Boffl's picture
Upload trainer_log.jsonl with huggingface_hub
35b016b verified
{"current_steps": 10, "total_steps": 847, "loss": 1.7524, "learning_rate": 5.294117647058824e-06, "epoch": 0.011802891708468575, "percentage": 1.18, "elapsed_time": "0:07:32", "remaining_time": "10:31:10"}
{"current_steps": 20, "total_steps": 847, "loss": 1.3315, "learning_rate": 1.1176470588235295e-05, "epoch": 0.02360578341693715, "percentage": 2.36, "elapsed_time": "0:15:07", "remaining_time": "10:25:41"}
{"current_steps": 30, "total_steps": 847, "loss": 1.1148, "learning_rate": 1.7058823529411767e-05, "epoch": 0.03540867512540572, "percentage": 3.54, "elapsed_time": "0:23:01", "remaining_time": "10:26:51"}
{"current_steps": 40, "total_steps": 847, "loss": 1.0196, "learning_rate": 2.235294117647059e-05, "epoch": 0.0472115668338743, "percentage": 4.72, "elapsed_time": "0:31:03", "remaining_time": "10:26:30"}
{"current_steps": 50, "total_steps": 847, "loss": 0.94, "learning_rate": 2.823529411764706e-05, "epoch": 0.05901445854234287, "percentage": 5.9, "elapsed_time": "0:38:34", "remaining_time": "10:14:55"}
{"current_steps": 60, "total_steps": 847, "loss": 0.8918, "learning_rate": 3.411764705882353e-05, "epoch": 0.07081735025081144, "percentage": 7.08, "elapsed_time": "0:45:58", "remaining_time": "10:03:02"}
{"current_steps": 70, "total_steps": 847, "loss": 0.8412, "learning_rate": 4e-05, "epoch": 0.08262024195928003, "percentage": 8.26, "elapsed_time": "0:54:04", "remaining_time": "10:00:14"}
{"current_steps": 80, "total_steps": 847, "loss": 0.887, "learning_rate": 4.588235294117647e-05, "epoch": 0.0944231336677486, "percentage": 9.45, "elapsed_time": "1:02:00", "remaining_time": "9:54:34"}
{"current_steps": 90, "total_steps": 847, "loss": 0.8888, "learning_rate": 4.9998087784700426e-05, "epoch": 0.10622602537621717, "percentage": 10.63, "elapsed_time": "1:09:41", "remaining_time": "9:46:09"}
{"current_steps": 100, "total_steps": 847, "loss": 0.8555, "learning_rate": 4.996410098317137e-05, "epoch": 0.11802891708468574, "percentage": 11.81, "elapsed_time": "1:16:55", "remaining_time": "9:34:38"}
{"current_steps": 110, "total_steps": 847, "loss": 0.8424, "learning_rate": 4.989723448187131e-05, "epoch": 0.1298318087931543, "percentage": 12.99, "elapsed_time": "1:24:32", "remaining_time": "9:26:24"}
{"current_steps": 120, "total_steps": 847, "loss": 0.8445, "learning_rate": 4.9845268462432916e-05, "epoch": 0.14163470050162288, "percentage": 14.17, "elapsed_time": "1:31:57", "remaining_time": "9:17:06"}
{"current_steps": 130, "total_steps": 847, "loss": 0.8377, "learning_rate": 4.970969070763177e-05, "epoch": 0.15343759221009148, "percentage": 15.35, "elapsed_time": "1:39:25", "remaining_time": "9:08:19"}
{"current_steps": 140, "total_steps": 847, "loss": 0.759, "learning_rate": 4.953211814536217e-05, "epoch": 0.16524048391856005, "percentage": 16.53, "elapsed_time": "1:46:44", "remaining_time": "8:59:03"}
{"current_steps": 150, "total_steps": 847, "loss": 0.8121, "learning_rate": 4.931285256513868e-05, "epoch": 0.17704337562702863, "percentage": 17.71, "elapsed_time": "1:54:12", "remaining_time": "8:50:42"}
{"current_steps": 160, "total_steps": 847, "loss": 0.7896, "learning_rate": 4.905226661492095e-05, "epoch": 0.1888462673354972, "percentage": 18.89, "elapsed_time": "2:01:36", "remaining_time": "8:42:08"}
{"current_steps": 170, "total_steps": 847, "loss": 0.8057, "learning_rate": 4.8750803167788136e-05, "epoch": 0.20064915904396577, "percentage": 20.07, "elapsed_time": "2:08:26", "remaining_time": "8:31:30"}
{"current_steps": 180, "total_steps": 847, "loss": 0.7724, "learning_rate": 4.840897456926373e-05, "epoch": 0.21245205075243434, "percentage": 21.25, "elapsed_time": "2:15:44", "remaining_time": "8:22:59"}
{"current_steps": 190, "total_steps": 847, "loss": 0.7458, "learning_rate": 4.8027361766570117e-05, "epoch": 0.2242549424609029, "percentage": 22.43, "elapsed_time": "2:23:00", "remaining_time": "8:14:31"}
{"current_steps": 200, "total_steps": 847, "loss": 0.7686, "learning_rate": 4.760661332129254e-05, "epoch": 0.23605783416937148, "percentage": 23.61, "elapsed_time": "2:30:37", "remaining_time": "8:07:16"}
{"current_steps": 210, "total_steps": 847, "loss": 0.769, "learning_rate": 4.7147444307130686e-05, "epoch": 0.24786072587784008, "percentage": 24.79, "elapsed_time": "2:37:59", "remaining_time": "7:59:13"}
{"current_steps": 220, "total_steps": 847, "loss": 0.7574, "learning_rate": 4.665063509461097e-05, "epoch": 0.2596636175863086, "percentage": 25.97, "elapsed_time": "2:45:11", "remaining_time": "7:50:46"}
{"current_steps": 230, "total_steps": 847, "loss": 0.7826, "learning_rate": 4.6117030024825114e-05, "epoch": 0.2714665092947772, "percentage": 27.15, "elapsed_time": "2:52:38", "remaining_time": "7:43:08"}
{"current_steps": 240, "total_steps": 847, "loss": 0.7954, "learning_rate": 4.554753597444896e-05, "epoch": 0.28326940100324577, "percentage": 28.34, "elapsed_time": "2:59:48", "remaining_time": "7:34:45"}
{"current_steps": 250, "total_steps": 847, "loss": 0.7527, "learning_rate": 4.494312081448029e-05, "epoch": 0.29507229271171437, "percentage": 29.52, "elapsed_time": "3:07:05", "remaining_time": "7:26:47"}
{"current_steps": 260, "total_steps": 847, "loss": 0.7321, "learning_rate": 4.4304811765315105e-05, "epoch": 0.30687518442018297, "percentage": 30.7, "elapsed_time": "3:15:12", "remaining_time": "7:20:42"}
{"current_steps": 270, "total_steps": 847, "loss": 0.7047, "learning_rate": 4.3633693650957976e-05, "epoch": 0.3186780761286515, "percentage": 31.88, "elapsed_time": "3:22:30", "remaining_time": "7:12:45"}
{"current_steps": 280, "total_steps": 847, "loss": 0.7431, "learning_rate": 4.293090705533342e-05, "epoch": 0.3304809678371201, "percentage": 33.06, "elapsed_time": "3:30:34", "remaining_time": "7:06:24"}
{"current_steps": 290, "total_steps": 847, "loss": 0.7177, "learning_rate": 4.219764638383177e-05, "epoch": 0.34228385954558865, "percentage": 34.24, "elapsed_time": "3:37:34", "remaining_time": "6:57:53"}
{"current_steps": 300, "total_steps": 847, "loss": 0.7128, "learning_rate": 4.1435157833383955e-05, "epoch": 0.35408675125405725, "percentage": 35.42, "elapsed_time": "3:45:03", "remaining_time": "6:50:21"}
{"current_steps": 310, "total_steps": 847, "loss": 0.7162, "learning_rate": 4.06447372745151e-05, "epoch": 0.3658896429625258, "percentage": 36.6, "elapsed_time": "3:52:26", "remaining_time": "6:42:39"}
{"current_steps": 320, "total_steps": 847, "loss": 0.7212, "learning_rate": 3.982772804897649e-05, "epoch": 0.3776925346709944, "percentage": 37.78, "elapsed_time": "4:00:01", "remaining_time": "6:35:18"}
{"current_steps": 330, "total_steps": 847, "loss": 0.716, "learning_rate": 3.898551868669883e-05, "epoch": 0.389495426379463, "percentage": 38.96, "elapsed_time": "4:07:01", "remaining_time": "6:26:59"}
{"current_steps": 340, "total_steps": 847, "loss": 0.733, "learning_rate": 3.811954054594702e-05, "epoch": 0.40129831808793154, "percentage": 40.14, "elapsed_time": "4:14:28", "remaining_time": "6:19:27"}
{"current_steps": 350, "total_steps": 847, "loss": 0.764, "learning_rate": 3.723126538068686e-05, "epoch": 0.41310120979640014, "percentage": 41.32, "elapsed_time": "4:21:26", "remaining_time": "6:11:14"}
{"current_steps": 360, "total_steps": 847, "loss": 0.7302, "learning_rate": 3.632220283929822e-05, "epoch": 0.4249041015048687, "percentage": 42.5, "elapsed_time": "4:29:30", "remaining_time": "6:04:35"}
{"current_steps": 370, "total_steps": 847, "loss": 0.7127, "learning_rate": 3.5393897898885606e-05, "epoch": 0.4367069932133373, "percentage": 43.68, "elapsed_time": "4:37:01", "remaining_time": "5:57:08"}
{"current_steps": 380, "total_steps": 847, "loss": 0.6933, "learning_rate": 3.444792823954651e-05, "epoch": 0.4485098849218058, "percentage": 44.86, "elapsed_time": "4:44:42", "remaining_time": "5:49:52"}
{"current_steps": 390, "total_steps": 847, "loss": 0.7012, "learning_rate": 3.348590156306017e-05, "epoch": 0.4603127766302744, "percentage": 46.04, "elapsed_time": "4:52:18", "remaining_time": "5:42:31"}
{"current_steps": 400, "total_steps": 847, "loss": 0.7006, "learning_rate": 3.25094528605536e-05, "epoch": 0.47211566833874297, "percentage": 47.23, "elapsed_time": "4:59:34", "remaining_time": "5:34:46"}
{"current_steps": 410, "total_steps": 847, "loss": 0.7159, "learning_rate": 3.152024163378867e-05, "epoch": 0.48391856004721157, "percentage": 48.41, "elapsed_time": "5:07:17", "remaining_time": "5:27:31"}
{"current_steps": 420, "total_steps": 847, "loss": 0.7002, "learning_rate": 3.051994907479265e-05, "epoch": 0.49572145175568016, "percentage": 49.59, "elapsed_time": "5:14:28", "remaining_time": "5:19:42"}
{"current_steps": 430, "total_steps": 847, "loss": 0.6721, "learning_rate": 2.9510275208625522e-05, "epoch": 0.5075243434641488, "percentage": 50.77, "elapsed_time": "5:22:01", "remaining_time": "5:12:16"}
{"current_steps": 440, "total_steps": 847, "loss": 0.6612, "learning_rate": 2.849293600414002e-05, "epoch": 0.5193272351726173, "percentage": 51.95, "elapsed_time": "5:29:22", "remaining_time": "5:04:40"}
{"current_steps": 450, "total_steps": 847, "loss": 0.6786, "learning_rate": 2.7469660457644857e-05, "epoch": 0.5311301268810859, "percentage": 53.13, "elapsed_time": "5:37:01", "remaining_time": "4:57:20"}
{"current_steps": 460, "total_steps": 847, "loss": 0.6731, "learning_rate": 2.644218765442728e-05, "epoch": 0.5429330185895545, "percentage": 54.31, "elapsed_time": "5:44:13", "remaining_time": "4:49:35"}
{"current_steps": 470, "total_steps": 847, "loss": 0.6876, "learning_rate": 2.541226381312924e-05, "epoch": 0.554735910298023, "percentage": 55.49, "elapsed_time": "5:51:14", "remaining_time": "4:41:44"}
{"current_steps": 480, "total_steps": 847, "loss": 0.6757, "learning_rate": 2.4381639318000126e-05, "epoch": 0.5665388020064915, "percentage": 56.67, "elapsed_time": "5:58:19", "remaining_time": "4:33:57"}
{"current_steps": 490, "total_steps": 847, "loss": 0.7128, "learning_rate": 2.3352065744070072e-05, "epoch": 0.5783416937149601, "percentage": 57.85, "elapsed_time": "6:06:22", "remaining_time": "4:26:55"}
{"current_steps": 500, "total_steps": 847, "loss": 0.6928, "learning_rate": 2.2325292880299335e-05, "epoch": 0.5901445854234287, "percentage": 59.03, "elapsed_time": "6:13:53", "remaining_time": "4:19:28"}
{"current_steps": 510, "total_steps": 847, "loss": 0.6482, "learning_rate": 2.1303065755763277e-05, "epoch": 0.6019474771318973, "percentage": 60.21, "elapsed_time": "6:21:38", "remaining_time": "4:12:11"}
{"current_steps": 520, "total_steps": 847, "loss": 0.6759, "learning_rate": 2.0287121673926828e-05, "epoch": 0.6137503688403659, "percentage": 61.39, "elapsed_time": "6:29:33", "remaining_time": "4:04:58"}
{"current_steps": 530, "total_steps": 847, "loss": 0.674, "learning_rate": 1.92791872600489e-05, "epoch": 0.6255532605488344, "percentage": 62.57, "elapsed_time": "6:37:30", "remaining_time": "3:57:45"}
{"current_steps": 540, "total_steps": 847, "loss": 0.6636, "learning_rate": 1.8280975526734657e-05, "epoch": 0.637356152257303, "percentage": 63.75, "elapsed_time": "6:45:00", "remaining_time": "3:50:15"}
{"current_steps": 550, "total_steps": 847, "loss": 0.658, "learning_rate": 1.7294182962622846e-05, "epoch": 0.6491590439657716, "percentage": 64.94, "elapsed_time": "6:52:27", "remaining_time": "3:42:43"}
{"current_steps": 560, "total_steps": 847, "loss": 0.6563, "learning_rate": 1.632048664915622e-05, "epoch": 0.6609619356742402, "percentage": 66.12, "elapsed_time": "7:00:20", "remaining_time": "3:35:25"}
{"current_steps": 570, "total_steps": 847, "loss": 0.6481, "learning_rate": 1.536154141033482e-05, "epoch": 0.6727648273827088, "percentage": 67.3, "elapsed_time": "7:07:18", "remaining_time": "3:27:39"}
{"current_steps": 580, "total_steps": 847, "loss": 0.681, "learning_rate": 1.4418977000296552e-05, "epoch": 0.6845677190911773, "percentage": 68.48, "elapsed_time": "7:14:35", "remaining_time": "3:20:03"}
{"current_steps": 590, "total_steps": 847, "loss": 0.655, "learning_rate": 1.3494395333504622e-05, "epoch": 0.6963706107996459, "percentage": 69.66, "elapsed_time": "7:21:37", "remaining_time": "3:12:22"}
{"current_steps": 600, "total_steps": 847, "loss": 0.6557, "learning_rate": 1.2589367762249347e-05, "epoch": 0.7081735025081145, "percentage": 70.84, "elapsed_time": "7:29:00", "remaining_time": "3:04:50"}
{"current_steps": 610, "total_steps": 847, "loss": 0.6504, "learning_rate": 1.1705432406091085e-05, "epoch": 0.7199763942165831, "percentage": 72.02, "elapsed_time": "7:36:32", "remaining_time": "2:57:22"}
{"current_steps": 620, "total_steps": 847, "loss": 0.6509, "learning_rate": 1.0844091537783316e-05, "epoch": 0.7317792859250516, "percentage": 73.2, "elapsed_time": "7:43:39", "remaining_time": "2:49:45"}
{"current_steps": 630, "total_steps": 847, "loss": 0.6619, "learning_rate": 1.0006809030118181e-05, "epoch": 0.7435821776335202, "percentage": 74.38, "elapsed_time": "7:51:35", "remaining_time": "2:42:26"}
{"current_steps": 640, "total_steps": 847, "loss": 0.6083, "learning_rate": 9.195007868033933e-06, "epoch": 0.7553850693419888, "percentage": 75.56, "elapsed_time": "7:58:57", "remaining_time": "2:34:54"}
{"current_steps": 650, "total_steps": 847, "loss": 0.6464, "learning_rate": 8.410067730212439e-06, "epoch": 0.7671879610504574, "percentage": 76.74, "elapsed_time": "8:06:25", "remaining_time": "2:27:25"}
{"current_steps": 660, "total_steps": 847, "loss": 0.6342, "learning_rate": 7.653322644276779e-06, "epoch": 0.778990852758926, "percentage": 77.92, "elapsed_time": "8:14:20", "remaining_time": "2:20:03"}
{"current_steps": 670, "total_steps": 847, "loss": 0.6492, "learning_rate": 6.926058719574207e-06, "epoch": 0.7907937444673945, "percentage": 79.1, "elapsed_time": "8:21:13", "remaining_time": "2:12:24"}
{"current_steps": 680, "total_steps": 847, "loss": 0.6294, "learning_rate": 6.229511961397455e-06, "epoch": 0.8025966361758631, "percentage": 80.28, "elapsed_time": "8:28:45", "remaining_time": "2:04:56"}
{"current_steps": 690, "total_steps": 847, "loss": 0.638, "learning_rate": 5.564866170359351e-06, "epoch": 0.8143995278843317, "percentage": 81.46, "elapsed_time": "8:35:51", "remaining_time": "1:57:22"}
{"current_steps": 700, "total_steps": 847, "loss": 0.6096, "learning_rate": 4.933250930490715e-06, "epoch": 0.8262024195928003, "percentage": 82.64, "elapsed_time": "8:43:22", "remaining_time": "1:49:54"}
{"current_steps": 710, "total_steps": 847, "loss": 0.6285, "learning_rate": 4.335739689480778e-06, "epoch": 0.8380053113012688, "percentage": 83.83, "elapsed_time": "8:50:46", "remaining_time": "1:42:24"}
{"current_steps": 720, "total_steps": 847, "loss": 0.6257, "learning_rate": 3.773347934323035e-06, "epoch": 0.8498082030097374, "percentage": 85.01, "elapsed_time": "8:58:36", "remaining_time": "1:35:00"}
{"current_steps": 730, "total_steps": 847, "loss": 0.6142, "learning_rate": 3.2470314654667487e-06, "epoch": 0.861611094718206, "percentage": 86.19, "elapsed_time": "9:06:17", "remaining_time": "1:27:33"}
{"current_steps": 740, "total_steps": 847, "loss": 0.6485, "learning_rate": 2.7576847724075123e-06, "epoch": 0.8734139864266746, "percentage": 87.37, "elapsed_time": "9:14:03", "remaining_time": "1:20:06"}
{"current_steps": 750, "total_steps": 847, "loss": 0.6407, "learning_rate": 2.3061395134774038e-06, "epoch": 0.8852168781351432, "percentage": 88.55, "elapsed_time": "9:21:04", "remaining_time": "1:12:33"}
{"current_steps": 760, "total_steps": 847, "loss": 0.6535, "learning_rate": 1.8931631024185327e-06, "epoch": 0.8970197698436116, "percentage": 89.73, "elapsed_time": "9:28:38", "remaining_time": "1:05:05"}
{"current_steps": 770, "total_steps": 847, "loss": 0.642, "learning_rate": 1.5194574041419802e-06, "epoch": 0.9088226615520802, "percentage": 90.91, "elapsed_time": "9:36:07", "remaining_time": "0:57:36"}
{"current_steps": 780, "total_steps": 847, "loss": 0.617, "learning_rate": 1.185657541888857e-06, "epoch": 0.9206255532605488, "percentage": 92.09, "elapsed_time": "9:43:34", "remaining_time": "0:50:07"}
{"current_steps": 790, "total_steps": 847, "loss": 0.6415, "learning_rate": 8.923308178206552e-07, "epoch": 0.9324284449690174, "percentage": 93.27, "elapsed_time": "9:52:14", "remaining_time": "0:42:43"}
{"current_steps": 800, "total_steps": 847, "loss": 0.6388, "learning_rate": 6.39975748873431e-07, "epoch": 0.9442313366774859, "percentage": 94.45, "elapsed_time": "10:00:23", "remaining_time": "0:35:16"}
{"current_steps": 810, "total_steps": 847, "loss": 0.6366, "learning_rate": 4.2902121951440834e-07, "epoch": 0.9560342283859545, "percentage": 95.63, "elapsed_time": "10:07:44", "remaining_time": "0:27:45"}
{"current_steps": 820, "total_steps": 847, "loss": 0.6735, "learning_rate": 2.5982575284084486e-07, "epoch": 0.9678371200944231, "percentage": 96.81, "elapsed_time": "10:15:20", "remaining_time": "0:20:15"}
{"current_steps": 830, "total_steps": 847, "loss": 0.6238, "learning_rate": 1.3267690126008425e-07, "epoch": 0.9796400118028917, "percentage": 97.99, "elapsed_time": "10:22:39", "remaining_time": "0:12:45"}
{"current_steps": 840, "total_steps": 847, "loss": 0.6613, "learning_rate": 4.779075778620079e-08, "epoch": 0.9914429035113603, "percentage": 99.17, "elapsed_time": "10:30:05", "remaining_time": "0:05:15"}
{"current_steps": 847, "total_steps": 847, "epoch": 0.9997049277072882, "percentage": 100.0, "elapsed_time": "10:35:32", "remaining_time": "0:00:00"}