Training in progress, step 21000, checkpoint

checkpoint-21000/model.safetensors

@@ -1,3 +1,3 @@
 version https://git-lfs.github.com/spec/v1
-oid sha256:8d88e4391b49b7f47122e6579b3d3acfbc9c067ea039fb9eabac0ab6772aff48
+oid sha256:72a89a70a546fa93a695ee50828261bac5a679fd24c3759739491a09f07bd247
 size 1711848436

checkpoint-21000/optimizer.pt

@@ -1,3 +1,3 @@
 version https://git-lfs.github.com/spec/v1
-oid sha256:8b8ba3e1769c9a5eb1594683a4c35c2c714234c4df59ce37892e4d1d138efb08
+oid sha256:26d4b6ad84abd058eec7b962669c3ffb54895bd1da51be1a4470c28401ab46dc
 size 3424043887

checkpoint-21000/scheduler.pt

@@ -1,3 +1,3 @@
 version https://git-lfs.github.com/spec/v1
-oid sha256:259cafb41160652c3d54806748b10cd7a9019d0dd9321fe4c5ddeee2dc391a9c
+oid sha256:e12e681abba97cf7af865a6caf3b69ac4c0829816b94ae393763ecc15339e8c4
 size 623

checkpoint-21000/trainer_state.json

@@ -1,7 +1,7 @@
 {
-  "best_metric": 0.5568962693214417,
-  "best_model_checkpoint": "./output/clip-finetuned-csu-p14-336-e4l57-l/checkpoint-18500",
-  "epoch": 2.3640661938534278,
+  "best_metric": 0.4872625172138214,
+  "best_model_checkpoint": "./output/clip-finetuned-csu-p14-336-e4l57-l/checkpoint-20000",
+  "epoch": 2.2397610921501707,
   "eval_steps": 500,
   "global_step": 21000,
   "is_hyper_param_search": false,
@@ -9,640 +9,640 @@
   "is_world_process_zero": true,
   "log_history": [
     {
-      "epoch": 0.056287290329843524,
-      "grad_norm": 150.1642303466797,
-      "learning_rate": 4.929640887087696e-07,
-      "loss": 0.3486,
+      "epoch": 0.05332764505119454,
+      "grad_norm": 414.2552795410156,
+      "learning_rate": 4.911120591581342e-07,
+      "loss": 0.3809,
       "step": 500
     },
     {
-      "epoch": 0.056287290329843524,
-      "eval_loss": 1.1265727281570435,
-      "eval_runtime": 123.7806,
-      "eval_samples_per_second": 15.948,
-      "eval_steps_per_second": 1.995,
+      "epoch": 0.05332764505119454,
+      "eval_loss": 1.1109352111816406,
+      "eval_runtime": 63.6734,
+      "eval_samples_per_second": 15.501,
+      "eval_steps_per_second": 1.947,
       "step": 500
     },
     {
-      "epoch": 0.11257458065968705,
-      "grad_norm": 366.6913146972656,
-      "learning_rate": 4.859281774175391e-07,
-      "loss": 0.2733,
+      "epoch": 0.10665529010238908,
+      "grad_norm": 45.880027770996094,
+      "learning_rate": 4.822241183162685e-07,
+      "loss": 0.2649,
       "step": 1000
     },
     {
-      "epoch": 0.11257458065968705,
-      "eval_loss": 0.9742079377174377,
-      "eval_runtime": 125.8047,
-      "eval_samples_per_second": 15.691,
-      "eval_steps_per_second": 1.963,
+      "epoch": 0.10665529010238908,
+      "eval_loss": 0.9588962197303772,
+      "eval_runtime": 62.6825,
+      "eval_samples_per_second": 15.746,
+      "eval_steps_per_second": 1.978,
       "step": 1000
     },
     {
-      "epoch": 0.16886187098953057,
-      "grad_norm": 555.5016479492188,
-      "learning_rate": 4.788922661263087e-07,
-      "loss": 0.1851,
+      "epoch": 0.1599829351535836,
+      "grad_norm": 432.9915466308594,
+      "learning_rate": 4.733361774744027e-07,
+      "loss": 0.2104,
       "step": 1500
     },
     {
-      "epoch": 0.16886187098953057,
-      "eval_loss": 0.9162159562110901,
-      "eval_runtime": 126.7606,
-      "eval_samples_per_second": 15.573,
-      "eval_steps_per_second": 1.949,
+      "epoch": 0.1599829351535836,
+      "eval_loss": 0.90110844373703,
+      "eval_runtime": 62.6757,
+      "eval_samples_per_second": 15.748,
+      "eval_steps_per_second": 1.978,
       "step": 1500
     },
     {
-      "epoch": 0.2251491613193741,
-      "grad_norm": 0.423260897397995,
-      "learning_rate": 4.7185635483507824e-07,
-      "loss": 0.1973,
+      "epoch": 0.21331058020477817,
+      "grad_norm": 0.014208819717168808,
+      "learning_rate": 4.6444823663253695e-07,
+      "loss": 0.1849,
       "step": 2000
     },
     {
-      "epoch": 0.2251491613193741,
-      "eval_loss": 0.8716973662376404,
-      "eval_runtime": 127.6955,
-      "eval_samples_per_second": 15.459,
-      "eval_steps_per_second": 1.934,
+      "epoch": 0.21331058020477817,
+      "eval_loss": 0.8570474982261658,
+      "eval_runtime": 63.7296,
+      "eval_samples_per_second": 15.487,
+      "eval_steps_per_second": 1.946,
       "step": 2000
     },
     {
-      "epoch": 0.2814364516492176,
-      "grad_norm": 6.840139389038086,
-      "learning_rate": 4.6482044354384774e-07,
-      "loss": 0.1881,
+      "epoch": 0.2666382252559727,
+      "grad_norm": 174.73324584960938,
+      "learning_rate": 4.5556029579067116e-07,
+      "loss": 0.2056,
       "step": 2500
     },
     {
-      "epoch": 0.2814364516492176,
-      "eval_loss": 0.830047070980072,
-      "eval_runtime": 128.0771,
-      "eval_samples_per_second": 15.413,
-      "eval_steps_per_second": 1.929,
+      "epoch": 0.2666382252559727,
+      "eval_loss": 0.7895939350128174,
+      "eval_runtime": 63.8387,
+      "eval_samples_per_second": 15.461,
+      "eval_steps_per_second": 1.942,
       "step": 2500
     },
     {
-      "epoch": 0.33772374197906113,
-      "grad_norm": 729.8377685546875,
-      "learning_rate": 4.577845322526173e-07,
-      "loss": 0.1695,
+      "epoch": 0.3199658703071672,
+      "grad_norm": 9.852073823424234e-09,
+      "learning_rate": 4.4667235494880547e-07,
+      "loss": 0.159,
       "step": 3000
     },
     {
-      "epoch": 0.33772374197906113,
-      "eval_loss": 0.8265627026557922,
-      "eval_runtime": 128.052,
-      "eval_samples_per_second": 15.416,
-      "eval_steps_per_second": 1.929,
+      "epoch": 0.3199658703071672,
+      "eval_loss": 0.7647623419761658,
+      "eval_runtime": 63.9155,
+      "eval_samples_per_second": 15.442,
+      "eval_steps_per_second": 1.94,
       "step": 3000
     },
     {
-      "epoch": 0.39401103230890466,
-      "grad_norm": 7.96999938756926e-06,
-      "learning_rate": 4.507486209613869e-07,
-      "loss": 0.155,
+      "epoch": 0.37329351535836175,
+      "grad_norm": 0.0025215104687958956,
+      "learning_rate": 4.377844141069397e-07,
+      "loss": 0.1696,
       "step": 3500
     },
     {
-      "epoch": 0.39401103230890466,
-      "eval_loss": 0.7721038460731506,
-      "eval_runtime": 128.1871,
-      "eval_samples_per_second": 15.399,
-      "eval_steps_per_second": 1.927,
+      "epoch": 0.37329351535836175,
+      "eval_loss": 0.7638036012649536,
+      "eval_runtime": 63.8003,
+      "eval_samples_per_second": 15.47,
+      "eval_steps_per_second": 1.944,
       "step": 3500
     },
     {
-      "epoch": 0.4502983226387482,
-      "grad_norm": 0.006555848754942417,
-      "learning_rate": 4.4371270967015645e-07,
-      "loss": 0.1216,
+      "epoch": 0.42662116040955633,
+      "grad_norm": 2.4611830711364746,
+      "learning_rate": 4.2889647326507393e-07,
+      "loss": 0.1125,
       "step": 4000
     },
     {
-      "epoch": 0.4502983226387482,
-      "eval_loss": 0.7490401864051819,
-      "eval_runtime": 127.5516,
-      "eval_samples_per_second": 15.476,
-      "eval_steps_per_second": 1.936,
+      "epoch": 0.42662116040955633,
+      "eval_loss": 0.7486374378204346,
+      "eval_runtime": 63.9847,
+      "eval_samples_per_second": 15.426,
+      "eval_steps_per_second": 1.938,
       "step": 4000
     },
     {
-      "epoch": 0.5065856129685917,
-      "grad_norm": 0.0004515685432124883,
-      "learning_rate": 4.36676798378926e-07,
-      "loss": 0.1531,
+      "epoch": 0.47994880546075086,
+      "grad_norm": 5.230295658111572,
+      "learning_rate": 4.2000853242320814e-07,
+      "loss": 0.1322,
       "step": 4500
     },
     {
-      "epoch": 0.5065856129685917,
-      "eval_loss": 0.7533182501792908,
-      "eval_runtime": 127.4797,
-      "eval_samples_per_second": 15.485,
-      "eval_steps_per_second": 1.938,
+      "epoch": 0.47994880546075086,
+      "eval_loss": 0.7251659035682678,
+      "eval_runtime": 64.4782,
+      "eval_samples_per_second": 15.308,
+      "eval_steps_per_second": 1.923,
       "step": 4500
     },
     {
-      "epoch": 0.5628729032984352,
-      "grad_norm": 4.4633176003117114e-05,
-      "learning_rate": 4.2964088708769556e-07,
-      "loss": 0.1154,
+      "epoch": 0.5332764505119454,
+      "grad_norm": 3.821002974291332e-05,
+      "learning_rate": 4.1112059158134245e-07,
+      "loss": 0.1633,
       "step": 5000
     },
     {
-      "epoch": 0.5628729032984352,
-      "eval_loss": 0.747530460357666,
-      "eval_runtime": 127.8675,
-      "eval_samples_per_second": 15.438,
-      "eval_steps_per_second": 1.932,
+      "epoch": 0.5332764505119454,
+      "eval_loss": 0.7035414576530457,
+      "eval_runtime": 62.6186,
+      "eval_samples_per_second": 15.762,
+      "eval_steps_per_second": 1.98,
       "step": 5000
     },
     {
-      "epoch": 0.6191601936282788,
-      "grad_norm": 4.158839702606201,
-      "learning_rate": 4.2260497579646517e-07,
-      "loss": 0.1407,
+      "epoch": 0.58660409556314,
+      "grad_norm": 5.916939244343666e-06,
+      "learning_rate": 4.0223265073947665e-07,
+      "loss": 0.125,
       "step": 5500
     },
     {
-      "epoch": 0.6191601936282788,
-      "eval_loss": 0.724854588508606,
-      "eval_runtime": 125.1629,
-      "eval_samples_per_second": 15.771,
-      "eval_steps_per_second": 1.973,
+      "epoch": 0.58660409556314,
+      "eval_loss": 0.6792380809783936,
+      "eval_runtime": 65.0551,
+      "eval_samples_per_second": 15.172,
+      "eval_steps_per_second": 1.906,
       "step": 5500
     },
     {
-      "epoch": 0.6754474839581223,
-      "grad_norm": 0.748630166053772,
-      "learning_rate": 4.155690645052347e-07,
-      "loss": 0.1447,
+      "epoch": 0.6399317406143344,
+      "grad_norm": 92.8386459350586,
+      "learning_rate": 3.933447098976109e-07,
+      "loss": 0.1297,
       "step": 6000
     },
     {
-      "epoch": 0.6754474839581223,
-      "eval_loss": 0.7091771960258484,
-      "eval_runtime": 125.2549,
-      "eval_samples_per_second": 15.76,
-      "eval_steps_per_second": 1.972,
+      "epoch": 0.6399317406143344,
+      "eval_loss": 0.6549482345581055,
+      "eval_runtime": 64.0827,
+      "eval_samples_per_second": 15.402,
+      "eval_steps_per_second": 1.935,
       "step": 6000
     },
     {
-      "epoch": 0.7317347742879657,
-      "grad_norm": 2.8120924980612472e-05,
-      "learning_rate": 4.0853315321400427e-07,
-      "loss": 0.1146,
+      "epoch": 0.693259385665529,
+      "grad_norm": 0.00014717792510055006,
+      "learning_rate": 3.8445676905574517e-07,
+      "loss": 0.1114,
       "step": 6500
     },
     {
-      "epoch": 0.7317347742879657,
-      "eval_loss": 0.7105826735496521,
-      "eval_runtime": 127.7786,
-      "eval_samples_per_second": 15.449,
-      "eval_steps_per_second": 1.933,
+      "epoch": 0.693259385665529,
+      "eval_loss": 0.6223254799842834,
+      "eval_runtime": 64.89,
+      "eval_samples_per_second": 15.21,
+      "eval_steps_per_second": 1.911,
       "step": 6500
     },
     {
-      "epoch": 0.7880220646178093,
-      "grad_norm": 1.4741635823156685e-07,
-      "learning_rate": 4.014972419227738e-07,
-      "loss": 0.171,
+      "epoch": 0.7465870307167235,
+      "grad_norm": 4.789559397977428e-07,
+      "learning_rate": 3.755688282138794e-07,
+      "loss": 0.144,
       "step": 7000
     },
     {
-      "epoch": 0.7880220646178093,
-      "eval_loss": 0.6978534460067749,
-      "eval_runtime": 125.2567,
-      "eval_samples_per_second": 15.76,
-      "eval_steps_per_second": 1.972,
+      "epoch": 0.7465870307167235,
+      "eval_loss": 0.6204637289047241,
+      "eval_runtime": 62.7674,
+      "eval_samples_per_second": 15.725,
+      "eval_steps_per_second": 1.976,
       "step": 7000
     },
     {
-      "epoch": 0.8443093549476528,
-      "grad_norm": 0.002975167240947485,
-      "learning_rate": 3.944613306315434e-07,
-      "loss": 0.1584,
+      "epoch": 0.7999146757679181,
+      "grad_norm": 1.7078508138656616,
+      "learning_rate": 3.6668088737201363e-07,
+      "loss": 0.1399,
       "step": 7500
     },
     {
-      "epoch": 0.8443093549476528,
-      "eval_loss": 0.6923494935035706,
-      "eval_runtime": 125.3384,
-      "eval_samples_per_second": 15.749,
-      "eval_steps_per_second": 1.971,
+      "epoch": 0.7999146757679181,
+      "eval_loss": 0.6074371933937073,
+      "eval_runtime": 63.4694,
+      "eval_samples_per_second": 15.551,
+      "eval_steps_per_second": 1.954,
       "step": 7500
     },
     {
-      "epoch": 0.9005966452774964,
-      "grad_norm": 9.225498797604814e-05,
-      "learning_rate": 3.8742541934031293e-07,
-      "loss": 0.1384,
+      "epoch": 0.8532423208191127,
+      "grad_norm": 870.1183471679688,
+      "learning_rate": 3.5779294653014783e-07,
+      "loss": 0.141,
       "step": 8000
     },
     {
-      "epoch": 0.9005966452774964,
-      "eval_loss": 0.6872902512550354,
-      "eval_runtime": 128.114,
-      "eval_samples_per_second": 15.408,
-      "eval_steps_per_second": 1.928,
+      "epoch": 0.8532423208191127,
+      "eval_loss": 0.5950366854667664,
+      "eval_runtime": 62.915,
+      "eval_samples_per_second": 15.688,
+      "eval_steps_per_second": 1.971,
       "step": 8000
     },
     {
-      "epoch": 0.9568839356073399,
-      "grad_norm": 1.9474915902151224e-08,
-      "learning_rate": 3.803895080490825e-07,
-      "loss": 0.1243,
+      "epoch": 0.9065699658703071,
+      "grad_norm": 5.781071013188921e-05,
+      "learning_rate": 3.4890500568828214e-07,
+      "loss": 0.1291,
       "step": 8500
     },
     {
-      "epoch": 0.9568839356073399,
-      "eval_loss": 0.6788680553436279,
-      "eval_runtime": 125.0525,
-      "eval_samples_per_second": 15.785,
-      "eval_steps_per_second": 1.975,
+      "epoch": 0.9065699658703071,
+      "eval_loss": 0.5800224542617798,
+      "eval_runtime": 62.3336,
+      "eval_samples_per_second": 15.834,
+      "eval_steps_per_second": 1.989,
       "step": 8500
     },
     {
-      "epoch": 1.0131712259371835,
-      "grad_norm": 5.165647506713867,
-      "learning_rate": 3.7335359675785204e-07,
-      "loss": 0.1102,
+      "epoch": 0.9598976109215017,
+      "grad_norm": 1.3669992685317993,
+      "learning_rate": 3.4001706484641635e-07,
+      "loss": 0.1246,
       "step": 9000
     },
     {
-      "epoch": 1.0131712259371835,
-      "eval_loss": 0.6788864135742188,
-      "eval_runtime": 124.7777,
-      "eval_samples_per_second": 15.82,
-      "eval_steps_per_second": 1.98,
+      "epoch": 0.9598976109215017,
+      "eval_loss": 0.5702486038208008,
+      "eval_runtime": 64.055,
+      "eval_samples_per_second": 15.409,
+      "eval_steps_per_second": 1.936,
       "step": 9000
     },
     {
-      "epoch": 1.069458516267027,
-      "grad_norm": 0.0005840375670231879,
-      "learning_rate": 3.6631768546662164e-07,
-      "loss": 0.0523,
+      "epoch": 1.0132252559726962,
+      "grad_norm": 644.68115234375,
+      "learning_rate": 3.311291240045506e-07,
+      "loss": 0.1257,
       "step": 9500
     },
     {
-      "epoch": 1.069458516267027,
-      "eval_loss": 0.6628187894821167,
-      "eval_runtime": 127.8018,
-      "eval_samples_per_second": 15.446,
-      "eval_steps_per_second": 1.933,
+      "epoch": 1.0132252559726962,
+      "eval_loss": 0.5807780623435974,
+      "eval_runtime": 62.5312,
+      "eval_samples_per_second": 15.784,
+      "eval_steps_per_second": 1.983,
       "step": 9500
     },
     {
-      "epoch": 1.1257458065968704,
-      "grad_norm": 0.001191094284877181,
-      "learning_rate": 3.592817741753912e-07,
-      "loss": 0.0617,
+      "epoch": 1.0665529010238908,
+      "grad_norm": 27.854568481445312,
+      "learning_rate": 3.2224118316268486e-07,
+      "loss": 0.0336,
       "step": 10000
     },
     {
-      "epoch": 1.1257458065968704,
-      "eval_loss": 0.6507942080497742,
-      "eval_runtime": 125.484,
-      "eval_samples_per_second": 15.731,
-      "eval_steps_per_second": 1.968,
+      "epoch": 1.0665529010238908,
+      "eval_loss": 0.5713164210319519,
+      "eval_runtime": 62.5622,
+      "eval_samples_per_second": 15.776,
+      "eval_steps_per_second": 1.982,
       "step": 10000
     },
     {
-      "epoch": 1.1820330969267139,
-      "grad_norm": 0.0003092484548687935,
-      "learning_rate": 3.5224586288416075e-07,
-      "loss": 0.0598,
+      "epoch": 1.1198805460750854,
+      "grad_norm": 0.00045933027286082506,
+      "learning_rate": 3.133532423208191e-07,
+      "loss": 0.0663,
       "step": 10500
     },
     {
-      "epoch": 1.1820330969267139,
-      "eval_loss": 0.6480408906936646,
-      "eval_runtime": 125.7661,
-      "eval_samples_per_second": 15.696,
-      "eval_steps_per_second": 1.964,
+      "epoch": 1.1198805460750854,
+      "eval_loss": 0.5601483583450317,
+      "eval_runtime": 63.4675,
+      "eval_samples_per_second": 15.551,
+      "eval_steps_per_second": 1.954,
       "step": 10500
     },
     {
-      "epoch": 1.2383203872565574,
-      "grad_norm": 5.508670710696606e-07,
-      "learning_rate": 3.452099515929303e-07,
-      "loss": 0.0782,
+      "epoch": 1.17320819112628,
+      "grad_norm": 0.14330387115478516,
+      "learning_rate": 3.044653014789533e-07,
+      "loss": 0.078,
       "step": 11000
     },
     {
-      "epoch": 1.2383203872565574,
-      "eval_loss": 0.6403237581253052,
-      "eval_runtime": 128.3947,
-      "eval_samples_per_second": 15.374,
-      "eval_steps_per_second": 1.924,
+      "epoch": 1.17320819112628,
+      "eval_loss": 0.5580261945724487,
+      "eval_runtime": 63.3927,
+      "eval_samples_per_second": 15.57,
+      "eval_steps_per_second": 1.956,
       "step": 11000
     },
     {
-      "epoch": 1.294607677586401,
-      "grad_norm": 0.1470593810081482,
-      "learning_rate": 3.3817404030169986e-07,
-      "loss": 0.0616,
+      "epoch": 1.2265358361774745,
+      "grad_norm": 4.066005518388316e-11,
+      "learning_rate": 2.955773606370876e-07,
+      "loss": 0.1118,
       "step": 11500
     },
     {
-      "epoch": 1.294607677586401,
-      "eval_loss": 0.6376460790634155,
-      "eval_runtime": 125.5391,
-      "eval_samples_per_second": 15.724,
-      "eval_steps_per_second": 1.968,
+      "epoch": 1.2265358361774745,
+      "eval_loss": 0.5491370558738708,
+      "eval_runtime": 63.7293,
+      "eval_samples_per_second": 15.487,
+      "eval_steps_per_second": 1.946,
       "step": 11500
     },
     {
-      "epoch": 1.3508949679162445,
-      "grad_norm": 1.8021532014245167e-05,
-      "learning_rate": 3.3113812901046946e-07,
-      "loss": 0.0723,
+      "epoch": 1.2798634812286689,
+      "grad_norm": 7.519358769059181e-05,
+      "learning_rate": 2.8668941979522184e-07,
+      "loss": 0.0534,
       "step": 12000
     },
     {
-      "epoch": 1.3508949679162445,
-      "eval_loss": 0.6378474831581116,
-      "eval_runtime": 125.5914,
-      "eval_samples_per_second": 15.718,
-      "eval_steps_per_second": 1.967,
+      "epoch": 1.2798634812286689,
+      "eval_loss": 0.5478764772415161,
+      "eval_runtime": 62.7547,
+      "eval_samples_per_second": 15.728,
+      "eval_steps_per_second": 1.976,
       "step": 12000
     },
     {
-      "epoch": 1.407182258246088,
-      "grad_norm": 0.00013483635848388076,
-      "learning_rate": 3.2410221771923896e-07,
-      "loss": 0.0459,
+      "epoch": 1.3331911262798635,
+      "grad_norm": 8.400617446113756e-08,
+      "learning_rate": 2.778014789533561e-07,
+      "loss": 0.0507,
       "step": 12500
     },
     {
-      "epoch": 1.407182258246088,
-      "eval_loss": 0.6378260850906372,
-      "eval_runtime": 128.2725,
-      "eval_samples_per_second": 15.389,
-      "eval_steps_per_second": 1.926,
+      "epoch": 1.3331911262798635,
+      "eval_loss": 0.576600968837738,
+      "eval_runtime": 64.301,
+      "eval_samples_per_second": 15.35,
+      "eval_steps_per_second": 1.928,
       "step": 12500
     },
     {
-      "epoch": 1.4634695485759315,
-      "grad_norm": 4.757418707868055e-07,
-      "learning_rate": 3.170663064280085e-07,
-      "loss": 0.0571,
+      "epoch": 1.386518771331058,
+      "grad_norm": 2.5385464130084356e-12,
+      "learning_rate": 2.689135381114903e-07,
+      "loss": 0.0784,
       "step": 13000
     },
     {
-      "epoch": 1.4634695485759315,
-      "eval_loss": 0.6366899609565735,
-      "eval_runtime": 125.6986,
-      "eval_samples_per_second": 15.704,
-      "eval_steps_per_second": 1.965,
+      "epoch": 1.386518771331058,
+      "eval_loss": 0.5729931592941284,
+      "eval_runtime": 63.652,
+      "eval_samples_per_second": 15.506,
+      "eval_steps_per_second": 1.948,
       "step": 13000
     },
     {
-      "epoch": 1.5197568389057752,
-      "grad_norm": 0.00028420978924259543,
-      "learning_rate": 3.1003039513677807e-07,
-      "loss": 0.0499,
+      "epoch": 1.4398464163822526,
+      "grad_norm": 0.08409273624420166,
+      "learning_rate": 2.600255972696246e-07,
+      "loss": 0.079,
       "step": 13500
     },
     {
-      "epoch": 1.5197568389057752,
-      "eval_loss": 0.6286123394966125,
-      "eval_runtime": 127.9165,
-      "eval_samples_per_second": 15.432,
-      "eval_steps_per_second": 1.931,
+      "epoch": 1.4398464163822526,
+      "eval_loss": 0.5694039463996887,
+      "eval_runtime": 62.8247,
+      "eval_samples_per_second": 15.71,
+      "eval_steps_per_second": 1.974,
       "step": 13500
     },
     {
-      "epoch": 1.5760441292356187,
-      "grad_norm": 2.1999912291903456e-07,
-      "learning_rate": 3.0299448384554767e-07,
-      "loss": 0.0679,
+      "epoch": 1.493174061433447,
+      "grad_norm": 3.7334253022436314e-08,
+      "learning_rate": 2.511376564277588e-07,
+      "loss": 0.0507,
       "step": 14000
     },
     {
-      "epoch": 1.5760441292356187,
-      "eval_loss": 0.634914755821228,
-      "eval_runtime": 125.0396,
-      "eval_samples_per_second": 15.787,
-      "eval_steps_per_second": 1.975,
+      "epoch": 1.493174061433447,
+      "eval_loss": 0.5713907480239868,
+      "eval_runtime": 62.3577,
+      "eval_samples_per_second": 15.828,
+      "eval_steps_per_second": 1.989,
       "step": 14000
     },
     {
-      "epoch": 1.6323314195654621,
-      "grad_norm": 0.00023505109129473567,
-      "learning_rate": 2.959585725543172e-07,
-      "loss": 0.0554,
+      "epoch": 1.5465017064846416,
+      "grad_norm": 1.5219894647598267,
+      "learning_rate": 2.422497155858931e-07,
+      "loss": 0.0774,
       "step": 14500
     },
     {
-      "epoch": 1.6323314195654621,
-      "eval_loss": 0.623322069644928,
-      "eval_runtime": 125.2429,
-      "eval_samples_per_second": 15.761,
-      "eval_steps_per_second": 1.972,
+      "epoch": 1.5465017064846416,
+      "eval_loss": 0.5674533843994141,
+      "eval_runtime": 62.3753,
+      "eval_samples_per_second": 15.824,
+      "eval_steps_per_second": 1.988,
       "step": 14500
     },
     {
-      "epoch": 1.6886187098953056,
-      "grad_norm": 11.346132278442383,
-      "learning_rate": 2.889226612630868e-07,
-      "loss": 0.0617,
+      "epoch": 1.5998293515358362,
+      "grad_norm": 879.1193237304688,
+      "learning_rate": 2.333617747440273e-07,
+      "loss": 0.0376,
       "step": 15000
     },
     {
-      "epoch": 1.6886187098953056,
-      "eval_loss": 0.6115825772285461,
-      "eval_runtime": 127.8813,
-      "eval_samples_per_second": 15.436,
-      "eval_steps_per_second": 1.931,
+      "epoch": 1.5998293515358362,
+      "eval_loss": 0.561253547668457,
+      "eval_runtime": 62.4788,
+      "eval_samples_per_second": 15.797,
+      "eval_steps_per_second": 1.985,
       "step": 15000
     },
     {
-      "epoch": 1.744906000225149,
-      "grad_norm": 308.43804931640625,
-      "learning_rate": 2.8188674997185633e-07,
-      "loss": 0.0561,
+      "epoch": 1.6531569965870307,
+      "grad_norm": 0.0006221308722160757,
+      "learning_rate": 2.2447383390216154e-07,
+      "loss": 0.0929,
       "step": 15500
     },
     {
-      "epoch": 1.744906000225149,
-      "eval_loss": 0.5951005220413208,
-      "eval_runtime": 125.2491,
-      "eval_samples_per_second": 15.761,
-      "eval_steps_per_second": 1.972,
+      "epoch": 1.6531569965870307,
+      "eval_loss": 0.5395999550819397,
+      "eval_runtime": 62.5651,
+      "eval_samples_per_second": 15.776,
+      "eval_steps_per_second": 1.982,
       "step": 15500
     },
     {
-      "epoch": 1.8011932905549926,
-      "grad_norm": 6.91558055265773e-09,
-      "learning_rate": 2.7485083868062594e-07,
-      "loss": 0.0672,
+      "epoch": 1.7064846416382253,
+      "grad_norm": 37.85109329223633,
+      "learning_rate": 2.155858930602958e-07,
+      "loss": 0.0442,
       "step": 16000
     },
     {
-      "epoch": 1.8011932905549926,
-      "eval_loss": 0.5809568166732788,
-      "eval_runtime": 125.2557,
-      "eval_samples_per_second": 15.76,
-      "eval_steps_per_second": 1.972,
+      "epoch": 1.7064846416382253,
+      "eval_loss": 0.5307884216308594,
+      "eval_runtime": 62.4681,
+      "eval_samples_per_second": 15.8,
+      "eval_steps_per_second": 1.985,
       "step": 16000
     },
     {
-      "epoch": 1.8574805808848363,
-      "grad_norm": 2.2250075915053458e-07,
-      "learning_rate": 2.678149273893955e-07,
-      "loss": 0.0812,
+      "epoch": 1.75981228668942,
+      "grad_norm": 9.600337885773413e-11,
+      "learning_rate": 2.0669795221843002e-07,
+      "loss": 0.0593,
       "step": 16500
     },
     {
-      "epoch": 1.8574805808848363,
-      "eval_loss": 0.5691242218017578,
-      "eval_runtime": 128.3032,
-      "eval_samples_per_second": 15.385,
-      "eval_steps_per_second": 1.925,
+      "epoch": 1.75981228668942,
+      "eval_loss": 0.5230685472488403,
+      "eval_runtime": 63.8124,
+      "eval_samples_per_second": 15.467,
+      "eval_steps_per_second": 1.943,
       "step": 16500
     },
     {
-      "epoch": 1.9137678712146797,
-      "grad_norm": 1.1724663972854614,
-      "learning_rate": 2.6077901609816504e-07,
-      "loss": 0.0614,
+      "epoch": 1.8131399317406145,
+      "grad_norm": 9.067448940129619e-13,
+      "learning_rate": 1.9781001137656428e-07,
+      "loss": 0.0578,
       "step": 17000
     },
     {
-      "epoch": 1.9137678712146797,
-      "eval_loss": 0.5643230080604553,
-      "eval_runtime": 125.7191,
-      "eval_samples_per_second": 15.702,
-      "eval_steps_per_second": 1.965,
+      "epoch": 1.8131399317406145,
+      "eval_loss": 0.5200281739234924,
+      "eval_runtime": 62.3767,
+      "eval_samples_per_second": 15.823,
+      "eval_steps_per_second": 1.988,
       "step": 17000
     },
     {
-      "epoch": 1.9700551615445232,
-      "grad_norm": 0.00030226208036765456,
-      "learning_rate": 2.5374310480693454e-07,
-      "loss": 0.0462,
+      "epoch": 1.8664675767918089,
+      "grad_norm": 0.00843018852174282,
+      "learning_rate": 1.889220705346985e-07,
+      "loss": 0.0928,
       "step": 17500
     },
     {
-      "epoch": 1.9700551615445232,
-      "eval_loss": 0.5657238364219666,
-      "eval_runtime": 124.9923,
-      "eval_samples_per_second": 15.793,
-      "eval_steps_per_second": 1.976,
+      "epoch": 1.8664675767918089,
+      "eval_loss": 0.509168803691864,
+      "eval_runtime": 62.4108,
+      "eval_samples_per_second": 15.815,
+      "eval_steps_per_second": 1.987,
       "step": 17500
     },
     {
-      "epoch": 2.026342451874367,
-      "grad_norm": 1.949754704355655e-07,
-      "learning_rate": 2.4670719351570415e-07,
-      "loss": 0.0565,
+      "epoch": 1.9197952218430034,
+      "grad_norm": 4.264631314754297e-09,
+      "learning_rate": 1.8003412969283277e-07,
+      "loss": 0.0633,
       "step": 18000
     },
     {
-      "epoch": 2.026342451874367,
-      "eval_loss": 0.5607953667640686,
-      "eval_runtime": 127.7187,
-      "eval_samples_per_second": 15.456,
-      "eval_steps_per_second": 1.934,
+      "epoch": 1.9197952218430034,
+      "eval_loss": 0.4954444468021393,
+      "eval_runtime": 63.3639,
+      "eval_samples_per_second": 15.577,
+      "eval_steps_per_second": 1.957,
       "step": 18000
     },
     {
-      "epoch": 2.0826297422042104,
-      "grad_norm": 14.2212553024292,
-      "learning_rate": 2.396712822244737e-07,
-      "loss": 0.0549,
+      "epoch": 1.9731228668941978,
+      "grad_norm": 459.5137939453125,
+      "learning_rate": 1.7114618885096697e-07,
+      "loss": 0.067,
       "step": 18500
     },
     {
-      "epoch": 2.0826297422042104,
-      "eval_loss": 0.5568962693214417,
-      "eval_runtime": 125.0503,
-      "eval_samples_per_second": 15.786,
-      "eval_steps_per_second": 1.975,
+      "epoch": 1.9731228668941978,
+      "eval_loss": 0.49844062328338623,
+      "eval_runtime": 63.6573,
+      "eval_samples_per_second": 15.505,
+      "eval_steps_per_second": 1.948,
       "step": 18500
     },
     {
-      "epoch": 2.138917032534054,
-      "grad_norm": 1.2045973539352417,
-      "learning_rate": 2.3263537093324326e-07,
-      "loss": 0.0245,
+      "epoch": 2.0264505119453924,
+      "grad_norm": 0.27590689063072205,
+      "learning_rate": 1.6225824800910123e-07,
+      "loss": 0.0525,
       "step": 19000
     },
     {
-      "epoch": 2.138917032534054,
-      "eval_loss": 0.5595241785049438,
-      "eval_runtime": 124.8527,
-      "eval_samples_per_second": 15.811,
-      "eval_steps_per_second": 1.978,
+      "epoch": 2.0264505119453924,
+      "eval_loss": 0.4999145269393921,
+      "eval_runtime": 63.1727,
+      "eval_samples_per_second": 15.624,
+      "eval_steps_per_second": 1.963,
       "step": 19000
     },
     {
-      "epoch": 2.1952043228638973,
-      "grad_norm": 8.770056410867255e-06,
-      "learning_rate": 2.2559945964201284e-07,
-      "loss": 0.0428,
+      "epoch": 2.079778156996587,
+      "grad_norm": 1.349623744317796e-06,
+      "learning_rate": 1.533703071672355e-07,
+      "loss": 0.0301,
       "step": 19500
     },
     {
-      "epoch": 2.1952043228638973,
-      "eval_loss": 0.5740877985954285,
-      "eval_runtime": 127.9211,
-      "eval_samples_per_second": 15.431,
-      "eval_steps_per_second": 1.931,
+      "epoch": 2.079778156996587,
+      "eval_loss": 0.4926412105560303,
+      "eval_runtime": 63.4421,
+      "eval_samples_per_second": 15.557,
+      "eval_steps_per_second": 1.955,
       "step": 19500
     },
     {
-      "epoch": 2.251491613193741,
-      "grad_norm": 6.243853079013206e-08,
-      "learning_rate": 2.185635483507824e-07,
-      "loss": 0.0301,
+      "epoch": 2.1331058020477816,
+      "grad_norm": 4.928138878312893e-06,
+      "learning_rate": 1.4448236632536972e-07,
+      "loss": 0.0408,
       "step": 20000
     },
     {
-      "epoch": 2.251491613193741,
-      "eval_loss": 0.5761672854423523,
-      "eval_runtime": 125.2829,
-      "eval_samples_per_second": 15.756,
-      "eval_steps_per_second": 1.972,
+      "epoch": 2.1331058020477816,
+      "eval_loss": 0.4872625172138214,
+      "eval_runtime": 63.2047,
+      "eval_samples_per_second": 15.616,
+      "eval_steps_per_second": 1.962,
       "step": 20000
     },
     {
-      "epoch": 2.3077789035235843,
-      "grad_norm": 2.299124268745345e-08,
-      "learning_rate": 2.1152763705955197e-07,
-      "loss": 0.04,
+      "epoch": 2.186433447098976,
+      "grad_norm": 0.003605826525017619,
+      "learning_rate": 1.3559442548350398e-07,
+      "loss": 0.0489,
       "step": 20500
     },
     {
-      "epoch": 2.3077789035235843,
-      "eval_loss": 0.5712103843688965,
-      "eval_runtime": 125.3797,
-      "eval_samples_per_second": 15.744,
-      "eval_steps_per_second": 1.97,
+      "epoch": 2.186433447098976,
+      "eval_loss": 0.48834049701690674,
+      "eval_runtime": 63.5831,
+      "eval_samples_per_second": 15.523,
+      "eval_steps_per_second": 1.95,
       "step": 20500
     },
     {
-      "epoch": 2.3640661938534278,
-      "grad_norm": 1.971009623957798e-06,
-      "learning_rate": 2.044917257683215e-07,
-      "loss": 0.0349,
+      "epoch": 2.2397610921501707,
+      "grad_norm": 0.0027391298208385706,
+      "learning_rate": 1.267064846416382e-07,
+      "loss": 0.0354,
       "step": 21000
     },
     {
-      "epoch": 2.3640661938534278,
-      "eval_loss": 0.5690667629241943,
-      "eval_runtime": 125.4733,
-      "eval_samples_per_second": 15.732,
-      "eval_steps_per_second": 1.969,
+      "epoch": 2.2397610921501707,
+      "eval_loss": 0.4887964129447937,
+      "eval_runtime": 63.0462,
+      "eval_samples_per_second": 15.655,
+      "eval_steps_per_second": 1.967,
       "step": 21000
     }
   ],
   "logging_steps": 500,
-  "max_steps": 35532,
+  "max_steps": 28128,
   "num_input_tokens_seen": 0,
-  "num_train_epochs": 4,
+  "num_train_epochs": 3,
   "save_steps": 500,
   "stateful_callbacks": {
     "TrainerControl": {
@@ -656,7 +656,7 @@
       "attributes": {}
     }
   },
-  "total_flos": 7555285134811260.0,
+  "total_flos": 7555644893162520.0,
   "train_batch_size": 2,
   "trial_name": null,
   "trial_params": null

checkpoint-21000/training_args.bin

@@ -1,3 +1,3 @@
 version https://git-lfs.github.com/spec/v1
-oid sha256:319be2a769ef7117b6826908976f080f9082f159f4f0ad75da1a151ad29d9bff
+oid sha256:057f6bbf1ff3e34abda0863071bbf5fdcc7edb1102f04dba05c7a15087bc3396
 size 4847