| { | |
| "epoch": 35.83061889250814, | |
| "grad_norm": 20.85877227783203, | |
| "learning_rate": 1.430921052631579e-05, | |
| "loss": 7.9641, | |
| "step": 22000 | |
| }, | |
| { | |
| "epoch": 36.0, | |
| "eval_loss": 7.176805019378662, | |
| "eval_map": 0.6014, | |
| "eval_map_0": 0.4537, | |
| "eval_map_1": 0.6438, | |
| "eval_map_2": 0.6762, | |
| "eval_map_3": 0.6317, | |
| "eval_map_50": 0.8479, | |
| "eval_map_75": 0.6898, | |
| "eval_map_car": -1.0, | |
| "eval_map_large": 0.628, | |
| "eval_map_medium": 0.7014, | |
| "eval_map_small": 0.4244, | |
| "eval_mar_1": 0.2941, | |
| "eval_mar_10": 0.6697, | |
| "eval_mar_100": 0.7533, | |
| "eval_mar_100_0": 0.6248, | |
| "eval_mar_100_1": 0.7733, | |
| "eval_mar_100_2": 0.8256, | |
| "eval_mar_100_3": 0.7893, | |
| "eval_mar_100_car": -1.0, | |
| "eval_mar_large": 0.8969, | |
| "eval_mar_medium": 0.8141, | |
| "eval_mar_small": 0.6553, | |
| "eval_runtime": 185.4828, | |
| "eval_samples_per_second": 7.893, | |
| "eval_steps_per_second": 0.987, | |
| "step": 22104 | |
| }, | |
| { | |
| "epoch": 36.644951140065146, | |
| "grad_norm": 19.830358505249023, | |
| "learning_rate": 1.3486842105263159e-05, | |
| "loss": 7.9716, | |
| "step": 22500 | |
| }, | |
| { | |
| "epoch": 37.0, | |
| "eval_loss": 7.289509296417236, | |
| "eval_map": 0.5936, | |
| "eval_map_0": 0.4326, | |
| "eval_map_1": 0.6268, | |
| "eval_map_2": 0.6731, | |
| "eval_map_3": 0.6418, | |
| "eval_map_50": 0.8345, | |
| "eval_map_75": 0.6807, | |
| "eval_map_car": -1.0, | |
| "eval_map_large": 0.63, | |
| "eval_map_medium": 0.6932, | |
| "eval_map_small": 0.4056, | |
| "eval_mar_1": 0.294, | |
| "eval_mar_10": 0.6678, | |
| "eval_mar_100": 0.7535, | |
| "eval_mar_100_0": 0.6129, | |
| "eval_mar_100_1": 0.7668, | |
| "eval_mar_100_2": 0.8366, | |
| "eval_mar_100_3": 0.7979, | |
| "eval_mar_100_car": -1.0, | |
| "eval_mar_large": 0.913, | |
| "eval_mar_medium": 0.8205, | |
| "eval_mar_small": 0.6519, | |
| "eval_runtime": 185.7485, | |
| "eval_samples_per_second": 7.882, | |
| "eval_steps_per_second": 0.985, | |
| "step": 22718 | |
| }, | |
| { | |
| "epoch": 37.45928338762215, | |
| "grad_norm": 29.213857650756836, | |
| "learning_rate": 1.2664473684210526e-05, | |
| "loss": 7.9311, | |
| "step": 23000 | |
| }, | |
| { | |
| "epoch": 38.0, | |
| "eval_loss": 7.100484371185303, | |
| "eval_map": 0.5974, | |
| "eval_map_0": 0.4456, | |
| "eval_map_1": 0.6221, | |
| "eval_map_2": 0.6801, | |
| "eval_map_3": 0.6419, | |
| "eval_map_50": 0.8438, | |
| "eval_map_75": 0.6817, | |
| "eval_map_car": -1.0, | |
| "eval_map_large": 0.6267, | |
| "eval_map_medium": 0.6929, | |
| "eval_map_small": 0.4084, | |
| "eval_mar_1": 0.2923, | |
| "eval_mar_10": 0.6664, | |
| "eval_mar_100": 0.7482, | |
| "eval_mar_100_0": 0.6127, | |
| "eval_mar_100_1": 0.7589, | |
| "eval_mar_100_2": 0.8305, | |
| "eval_mar_100_3": 0.7904, | |
| "eval_mar_100_car": -1.0, | |
| "eval_mar_large": 0.8973, | |
| "eval_mar_medium": 0.8147, | |
| "eval_mar_small": 0.6449, | |
| "eval_runtime": 186.6631, | |
| "eval_samples_per_second": 7.843, | |
| "eval_steps_per_second": 0.98, | |
| "step": 23332 | |
| }, | |
| { | |
| "epoch": 38.273615635179155, | |
| "grad_norm": 14.58039379119873, | |
| "learning_rate": 1.1842105263157895e-05, | |
| "loss": 7.9469, | |
| "step": 23500 | |
| }, | |
| { | |
| "epoch": 39.0, | |
| "eval_loss": 7.078758716583252, | |
| "eval_map": 0.6061, | |
| "eval_map_0": 0.4562, | |
| "eval_map_1": 0.6347, | |
| "eval_map_2": 0.687, | |
| "eval_map_3": 0.6465, | |
| "eval_map_50": 0.8472, | |
| "eval_map_75": 0.6936, | |
| "eval_map_car": -1.0, | |
| "eval_map_large": 0.6352, | |
| "eval_map_medium": 0.7063, | |
| "eval_map_small": 0.4165, | |
| "eval_mar_1": 0.2956, | |
| "eval_mar_10": 0.6718, | |
| "eval_mar_100": 0.7541, | |
| "eval_mar_100_0": 0.6199, | |
| "eval_mar_100_1": 0.7679, | |
| "eval_mar_100_2": 0.8329, | |
| "eval_mar_100_3": 0.7955, | |
| "eval_mar_100_car": -1.0, | |
| "eval_mar_large": 0.8999, | |
| "eval_mar_medium": 0.8198, | |
| "eval_mar_small": 0.6504, | |
| "eval_runtime": 184.1637, | |
| "eval_samples_per_second": 7.949, | |
| "eval_steps_per_second": 0.994, | |
| "step": 23946 | |
| }, | |
| { | |
| "epoch": 39.08794788273616, | |
| "grad_norm": 17.678525924682617, | |
| "learning_rate": 1.1019736842105263e-05, | |
| "loss": 7.932, | |
| "step": 24000 | |
| }, | |
| { | |
| "epoch": 39.90228013029316, | |
| "grad_norm": 23.1060733795166, | |
| "learning_rate": 1.0197368421052632e-05, | |
| "loss": 7.8875, | |
| "step": 24500 | |
| }, | |
| { | |
| "epoch": 40.0, | |
| "eval_loss": 7.225273609161377, | |
| "eval_map": 0.5976, | |
| "eval_map_0": 0.4381, | |
| "eval_map_1": 0.6228, | |
| "eval_map_2": 0.6892, | |
| "eval_map_3": 0.6401, | |
| "eval_map_50": 0.8335, | |
| "eval_map_75": 0.6871, | |
| "eval_map_car": -1.0, | |
| "eval_map_large": 0.6327, | |
| "eval_map_medium": 0.702, | |
| "eval_map_small": 0.4066, | |
| "eval_mar_1": 0.2922, | |
| "eval_mar_10": 0.6699, | |
| "eval_mar_100": 0.7504, | |
| "eval_mar_100_0": 0.6031, | |
| "eval_mar_100_1": 0.7638, | |
| "eval_mar_100_2": 0.8397, | |
| "eval_mar_100_3": 0.7948, | |
| "eval_mar_100_car": -1.0, | |
| "eval_mar_large": 0.9017, | |
| "eval_mar_medium": 0.8204, | |
| "eval_mar_small": 0.6474, | |
| "eval_runtime": 183.9676, | |
| "eval_samples_per_second": 7.958, | |
| "eval_steps_per_second": 0.995, | |
| "step": 24560 | |
| }, | |
| { | |
| "epoch": 40.71661237785016, | |
| "grad_norm": 17.05281639099121, | |
| "learning_rate": 9.375000000000001e-06, | |
| "loss": 7.9047, | |
| "step": 25000 | |
| }, | |
| { | |
| "epoch": 41.0, | |
| "eval_loss": 6.954546928405762, | |
| "eval_map": 0.6108, | |
| "eval_map_0": 0.4489, | |
| "eval_map_1": 0.6488, | |
| "eval_map_2": 0.6987, | |
| "eval_map_3": 0.647, | |
| "eval_map_50": 0.8507, | |
| "eval_map_75": 0.6992, | |
| "eval_map_car": -1.0, | |
| "eval_map_large": 0.6383, | |
| "eval_map_medium": 0.7152, | |
| "eval_map_small": 0.4209, | |
| "eval_mar_1": 0.2955, | |
| "eval_mar_10": 0.672, | |
| "eval_mar_100": 0.7512, | |
| "eval_mar_100_0": 0.6051, | |
| "eval_mar_100_1": 0.7714, | |
| "eval_mar_100_2": 0.8358, | |
| "eval_mar_100_3": 0.7924, | |
| "eval_mar_100_car": -1.0, | |
| "eval_mar_large": 0.8977, | |
| "eval_mar_medium": 0.8201, | |
| "eval_mar_small": 0.6508, | |
| "eval_runtime": 184.3392, | |
| "eval_samples_per_second": 7.942, | |
| "eval_steps_per_second": 0.993, | |
| "step": 25174 | |
| }, | |
| { | |
| "epoch": 41.530944625407166, | |
| "grad_norm": 19.473421096801758, | |
| "learning_rate": 8.552631578947368e-06, | |
| "loss": 7.8802, | |
| "step": 25500 | |
| }, | |
| { | |
| "epoch": 42.0, | |
| "eval_loss": 6.817875862121582, | |
| "eval_map": 0.6087, | |
| "eval_map_0": 0.4591, | |
| "eval_map_1": 0.642, | |
| "eval_map_2": 0.6811, | |
| "eval_map_3": 0.6526, | |
| "eval_map_50": 0.8471, | |
| "eval_map_75": 0.6986, | |
| "eval_map_car": -1.0, | |
| "eval_map_large": 0.6409, | |
| "eval_map_medium": 0.7147, | |
| "eval_map_small": 0.4211, | |
| "eval_mar_1": 0.2962, | |
| "eval_mar_10": 0.6743, | |
| "eval_mar_100": 0.7514, | |
| "eval_mar_100_0": 0.6036, | |
| "eval_mar_100_1": 0.7717, | |
| "eval_mar_100_2": 0.8332, | |
| "eval_mar_100_3": 0.7971, | |
| "eval_mar_100_car": -1.0, | |
| "eval_mar_large": 0.8882, | |
| "eval_mar_medium": 0.8233, | |
| "eval_mar_small": 0.6452, | |
| "eval_runtime": 186.7464, | |
| "eval_samples_per_second": 7.84, | |
| "eval_steps_per_second": 0.98, | |
| "step": 25788 | |
| }, | |
| { | |
| "epoch": 42.34527687296417, | |
| "grad_norm": 18.864524841308594, | |
| "learning_rate": 7.730263157894737e-06, | |
| "loss": 7.8638, | |
| "step": 26000 | |
| }, | |
| { | |
| "epoch": 43.0, | |
| "eval_loss": 6.949310302734375, | |
| "eval_map": 0.6063, | |
| "eval_map_0": 0.45, | |
| "eval_map_1": 0.6433, | |
| "eval_map_2": 0.6873, | |
| "eval_map_3": 0.6445, | |
| "eval_map_50": 0.8418, | |
| "eval_map_75": 0.6956, | |
| "eval_map_car": -1.0, | |
| "eval_map_large": 0.644, | |
| "eval_map_medium": 0.7116, | |
| "eval_map_small": 0.4125, | |
| "eval_mar_1": 0.2934, | |
| "eval_mar_10": 0.6727, | |
| "eval_mar_100": 0.7507, | |
| "eval_mar_100_0": 0.605, | |
| "eval_mar_100_1": 0.7675, | |
| "eval_mar_100_2": 0.8324, | |
| "eval_mar_100_3": 0.7979, | |
| "eval_mar_100_car": -1.0, | |
| "eval_mar_large": 0.9027, | |
| "eval_mar_medium": 0.8234, | |
| "eval_mar_small": 0.6437, | |
| "eval_runtime": 185.175, | |
| "eval_samples_per_second": 7.906, | |
| "eval_steps_per_second": 0.988, | |
| "step": 26402 | |
| }, | |
| { | |
| "epoch": 43.159609120521175, | |
| "grad_norm": 16.354448318481445, | |
| "learning_rate": 6.9078947368421065e-06, | |
| "loss": 7.899, | |
| "step": 26500 | |
| }, | |
| { | |
| "epoch": 43.97394136807817, | |
| "grad_norm": 21.469823837280273, | |
| "learning_rate": 6.085526315789474e-06, | |
| "loss": 7.83, | |
| "step": 27000 | |
| }, | |
| { | |
| "epoch": 44.0, | |
| "eval_loss": 6.987934589385986, | |
| "eval_map": 0.6019, | |
| "eval_map_0": 0.442, | |
| "eval_map_1": 0.6391, | |
| "eval_map_2": 0.6782, | |
| "eval_map_3": 0.6484, | |
| "eval_map_50": 0.8386, | |
| "eval_map_75": 0.6914, | |
| "eval_map_car": -1.0, | |
| "eval_map_large": 0.635, | |
| "eval_map_medium": 0.7041, | |
| "eval_map_small": 0.4141, | |
| "eval_mar_1": 0.2955, | |
| "eval_mar_10": 0.6728, | |
| "eval_mar_100": 0.7526, | |
| "eval_mar_100_0": 0.605, | |
| "eval_mar_100_1": 0.7712, | |
| "eval_mar_100_2": 0.8352, | |
| "eval_mar_100_3": 0.7989, | |
| "eval_mar_100_car": -1.0, | |
| "eval_mar_large": 0.9011, | |
| "eval_mar_medium": 0.8196, | |
| "eval_mar_small": 0.6504, | |
| "eval_runtime": 184.0359, | |
| "eval_samples_per_second": 7.955, | |
| "eval_steps_per_second": 0.994, | |
| "step": 27016 | |
| } | |
| ], | |
| "logging_steps": 500, | |
| "max_steps": 30700, | |
| "num_input_tokens_seen": 0, | |
| "num_train_epochs": 50, | |
| "save_steps": 500, | |
| "stateful_callbacks": { | |
| "TrainerControl": { | |
| "args": { | |
| "should_epoch_stop": false, | |
| "should_evaluate": false, | |
| "should_log": false, | |
| "should_save": true, | |
| "should_training_stop": false | |
| }, | |
| "attributes": {} | |
| } | |
| }, | |
| "total_flos": 1.3676452856032679e+20, | |
| "train_batch_size": 16, | |
| "trial_name": null, | |
| "trial_params": null | |
| } | |