{ "best_metric": 14.025749206542969, "best_model_checkpoint": "./qwen2.5-0.5b/qwen2.5-0.5b-expo-L2EXPO-ES-1000/checkpoint-900", "epoch": 3.689182805857345, "eval_steps": 50, "global_step": 1300, "is_hyper_param_search": false, "is_local_process_zero": true, "is_world_process_zero": true, "log_history": [ { "dpo_loss": 0.6931471824645996, "epoch": 0.002834199338686821, "grad_norm": 368848.22474206856, "learning_rate": 2.840909090909091e-08, "logits": -1.359458565711975, "logps": -84.69721221923828, "loss": 0.3913, "objective": 0.3618059456348419, "ranking_idealized": 0.5833333134651184, "ranking_idealized_expo": 0.5833333134651184, "ranking_simple": 0.5833333134651184, "regularize": 0.3618059456348419, "step": 1, "wo_beta": 14.830931663513184 }, { "dpo_loss": 200.59906005859375, "epoch": 0.14170996693434104, "grad_norm": 364227.60279874475, "learning_rate": 1.4204545454545458e-06, "logits": -1.4551467895507812, "logps": -84.4092788696289, "loss": 423.0596, "objective": 404.94305419921875, "ranking_idealized": 0.5216836929321289, "ranking_idealized_expo": 0.5216836929321289, "ranking_simple": 0.5204081535339355, "regularize": 404.94305419921875, "step": 50, "wo_beta": 15.620062828063965 }, { "epoch": 0.14170996693434104, "eval_dpo_loss": 269.6528015136719, "eval_logits": -1.399541974067688, "eval_logps": -90.7221450805664, "eval_loss": 547.3512573242188, "eval_objective": 547.0781860351562, "eval_ranking_idealized": 0.5212215185165405, "eval_ranking_idealized_expo": 0.5212215185165405, "eval_ranking_simple": 0.523809552192688, "eval_regularize": 547.0781860351562, "eval_runtime": 309.3398, "eval_samples_per_second": 18.717, "eval_steps_per_second": 1.561, "eval_wo_beta": 16.238561630249023, "step": 50 }, { "dpo_loss": 905.3652954101562, "epoch": 0.2834199338686821, "grad_norm": 272683.76538272557, "learning_rate": 2.8409090909090916e-06, "logits": -1.356760025024414, "logps": -82.99842071533203, "loss": 1707.6314, "objective": 1747.880615234375, "ranking_idealized": 0.5137500166893005, "ranking_idealized_expo": 0.5137500166893005, "ranking_simple": 0.5120833516120911, "regularize": 1747.880615234375, "step": 100, "wo_beta": 15.263524055480957 }, { "epoch": 0.2834199338686821, "eval_dpo_loss": 848.0440063476562, "eval_logits": -1.3187600374221802, "eval_logps": -86.97607421875, "eval_loss": 1694.1810302734375, "eval_objective": 1675.62744140625, "eval_ranking_idealized": 0.5212215185165405, "eval_ranking_idealized_expo": 0.5212215185165405, "eval_ranking_simple": 0.5201863646507263, "eval_regularize": 1675.62744140625, "eval_runtime": 309.6783, "eval_samples_per_second": 18.697, "eval_steps_per_second": 1.56, "eval_wo_beta": 15.640727996826172, "step": 100 }, { "dpo_loss": 1518.7205810546875, "epoch": 0.42512990080302315, "grad_norm": 216991.1383206667, "learning_rate": 4.2613636363636365e-06, "logits": -1.2785295248031616, "logps": -78.66228485107422, "loss": 2824.1562, "objective": 2852.929443359375, "ranking_idealized": 0.527916669845581, "ranking_idealized_expo": 0.527916669845581, "ranking_simple": 0.5241666436195374, "regularize": 2852.929443359375, "step": 150, "wo_beta": 15.12118148803711 }, { "epoch": 0.42512990080302315, "eval_dpo_loss": 1511.363037109375, "eval_logits": -1.2861689329147339, "eval_logps": -82.21775817871094, "eval_loss": 3014.146484375, "eval_objective": 2968.77197265625, "eval_ranking_idealized": 0.5212215185165405, "eval_ranking_idealized_expo": 0.5212215185165405, "eval_ranking_simple": 0.5274327397346497, "eval_regularize": 2968.77197265625, "eval_runtime": 313.2906, "eval_samples_per_second": 18.481, "eval_steps_per_second": 1.542, "eval_wo_beta": 15.032276153564453, "step": 150 }, { "dpo_loss": 1896.1614990234375, "epoch": 0.5668398677373642, "grad_norm": 205665.2147772606, "learning_rate": 4.997168347957521e-06, "logits": -0.987150251865387, "logps": -77.89788055419922, "loss": 3551.5363, "objective": 3598.224609375, "ranking_idealized": 0.51541668176651, "ranking_idealized_expo": 0.51541668176651, "ranking_simple": 0.5062500238418579, "regularize": 3598.224609375, "step": 200, "wo_beta": 15.392863273620605 }, { "epoch": 0.5668398677373642, "eval_dpo_loss": 1970.65869140625, "eval_logits": -0.7799269556999207, "eval_logps": -81.00556945800781, "eval_loss": 3928.583740234375, "eval_objective": 3925.496826171875, "eval_ranking_idealized": 0.5212215185165405, "eval_ranking_idealized_expo": 0.5212215185165405, "eval_ranking_simple": 0.5248447060585022, "eval_regularize": 3925.496826171875, "eval_runtime": 308.6708, "eval_samples_per_second": 18.758, "eval_steps_per_second": 1.565, "eval_wo_beta": 14.613198280334473, "step": 200 }, { "dpo_loss": 1941.4073486328125, "epoch": 0.7085498346717053, "grad_norm": 204108.9756559366, "learning_rate": 4.973122855144066e-06, "logits": -0.7714519500732422, "logps": -75.8036117553711, "loss": 3769.9247, "objective": 3796.307861328125, "ranking_idealized": 0.5162500143051147, "ranking_idealized_expo": 0.5162500143051147, "ranking_simple": 0.5104166865348816, "regularize": 3796.307861328125, "step": 250, "wo_beta": 15.552165985107422 }, { "epoch": 0.7085498346717053, "eval_dpo_loss": 2167.546630859375, "eval_logits": -0.7279743552207947, "eval_logps": -80.5808334350586, "eval_loss": 4317.60498046875, "eval_objective": 4303.1142578125, "eval_ranking_idealized": 0.5212215185165405, "eval_ranking_idealized_expo": 0.5212215185165405, "eval_ranking_simple": 0.5269151329994202, "eval_regularize": 4303.1142578125, "eval_runtime": 309.0835, "eval_samples_per_second": 18.733, "eval_steps_per_second": 1.563, "eval_wo_beta": 14.582874298095703, "step": 250 }, { "dpo_loss": 1839.3994140625, "epoch": 0.8502598016060463, "grad_norm": 180485.23157207714, "learning_rate": 4.924776641419513e-06, "logits": -0.5737716555595398, "logps": -78.0151138305664, "loss": 3591.3281, "objective": 3690.973388671875, "ranking_idealized": 0.4950000047683716, "ranking_idealized_expo": 0.4950000047683716, "ranking_simple": 0.5079166889190674, "regularize": 3690.973388671875, "step": 300, "wo_beta": 15.1497163772583 }, { "epoch": 0.8502598016060463, "eval_dpo_loss": 2308.43505859375, "eval_logits": -0.584596574306488, "eval_logps": -82.77127075195312, "eval_loss": 4553.76318359375, "eval_objective": 4559.634765625, "eval_ranking_idealized": 0.5212215185165405, "eval_ranking_idealized_expo": 0.5212215185165405, "eval_ranking_simple": 0.5248447060585022, "eval_regularize": 4559.634765625, "eval_runtime": 308.87, "eval_samples_per_second": 18.746, "eval_steps_per_second": 1.564, "eval_wo_beta": 14.591386795043945, "step": 300 }, { "dpo_loss": 1713.5703125, "epoch": 0.9919697685403873, "grad_norm": 169039.33062429514, "learning_rate": 4.8526047530778175e-06, "logits": -0.629047155380249, "logps": -77.71513366699219, "loss": 3315.5613, "objective": 3303.522705078125, "ranking_idealized": 0.5254166722297668, "ranking_idealized_expo": 0.5254166722297668, "ranking_simple": 0.5233333110809326, "regularize": 3303.522705078125, "step": 350, "wo_beta": 15.061549186706543 }, { "epoch": 0.9919697685403873, "eval_dpo_loss": 2326.014404296875, "eval_logits": -0.7541128396987915, "eval_logps": -80.80506896972656, "eval_loss": 4667.9404296875, "eval_objective": 4670.26171875, "eval_ranking_idealized": 0.5212215185165405, "eval_ranking_idealized_expo": 0.5212215185165405, "eval_ranking_simple": 0.5331262946128845, "eval_regularize": 4670.26171875, "eval_runtime": 308.7398, "eval_samples_per_second": 18.754, "eval_steps_per_second": 1.564, "eval_wo_beta": 14.305211067199707, "step": 350 }, { "dpo_loss": 1624.2042236328125, "epoch": 1.1336797354747283, "grad_norm": 165220.9620293089, "learning_rate": 4.757316345716554e-06, "logits": -0.6056586503982544, "logps": -78.14913940429688, "loss": 3140.2284, "objective": 3177.530029296875, "ranking_idealized": 0.5320833325386047, "ranking_idealized_expo": 0.5320833325386047, "ranking_simple": 0.5299999713897705, "regularize": 3177.530029296875, "step": 400, "wo_beta": 15.318251609802246 }, { "epoch": 1.1336797354747283, "eval_dpo_loss": 2524.619140625, "eval_logits": -0.6474378108978271, "eval_logps": -81.57711029052734, "eval_loss": 4876.318359375, "eval_objective": 4879.08154296875, "eval_ranking_idealized": 0.5212215185165405, "eval_ranking_idealized_expo": 0.5212215185165405, "eval_ranking_simple": 0.522774338722229, "eval_regularize": 4879.08154296875, "eval_runtime": 308.5826, "eval_samples_per_second": 18.763, "eval_steps_per_second": 1.565, "eval_wo_beta": 14.327096939086914, "step": 400 }, { "dpo_loss": 1540.3419189453125, "epoch": 1.2753897024090695, "grad_norm": 171392.04304109752, "learning_rate": 4.639847716126855e-06, "logits": -0.636415421962738, "logps": -79.45496368408203, "loss": 2984.025, "objective": 3080.5654296875, "ranking_idealized": 0.5191666483879089, "ranking_idealized_expo": 0.5191666483879089, "ranking_simple": 0.5229166746139526, "regularize": 3080.5654296875, "step": 450, "wo_beta": 15.8497314453125 }, { "epoch": 1.2753897024090695, "eval_dpo_loss": 2466.713134765625, "eval_logits": -0.79082190990448, "eval_logps": -84.27051544189453, "eval_loss": 4773.4326171875, "eval_objective": 4785.75341796875, "eval_ranking_idealized": 0.5212215185165405, "eval_ranking_idealized_expo": 0.5212215185165405, "eval_ranking_simple": 0.5248447060585022, "eval_regularize": 4785.75341796875, "eval_runtime": 309.0671, "eval_samples_per_second": 18.734, "eval_steps_per_second": 1.563, "eval_wo_beta": 14.321349143981934, "step": 450 }, { "dpo_loss": 1396.68310546875, "epoch": 1.4170996693434104, "grad_norm": 149255.8927126336, "learning_rate": 4.501353102310901e-06, "logits": -0.6039130687713623, "logps": -79.58264923095703, "loss": 2769.3719, "objective": 2762.444580078125, "ranking_idealized": 0.4970833361148834, "ranking_idealized_expo": 0.4970833361148834, "ranking_simple": 0.49916666746139526, "regularize": 2762.444580078125, "step": 500, "wo_beta": 15.438196182250977 }, { "epoch": 1.4170996693434104, "eval_dpo_loss": 2513.819091796875, "eval_logits": -0.7098349332809448, "eval_logps": -81.49170684814453, "eval_loss": 4863.71484375, "eval_objective": 4866.62353515625, "eval_ranking_idealized": 0.5212215185165405, "eval_ranking_idealized_expo": 0.5212215185165405, "eval_ranking_simple": 0.5191511511802673, "eval_regularize": 4866.62353515625, "eval_runtime": 308.6488, "eval_samples_per_second": 18.759, "eval_steps_per_second": 1.565, "eval_wo_beta": 14.19344425201416, "step": 500 }, { "dpo_loss": 1364.8533935546875, "epoch": 1.5588096362777515, "grad_norm": 159666.559697962, "learning_rate": 4.34319334202531e-06, "logits": -0.5462002158164978, "logps": -79.02467346191406, "loss": 2620.0086, "objective": 2664.316162109375, "ranking_idealized": 0.5104166865348816, "ranking_idealized_expo": 0.5104166865348816, "ranking_simple": 0.5095833539962769, "regularize": 2664.316162109375, "step": 550, "wo_beta": 15.136716842651367 }, { "epoch": 1.5588096362777515, "eval_dpo_loss": 2463.116943359375, "eval_logits": -0.5653336644172668, "eval_logps": -81.83069610595703, "eval_loss": 4887.2939453125, "eval_objective": 4877.46826171875, "eval_ranking_idealized": 0.5212215185165405, "eval_ranking_idealized_expo": 0.5212215185165405, "eval_ranking_simple": 0.5248447060585022, "eval_regularize": 4877.46826171875, "eval_runtime": 308.8536, "eval_samples_per_second": 18.747, "eval_steps_per_second": 1.564, "eval_wo_beta": 14.175724029541016, "step": 550 }, { "dpo_loss": 1253.067138671875, "epoch": 1.7005196032120926, "grad_norm": 158366.9559774092, "learning_rate": 4.16692250129073e-06, "logits": -0.4213114380836487, "logps": -78.87406921386719, "loss": 2530.9462, "objective": 2507.11328125, "ranking_idealized": 0.5149999856948853, "ranking_idealized_expo": 0.5149999856948853, "ranking_simple": 0.5095833539962769, "regularize": 2507.11328125, "step": 600, "wo_beta": 15.036109924316406 }, { "epoch": 1.7005196032120926, "eval_dpo_loss": 2522.071533203125, "eval_logits": -0.48861420154571533, "eval_logps": -82.87271881103516, "eval_loss": 4965.42333984375, "eval_objective": 5013.287109375, "eval_ranking_idealized": 0.5212215185165405, "eval_ranking_idealized_expo": 0.5212215185165405, "eval_ranking_simple": 0.5232919454574585, "eval_regularize": 5013.287109375, "eval_runtime": 308.7234, "eval_samples_per_second": 18.755, "eval_steps_per_second": 1.565, "eval_wo_beta": 14.257344245910645, "step": 600 }, { "dpo_loss": 1245.341064453125, "epoch": 1.8422295701464337, "grad_norm": 159825.03654087533, "learning_rate": 3.974272604254906e-06, "logits": -0.44398489594459534, "logps": -79.5468521118164, "loss": 2445.0009, "objective": 2442.403076171875, "ranking_idealized": 0.527916669845581, "ranking_idealized_expo": 0.527916669845581, "ranking_simple": 0.5249999761581421, "regularize": 2442.403076171875, "step": 650, "wo_beta": 15.771297454833984 }, { "epoch": 1.8422295701464337, "eval_dpo_loss": 2509.764404296875, "eval_logits": -0.5172779560089111, "eval_logps": -81.83026123046875, "eval_loss": 4964.3994140625, "eval_objective": 4986.9541015625, "eval_ranking_idealized": 0.5212215185165405, "eval_ranking_idealized_expo": 0.5212215185165405, "eval_ranking_simple": 0.5243270993232727, "eval_regularize": 4986.9541015625, "eval_runtime": 309.186, "eval_samples_per_second": 18.727, "eval_steps_per_second": 1.562, "eval_wo_beta": 14.255688667297363, "step": 650 }, { "dpo_loss": 1206.533203125, "epoch": 1.9839395370807746, "grad_norm": 156012.46785149653, "learning_rate": 3.767136614452458e-06, "logits": -0.4590357840061188, "logps": -80.01580047607422, "loss": 2287.7192, "objective": 2330.523193359375, "ranking_idealized": 0.5108333230018616, "ranking_idealized_expo": 0.5108333230018616, "ranking_simple": 0.5108333230018616, "regularize": 2330.523193359375, "step": 700, "wo_beta": 15.533180236816406 }, { "epoch": 1.9839395370807746, "eval_dpo_loss": 2561.16015625, "eval_logits": -0.5353549718856812, "eval_logps": -83.87384033203125, "eval_loss": 5034.0654296875, "eval_objective": 5065.85205078125, "eval_ranking_idealized": 0.5212215185165405, "eval_ranking_idealized_expo": 0.5212215185165405, "eval_ranking_simple": 0.52173912525177, "eval_regularize": 5065.85205078125, "eval_runtime": 309.3567, "eval_samples_per_second": 18.716, "eval_steps_per_second": 1.561, "eval_wo_beta": 14.084709167480469, "step": 700 }, { "dpo_loss": 1036.373779296875, "epoch": 2.1256495040151155, "grad_norm": 147121.88795994341, "learning_rate": 3.547549834686222e-06, "logits": -0.47968342900276184, "logps": -79.23291015625, "loss": 2066.9519, "objective": 2005.8623046875, "ranking_idealized": 0.5112500190734863, "ranking_idealized_expo": 0.5112500190734863, "ranking_simple": 0.5066666603088379, "regularize": 2005.8623046875, "step": 750, "wo_beta": 15.400221824645996 }, { "epoch": 2.1256495040151155, "eval_dpo_loss": 2654.179443359375, "eval_logits": -0.49486637115478516, "eval_logps": -82.19441223144531, "eval_loss": 5229.88525390625, "eval_objective": 5264.4931640625, "eval_ranking_idealized": 0.5212215185165405, "eval_ranking_idealized_expo": 0.5212215185165405, "eval_ranking_simple": 0.5253623127937317, "eval_regularize": 5264.4931640625, "eval_runtime": 308.7479, "eval_samples_per_second": 18.753, "eval_steps_per_second": 1.564, "eval_wo_beta": 14.098138809204102, "step": 750 }, { "dpo_loss": 994.4859619140625, "epoch": 2.2673594709494567, "grad_norm": 143685.9375211257, "learning_rate": 3.3176699082935546e-06, "logits": -0.466074675321579, "logps": -79.23563385009766, "loss": 1963.7713, "objective": 1942.085205078125, "ranking_idealized": 0.512499988079071, "ranking_idealized_expo": 0.512499988079071, "ranking_simple": 0.5108333230018616, "regularize": 1942.085205078125, "step": 800, "wo_beta": 14.980566024780273 }, { "epoch": 2.2673594709494567, "eval_dpo_loss": 2636.38330078125, "eval_logits": -0.47902175784111023, "eval_logps": -82.23072052001953, "eval_loss": 5180.23876953125, "eval_objective": 5235.76953125, "eval_ranking_idealized": 0.5212215185165405, "eval_ranking_idealized_expo": 0.5212215185165405, "eval_ranking_simple": 0.5243270993232727, "eval_regularize": 5235.76953125, "eval_runtime": 308.7297, "eval_samples_per_second": 18.754, "eval_steps_per_second": 1.564, "eval_wo_beta": 14.037761688232422, "step": 800 }, { "dpo_loss": 922.8798828125, "epoch": 2.409069437883798, "grad_norm": 142497.46137454137, "learning_rate": 3.0797556183036582e-06, "logits": -0.435017853975296, "logps": -80.0512466430664, "loss": 1854.7628, "objective": 1856.9925537109375, "ranking_idealized": 0.5133333206176758, "ranking_idealized_expo": 0.5133333206176758, "ranking_simple": 0.5149999856948853, "regularize": 1856.9925537109375, "step": 850, "wo_beta": 15.263174057006836 }, { "epoch": 2.409069437883798, "eval_dpo_loss": 2612.387451171875, "eval_logits": -0.4899817407131195, "eval_logps": -82.96641540527344, "eval_loss": 5130.60693359375, "eval_objective": 5171.9189453125, "eval_ranking_idealized": 0.5212215185165405, "eval_ranking_idealized_expo": 0.5212215185165405, "eval_ranking_simple": 0.5269151329994202, "eval_regularize": 5171.9189453125, "eval_runtime": 313.09, "eval_samples_per_second": 18.493, "eval_steps_per_second": 1.543, "eval_wo_beta": 14.114248275756836, "step": 850 }, { "dpo_loss": 880.90771484375, "epoch": 2.550779404818139, "grad_norm": 149283.81145436864, "learning_rate": 2.8361446928038298e-06, "logits": -0.4352048635482788, "logps": -80.65638732910156, "loss": 1711.9678, "objective": 1688.95556640625, "ranking_idealized": 0.5183333158493042, "ranking_idealized_expo": 0.5179166793823242, "ranking_simple": 0.5204166769981384, "regularize": 1688.95556640625, "step": 900, "wo_beta": 15.470394134521484 }, { "epoch": 2.550779404818139, "eval_dpo_loss": 2703.03125, "eval_logits": -0.4645061194896698, "eval_logps": -85.19200134277344, "eval_loss": 5280.861328125, "eval_objective": 5329.05712890625, "eval_ranking_idealized": 0.5212215185165405, "eval_ranking_idealized_expo": 0.5212215185165405, "eval_ranking_simple": 0.5263975262641907, "eval_regularize": 5329.05712890625, "eval_runtime": 309.3219, "eval_samples_per_second": 18.718, "eval_steps_per_second": 1.561, "eval_wo_beta": 14.025749206542969, "step": 900 }, { "dpo_loss": 851.3322143554688, "epoch": 2.69248937175248, "grad_norm": 149343.91741290496, "learning_rate": 2.5892308345974517e-06, "logits": -0.4298136830329895, "logps": -81.61705017089844, "loss": 1682.3781, "objective": 1668.57568359375, "ranking_idealized": 0.5058333277702332, "ranking_idealized_expo": 0.5058333277702332, "ranking_simple": 0.5095833539962769, "regularize": 1668.57568359375, "step": 950, "wo_beta": 15.541181564331055 }, { "epoch": 2.69248937175248, "eval_dpo_loss": 2644.848388671875, "eval_logits": -0.4320021867752075, "eval_logps": -83.93761444091797, "eval_loss": 5177.08154296875, "eval_objective": 5195.845703125, "eval_ranking_idealized": 0.5212215185165405, "eval_ranking_idealized_expo": 0.5212215185165405, "eval_ranking_simple": 0.5253623127937317, "eval_regularize": 5195.845703125, "eval_runtime": 308.7152, "eval_samples_per_second": 18.755, "eval_steps_per_second": 1.565, "eval_wo_beta": 14.169083595275879, "step": 950 }, { "dpo_loss": 776.9803466796875, "epoch": 2.8341993386868207, "grad_norm": 137448.18146621465, "learning_rate": 2.341440200858589e-06, "logits": -0.4145285189151764, "logps": -79.74474334716797, "loss": 1508.6941, "objective": 1524.09228515625, "ranking_idealized": 0.5095833539962769, "ranking_idealized_expo": 0.5095833539962769, "ranking_simple": 0.5054166913032532, "regularize": 1524.09228515625, "step": 1000, "wo_beta": 15.0802640914917 }, { "epoch": 2.8341993386868207, "eval_dpo_loss": 2632.400634765625, "eval_logits": -0.5013601779937744, "eval_logps": -83.42354583740234, "eval_loss": 5124.71435546875, "eval_objective": 5131.57275390625, "eval_ranking_idealized": 0.5212215185165405, "eval_ranking_idealized_expo": 0.5212215185165405, "eval_ranking_simple": 0.5243270993232727, "eval_regularize": 5131.57275390625, "eval_runtime": 311.7146, "eval_samples_per_second": 18.575, "eval_steps_per_second": 1.549, "eval_wo_beta": 14.1500883102417, "step": 1000 }, { "dpo_loss": 708.8336181640625, "epoch": 2.975909305621162, "grad_norm": 147550.24239683093, "learning_rate": 2.0952075638923656e-06, "logits": -0.49126318097114563, "logps": -79.99681854248047, "loss": 1432.2169, "objective": 1416.6346435546875, "ranking_idealized": 0.5170833468437195, "ranking_idealized_expo": 0.5170833468437195, "ranking_simple": 0.5212500095367432, "regularize": 1416.6346435546875, "step": 1050, "wo_beta": 15.470003128051758 }, { "epoch": 2.975909305621162, "eval_dpo_loss": 2638.496337890625, "eval_logits": -0.46873465180397034, "eval_logps": -83.80735778808594, "eval_loss": 5215.619140625, "eval_objective": 5232.5947265625, "eval_ranking_idealized": 0.5212215185165405, "eval_ranking_idealized_expo": 0.5212215185165405, "eval_ranking_simple": 0.5295031070709229, "eval_regularize": 5232.5947265625, "eval_runtime": 308.5567, "eval_samples_per_second": 18.765, "eval_steps_per_second": 1.565, "eval_wo_beta": 14.238861083984375, "step": 1050 }, { "dpo_loss": 630.0592651367188, "epoch": 3.122342938119981, "grad_norm": 146194.44633521218, "learning_rate": 1.852952387243698e-06, "logits": -0.47443822026252747, "logps": -80.19267272949219, "loss": 1247.6562, "objective": 1267.6502685546875, "ranking_idealized": 0.528333306312561, "ranking_idealized_expo": 0.528333306312561, "ranking_simple": 0.5299999713897705, "regularize": 1267.6502685546875, "step": 1100, "wo_beta": 15.557552337646484 }, { "epoch": 3.122342938119981, "eval_dpo_loss": 2631.76611328125, "eval_logits": -0.5460559129714966, "eval_logps": -84.16140747070312, "eval_loss": 5184.86962890625, "eval_objective": 5190.6357421875, "eval_ranking_idealized": 0.5212215185165405, "eval_ranking_idealized_expo": 0.5212215185165405, "eval_ranking_simple": 0.5263975262641907, "eval_regularize": 5190.6357421875, "eval_runtime": 310.23, "eval_samples_per_second": 18.664, "eval_steps_per_second": 1.557, "eval_wo_beta": 14.152945518493652, "step": 1100 }, { "dpo_loss": 565.9061889648438, "epoch": 3.264052905054322, "grad_norm": 138842.10059998793, "learning_rate": 1.617055052228768e-06, "logits": -0.4821857213973999, "logps": -80.27399444580078, "loss": 1136.2859, "objective": 1102.266845703125, "ranking_idealized": 0.5083333253860474, "ranking_idealized_expo": 0.5083333253860474, "ranking_simple": 0.5137500166893005, "regularize": 1102.266845703125, "step": 1150, "wo_beta": 15.317670822143555 }, { "epoch": 3.264052905054322, "eval_dpo_loss": 2590.183837890625, "eval_logits": -0.5632494688034058, "eval_logps": -83.88517761230469, "eval_loss": 5110.20556640625, "eval_objective": 5112.02783203125, "eval_ranking_idealized": 0.5212215185165405, "eval_ranking_idealized_expo": 0.5212215185165405, "eval_ranking_simple": 0.5279502868652344, "eval_regularize": 5112.02783203125, "eval_runtime": 308.5625, "eval_samples_per_second": 18.764, "eval_steps_per_second": 1.565, "eval_wo_beta": 14.093263626098633, "step": 1150 }, { "dpo_loss": 526.139404296875, "epoch": 3.4057628719886632, "grad_norm": 142396.22516348853, "learning_rate": 1.3898334684855647e-06, "logits": -0.49703070521354675, "logps": -81.49961853027344, "loss": 1042.7762, "objective": 1025.806884765625, "ranking_idealized": 0.5079166889190674, "ranking_idealized_expo": 0.5079166889190674, "ranking_simple": 0.5054166913032532, "regularize": 1025.806884765625, "step": 1200, "wo_beta": 15.595675468444824 }, { "epoch": 3.4057628719886632, "eval_dpo_loss": 2612.26611328125, "eval_logits": -0.5505208969116211, "eval_logps": -83.96298217773438, "eval_loss": 5146.40771484375, "eval_objective": 5162.16650390625, "eval_ranking_idealized": 0.5212215185165405, "eval_ranking_idealized_expo": 0.5212215185165405, "eval_ranking_simple": 0.5274327397346497, "eval_regularize": 5162.16650390625, "eval_runtime": 313.9682, "eval_samples_per_second": 18.441, "eval_steps_per_second": 1.538, "eval_wo_beta": 14.112215042114258, "step": 1200 }, { "dpo_loss": 476.4833984375, "epoch": 3.5474728389230044, "grad_norm": 138317.9821042345, "learning_rate": 1.1735202983664803e-06, "logits": -0.45970824360847473, "logps": -81.46965789794922, "loss": 978.7787, "objective": 968.6390380859375, "ranking_idealized": 0.5141666531562805, "ranking_idealized_expo": 0.5141666531562805, "ranking_simple": 0.5137500166893005, "regularize": 968.6390380859375, "step": 1250, "wo_beta": 15.359482765197754 }, { "epoch": 3.5474728389230044, "eval_dpo_loss": 2605.342041015625, "eval_logits": -0.4993049204349518, "eval_logps": -83.89867401123047, "eval_loss": 5115.50927734375, "eval_objective": 5140.92578125, "eval_ranking_idealized": 0.5212215185165405, "eval_ranking_idealized_expo": 0.5212215185165405, "eval_ranking_simple": 0.5279502868652344, "eval_regularize": 5140.92578125, "eval_runtime": 308.6628, "eval_samples_per_second": 18.758, "eval_steps_per_second": 1.565, "eval_wo_beta": 14.127862930297852, "step": 1250 }, { "dpo_loss": 455.1197814941406, "epoch": 3.689182805857345, "grad_norm": 136461.83557945368, "learning_rate": 9.702410189643838e-07, "logits": -0.4638132154941559, "logps": -80.26090240478516, "loss": 864.8715, "objective": 869.080810546875, "ranking_idealized": 0.49958333373069763, "ranking_idealized_expo": 0.49958333373069763, "ranking_simple": 0.4983333349227905, "regularize": 869.080810546875, "step": 1300, "wo_beta": 15.24673843383789 }, { "epoch": 3.689182805857345, "eval_dpo_loss": 2621.07275390625, "eval_logits": -0.5245055556297302, "eval_logps": -84.29293060302734, "eval_loss": 5143.4609375, "eval_objective": 5173.7548828125, "eval_ranking_idealized": 0.5212215185165405, "eval_ranking_idealized_expo": 0.5212215185165405, "eval_ranking_simple": 0.5258799195289612, "eval_regularize": 5173.7548828125, "eval_runtime": 308.7148, "eval_samples_per_second": 18.755, "eval_steps_per_second": 1.565, "eval_wo_beta": 14.158381462097168, "step": 1300 }, { "epoch": 3.689182805857345, "step": 1300, "total_flos": 0.0, "train_loss": 202.7064813701923, "train_runtime": 6887.9795, "train_samples_per_second": 36.877, "train_steps_per_second": 0.256 } ], "logging_steps": 50, "max_steps": 1760, "num_input_tokens_seen": 0, "num_train_epochs": 5, "save_steps": 50, "stateful_callbacks": { "EarlyStoppingCallback": { "args": { "early_stopping_patience": 5, "early_stopping_threshold": 0.0 }, "attributes": { "early_stopping_patience_counter": 0 } }, "TrainerControl": { "args": { "should_epoch_stop": false, "should_evaluate": false, "should_log": false, "should_save": true, "should_training_stop": true }, "attributes": {} } }, "total_flos": 0.0, "train_batch_size": 4, "trial_name": null, "trial_params": null }