| { | |
| "best_metric": 0.020110823214054108, | |
| "best_model_checkpoint": "./deberta-v3-base_finetuned_bluegennx_run2.19_2e/checkpoint-30642", | |
| "epoch": 2.0, | |
| "eval_steps": 500, | |
| "global_step": 30642, | |
| "is_hyper_param_search": false, | |
| "is_local_process_zero": true, | |
| "is_world_process_zero": true, | |
| "log_history": [ | |
| { | |
| "epoch": 0.03, | |
| "grad_norm": 2.541147470474243, | |
| "learning_rate": 4.078968836678088e-06, | |
| "loss": 1.5263, | |
| "step": 500 | |
| }, | |
| { | |
| "epoch": 0.07, | |
| "grad_norm": 1.968030571937561, | |
| "learning_rate": 8.157937673356177e-06, | |
| "loss": 0.2859, | |
| "step": 1000 | |
| }, | |
| { | |
| "epoch": 0.1, | |
| "grad_norm": 0.2945529818534851, | |
| "learning_rate": 1.2236906510034265e-05, | |
| "loss": 0.1322, | |
| "step": 1500 | |
| }, | |
| { | |
| "epoch": 0.13, | |
| "grad_norm": 1.4127711057662964, | |
| "learning_rate": 1.6315875346712353e-05, | |
| "loss": 0.0928, | |
| "step": 2000 | |
| }, | |
| { | |
| "epoch": 0.16, | |
| "grad_norm": 1.8087753057479858, | |
| "learning_rate": 2.039484418339044e-05, | |
| "loss": 0.0827, | |
| "step": 2500 | |
| }, | |
| { | |
| "epoch": 0.2, | |
| "grad_norm": 0.4666314125061035, | |
| "learning_rate": 2.447381302006853e-05, | |
| "loss": 0.0665, | |
| "step": 3000 | |
| }, | |
| { | |
| "epoch": 0.23, | |
| "grad_norm": 2.213435411453247, | |
| "learning_rate": 2.8552781856746613e-05, | |
| "loss": 0.0621, | |
| "step": 3500 | |
| }, | |
| { | |
| "epoch": 0.26, | |
| "grad_norm": 0.5923435091972351, | |
| "learning_rate": 3.2631750693424706e-05, | |
| "loss": 0.0622, | |
| "step": 4000 | |
| }, | |
| { | |
| "epoch": 0.29, | |
| "grad_norm": 0.3192698657512665, | |
| "learning_rate": 3.6710719530102796e-05, | |
| "loss": 0.0545, | |
| "step": 4500 | |
| }, | |
| { | |
| "epoch": 0.33, | |
| "grad_norm": 0.21803532540798187, | |
| "learning_rate": 4.078968836678088e-05, | |
| "loss": 0.0557, | |
| "step": 5000 | |
| }, | |
| { | |
| "epoch": 0.36, | |
| "grad_norm": 0.024202294647693634, | |
| "learning_rate": 4.486865720345897e-05, | |
| "loss": 0.0468, | |
| "step": 5500 | |
| }, | |
| { | |
| "epoch": 0.39, | |
| "grad_norm": 0.12951496243476868, | |
| "learning_rate": 4.894762604013706e-05, | |
| "loss": 0.0501, | |
| "step": 6000 | |
| }, | |
| { | |
| "epoch": 0.42, | |
| "grad_norm": 0.19036373496055603, | |
| "learning_rate": 4.9971745812299085e-05, | |
| "loss": 0.0521, | |
| "step": 6500 | |
| }, | |
| { | |
| "epoch": 0.46, | |
| "grad_norm": 1.3507754802703857, | |
| "learning_rate": 4.984440267634476e-05, | |
| "loss": 0.048, | |
| "step": 7000 | |
| }, | |
| { | |
| "epoch": 0.49, | |
| "grad_norm": 0.057900320738554, | |
| "learning_rate": 4.9615076778917275e-05, | |
| "loss": 0.0448, | |
| "step": 7500 | |
| }, | |
| { | |
| "epoch": 0.52, | |
| "grad_norm": 0.3218132257461548, | |
| "learning_rate": 4.928470947041259e-05, | |
| "loss": 0.0466, | |
| "step": 8000 | |
| }, | |
| { | |
| "epoch": 0.55, | |
| "grad_norm": 0.9866151213645935, | |
| "learning_rate": 4.885465686193794e-05, | |
| "loss": 0.0441, | |
| "step": 8500 | |
| }, | |
| { | |
| "epoch": 0.59, | |
| "grad_norm": 1.0197815895080566, | |
| "learning_rate": 4.832668425866738e-05, | |
| "loss": 0.0441, | |
| "step": 9000 | |
| }, | |
| { | |
| "epoch": 0.62, | |
| "grad_norm": 0.9317641854286194, | |
| "learning_rate": 4.7702958913513574e-05, | |
| "loss": 0.0445, | |
| "step": 9500 | |
| }, | |
| { | |
| "epoch": 0.65, | |
| "grad_norm": 0.06256090104579926, | |
| "learning_rate": 4.69860411308609e-05, | |
| "loss": 0.0407, | |
| "step": 10000 | |
| }, | |
| { | |
| "epoch": 0.69, | |
| "grad_norm": 0.36358851194381714, | |
| "learning_rate": 4.6178873756877836e-05, | |
| "loss": 0.0421, | |
| "step": 10500 | |
| }, | |
| { | |
| "epoch": 0.72, | |
| "grad_norm": 0.4968320429325104, | |
| "learning_rate": 4.528477009954938e-05, | |
| "loss": 0.0304, | |
| "step": 11000 | |
| }, | |
| { | |
| "epoch": 0.75, | |
| "grad_norm": 1.2788376808166504, | |
| "learning_rate": 4.4307400328016045e-05, | |
| "loss": 0.0411, | |
| "step": 11500 | |
| }, | |
| { | |
| "epoch": 0.78, | |
| "grad_norm": 0.6715700626373291, | |
| "learning_rate": 4.325077640704834e-05, | |
| "loss": 0.0336, | |
| "step": 12000 | |
| }, | |
| { | |
| "epoch": 0.82, | |
| "grad_norm": 1.854251503944397, | |
| "learning_rate": 4.211923562849836e-05, | |
| "loss": 0.0353, | |
| "step": 12500 | |
| }, | |
| { | |
| "epoch": 0.85, | |
| "grad_norm": 0.044830091297626495, | |
| "learning_rate": 4.0917422807329866e-05, | |
| "loss": 0.0361, | |
| "step": 13000 | |
| }, | |
| { | |
| "epoch": 0.88, | |
| "grad_norm": 0.010744288563728333, | |
| "learning_rate": 3.9650271215309246e-05, | |
| "loss": 0.0347, | |
| "step": 13500 | |
| }, | |
| { | |
| "epoch": 0.91, | |
| "grad_norm": 2.754666566848755, | |
| "learning_rate": 3.832298233062228e-05, | |
| "loss": 0.0316, | |
| "step": 14000 | |
| }, | |
| { | |
| "epoch": 0.95, | |
| "grad_norm": 1.647675633430481, | |
| "learning_rate": 3.6941004486541304e-05, | |
| "loss": 0.0273, | |
| "step": 14500 | |
| }, | |
| { | |
| "epoch": 0.98, | |
| "grad_norm": 0.013370398432016373, | |
| "learning_rate": 3.5510010506787223e-05, | |
| "loss": 0.0261, | |
| "step": 15000 | |
| }, | |
| { | |
| "epoch": 1.0, | |
| "eval_AADHAR_CARD_f1": 0.9613259668508287, | |
| "eval_AGE_f1": 0.9462576687116565, | |
| "eval_CITY_f1": 0.9541174456428695, | |
| "eval_COUNTRY_f1": 0.983173076923077, | |
| "eval_CREDITCARDCVV_f1": 0.9793072424651371, | |
| "eval_CREDITCARDNUMBER_f1": 0.926996918701114, | |
| "eval_DATEOFBIRTH_f1": 0.8766835016835016, | |
| "eval_DATE_f1": 0.9481049562682217, | |
| "eval_EMAIL_f1": 0.9793307086614174, | |
| "eval_EXPIRYDATE_f1": 0.9809104258443465, | |
| "eval_ORGANIZATION_f1": 0.9881575603557815, | |
| "eval_PAN_CARD_f1": 0.9750609413088318, | |
| "eval_PERSON_f1": 0.984000540066158, | |
| "eval_PHONENUMBER_f1": 0.9747039040795439, | |
| "eval_PINCODE_f1": 0.9834765366820886, | |
| "eval_SECONDARYADDRESS_f1": 0.9830883833299778, | |
| "eval_STATE_f1": 0.9619748161077173, | |
| "eval_TIME_f1": 0.9779588812743101, | |
| "eval_URL_f1": 0.9872625698324023, | |
| "eval_loss": 0.028734495863318443, | |
| "eval_overall_accuracy": 0.9933771207589489, | |
| "eval_overall_f1": 0.9699552059434065, | |
| "eval_overall_precision": 0.9619155967278834, | |
| "eval_overall_recall": 0.9781303365834848, | |
| "eval_runtime": 285.0273, | |
| "eval_samples_per_second": 53.753, | |
| "eval_steps_per_second": 13.441, | |
| "step": 15321 | |
| }, | |
| { | |
| "epoch": 1.01, | |
| "grad_norm": 0.2590649425983429, | |
| "learning_rate": 3.403587441938993e-05, | |
| "loss": 0.0302, | |
| "step": 15500 | |
| }, | |
| { | |
| "epoch": 1.04, | |
| "grad_norm": 0.24032843112945557, | |
| "learning_rate": 3.2524647344633493e-05, | |
| "loss": 0.0267, | |
| "step": 16000 | |
| }, | |
| { | |
| "epoch": 1.08, | |
| "grad_norm": 0.031069770455360413, | |
| "learning_rate": 3.0982532656062604e-05, | |
| "loss": 0.0244, | |
| "step": 16500 | |
| }, | |
| { | |
| "epoch": 1.11, | |
| "grad_norm": 0.65534907579422, | |
| "learning_rate": 2.9415860516511033e-05, | |
| "loss": 0.024, | |
| "step": 17000 | |
| }, | |
| { | |
| "epoch": 1.14, | |
| "grad_norm": 1.3654996156692505, | |
| "learning_rate": 2.783106189367786e-05, | |
| "loss": 0.0227, | |
| "step": 17500 | |
| }, | |
| { | |
| "epoch": 1.17, | |
| "grad_norm": 0.11274850368499756, | |
| "learning_rate": 2.623464216191404e-05, | |
| "loss": 0.0225, | |
| "step": 18000 | |
| }, | |
| { | |
| "epoch": 1.21, | |
| "grad_norm": 3.295466661453247, | |
| "learning_rate": 2.463315439858015e-05, | |
| "loss": 0.0208, | |
| "step": 18500 | |
| }, | |
| { | |
| "epoch": 1.24, | |
| "grad_norm": 1.3590203523635864, | |
| "learning_rate": 2.303317248459006e-05, | |
| "loss": 0.0246, | |
| "step": 19000 | |
| }, | |
| { | |
| "epoch": 1.27, | |
| "grad_norm": 0.02803308703005314, | |
| "learning_rate": 2.1441264119558903e-05, | |
| "loss": 0.022, | |
| "step": 19500 | |
| }, | |
| { | |
| "epoch": 1.31, | |
| "grad_norm": 0.3329419493675232, | |
| "learning_rate": 1.9863963862324454e-05, | |
| "loss": 0.0215, | |
| "step": 20000 | |
| }, | |
| { | |
| "epoch": 1.34, | |
| "grad_norm": 1.7840496301651, | |
| "learning_rate": 1.830774630750671e-05, | |
| "loss": 0.021, | |
| "step": 20500 | |
| }, | |
| { | |
| "epoch": 1.37, | |
| "grad_norm": 0.012373638339340687, | |
| "learning_rate": 1.6778999508212068e-05, | |
| "loss": 0.0192, | |
| "step": 21000 | |
| }, | |
| { | |
| "epoch": 1.4, | |
| "grad_norm": 0.7150644063949585, | |
| "learning_rate": 1.528399875397824e-05, | |
| "loss": 0.0193, | |
| "step": 21500 | |
| }, | |
| { | |
| "epoch": 1.44, | |
| "grad_norm": 1.5099574327468872, | |
| "learning_rate": 1.3828880811597606e-05, | |
| "loss": 0.0181, | |
| "step": 22000 | |
| }, | |
| { | |
| "epoch": 1.47, | |
| "grad_norm": 0.07326490432024002, | |
| "learning_rate": 1.2419618734556984e-05, | |
| "loss": 0.0184, | |
| "step": 22500 | |
| }, | |
| { | |
| "epoch": 1.5, | |
| "grad_norm": 0.23956900835037231, | |
| "learning_rate": 1.1061997344497197e-05, | |
| "loss": 0.0174, | |
| "step": 23000 | |
| }, | |
| { | |
| "epoch": 1.53, | |
| "grad_norm": 0.00975050963461399, | |
| "learning_rate": 9.761589485337946e-06, | |
| "loss": 0.0181, | |
| "step": 23500 | |
| }, | |
| { | |
| "epoch": 1.57, | |
| "grad_norm": 0.04456551373004913, | |
| "learning_rate": 8.523733147541155e-06, | |
| "loss": 0.0155, | |
| "step": 24000 | |
| }, | |
| { | |
| "epoch": 1.6, | |
| "grad_norm": 0.02919822372496128, | |
| "learning_rate": 7.3535095564145976e-06, | |
| "loss": 0.0181, | |
| "step": 24500 | |
| }, | |
| { | |
| "epoch": 1.63, | |
| "grad_norm": 0.07429083436727524, | |
| "learning_rate": 6.2557223144002006e-06, | |
| "loss": 0.0211, | |
| "step": 25000 | |
| }, | |
| { | |
| "epoch": 1.66, | |
| "grad_norm": 0.7385916113853455, | |
| "learning_rate": 5.234877682964934e-06, | |
| "loss": 0.0158, | |
| "step": 25500 | |
| }, | |
| { | |
| "epoch": 1.7, | |
| "grad_norm": 0.06193115562200546, | |
| "learning_rate": 4.295166085034444e-06, | |
| "loss": 0.0152, | |
| "step": 26000 | |
| }, | |
| { | |
| "epoch": 1.73, | |
| "grad_norm": 0.24567507207393646, | |
| "learning_rate": 3.440444903899198e-06, | |
| "loss": 0.0191, | |
| "step": 26500 | |
| }, | |
| { | |
| "epoch": 1.76, | |
| "grad_norm": 0.2641615867614746, | |
| "learning_rate": 2.6742226492012856e-06, | |
| "loss": 0.0145, | |
| "step": 27000 | |
| }, | |
| { | |
| "epoch": 1.79, | |
| "grad_norm": 0.020536798983812332, | |
| "learning_rate": 1.999644554998209e-06, | |
| "loss": 0.0151, | |
| "step": 27500 | |
| }, | |
| { | |
| "epoch": 1.83, | |
| "grad_norm": 0.5791102051734924, | |
| "learning_rate": 1.4194796690216673e-06, | |
| "loss": 0.0162, | |
| "step": 28000 | |
| }, | |
| { | |
| "epoch": 1.86, | |
| "grad_norm": 0.14257295429706573, | |
| "learning_rate": 9.361094861279646e-07, | |
| "loss": 0.0142, | |
| "step": 28500 | |
| }, | |
| { | |
| "epoch": 1.89, | |
| "grad_norm": 0.7881425619125366, | |
| "learning_rate": 5.515181725982282e-07, | |
| "loss": 0.0151, | |
| "step": 29000 | |
| }, | |
| { | |
| "epoch": 1.93, | |
| "grad_norm": 0.030007589608430862, | |
| "learning_rate": 2.672844214162862e-07, | |
| "loss": 0.0153, | |
| "step": 29500 | |
| }, | |
| { | |
| "epoch": 1.96, | |
| "grad_norm": 0.5309509634971619, | |
| "learning_rate": 8.457497195707431e-08, | |
| "loss": 0.0177, | |
| "step": 30000 | |
| }, | |
| { | |
| "epoch": 1.99, | |
| "grad_norm": 1.1870931386947632, | |
| "learning_rate": 4.139820686432927e-09, | |
| "loss": 0.0152, | |
| "step": 30500 | |
| }, | |
| { | |
| "epoch": 2.0, | |
| "eval_AADHAR_CARD_f1": 0.9837270341207349, | |
| "eval_AGE_f1": 0.9633416458852868, | |
| "eval_CITY_f1": 0.9842361227570016, | |
| "eval_COUNTRY_f1": 0.9843467790487658, | |
| "eval_CREDITCARDCVV_f1": 0.9878760664571171, | |
| "eval_CREDITCARDNUMBER_f1": 0.9416398138202648, | |
| "eval_DATEOFBIRTH_f1": 0.9023332645054718, | |
| "eval_DATE_f1": 0.9600118046333186, | |
| "eval_EMAIL_f1": 0.990012854741422, | |
| "eval_EXPIRYDATE_f1": 0.9912280701754386, | |
| "eval_ORGANIZATION_f1": 0.991032304086416, | |
| "eval_PAN_CARD_f1": 0.9867424242424242, | |
| "eval_PERSON_f1": 0.9877905928996216, | |
| "eval_PHONENUMBER_f1": 0.9857583321098224, | |
| "eval_PINCODE_f1": 0.9907161803713527, | |
| "eval_SECONDARYADDRESS_f1": 0.9877938061131848, | |
| "eval_STATE_f1": 0.9909125815947779, | |
| "eval_TIME_f1": 0.9819761530640541, | |
| "eval_URL_f1": 0.9948626312262676, | |
| "eval_loss": 0.020110823214054108, | |
| "eval_overall_accuracy": 0.9951943362620375, | |
| "eval_overall_f1": 0.9803088380243128, | |
| "eval_overall_precision": 0.9744924065102607, | |
| "eval_overall_recall": 0.9861951192640335, | |
| "eval_runtime": 251.2008, | |
| "eval_samples_per_second": 60.991, | |
| "eval_steps_per_second": 15.251, | |
| "step": 30642 | |
| }, | |
| { | |
| "epoch": 2.0, | |
| "step": 30642, | |
| "total_flos": 1.1909329832222172e+16, | |
| "train_loss": 0.06240972641763811, | |
| "train_runtime": 5473.6718, | |
| "train_samples_per_second": 22.391, | |
| "train_steps_per_second": 5.598 | |
| }, | |
| { | |
| "epoch": 2.0, | |
| "eval_AADHAR_CARD_f1": 0.9837270341207349, | |
| "eval_AGE_f1": 0.9633416458852868, | |
| "eval_CITY_f1": 0.9842361227570016, | |
| "eval_COUNTRY_f1": 0.9843467790487658, | |
| "eval_CREDITCARDCVV_f1": 0.9878760664571171, | |
| "eval_CREDITCARDNUMBER_f1": 0.9416398138202648, | |
| "eval_DATEOFBIRTH_f1": 0.9023332645054718, | |
| "eval_DATE_f1": 0.9600118046333186, | |
| "eval_EMAIL_f1": 0.990012854741422, | |
| "eval_EXPIRYDATE_f1": 0.9912280701754386, | |
| "eval_ORGANIZATION_f1": 0.991032304086416, | |
| "eval_PAN_CARD_f1": 0.9867424242424242, | |
| "eval_PERSON_f1": 0.9877905928996216, | |
| "eval_PHONENUMBER_f1": 0.9857583321098224, | |
| "eval_PINCODE_f1": 0.9907161803713527, | |
| "eval_SECONDARYADDRESS_f1": 0.9877938061131848, | |
| "eval_STATE_f1": 0.9909125815947779, | |
| "eval_TIME_f1": 0.9819761530640541, | |
| "eval_URL_f1": 0.9948626312262676, | |
| "eval_loss": 0.020110823214054108, | |
| "eval_overall_accuracy": 0.9951943362620375, | |
| "eval_overall_f1": 0.9803088380243128, | |
| "eval_overall_precision": 0.9744924065102607, | |
| "eval_overall_recall": 0.9861951192640335, | |
| "eval_runtime": 249.9695, | |
| "eval_samples_per_second": 61.291, | |
| "eval_steps_per_second": 15.326, | |
| "step": 30642 | |
| } | |
| ], | |
| "logging_steps": 500, | |
| "max_steps": 30642, | |
| "num_input_tokens_seen": 0, | |
| "num_train_epochs": 2, | |
| "save_steps": 500, | |
| "total_flos": 1.1909329832222172e+16, | |
| "train_batch_size": 4, | |
| "trial_name": null, | |
| "trial_params": null | |
| } | |