| { |
| "best_metric": 0.0025361417792737484, |
| "best_model_checkpoint": "./codebert-vuln-tokcls/checkpoint-2875", |
| "epoch": 1.0, |
| "eval_steps": 500, |
| "global_step": 2875, |
| "is_hyper_param_search": false, |
| "is_local_process_zero": true, |
| "is_world_process_zero": true, |
| "log_history": [ |
| { |
| "epoch": 0.017391304347826087, |
| "grad_norm": 0.41397327184677124, |
| "learning_rate": 4.9710144927536237e-05, |
| "loss": 0.1154, |
| "step": 50 |
| }, |
| { |
| "epoch": 0.034782608695652174, |
| "grad_norm": 0.555982768535614, |
| "learning_rate": 4.9420289855072464e-05, |
| "loss": 0.0386, |
| "step": 100 |
| }, |
| { |
| "epoch": 0.05217391304347826, |
| "grad_norm": 0.3676406145095825, |
| "learning_rate": 4.91304347826087e-05, |
| "loss": 0.0315, |
| "step": 150 |
| }, |
| { |
| "epoch": 0.06956521739130435, |
| "grad_norm": 0.574974536895752, |
| "learning_rate": 4.884057971014493e-05, |
| "loss": 0.028, |
| "step": 200 |
| }, |
| { |
| "epoch": 0.08695652173913043, |
| "grad_norm": 0.41208797693252563, |
| "learning_rate": 4.855072463768116e-05, |
| "loss": 0.0252, |
| "step": 250 |
| }, |
| { |
| "epoch": 0.10434782608695652, |
| "grad_norm": 0.34254035353660583, |
| "learning_rate": 4.8260869565217394e-05, |
| "loss": 0.0231, |
| "step": 300 |
| }, |
| { |
| "epoch": 0.12173913043478261, |
| "grad_norm": 0.17080576717853546, |
| "learning_rate": 4.797101449275362e-05, |
| "loss": 0.0232, |
| "step": 350 |
| }, |
| { |
| "epoch": 0.1391304347826087, |
| "grad_norm": 0.1468990445137024, |
| "learning_rate": 4.7681159420289855e-05, |
| "loss": 0.0194, |
| "step": 400 |
| }, |
| { |
| "epoch": 0.1565217391304348, |
| "grad_norm": 0.20004335045814514, |
| "learning_rate": 4.739130434782609e-05, |
| "loss": 0.0198, |
| "step": 450 |
| }, |
| { |
| "epoch": 0.17391304347826086, |
| "grad_norm": 0.6127797961235046, |
| "learning_rate": 4.710144927536232e-05, |
| "loss": 0.0167, |
| "step": 500 |
| }, |
| { |
| "epoch": 0.19130434782608696, |
| "grad_norm": 0.2098260074853897, |
| "learning_rate": 4.681159420289855e-05, |
| "loss": 0.0195, |
| "step": 550 |
| }, |
| { |
| "epoch": 0.20869565217391303, |
| "grad_norm": 0.26583606004714966, |
| "learning_rate": 4.6521739130434785e-05, |
| "loss": 0.0158, |
| "step": 600 |
| }, |
| { |
| "epoch": 0.22608695652173913, |
| "grad_norm": 0.30250421166419983, |
| "learning_rate": 4.623188405797101e-05, |
| "loss": 0.0163, |
| "step": 650 |
| }, |
| { |
| "epoch": 0.24347826086956523, |
| "grad_norm": 0.2696796655654907, |
| "learning_rate": 4.594202898550725e-05, |
| "loss": 0.0131, |
| "step": 700 |
| }, |
| { |
| "epoch": 0.2608695652173913, |
| "grad_norm": 0.4439266324043274, |
| "learning_rate": 4.565217391304348e-05, |
| "loss": 0.0136, |
| "step": 750 |
| }, |
| { |
| "epoch": 0.2782608695652174, |
| "grad_norm": 0.4392126798629761, |
| "learning_rate": 4.5362318840579715e-05, |
| "loss": 0.0137, |
| "step": 800 |
| }, |
| { |
| "epoch": 0.2956521739130435, |
| "grad_norm": 0.36367908120155334, |
| "learning_rate": 4.507246376811595e-05, |
| "loss": 0.0116, |
| "step": 850 |
| }, |
| { |
| "epoch": 0.3130434782608696, |
| "grad_norm": 0.3119848072528839, |
| "learning_rate": 4.478260869565218e-05, |
| "loss": 0.0129, |
| "step": 900 |
| }, |
| { |
| "epoch": 0.33043478260869563, |
| "grad_norm": 0.3177811801433563, |
| "learning_rate": 4.449275362318841e-05, |
| "loss": 0.0095, |
| "step": 950 |
| }, |
| { |
| "epoch": 0.34782608695652173, |
| "grad_norm": 0.9409816265106201, |
| "learning_rate": 4.4202898550724645e-05, |
| "loss": 0.0087, |
| "step": 1000 |
| }, |
| { |
| "epoch": 0.3652173913043478, |
| "grad_norm": 0.19778360426425934, |
| "learning_rate": 4.391304347826087e-05, |
| "loss": 0.011, |
| "step": 1050 |
| }, |
| { |
| "epoch": 0.3826086956521739, |
| "grad_norm": 0.21794016659259796, |
| "learning_rate": 4.362318840579711e-05, |
| "loss": 0.009, |
| "step": 1100 |
| }, |
| { |
| "epoch": 0.4, |
| "grad_norm": 0.3412404954433441, |
| "learning_rate": 4.3333333333333334e-05, |
| "loss": 0.0096, |
| "step": 1150 |
| }, |
| { |
| "epoch": 0.41739130434782606, |
| "grad_norm": 0.27317434549331665, |
| "learning_rate": 4.304347826086957e-05, |
| "loss": 0.0081, |
| "step": 1200 |
| }, |
| { |
| "epoch": 0.43478260869565216, |
| "grad_norm": 0.0843023881316185, |
| "learning_rate": 4.27536231884058e-05, |
| "loss": 0.0082, |
| "step": 1250 |
| }, |
| { |
| "epoch": 0.45217391304347826, |
| "grad_norm": 0.17682024836540222, |
| "learning_rate": 4.246376811594203e-05, |
| "loss": 0.0077, |
| "step": 1300 |
| }, |
| { |
| "epoch": 0.46956521739130436, |
| "grad_norm": 0.41122502088546753, |
| "learning_rate": 4.2173913043478264e-05, |
| "loss": 0.0059, |
| "step": 1350 |
| }, |
| { |
| "epoch": 0.48695652173913045, |
| "grad_norm": 0.2303285449743271, |
| "learning_rate": 4.18840579710145e-05, |
| "loss": 0.0075, |
| "step": 1400 |
| }, |
| { |
| "epoch": 0.5043478260869565, |
| "grad_norm": 0.1975090056657791, |
| "learning_rate": 4.1594202898550726e-05, |
| "loss": 0.0068, |
| "step": 1450 |
| }, |
| { |
| "epoch": 0.5217391304347826, |
| "grad_norm": 0.30486345291137695, |
| "learning_rate": 4.130434782608696e-05, |
| "loss": 0.0056, |
| "step": 1500 |
| }, |
| { |
| "epoch": 0.5391304347826087, |
| "grad_norm": 0.08862517029047012, |
| "learning_rate": 4.101449275362319e-05, |
| "loss": 0.0054, |
| "step": 1550 |
| }, |
| { |
| "epoch": 0.5565217391304348, |
| "grad_norm": 0.3306777775287628, |
| "learning_rate": 4.072463768115942e-05, |
| "loss": 0.0055, |
| "step": 1600 |
| }, |
| { |
| "epoch": 0.5739130434782609, |
| "grad_norm": 0.15247593820095062, |
| "learning_rate": 4.0434782608695655e-05, |
| "loss": 0.0045, |
| "step": 1650 |
| }, |
| { |
| "epoch": 0.591304347826087, |
| "grad_norm": 0.2155187875032425, |
| "learning_rate": 4.014492753623188e-05, |
| "loss": 0.0074, |
| "step": 1700 |
| }, |
| { |
| "epoch": 0.6086956521739131, |
| "grad_norm": 0.14913377165794373, |
| "learning_rate": 3.985507246376812e-05, |
| "loss": 0.0051, |
| "step": 1750 |
| }, |
| { |
| "epoch": 0.6260869565217392, |
| "grad_norm": 0.10386908054351807, |
| "learning_rate": 3.956521739130435e-05, |
| "loss": 0.0058, |
| "step": 1800 |
| }, |
| { |
| "epoch": 0.6434782608695652, |
| "grad_norm": 0.25304195284843445, |
| "learning_rate": 3.927536231884058e-05, |
| "loss": 0.0063, |
| "step": 1850 |
| }, |
| { |
| "epoch": 0.6608695652173913, |
| "grad_norm": 0.18636035919189453, |
| "learning_rate": 3.898550724637681e-05, |
| "loss": 0.0054, |
| "step": 1900 |
| }, |
| { |
| "epoch": 0.6782608695652174, |
| "grad_norm": 0.26325523853302, |
| "learning_rate": 3.869565217391305e-05, |
| "loss": 0.004, |
| "step": 1950 |
| }, |
| { |
| "epoch": 0.6956521739130435, |
| "grad_norm": 0.21789908409118652, |
| "learning_rate": 3.8405797101449274e-05, |
| "loss": 0.0045, |
| "step": 2000 |
| }, |
| { |
| "epoch": 0.7130434782608696, |
| "grad_norm": 0.3563387393951416, |
| "learning_rate": 3.811594202898551e-05, |
| "loss": 0.0048, |
| "step": 2050 |
| }, |
| { |
| "epoch": 0.7304347826086957, |
| "grad_norm": 0.09919373691082001, |
| "learning_rate": 3.7826086956521736e-05, |
| "loss": 0.0049, |
| "step": 2100 |
| }, |
| { |
| "epoch": 0.7478260869565218, |
| "grad_norm": 0.7440153360366821, |
| "learning_rate": 3.753623188405797e-05, |
| "loss": 0.0054, |
| "step": 2150 |
| }, |
| { |
| "epoch": 0.7652173913043478, |
| "grad_norm": 0.14773143827915192, |
| "learning_rate": 3.7246376811594204e-05, |
| "loss": 0.0039, |
| "step": 2200 |
| }, |
| { |
| "epoch": 0.782608695652174, |
| "grad_norm": 0.09251661598682404, |
| "learning_rate": 3.695652173913043e-05, |
| "loss": 0.0039, |
| "step": 2250 |
| }, |
| { |
| "epoch": 0.8, |
| "grad_norm": 0.02156330831348896, |
| "learning_rate": 3.6666666666666666e-05, |
| "loss": 0.004, |
| "step": 2300 |
| }, |
| { |
| "epoch": 0.8173913043478261, |
| "grad_norm": 0.30019891262054443, |
| "learning_rate": 3.63768115942029e-05, |
| "loss": 0.0038, |
| "step": 2350 |
| }, |
| { |
| "epoch": 0.8347826086956521, |
| "grad_norm": 0.08380095660686493, |
| "learning_rate": 3.6086956521739134e-05, |
| "loss": 0.0047, |
| "step": 2400 |
| }, |
| { |
| "epoch": 0.8521739130434782, |
| "grad_norm": 0.17295339703559875, |
| "learning_rate": 3.579710144927537e-05, |
| "loss": 0.0037, |
| "step": 2450 |
| }, |
| { |
| "epoch": 0.8695652173913043, |
| "grad_norm": 0.02843887358903885, |
| "learning_rate": 3.5507246376811596e-05, |
| "loss": 0.0033, |
| "step": 2500 |
| }, |
| { |
| "epoch": 0.8869565217391304, |
| "grad_norm": 0.35363340377807617, |
| "learning_rate": 3.521739130434783e-05, |
| "loss": 0.0042, |
| "step": 2550 |
| }, |
| { |
| "epoch": 0.9043478260869565, |
| "grad_norm": 0.6101410984992981, |
| "learning_rate": 3.4927536231884064e-05, |
| "loss": 0.0035, |
| "step": 2600 |
| }, |
| { |
| "epoch": 0.9217391304347826, |
| "grad_norm": 0.043613508343696594, |
| "learning_rate": 3.463768115942029e-05, |
| "loss": 0.0041, |
| "step": 2650 |
| }, |
| { |
| "epoch": 0.9391304347826087, |
| "grad_norm": 0.12494798004627228, |
| "learning_rate": 3.4347826086956526e-05, |
| "loss": 0.0038, |
| "step": 2700 |
| }, |
| { |
| "epoch": 0.9565217391304348, |
| "grad_norm": 0.03580565005540848, |
| "learning_rate": 3.405797101449276e-05, |
| "loss": 0.0044, |
| "step": 2750 |
| }, |
| { |
| "epoch": 0.9739130434782609, |
| "grad_norm": 0.12163197249174118, |
| "learning_rate": 3.376811594202899e-05, |
| "loss": 0.0029, |
| "step": 2800 |
| }, |
| { |
| "epoch": 0.991304347826087, |
| "grad_norm": 0.09125994145870209, |
| "learning_rate": 3.347826086956522e-05, |
| "loss": 0.0035, |
| "step": 2850 |
| }, |
| { |
| "epoch": 1.0, |
| "eval_loss": 0.0025361417792737484, |
| "eval_runtime": 96.0353, |
| "eval_samples_per_second": 29.937, |
| "eval_steps_per_second": 3.749, |
| "step": 2875 |
| } |
| ], |
| "logging_steps": 50, |
| "max_steps": 8625, |
| "num_input_tokens_seen": 0, |
| "num_train_epochs": 3, |
| "save_steps": 500, |
| "total_flos": 6009564108171264.0, |
| "train_batch_size": 8, |
| "trial_name": null, |
| "trial_params": null |
| } |
|
|
