| { | |
| "best_metric": 0.9063403443920213, | |
| "best_model_checkpoint": "./saved_models/climamba_prompt_sbdh_gpt4_v2_0/checkpoint-192", | |
| "epoch": 8.0, | |
| "eval_steps": 500, | |
| "global_step": 192, | |
| "is_hyper_param_search": false, | |
| "is_local_process_zero": true, | |
| "is_world_process_zero": true, | |
| "log_history": [ | |
| { | |
| "epoch": 1.0, | |
| "grad_norm": 2.0851898193359375, | |
| "learning_rate": 0.0003, | |
| "loss": 1.4049, | |
| "step": 24 | |
| }, | |
| { | |
| "epoch": 1.0, | |
| "eval_acc_macro": 0.05554596988321858, | |
| "eval_acc_micro": 0.17600550017187433, | |
| "eval_auc_macro": 0.6058912515968368, | |
| "eval_auc_micro": 0.6488212280527301, | |
| "eval_f1_at_5": 0.25079507780987353, | |
| "eval_f1_at_8": 0.24432286984749932, | |
| "eval_f1_macro": 0.09066299141125865, | |
| "eval_f1_micro": 0.2993276819643204, | |
| "eval_loss": 0.394575834274292, | |
| "eval_prec_at_5": 0.16187214611872147, | |
| "eval_prec_at_8": 0.1436929223744292, | |
| "eval_prec_macro": 0.05666691972017087, | |
| "eval_prec_micro": 0.23115124153497826, | |
| "eval_rec_at_5": 0.5565068493150684, | |
| "eval_rec_at_8": 0.8152587519025876, | |
| "eval_rec_macro": 0.22834645669278442, | |
| "eval_rec_micro": 0.4245439469319714, | |
| "eval_runtime": 6.3514, | |
| "eval_samples_per_second": 137.922, | |
| "eval_steps_per_second": 17.319, | |
| "step": 24 | |
| }, | |
| { | |
| "epoch": 2.0, | |
| "grad_norm": 0.4896126985549927, | |
| "learning_rate": 0.0003, | |
| "loss": 0.3397, | |
| "step": 48 | |
| }, | |
| { | |
| "epoch": 2.0, | |
| "eval_acc_macro": 0.5307003853873716, | |
| "eval_acc_micro": 0.5814606741572625, | |
| "eval_auc_macro": 0.9303990338907523, | |
| "eval_auc_micro": 0.9371828621061181, | |
| "eval_f1_at_5": 0.4088221844905389, | |
| "eval_f1_at_8": 0.28901449364842885, | |
| "eval_f1_macro": 0.6465579930529745, | |
| "eval_f1_micro": 0.7353463587921194, | |
| "eval_loss": 0.18900679051876068, | |
| "eval_prec_at_5": 0.2614155251141553, | |
| "eval_prec_at_8": 0.16980593607305935, | |
| "eval_prec_macro": 0.7313857838097021, | |
| "eval_prec_micro": 0.7915869980878784, | |
| "eval_rec_at_5": 0.9374048706240486, | |
| "eval_rec_at_8": 0.9699391171993911, | |
| "eval_rec_macro": 0.6410238611610916, | |
| "eval_rec_micro": 0.6865671641790475, | |
| "eval_runtime": 6.4111, | |
| "eval_samples_per_second": 136.637, | |
| "eval_steps_per_second": 17.158, | |
| "step": 48 | |
| }, | |
| { | |
| "epoch": 3.0, | |
| "grad_norm": 0.1705099642276764, | |
| "learning_rate": 0.0003, | |
| "loss": 0.0916, | |
| "step": 72 | |
| }, | |
| { | |
| "epoch": 3.0, | |
| "eval_acc_macro": 0.8244239681810236, | |
| "eval_acc_micro": 0.8311011904761286, | |
| "eval_auc_macro": 0.9919756925761466, | |
| "eval_auc_micro": 0.9920428839397825, | |
| "eval_f1_at_5": 0.42784762619628447, | |
| "eval_f1_at_8": 0.29274329072668936, | |
| "eval_f1_macro": 0.9019972941167671, | |
| "eval_f1_micro": 0.907761072734587, | |
| "eval_loss": 0.0666864886879921, | |
| "eval_prec_at_5": 0.273972602739726, | |
| "eval_prec_at_8": 0.1720890410958904, | |
| "eval_prec_macro": 0.8888004919536745, | |
| "eval_prec_micro": 0.8900398406373793, | |
| "eval_rec_at_5": 0.976027397260274, | |
| "eval_rec_at_8": 0.9794520547945206, | |
| "eval_rec_macro": 0.9176060797993615, | |
| "eval_rec_micro": 0.9262023217246329, | |
| "eval_runtime": 6.4724, | |
| "eval_samples_per_second": 135.345, | |
| "eval_steps_per_second": 16.995, | |
| "step": 72 | |
| }, | |
| { | |
| "epoch": 4.0, | |
| "grad_norm": 0.16448086500167847, | |
| "learning_rate": 0.0003, | |
| "loss": 0.0471, | |
| "step": 96 | |
| }, | |
| { | |
| "epoch": 4.0, | |
| "eval_acc_macro": 0.8271437680276396, | |
| "eval_acc_micro": 0.8323529411764093, | |
| "eval_auc_macro": 0.993671108576282, | |
| "eval_auc_micro": 0.9937186337101654, | |
| "eval_f1_at_5": 0.4285323653750887, | |
| "eval_f1_at_8": 0.29274329072668936, | |
| "eval_f1_macro": 0.9038769548624298, | |
| "eval_f1_micro": 0.9085072231138916, | |
| "eval_loss": 0.06305181980133057, | |
| "eval_prec_at_5": 0.27442922374429224, | |
| "eval_prec_at_8": 0.1720890410958904, | |
| "eval_prec_macro": 0.8786407478949387, | |
| "eval_prec_micro": 0.8802488335924665, | |
| "eval_rec_at_5": 0.977359208523592, | |
| "eval_rec_at_8": 0.9794520547945206, | |
| "eval_rec_macro": 0.9344636739690416, | |
| "eval_rec_micro": 0.9386401326699055, | |
| "eval_runtime": 6.3196, | |
| "eval_samples_per_second": 138.617, | |
| "eval_steps_per_second": 17.406, | |
| "step": 96 | |
| }, | |
| { | |
| "epoch": 5.0, | |
| "grad_norm": 0.22940804064273834, | |
| "learning_rate": 0.0003, | |
| "loss": 0.0292, | |
| "step": 120 | |
| }, | |
| { | |
| "epoch": 5.0, | |
| "eval_acc_macro": 0.8308612963912455, | |
| "eval_acc_micro": 0.8424710424709774, | |
| "eval_auc_macro": 0.9930217118288992, | |
| "eval_auc_micro": 0.9944139892271574, | |
| "eval_f1_at_5": 0.4295136788854873, | |
| "eval_f1_at_8": 0.29274329072668936, | |
| "eval_f1_macro": 0.90558341223627, | |
| "eval_f1_micro": 0.9145012573343744, | |
| "eval_loss": 0.06955315172672272, | |
| "eval_prec_at_5": 0.2751141552511416, | |
| "eval_prec_at_8": 0.1720890410958904, | |
| "eval_prec_macro": 0.9134147654048902, | |
| "eval_prec_micro": 0.9245762711863623, | |
| "eval_rec_at_5": 0.9788812785388128, | |
| "eval_rec_at_8": 0.9794520547945206, | |
| "eval_rec_macro": 0.9000407679288313, | |
| "eval_rec_micro": 0.9046434494194938, | |
| "eval_runtime": 6.3259, | |
| "eval_samples_per_second": 138.479, | |
| "eval_steps_per_second": 17.389, | |
| "step": 120 | |
| }, | |
| { | |
| "epoch": 6.0, | |
| "grad_norm": 0.14079643785953522, | |
| "learning_rate": 0.0003, | |
| "loss": 0.0169, | |
| "step": 144 | |
| }, | |
| { | |
| "epoch": 6.0, | |
| "eval_acc_macro": 0.8220596818065521, | |
| "eval_acc_micro": 0.8319327731091801, | |
| "eval_auc_macro": 0.9927035556143388, | |
| "eval_auc_micro": 0.9939625962172801, | |
| "eval_f1_at_5": 0.4277378477874334, | |
| "eval_f1_at_8": 0.29274329072668936, | |
| "eval_f1_macro": 0.8993295914850711, | |
| "eval_f1_micro": 0.9082568807338691, | |
| "eval_loss": 0.07188171148300171, | |
| "eval_prec_at_5": 0.27397260273972607, | |
| "eval_prec_at_8": 0.1720890410958904, | |
| "eval_prec_macro": 0.9063614647378323, | |
| "eval_prec_micro": 0.913590604026769, | |
| "eval_rec_at_5": 0.9748858447488584, | |
| "eval_rec_at_8": 0.9794520547945206, | |
| "eval_rec_macro": 0.8937033781116202, | |
| "eval_rec_micro": 0.9029850746267908, | |
| "eval_runtime": 6.3449, | |
| "eval_samples_per_second": 138.064, | |
| "eval_steps_per_second": 17.337, | |
| "step": 144 | |
| }, | |
| { | |
| "epoch": 7.0, | |
| "grad_norm": 0.12200725823640823, | |
| "learning_rate": 0.0003, | |
| "loss": 0.0095, | |
| "step": 168 | |
| }, | |
| { | |
| "epoch": 7.0, | |
| "eval_acc_macro": 0.8261283592776246, | |
| "eval_acc_micro": 0.8320433436531863, | |
| "eval_auc_macro": 0.9914073448442141, | |
| "eval_auc_micro": 0.9922562843066707, | |
| "eval_f1_at_5": 0.42747783226694747, | |
| "eval_f1_at_8": 0.2925113305987028, | |
| "eval_f1_macro": 0.9026710595366999, | |
| "eval_f1_micro": 0.9083227714405654, | |
| "eval_loss": 0.09724259376525879, | |
| "eval_prec_at_5": 0.27374429223744295, | |
| "eval_prec_at_8": 0.17194634703196346, | |
| "eval_prec_macro": 0.9186608707729786, | |
| "eval_prec_micro": 0.9259259259258461, | |
| "eval_rec_at_5": 0.975076103500761, | |
| "eval_rec_at_8": 0.9788812785388128, | |
| "eval_rec_macro": 0.8890001231603133, | |
| "eval_rec_micro": 0.8913764510778697, | |
| "eval_runtime": 6.3561, | |
| "eval_samples_per_second": 137.82, | |
| "eval_steps_per_second": 17.306, | |
| "step": 168 | |
| }, | |
| { | |
| "epoch": 8.0, | |
| "grad_norm": 0.1273776739835739, | |
| "learning_rate": 0.0003, | |
| "loss": 0.0061, | |
| "step": 192 | |
| }, | |
| { | |
| "epoch": 8.0, | |
| "eval_acc_macro": 0.8331638553801589, | |
| "eval_acc_micro": 0.8417047184169831, | |
| "eval_auc_macro": 0.9887523718801591, | |
| "eval_auc_micro": 0.9910290762677764, | |
| "eval_f1_at_5": 0.42681135828205624, | |
| "eval_f1_at_8": 0.29225388037164507, | |
| "eval_f1_macro": 0.9063403443920213, | |
| "eval_f1_micro": 0.9140495867767839, | |
| "eval_loss": 0.09798464179039001, | |
| "eval_prec_at_5": 0.2732876712328767, | |
| "eval_prec_at_8": 0.17180365296803654, | |
| "eval_prec_macro": 0.9010794509609382, | |
| "eval_prec_micro": 0.9110378912684587, | |
| "eval_rec_at_5": 0.9739345509893454, | |
| "eval_rec_at_8": 0.9777397260273972, | |
| "eval_rec_macro": 0.9130526960570995, | |
| "eval_rec_micro": 0.9170812603647663, | |
| "eval_runtime": 6.4145, | |
| "eval_samples_per_second": 136.565, | |
| "eval_steps_per_second": 17.149, | |
| "step": 192 | |
| } | |
| ], | |
| "logging_steps": 500, | |
| "max_steps": 960, | |
| "num_input_tokens_seen": 0, | |
| "num_train_epochs": 40, | |
| "save_steps": 500, | |
| "total_flos": 0.0, | |
| "train_batch_size": 32, | |
| "trial_name": null, | |
| "trial_params": null | |
| } | |