| { | |
| "best_metric": 0.4074449594438007, | |
| "best_model_checkpoint": "audio/train/checkpoint/wav2vec2-base/superb_si_42/checkpoint-43240", | |
| "epoch": 10.0, | |
| "eval_steps": 500, | |
| "global_step": 43240, | |
| "is_hyper_param_search": false, | |
| "is_local_process_zero": true, | |
| "is_world_process_zero": true, | |
| "log_history": [ | |
| { | |
| "epoch": 0.23126734505087881, | |
| "grad_norm": 1.405583381652832, | |
| "learning_rate": 1.1563367252543942e-05, | |
| "loss": 7.1089, | |
| "step": 1000 | |
| }, | |
| { | |
| "epoch": 0.46253469010175763, | |
| "grad_norm": 3.354081869125366, | |
| "learning_rate": 2.3126734505087884e-05, | |
| "loss": 6.7944, | |
| "step": 2000 | |
| }, | |
| { | |
| "epoch": 0.6938020351526365, | |
| "grad_norm": 4.870436668395996, | |
| "learning_rate": 3.469010175763183e-05, | |
| "loss": 6.3872, | |
| "step": 3000 | |
| }, | |
| { | |
| "epoch": 0.9250693802035153, | |
| "grad_norm": 6.949980735778809, | |
| "learning_rate": 4.625346901017577e-05, | |
| "loss": 6.1211, | |
| "step": 4000 | |
| }, | |
| { | |
| "epoch": 1.0, | |
| "eval_accuracy": 0.005069524913093859, | |
| "eval_loss": 6.53181791305542, | |
| "eval_runtime": 122.9456, | |
| "eval_samples_per_second": 56.155, | |
| "eval_steps_per_second": 3.514, | |
| "step": 4324 | |
| }, | |
| { | |
| "epoch": 1.1563367252543941, | |
| "grad_norm": 6.596773147583008, | |
| "learning_rate": 4.913146263747559e-05, | |
| "loss": 5.8538, | |
| "step": 5000 | |
| }, | |
| { | |
| "epoch": 1.3876040703052728, | |
| "grad_norm": 7.879302501678467, | |
| "learning_rate": 4.78466440538596e-05, | |
| "loss": 5.6256, | |
| "step": 6000 | |
| }, | |
| { | |
| "epoch": 1.6188714153561516, | |
| "grad_norm": 10.183388710021973, | |
| "learning_rate": 4.65618254702436e-05, | |
| "loss": 5.4321, | |
| "step": 7000 | |
| }, | |
| { | |
| "epoch": 1.8501387604070305, | |
| "grad_norm": 9.450615882873535, | |
| "learning_rate": 4.527700688662761e-05, | |
| "loss": 5.2457, | |
| "step": 8000 | |
| }, | |
| { | |
| "epoch": 2.0, | |
| "eval_accuracy": 0.03765932792584009, | |
| "eval_loss": 5.4292426109313965, | |
| "eval_runtime": 119.8822, | |
| "eval_samples_per_second": 57.59, | |
| "eval_steps_per_second": 3.604, | |
| "step": 8648 | |
| }, | |
| { | |
| "epoch": 2.0814061054579094, | |
| "grad_norm": 12.032964706420898, | |
| "learning_rate": 4.399218830301162e-05, | |
| "loss": 5.0305, | |
| "step": 9000 | |
| }, | |
| { | |
| "epoch": 2.3126734505087883, | |
| "grad_norm": 12.4326810836792, | |
| "learning_rate": 4.2707369719395625e-05, | |
| "loss": 4.8168, | |
| "step": 10000 | |
| }, | |
| { | |
| "epoch": 2.543940795559667, | |
| "grad_norm": 13.46068286895752, | |
| "learning_rate": 4.142255113577963e-05, | |
| "loss": 4.6195, | |
| "step": 11000 | |
| }, | |
| { | |
| "epoch": 2.7752081406105455, | |
| "grad_norm": 16.406736373901367, | |
| "learning_rate": 4.0137732552163636e-05, | |
| "loss": 4.4561, | |
| "step": 12000 | |
| }, | |
| { | |
| "epoch": 3.0, | |
| "eval_accuracy": 0.09154113557358054, | |
| "eval_loss": 4.708800315856934, | |
| "eval_runtime": 119.9183, | |
| "eval_samples_per_second": 57.573, | |
| "eval_steps_per_second": 3.602, | |
| "step": 12972 | |
| }, | |
| { | |
| "epoch": 3.0064754856614244, | |
| "grad_norm": 14.04348373413086, | |
| "learning_rate": 3.885291396854764e-05, | |
| "loss": 4.2699, | |
| "step": 13000 | |
| }, | |
| { | |
| "epoch": 3.2377428307123033, | |
| "grad_norm": 14.909345626831055, | |
| "learning_rate": 3.756809538493165e-05, | |
| "loss": 4.1135, | |
| "step": 14000 | |
| }, | |
| { | |
| "epoch": 3.469010175763182, | |
| "grad_norm": 13.025561332702637, | |
| "learning_rate": 3.628327680131566e-05, | |
| "loss": 3.9814, | |
| "step": 15000 | |
| }, | |
| { | |
| "epoch": 3.700277520814061, | |
| "grad_norm": 13.69780445098877, | |
| "learning_rate": 3.4998458217699664e-05, | |
| "loss": 3.8563, | |
| "step": 16000 | |
| }, | |
| { | |
| "epoch": 3.93154486586494, | |
| "grad_norm": 15.66323471069336, | |
| "learning_rate": 3.371363963408367e-05, | |
| "loss": 3.7443, | |
| "step": 17000 | |
| }, | |
| { | |
| "epoch": 4.0, | |
| "eval_accuracy": 0.15961761297798377, | |
| "eval_loss": 4.159976959228516, | |
| "eval_runtime": 119.8089, | |
| "eval_samples_per_second": 57.625, | |
| "eval_steps_per_second": 3.606, | |
| "step": 17296 | |
| }, | |
| { | |
| "epoch": 4.162812210915819, | |
| "grad_norm": 17.160215377807617, | |
| "learning_rate": 3.242882105046768e-05, | |
| "loss": 3.6263, | |
| "step": 18000 | |
| }, | |
| { | |
| "epoch": 4.394079555966697, | |
| "grad_norm": 16.623626708984375, | |
| "learning_rate": 3.114400246685168e-05, | |
| "loss": 3.5286, | |
| "step": 19000 | |
| }, | |
| { | |
| "epoch": 4.6253469010175765, | |
| "grad_norm": 18.177021026611328, | |
| "learning_rate": 2.9859183883235685e-05, | |
| "loss": 3.4055, | |
| "step": 20000 | |
| }, | |
| { | |
| "epoch": 4.856614246068455, | |
| "grad_norm": 16.725624084472656, | |
| "learning_rate": 2.8574365299619694e-05, | |
| "loss": 3.3365, | |
| "step": 21000 | |
| }, | |
| { | |
| "epoch": 5.0, | |
| "eval_accuracy": 0.2071263035921205, | |
| "eval_loss": 3.8531501293182373, | |
| "eval_runtime": 119.8981, | |
| "eval_samples_per_second": 57.582, | |
| "eval_steps_per_second": 3.603, | |
| "step": 21620 | |
| }, | |
| { | |
| "epoch": 5.087881591119334, | |
| "grad_norm": 17.34439468383789, | |
| "learning_rate": 2.72895467160037e-05, | |
| "loss": 3.2436, | |
| "step": 22000 | |
| }, | |
| { | |
| "epoch": 5.319148936170213, | |
| "grad_norm": 17.53919219970703, | |
| "learning_rate": 2.6004728132387708e-05, | |
| "loss": 3.1418, | |
| "step": 23000 | |
| }, | |
| { | |
| "epoch": 5.550416281221091, | |
| "grad_norm": 15.603684425354004, | |
| "learning_rate": 2.4719909548771713e-05, | |
| "loss": 3.0568, | |
| "step": 24000 | |
| }, | |
| { | |
| "epoch": 5.78168362627197, | |
| "grad_norm": 17.67478370666504, | |
| "learning_rate": 2.343509096515572e-05, | |
| "loss": 3.0029, | |
| "step": 25000 | |
| }, | |
| { | |
| "epoch": 6.0, | |
| "eval_accuracy": 0.2820104287369641, | |
| "eval_loss": 3.3280937671661377, | |
| "eval_runtime": 120.3561, | |
| "eval_samples_per_second": 57.363, | |
| "eval_steps_per_second": 3.589, | |
| "step": 25944 | |
| }, | |
| { | |
| "epoch": 6.012950971322849, | |
| "grad_norm": 17.622488021850586, | |
| "learning_rate": 2.2150272381539727e-05, | |
| "loss": 2.9546, | |
| "step": 26000 | |
| }, | |
| { | |
| "epoch": 6.244218316373728, | |
| "grad_norm": 19.121681213378906, | |
| "learning_rate": 2.0865453797923736e-05, | |
| "loss": 2.8432, | |
| "step": 27000 | |
| }, | |
| { | |
| "epoch": 6.475485661424607, | |
| "grad_norm": 19.73898696899414, | |
| "learning_rate": 1.958063521430774e-05, | |
| "loss": 2.7844, | |
| "step": 28000 | |
| }, | |
| { | |
| "epoch": 6.706753006475486, | |
| "grad_norm": 18.709022521972656, | |
| "learning_rate": 1.8295816630691746e-05, | |
| "loss": 2.734, | |
| "step": 29000 | |
| }, | |
| { | |
| "epoch": 6.938020351526364, | |
| "grad_norm": 17.50193214416504, | |
| "learning_rate": 1.7010998047075755e-05, | |
| "loss": 2.6762, | |
| "step": 30000 | |
| }, | |
| { | |
| "epoch": 7.0, | |
| "eval_accuracy": 0.3422653534183082, | |
| "eval_loss": 3.0051939487457275, | |
| "eval_runtime": 120.0955, | |
| "eval_samples_per_second": 57.488, | |
| "eval_steps_per_second": 3.597, | |
| "step": 30268 | |
| }, | |
| { | |
| "epoch": 7.169287696577244, | |
| "grad_norm": 18.220510482788086, | |
| "learning_rate": 1.572617946345976e-05, | |
| "loss": 2.5939, | |
| "step": 31000 | |
| }, | |
| { | |
| "epoch": 7.400555041628122, | |
| "grad_norm": 17.59881591796875, | |
| "learning_rate": 1.4441360879843766e-05, | |
| "loss": 2.5541, | |
| "step": 32000 | |
| }, | |
| { | |
| "epoch": 7.631822386679001, | |
| "grad_norm": 15.690144538879395, | |
| "learning_rate": 1.3156542296227773e-05, | |
| "loss": 2.5171, | |
| "step": 33000 | |
| }, | |
| { | |
| "epoch": 7.86308973172988, | |
| "grad_norm": 17.54080581665039, | |
| "learning_rate": 1.187172371261178e-05, | |
| "loss": 2.4949, | |
| "step": 34000 | |
| }, | |
| { | |
| "epoch": 8.0, | |
| "eval_accuracy": 0.37181344148319817, | |
| "eval_loss": 2.9019951820373535, | |
| "eval_runtime": 120.2932, | |
| "eval_samples_per_second": 57.393, | |
| "eval_steps_per_second": 3.591, | |
| "step": 34592 | |
| }, | |
| { | |
| "epoch": 8.094357076780758, | |
| "grad_norm": 19.050064086914062, | |
| "learning_rate": 1.0586905128995787e-05, | |
| "loss": 2.4261, | |
| "step": 35000 | |
| }, | |
| { | |
| "epoch": 8.325624421831638, | |
| "grad_norm": 18.02264404296875, | |
| "learning_rate": 9.302086545379794e-06, | |
| "loss": 2.4037, | |
| "step": 36000 | |
| }, | |
| { | |
| "epoch": 8.556891766882517, | |
| "grad_norm": 16.733930587768555, | |
| "learning_rate": 8.017267961763799e-06, | |
| "loss": 2.3584, | |
| "step": 37000 | |
| }, | |
| { | |
| "epoch": 8.788159111933394, | |
| "grad_norm": 20.43987464904785, | |
| "learning_rate": 6.732449378147807e-06, | |
| "loss": 2.3192, | |
| "step": 38000 | |
| }, | |
| { | |
| "epoch": 9.0, | |
| "eval_accuracy": 0.3952780996523754, | |
| "eval_loss": 2.7638044357299805, | |
| "eval_runtime": 120.1123, | |
| "eval_samples_per_second": 57.48, | |
| "eval_steps_per_second": 3.597, | |
| "step": 38916 | |
| }, | |
| { | |
| "epoch": 9.019426456984274, | |
| "grad_norm": 18.636188507080078, | |
| "learning_rate": 5.447630794531813e-06, | |
| "loss": 2.3129, | |
| "step": 39000 | |
| }, | |
| { | |
| "epoch": 9.250693802035153, | |
| "grad_norm": 21.51548194885254, | |
| "learning_rate": 4.162812210915819e-06, | |
| "loss": 2.2573, | |
| "step": 40000 | |
| }, | |
| { | |
| "epoch": 9.481961147086032, | |
| "grad_norm": 18.282093048095703, | |
| "learning_rate": 2.877993627299825e-06, | |
| "loss": 2.2447, | |
| "step": 41000 | |
| }, | |
| { | |
| "epoch": 9.71322849213691, | |
| "grad_norm": 21.595134735107422, | |
| "learning_rate": 1.5931750436838318e-06, | |
| "loss": 2.2316, | |
| "step": 42000 | |
| }, | |
| { | |
| "epoch": 9.94449583718779, | |
| "grad_norm": 21.02860450744629, | |
| "learning_rate": 3.0835646006783846e-07, | |
| "loss": 2.2312, | |
| "step": 43000 | |
| }, | |
| { | |
| "epoch": 10.0, | |
| "eval_accuracy": 0.4074449594438007, | |
| "eval_loss": 2.7118172645568848, | |
| "eval_runtime": 120.1679, | |
| "eval_samples_per_second": 57.453, | |
| "eval_steps_per_second": 3.595, | |
| "step": 43240 | |
| }, | |
| { | |
| "epoch": 10.0, | |
| "step": 43240, | |
| "total_flos": 1.260392330300448e+19, | |
| "train_loss": 3.683282250502284, | |
| "train_runtime": 4693.6982, | |
| "train_samples_per_second": 294.78, | |
| "train_steps_per_second": 9.212 | |
| } | |
| ], | |
| "logging_steps": 1000, | |
| "max_steps": 43240, | |
| "num_input_tokens_seen": 0, | |
| "num_train_epochs": 10, | |
| "save_steps": 500, | |
| "total_flos": 1.260392330300448e+19, | |
| "train_batch_size": 32, | |
| "trial_name": null, | |
| "trial_params": null | |
| } | |