Training in progress, step 21000, checkpoint

checkpoint-21000/model.safetensors

@@ -1,3 +1,3 @@
 version https://git-lfs.github.com/spec/v1
-oid sha256:ea50fac774db3c9cb31abb6664787287909cb07e28db1e469d90f49d2c7441b6
+oid sha256:8d88e4391b49b7f47122e6579b3d3acfbc9c067ea039fb9eabac0ab6772aff48
 size 1711848436

checkpoint-21000/optimizer.pt

@@ -1,3 +1,3 @@
 version https://git-lfs.github.com/spec/v1
-oid sha256:6ce3ec4e146af87c30698f675f685d1d1a15658b05ea253d024c37da6b2163f7
+oid sha256:8b8ba3e1769c9a5eb1594683a4c35c2c714234c4df59ce37892e4d1d138efb08
 size 3424043887

checkpoint-21000/rng_state.pth

@@ -1,3 +1,3 @@
 version https://git-lfs.github.com/spec/v1
-oid sha256:affd57ddcceffdb912ac3fa26b73d1fcfe8f6aab8df2dcc9dd1f5986cf46914c
+oid sha256:551bd9fefe21a0210b8f009f7e379124ea37132a2e134634bfee8fc680d70fd3
 size 14503

checkpoint-21000/scheduler.pt

@@ -1,3 +1,3 @@
 version https://git-lfs.github.com/spec/v1
-oid sha256:3069ad9714a7c357221ee9aea0d2606ad0c81ae4e83d04a48cc102d867288b84
+oid sha256:259cafb41160652c3d54806748b10cd7a9019d0dd9321fe4c5ddeee2dc391a9c
 size 623

checkpoint-21000/trainer_state.json

@@ -1,7 +1,7 @@
 {
-  "best_metric": 0.49561700224876404,
-  "best_model_checkpoint": "./output/clip-finetuned-csu-p14-336-e4l57-l/checkpoint-21000",
-  "epoch": 3.868828297715549,
+  "best_metric": 0.5568962693214417,
+  "best_model_checkpoint": "./output/clip-finetuned-csu-p14-336-e4l57-l/checkpoint-18500",
+  "epoch": 2.3640661938534278,
   "eval_steps": 500,
   "global_step": 21000,
   "is_hyper_param_search": false,
@@ -9,638 +9,638 @@
   "is_world_process_zero": true,
   "log_history": [
     {
-      "epoch": 0.09211495946941783,
-      "grad_norm": 3.0293657779693604,
-      "learning_rate": 4.884856300663227e-07,
-      "loss": 0.271,
+      "epoch": 0.056287290329843524,
+      "grad_norm": 150.1642303466797,
+      "learning_rate": 4.929640887087696e-07,
+      "loss": 0.3486,
       "step": 500
     },
     {
-      "epoch": 0.09211495946941783,
-      "eval_loss": 1.069305419921875,
-      "eval_runtime": 75.1409,
-      "eval_samples_per_second": 16.063,
-      "eval_steps_per_second": 2.01,
+      "epoch": 0.056287290329843524,
+      "eval_loss": 1.1265727281570435,
+      "eval_runtime": 123.7806,
+      "eval_samples_per_second": 15.948,
+      "eval_steps_per_second": 1.995,
       "step": 500
     },
     {
-      "epoch": 0.18422991893883567,
-      "grad_norm": 7.3040852546691895,
-      "learning_rate": 4.769712601326456e-07,
-      "loss": 0.2493,
+      "epoch": 0.11257458065968705,
+      "grad_norm": 366.6913146972656,
+      "learning_rate": 4.859281774175391e-07,
+      "loss": 0.2733,
       "step": 1000
     },
     {
-      "epoch": 0.18422991893883567,
-      "eval_loss": 0.9427077770233154,
-      "eval_runtime": 74.8926,
-      "eval_samples_per_second": 16.116,
-      "eval_steps_per_second": 2.016,
+      "epoch": 0.11257458065968705,
+      "eval_loss": 0.9742079377174377,
+      "eval_runtime": 125.8047,
+      "eval_samples_per_second": 15.691,
+      "eval_steps_per_second": 1.963,
       "step": 1000
     },
     {
-      "epoch": 0.2763448784082535,
-      "grad_norm": 0.020736657083034515,
-      "learning_rate": 4.654568901989683e-07,
-      "loss": 0.2348,
+      "epoch": 0.16886187098953057,
+      "grad_norm": 555.5016479492188,
+      "learning_rate": 4.788922661263087e-07,
+      "loss": 0.1851,
       "step": 1500
     },
     {
-      "epoch": 0.2763448784082535,
-      "eval_loss": 0.8726764917373657,
-      "eval_runtime": 76.6621,
-      "eval_samples_per_second": 15.744,
-      "eval_steps_per_second": 1.97,
+      "epoch": 0.16886187098953057,
+      "eval_loss": 0.9162159562110901,
+      "eval_runtime": 126.7606,
+      "eval_samples_per_second": 15.573,
+      "eval_steps_per_second": 1.949,
       "step": 1500
     },
     {
-      "epoch": 0.36845983787767134,
-      "grad_norm": 5.3738909628009424e-05,
-      "learning_rate": 4.5394252026529107e-07,
-      "loss": 0.1552,
+      "epoch": 0.2251491613193741,
+      "grad_norm": 0.423260897397995,
+      "learning_rate": 4.7185635483507824e-07,
+      "loss": 0.1973,
       "step": 2000
     },
     {
-      "epoch": 0.36845983787767134,
-      "eval_loss": 0.8325753808021545,
-      "eval_runtime": 76.6127,
-      "eval_samples_per_second": 15.755,
-      "eval_steps_per_second": 1.971,
+      "epoch": 0.2251491613193741,
+      "eval_loss": 0.8716973662376404,
+      "eval_runtime": 127.6955,
+      "eval_samples_per_second": 15.459,
+      "eval_steps_per_second": 1.934,
       "step": 2000
     },
     {
-      "epoch": 0.46057479734708917,
-      "grad_norm": 404.510009765625,
-      "learning_rate": 4.4242815033161386e-07,
-      "loss": 0.1753,
+      "epoch": 0.2814364516492176,
+      "grad_norm": 6.840139389038086,
+      "learning_rate": 4.6482044354384774e-07,
+      "loss": 0.1881,
       "step": 2500
     },
     {
-      "epoch": 0.46057479734708917,
-      "eval_loss": 0.7550467848777771,
-      "eval_runtime": 76.8024,
-      "eval_samples_per_second": 15.716,
-      "eval_steps_per_second": 1.966,
+      "epoch": 0.2814364516492176,
+      "eval_loss": 0.830047070980072,
+      "eval_runtime": 128.0771,
+      "eval_samples_per_second": 15.413,
+      "eval_steps_per_second": 1.929,
       "step": 2500
     },
     {
-      "epoch": 0.552689756816507,
-      "grad_norm": 370.7579650878906,
-      "learning_rate": 4.3091378039793665e-07,
-      "loss": 0.1659,
+      "epoch": 0.33772374197906113,
+      "grad_norm": 729.8377685546875,
+      "learning_rate": 4.577845322526173e-07,
+      "loss": 0.1695,
       "step": 3000
     },
     {
-      "epoch": 0.552689756816507,
-      "eval_loss": 0.719167172908783,
-      "eval_runtime": 76.6327,
-      "eval_samples_per_second": 15.75,
-      "eval_steps_per_second": 1.97,
+      "epoch": 0.33772374197906113,
+      "eval_loss": 0.8265627026557922,
+      "eval_runtime": 128.052,
+      "eval_samples_per_second": 15.416,
+      "eval_steps_per_second": 1.929,
       "step": 3000
     },
     {
-      "epoch": 0.6448047162859248,
-      "grad_norm": 24.97766876220703,
-      "learning_rate": 4.193994104642594e-07,
-      "loss": 0.105,
+      "epoch": 0.39401103230890466,
+      "grad_norm": 7.96999938756926e-06,
+      "learning_rate": 4.507486209613869e-07,
+      "loss": 0.155,
       "step": 3500
     },
     {
-      "epoch": 0.6448047162859248,
-      "eval_loss": 0.7117515802383423,
-      "eval_runtime": 77.1286,
-      "eval_samples_per_second": 15.649,
-      "eval_steps_per_second": 1.958,
+      "epoch": 0.39401103230890466,
+      "eval_loss": 0.7721038460731506,
+      "eval_runtime": 128.1871,
+      "eval_samples_per_second": 15.399,
+      "eval_steps_per_second": 1.927,
       "step": 3500
     },
     {
-      "epoch": 0.7369196757553427,
-      "grad_norm": 305.62060546875,
-      "learning_rate": 4.0788504053058217e-07,
-      "loss": 0.1336,
+      "epoch": 0.4502983226387482,
+      "grad_norm": 0.006555848754942417,
+      "learning_rate": 4.4371270967015645e-07,
+      "loss": 0.1216,
       "step": 4000
     },
     {
-      "epoch": 0.7369196757553427,
-      "eval_loss": 0.6952972412109375,
-      "eval_runtime": 76.7253,
-      "eval_samples_per_second": 15.731,
-      "eval_steps_per_second": 1.968,
+      "epoch": 0.4502983226387482,
+      "eval_loss": 0.7490401864051819,
+      "eval_runtime": 127.5516,
+      "eval_samples_per_second": 15.476,
+      "eval_steps_per_second": 1.936,
       "step": 4000
     },
     {
-      "epoch": 0.8290346352247605,
-      "grad_norm": 506.12237548828125,
-      "learning_rate": 3.9637067059690496e-07,
-      "loss": 0.1154,
+      "epoch": 0.5065856129685917,
+      "grad_norm": 0.0004515685432124883,
+      "learning_rate": 4.36676798378926e-07,
+      "loss": 0.1531,
       "step": 4500
     },
     {
-      "epoch": 0.8290346352247605,
-      "eval_loss": 0.6744509935379028,
-      "eval_runtime": 76.9493,
-      "eval_samples_per_second": 15.686,
-      "eval_steps_per_second": 1.962,
+      "epoch": 0.5065856129685917,
+      "eval_loss": 0.7533182501792908,
+      "eval_runtime": 127.4797,
+      "eval_samples_per_second": 15.485,
+      "eval_steps_per_second": 1.938,
       "step": 4500
     },
     {
-      "epoch": 0.9211495946941783,
-      "grad_norm": 8.823301322991028e-05,
-      "learning_rate": 3.8485630066322774e-07,
-      "loss": 0.108,
+      "epoch": 0.5628729032984352,
+      "grad_norm": 4.4633176003117114e-05,
+      "learning_rate": 4.2964088708769556e-07,
+      "loss": 0.1154,
       "step": 5000
     },
     {
-      "epoch": 0.9211495946941783,
-      "eval_loss": 0.6559990048408508,
-      "eval_runtime": 75.2391,
-      "eval_samples_per_second": 16.042,
-      "eval_steps_per_second": 2.007,
+      "epoch": 0.5628729032984352,
+      "eval_loss": 0.747530460357666,
+      "eval_runtime": 127.8675,
+      "eval_samples_per_second": 15.438,
+      "eval_steps_per_second": 1.932,
       "step": 5000
     },
     {
-      "epoch": 1.013264554163596,
-      "grad_norm": 0.0008560372516512871,
-      "learning_rate": 3.733419307295504e-07,
-      "loss": 0.1106,
+      "epoch": 0.6191601936282788,
+      "grad_norm": 4.158839702606201,
+      "learning_rate": 4.2260497579646517e-07,
+      "loss": 0.1407,
       "step": 5500
     },
     {
-      "epoch": 1.013264554163596,
-      "eval_loss": 0.6367496848106384,
-      "eval_runtime": 77.0031,
-      "eval_samples_per_second": 15.675,
-      "eval_steps_per_second": 1.961,
+      "epoch": 0.6191601936282788,
+      "eval_loss": 0.724854588508606,
+      "eval_runtime": 125.1629,
+      "eval_samples_per_second": 15.771,
+      "eval_steps_per_second": 1.973,
       "step": 5500
     },
     {
-      "epoch": 1.105379513633014,
-      "grad_norm": 0.04510360211133957,
-      "learning_rate": 3.618275607958732e-07,
-      "loss": 0.0591,
+      "epoch": 0.6754474839581223,
+      "grad_norm": 0.748630166053772,
+      "learning_rate": 4.155690645052347e-07,
+      "loss": 0.1447,
       "step": 6000
     },
     {
-      "epoch": 1.105379513633014,
-      "eval_loss": 0.6259325742721558,
-      "eval_runtime": 75.2787,
-      "eval_samples_per_second": 16.034,
-      "eval_steps_per_second": 2.006,
+      "epoch": 0.6754474839581223,
+      "eval_loss": 0.7091771960258484,
+      "eval_runtime": 125.2549,
+      "eval_samples_per_second": 15.76,
+      "eval_steps_per_second": 1.972,
       "step": 6000
     },
     {
-      "epoch": 1.1974944731024317,
-      "grad_norm": 0.0005158882704563439,
-      "learning_rate": 3.50313190862196e-07,
-      "loss": 0.0745,
+      "epoch": 0.7317347742879657,
+      "grad_norm": 2.8120924980612472e-05,
+      "learning_rate": 4.0853315321400427e-07,
+      "loss": 0.1146,
       "step": 6500
     },
     {
-      "epoch": 1.1974944731024317,
-      "eval_loss": 0.6209793090820312,
-      "eval_runtime": 77.1989,
-      "eval_samples_per_second": 15.635,
-      "eval_steps_per_second": 1.956,
+      "epoch": 0.7317347742879657,
+      "eval_loss": 0.7105826735496521,
+      "eval_runtime": 127.7786,
+      "eval_samples_per_second": 15.449,
+      "eval_steps_per_second": 1.933,
       "step": 6500
     },
     {
-      "epoch": 1.2896094325718497,
-      "grad_norm": 3.5883346072296263e-07,
-      "learning_rate": 3.387988209285188e-07,
-      "loss": 0.0502,
+      "epoch": 0.7880220646178093,
+      "grad_norm": 1.4741635823156685e-07,
+      "learning_rate": 4.014972419227738e-07,
+      "loss": 0.171,
       "step": 7000
     },
     {
-      "epoch": 1.2896094325718497,
-      "eval_loss": 0.6133300065994263,
-      "eval_runtime": 75.1116,
-      "eval_samples_per_second": 16.069,
-      "eval_steps_per_second": 2.01,
+      "epoch": 0.7880220646178093,
+      "eval_loss": 0.6978534460067749,
+      "eval_runtime": 125.2567,
+      "eval_samples_per_second": 15.76,
+      "eval_steps_per_second": 1.972,
       "step": 7000
     },
     {
-      "epoch": 1.3817243920412676,
-      "grad_norm": 21.66597557067871,
-      "learning_rate": 3.272844509948415e-07,
-      "loss": 0.079,
+      "epoch": 0.8443093549476528,
+      "grad_norm": 0.002975167240947485,
+      "learning_rate": 3.944613306315434e-07,
+      "loss": 0.1584,
       "step": 7500
     },
     {
-      "epoch": 1.3817243920412676,
-      "eval_loss": 0.6006803512573242,
-      "eval_runtime": 75.0086,
-      "eval_samples_per_second": 16.091,
-      "eval_steps_per_second": 2.013,
+      "epoch": 0.8443093549476528,
+      "eval_loss": 0.6923494935035706,
+      "eval_runtime": 125.3384,
+      "eval_samples_per_second": 15.749,
+      "eval_steps_per_second": 1.971,
       "step": 7500
     },
     {
-      "epoch": 1.4738393515106853,
-      "grad_norm": 7.992870814632624e-05,
-      "learning_rate": 3.157700810611643e-07,
-      "loss": 0.0776,
+      "epoch": 0.9005966452774964,
+      "grad_norm": 9.225498797604814e-05,
+      "learning_rate": 3.8742541934031293e-07,
+      "loss": 0.1384,
       "step": 8000
     },
     {
-      "epoch": 1.4738393515106853,
-      "eval_loss": 0.5866427421569824,
-      "eval_runtime": 75.6702,
-      "eval_samples_per_second": 15.951,
-      "eval_steps_per_second": 1.996,
+      "epoch": 0.9005966452774964,
+      "eval_loss": 0.6872902512550354,
+      "eval_runtime": 128.114,
+      "eval_samples_per_second": 15.408,
+      "eval_steps_per_second": 1.928,
       "step": 8000
     },
     {
-      "epoch": 1.565954310980103,
-      "grad_norm": 0.0010381735628470778,
-      "learning_rate": 3.042557111274871e-07,
-      "loss": 0.0492,
+      "epoch": 0.9568839356073399,
+      "grad_norm": 1.9474915902151224e-08,
+      "learning_rate": 3.803895080490825e-07,
+      "loss": 0.1243,
       "step": 8500
     },
     {
-      "epoch": 1.565954310980103,
-      "eval_loss": 0.5679212212562561,
-      "eval_runtime": 75.0643,
-      "eval_samples_per_second": 16.08,
-      "eval_steps_per_second": 2.012,
+      "epoch": 0.9568839356073399,
+      "eval_loss": 0.6788680553436279,
+      "eval_runtime": 125.0525,
+      "eval_samples_per_second": 15.785,
+      "eval_steps_per_second": 1.975,
       "step": 8500
     },
     {
-      "epoch": 1.658069270449521,
-      "grad_norm": 1.733377956725235e-07,
-      "learning_rate": 2.927413411938099e-07,
-      "loss": 0.0794,
+      "epoch": 1.0131712259371835,
+      "grad_norm": 5.165647506713867,
+      "learning_rate": 3.7335359675785204e-07,
+      "loss": 0.1102,
       "step": 9000
     },
     {
-      "epoch": 1.658069270449521,
-      "eval_loss": 0.5761673450469971,
-      "eval_runtime": 76.2738,
-      "eval_samples_per_second": 15.825,
+      "epoch": 1.0131712259371835,
+      "eval_loss": 0.6788864135742188,
+      "eval_runtime": 124.7777,
+      "eval_samples_per_second": 15.82,
       "eval_steps_per_second": 1.98,
       "step": 9000
     },
     {
-      "epoch": 1.750184229918939,
-      "grad_norm": 3.1063287359023306e-11,
-      "learning_rate": 2.812269712601326e-07,
-      "loss": 0.0677,
+      "epoch": 1.069458516267027,
+      "grad_norm": 0.0005840375670231879,
+      "learning_rate": 3.6631768546662164e-07,
+      "loss": 0.0523,
       "step": 9500
     },
     {
-      "epoch": 1.750184229918939,
-      "eval_loss": 0.5565799474716187,
-      "eval_runtime": 75.6897,
-      "eval_samples_per_second": 15.947,
-      "eval_steps_per_second": 1.995,
+      "epoch": 1.069458516267027,
+      "eval_loss": 0.6628187894821167,
+      "eval_runtime": 127.8018,
+      "eval_samples_per_second": 15.446,
+      "eval_steps_per_second": 1.933,
       "step": 9500
     },
     {
-      "epoch": 1.8422991893883567,
-      "grad_norm": 0.00010032750287791714,
-      "learning_rate": 2.697126013264554e-07,
-      "loss": 0.0566,
+      "epoch": 1.1257458065968704,
+      "grad_norm": 0.001191094284877181,
+      "learning_rate": 3.592817741753912e-07,
+      "loss": 0.0617,
       "step": 10000
     },
     {
-      "epoch": 1.8422991893883567,
-      "eval_loss": 0.5482199192047119,
-      "eval_runtime": 75.0029,
-      "eval_samples_per_second": 16.093,
-      "eval_steps_per_second": 2.013,
+      "epoch": 1.1257458065968704,
+      "eval_loss": 0.6507942080497742,
+      "eval_runtime": 125.484,
+      "eval_samples_per_second": 15.731,
+      "eval_steps_per_second": 1.968,
       "step": 10000
     },
     {
-      "epoch": 1.9344141488577744,
-      "grad_norm": 5.553330595375883e-08,
-      "learning_rate": 2.581982313927782e-07,
-      "loss": 0.0828,
+      "epoch": 1.1820330969267139,
+      "grad_norm": 0.0003092484548687935,
+      "learning_rate": 3.5224586288416075e-07,
+      "loss": 0.0598,
       "step": 10500
     },
     {
-      "epoch": 1.9344141488577744,
-      "eval_loss": 0.5500391125679016,
-      "eval_runtime": 76.2612,
-      "eval_samples_per_second": 15.827,
-      "eval_steps_per_second": 1.98,
+      "epoch": 1.1820330969267139,
+      "eval_loss": 0.6480408906936646,
+      "eval_runtime": 125.7661,
+      "eval_samples_per_second": 15.696,
+      "eval_steps_per_second": 1.964,
       "step": 10500
     },
     {
-      "epoch": 2.026529108327192,
-      "grad_norm": 1.5537947319899104e-06,
-      "learning_rate": 2.466838614591009e-07,
-      "loss": 0.0573,
+      "epoch": 1.2383203872565574,
+      "grad_norm": 5.508670710696606e-07,
+      "learning_rate": 3.452099515929303e-07,
+      "loss": 0.0782,
       "step": 11000
     },
     {
-      "epoch": 2.026529108327192,
-      "eval_loss": 0.5341849327087402,
-      "eval_runtime": 75.6131,
-      "eval_samples_per_second": 15.963,
-      "eval_steps_per_second": 1.997,
+      "epoch": 1.2383203872565574,
+      "eval_loss": 0.6403237581253052,
+      "eval_runtime": 128.3947,
+      "eval_samples_per_second": 15.374,
+      "eval_steps_per_second": 1.924,
       "step": 11000
     },
     {
-      "epoch": 2.1186440677966103,
-      "grad_norm": 36.10622024536133,
-      "learning_rate": 2.3516949152542374e-07,
-      "loss": 0.0401,
+      "epoch": 1.294607677586401,
+      "grad_norm": 0.1470593810081482,
+      "learning_rate": 3.3817404030169986e-07,
+      "loss": 0.0616,
       "step": 11500
     },
     {
-      "epoch": 2.1186440677966103,
-      "eval_loss": 0.5351412296295166,
-      "eval_runtime": 74.8873,
-      "eval_samples_per_second": 16.118,
-      "eval_steps_per_second": 2.016,
+      "epoch": 1.294607677586401,
+      "eval_loss": 0.6376460790634155,
+      "eval_runtime": 125.5391,
+      "eval_samples_per_second": 15.724,
+      "eval_steps_per_second": 1.968,
       "step": 11500
     },
     {
-      "epoch": 2.210759027266028,
-      "grad_norm": 0.49580878019332886,
-      "learning_rate": 2.236551215917465e-07,
-      "loss": 0.0152,
+      "epoch": 1.3508949679162445,
+      "grad_norm": 1.8021532014245167e-05,
+      "learning_rate": 3.3113812901046946e-07,
+      "loss": 0.0723,
       "step": 12000
     },
     {
-      "epoch": 2.210759027266028,
-      "eval_loss": 0.5349487662315369,
-      "eval_runtime": 76.3109,
-      "eval_samples_per_second": 15.817,
-      "eval_steps_per_second": 1.979,
+      "epoch": 1.3508949679162445,
+      "eval_loss": 0.6378474831581116,
+      "eval_runtime": 125.5914,
+      "eval_samples_per_second": 15.718,
+      "eval_steps_per_second": 1.967,
       "step": 12000
     },
     {
-      "epoch": 2.3028739867354457,
-      "grad_norm": 1.7469781637191772,
-      "learning_rate": 2.1214075165806928e-07,
-      "loss": 0.0638,
+      "epoch": 1.407182258246088,
+      "grad_norm": 0.00013483635848388076,
+      "learning_rate": 3.2410221771923896e-07,
+      "loss": 0.0459,
       "step": 12500
     },
     {
-      "epoch": 2.3028739867354457,
-      "eval_loss": 0.531768798828125,
-      "eval_runtime": 75.5712,
-      "eval_samples_per_second": 15.972,
-      "eval_steps_per_second": 1.998,
+      "epoch": 1.407182258246088,
+      "eval_loss": 0.6378260850906372,
+      "eval_runtime": 128.2725,
+      "eval_samples_per_second": 15.389,
+      "eval_steps_per_second": 1.926,
       "step": 12500
     },
     {
-      "epoch": 2.3949889462048635,
-      "grad_norm": 0.0006845183088444173,
-      "learning_rate": 2.0062638172439202e-07,
-      "loss": 0.0488,
+      "epoch": 1.4634695485759315,
+      "grad_norm": 4.757418707868055e-07,
+      "learning_rate": 3.170663064280085e-07,
+      "loss": 0.0571,
       "step": 13000
     },
     {
-      "epoch": 2.3949889462048635,
-      "eval_loss": 0.530552089214325,
-      "eval_runtime": 75.6479,
-      "eval_samples_per_second": 15.955,
-      "eval_steps_per_second": 1.996,
+      "epoch": 1.4634695485759315,
+      "eval_loss": 0.6366899609565735,
+      "eval_runtime": 125.6986,
+      "eval_samples_per_second": 15.704,
+      "eval_steps_per_second": 1.965,
       "step": 13000
     },
     {
-      "epoch": 2.4871039056742816,
-      "grad_norm": 0.0002451306500006467,
-      "learning_rate": 1.891120117907148e-07,
-      "loss": 0.0456,
+      "epoch": 1.5197568389057752,
+      "grad_norm": 0.00028420978924259543,
+      "learning_rate": 3.1003039513677807e-07,
+      "loss": 0.0499,
       "step": 13500
     },
     {
-      "epoch": 2.4871039056742816,
-      "eval_loss": 0.5211473107337952,
-      "eval_runtime": 76.4401,
-      "eval_samples_per_second": 15.79,
-      "eval_steps_per_second": 1.975,
+      "epoch": 1.5197568389057752,
+      "eval_loss": 0.6286123394966125,
+      "eval_runtime": 127.9165,
+      "eval_samples_per_second": 15.432,
+      "eval_steps_per_second": 1.931,
       "step": 13500
     },
     {
-      "epoch": 2.5792188651436994,
-      "grad_norm": 0.003609905717894435,
-      "learning_rate": 1.7759764185703757e-07,
-      "loss": 0.0264,
+      "epoch": 1.5760441292356187,
+      "grad_norm": 2.1999912291903456e-07,
+      "learning_rate": 3.0299448384554767e-07,
+      "loss": 0.0679,
       "step": 14000
     },
     {
-      "epoch": 2.5792188651436994,
-      "eval_loss": 0.5193639993667603,
-      "eval_runtime": 74.595,
-      "eval_samples_per_second": 16.181,
-      "eval_steps_per_second": 2.024,
+      "epoch": 1.5760441292356187,
+      "eval_loss": 0.634914755821228,
+      "eval_runtime": 125.0396,
+      "eval_samples_per_second": 15.787,
+      "eval_steps_per_second": 1.975,
       "step": 14000
     },
     {
-      "epoch": 2.671333824613117,
-      "grad_norm": 6.751325276610487e-09,
-      "learning_rate": 1.6608327192336035e-07,
-      "loss": 0.0381,
+      "epoch": 1.6323314195654621,
+      "grad_norm": 0.00023505109129473567,
+      "learning_rate": 2.959585725543172e-07,
+      "loss": 0.0554,
       "step": 14500
     },
     {
-      "epoch": 2.671333824613117,
-      "eval_loss": 0.520633339881897,
-      "eval_runtime": 75.2937,
-      "eval_samples_per_second": 16.031,
-      "eval_steps_per_second": 2.005,
+      "epoch": 1.6323314195654621,
+      "eval_loss": 0.623322069644928,
+      "eval_runtime": 125.2429,
+      "eval_samples_per_second": 15.761,
+      "eval_steps_per_second": 1.972,
       "step": 14500
     },
     {
-      "epoch": 2.7634487840825352,
-      "grad_norm": 8.452875022157968e-07,
-      "learning_rate": 1.5456890198968311e-07,
-      "loss": 0.0413,
+      "epoch": 1.6886187098953056,
+      "grad_norm": 11.346132278442383,
+      "learning_rate": 2.889226612630868e-07,
+      "loss": 0.0617,
       "step": 15000
     },
     {
-      "epoch": 2.7634487840825352,
-      "eval_loss": 0.5167532563209534,
-      "eval_runtime": 75.615,
-      "eval_samples_per_second": 15.962,
-      "eval_steps_per_second": 1.997,
+      "epoch": 1.6886187098953056,
+      "eval_loss": 0.6115825772285461,
+      "eval_runtime": 127.8813,
+      "eval_samples_per_second": 15.436,
+      "eval_steps_per_second": 1.931,
       "step": 15000
     },
     {
-      "epoch": 2.855563743551953,
-      "grad_norm": 2.365756202049596e-11,
-      "learning_rate": 1.430545320560059e-07,
-      "loss": 0.0392,
+      "epoch": 1.744906000225149,
+      "grad_norm": 308.43804931640625,
+      "learning_rate": 2.8188674997185633e-07,
+      "loss": 0.0561,
       "step": 15500
     },
     {
-      "epoch": 2.855563743551953,
-      "eval_loss": 0.5148531198501587,
-      "eval_runtime": 74.6217,
-      "eval_samples_per_second": 16.175,
-      "eval_steps_per_second": 2.024,
+      "epoch": 1.744906000225149,
+      "eval_loss": 0.5951005220413208,
+      "eval_runtime": 125.2491,
+      "eval_samples_per_second": 15.761,
+      "eval_steps_per_second": 1.972,
       "step": 15500
     },
     {
-      "epoch": 2.9476787030213707,
-      "grad_norm": 0.32707780599594116,
-      "learning_rate": 1.3154016212232866e-07,
-      "loss": 0.0352,
+      "epoch": 1.8011932905549926,
+      "grad_norm": 6.91558055265773e-09,
+      "learning_rate": 2.7485083868062594e-07,
+      "loss": 0.0672,
       "step": 16000
     },
     {
-      "epoch": 2.9476787030213707,
-      "eval_loss": 0.5112494826316833,
-      "eval_runtime": 75.142,
-      "eval_samples_per_second": 16.063,
-      "eval_steps_per_second": 2.01,
+      "epoch": 1.8011932905549926,
+      "eval_loss": 0.5809568166732788,
+      "eval_runtime": 125.2557,
+      "eval_samples_per_second": 15.76,
+      "eval_steps_per_second": 1.972,
       "step": 16000
     },
     {
-      "epoch": 3.0397936624907884,
-      "grad_norm": 4.2568670810538833e-11,
-      "learning_rate": 1.2002579218865142e-07,
-      "loss": 0.0467,
+      "epoch": 1.8574805808848363,
+      "grad_norm": 2.2250075915053458e-07,
+      "learning_rate": 2.678149273893955e-07,
+      "loss": 0.0812,
       "step": 16500
     },
     {
-      "epoch": 3.0397936624907884,
-      "eval_loss": 0.5097618699073792,
-      "eval_runtime": 75.705,
-      "eval_samples_per_second": 15.943,
-      "eval_steps_per_second": 1.995,
+      "epoch": 1.8574805808848363,
+      "eval_loss": 0.5691242218017578,
+      "eval_runtime": 128.3032,
+      "eval_samples_per_second": 15.385,
+      "eval_steps_per_second": 1.925,
       "step": 16500
     },
     {
-      "epoch": 3.131908621960206,
-      "grad_norm": 7.157668679980134e-09,
-      "learning_rate": 1.085114222549742e-07,
-      "loss": 0.0366,
+      "epoch": 1.9137678712146797,
+      "grad_norm": 1.1724663972854614,
+      "learning_rate": 2.6077901609816504e-07,
+      "loss": 0.0614,
       "step": 17000
     },
     {
-      "epoch": 3.131908621960206,
-      "eval_loss": 0.5088639855384827,
-      "eval_runtime": 74.5863,
-      "eval_samples_per_second": 16.183,
-      "eval_steps_per_second": 2.025,
+      "epoch": 1.9137678712146797,
+      "eval_loss": 0.5643230080604553,
+      "eval_runtime": 125.7191,
+      "eval_samples_per_second": 15.702,
+      "eval_steps_per_second": 1.965,
       "step": 17000
     },
     {
-      "epoch": 3.2240235814296243,
-      "grad_norm": 0.5768880248069763,
-      "learning_rate": 9.699705232129697e-08,
-      "loss": 0.0454,
+      "epoch": 1.9700551615445232,
+      "grad_norm": 0.00030226208036765456,
+      "learning_rate": 2.5374310480693454e-07,
+      "loss": 0.0462,
       "step": 17500
     },
     {
-      "epoch": 3.2240235814296243,
-      "eval_loss": 0.5103679299354553,
-      "eval_runtime": 74.9953,
-      "eval_samples_per_second": 16.094,
-      "eval_steps_per_second": 2.013,
+      "epoch": 1.9700551615445232,
+      "eval_loss": 0.5657238364219666,
+      "eval_runtime": 124.9923,
+      "eval_samples_per_second": 15.793,
+      "eval_steps_per_second": 1.976,
       "step": 17500
     },
     {
-      "epoch": 3.316138540899042,
-      "grad_norm": 0.029806463047862053,
-      "learning_rate": 8.548268238761974e-08,
-      "loss": 0.0209,
+      "epoch": 2.026342451874367,
+      "grad_norm": 1.949754704355655e-07,
+      "learning_rate": 2.4670719351570415e-07,
+      "loss": 0.0565,
       "step": 18000
     },
     {
-      "epoch": 3.316138540899042,
-      "eval_loss": 0.5070606470108032,
-      "eval_runtime": 75.7962,
-      "eval_samples_per_second": 15.924,
-      "eval_steps_per_second": 1.992,
+      "epoch": 2.026342451874367,
+      "eval_loss": 0.5607953667640686,
+      "eval_runtime": 127.7187,
+      "eval_samples_per_second": 15.456,
+      "eval_steps_per_second": 1.934,
       "step": 18000
     },
     {
-      "epoch": 3.4082535003684598,
-      "grad_norm": 0.09116169810295105,
-      "learning_rate": 7.396831245394252e-08,
-      "loss": 0.0636,
+      "epoch": 2.0826297422042104,
+      "grad_norm": 14.2212553024292,
+      "learning_rate": 2.396712822244737e-07,
+      "loss": 0.0549,
       "step": 18500
     },
     {
-      "epoch": 3.4082535003684598,
-      "eval_loss": 0.5044638514518738,
-      "eval_runtime": 74.7518,
-      "eval_samples_per_second": 16.147,
-      "eval_steps_per_second": 2.02,
+      "epoch": 2.0826297422042104,
+      "eval_loss": 0.5568962693214417,
+      "eval_runtime": 125.0503,
+      "eval_samples_per_second": 15.786,
+      "eval_steps_per_second": 1.975,
       "step": 18500
     },
     {
-      "epoch": 3.500368459837878,
-      "grad_norm": 0.13625642657279968,
-      "learning_rate": 6.245394252026529e-08,
-      "loss": 0.0159,
+      "epoch": 2.138917032534054,
+      "grad_norm": 1.2045973539352417,
+      "learning_rate": 2.3263537093324326e-07,
+      "loss": 0.0245,
       "step": 19000
     },
     {
-      "epoch": 3.500368459837878,
-      "eval_loss": 0.5019211769104004,
-      "eval_runtime": 75.0004,
-      "eval_samples_per_second": 16.093,
-      "eval_steps_per_second": 2.013,
+      "epoch": 2.138917032534054,
+      "eval_loss": 0.5595241785049438,
+      "eval_runtime": 124.8527,
+      "eval_samples_per_second": 15.811,
+      "eval_steps_per_second": 1.978,
       "step": 19000
     },
     {
-      "epoch": 3.5924834193072956,
-      "grad_norm": 2.3882080313342158e-06,
-      "learning_rate": 5.0939572586588065e-08,
-      "loss": 0.0303,
+      "epoch": 2.1952043228638973,
+      "grad_norm": 8.770056410867255e-06,
+      "learning_rate": 2.2559945964201284e-07,
+      "loss": 0.0428,
       "step": 19500
     },
     {
-      "epoch": 3.5924834193072956,
-      "eval_loss": 0.49851301312446594,
-      "eval_runtime": 75.7613,
-      "eval_samples_per_second": 15.932,
-      "eval_steps_per_second": 1.993,
+      "epoch": 2.1952043228638973,
+      "eval_loss": 0.5740877985954285,
+      "eval_runtime": 127.9211,
+      "eval_samples_per_second": 15.431,
+      "eval_steps_per_second": 1.931,
       "step": 19500
     },
     {
-      "epoch": 3.6845983787767134,
-      "grad_norm": 8.883699109674126e-09,
-      "learning_rate": 3.942520265291083e-08,
-      "loss": 0.0353,
+      "epoch": 2.251491613193741,
+      "grad_norm": 6.243853079013206e-08,
+      "learning_rate": 2.185635483507824e-07,
+      "loss": 0.0301,
       "step": 20000
     },
     {
-      "epoch": 3.6845983787767134,
-      "eval_loss": 0.49747875332832336,
-      "eval_runtime": 74.5978,
-      "eval_samples_per_second": 16.18,
-      "eval_steps_per_second": 2.024,
+      "epoch": 2.251491613193741,
+      "eval_loss": 0.5761672854423523,
+      "eval_runtime": 125.2829,
+      "eval_samples_per_second": 15.756,
+      "eval_steps_per_second": 1.972,
       "step": 20000
     },
     {
-      "epoch": 3.776713338246131,
-      "grad_norm": 0.12466106563806534,
-      "learning_rate": 2.7910832719233602e-08,
-      "loss": 0.0261,
+      "epoch": 2.3077789035235843,
+      "grad_norm": 2.299124268745345e-08,
+      "learning_rate": 2.1152763705955197e-07,
+      "loss": 0.04,
       "step": 20500
     },
     {
-      "epoch": 3.776713338246131,
-      "eval_loss": 0.4962019622325897,
-      "eval_runtime": 75.2409,
-      "eval_samples_per_second": 16.042,
-      "eval_steps_per_second": 2.007,
+      "epoch": 2.3077789035235843,
+      "eval_loss": 0.5712103843688965,
+      "eval_runtime": 125.3797,
+      "eval_samples_per_second": 15.744,
+      "eval_steps_per_second": 1.97,
       "step": 20500
     },
     {
-      "epoch": 3.868828297715549,
-      "grad_norm": 1.0646399317693067e-07,
-      "learning_rate": 1.6396462785556373e-08,
-      "loss": 0.0291,
+      "epoch": 2.3640661938534278,
+      "grad_norm": 1.971009623957798e-06,
+      "learning_rate": 2.044917257683215e-07,
+      "loss": 0.0349,
       "step": 21000
     },
     {
-      "epoch": 3.868828297715549,
-      "eval_loss": 0.49561700224876404,
-      "eval_runtime": 75.8041,
-      "eval_samples_per_second": 15.923,
-      "eval_steps_per_second": 1.992,
+      "epoch": 2.3640661938534278,
+      "eval_loss": 0.5690667629241943,
+      "eval_runtime": 125.4733,
+      "eval_samples_per_second": 15.732,
+      "eval_steps_per_second": 1.969,
       "step": 21000
     }
   ],
   "logging_steps": 500,
-  "max_steps": 21712,
+  "max_steps": 35532,
   "num_input_tokens_seen": 0,
   "num_train_epochs": 4,
   "save_steps": 500,
@@ -656,7 +656,7 @@
       "attributes": {}
     }
   },
-  "total_flos": 7555644893162520.0,
+  "total_flos": 7555285134811260.0,
   "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:bf76416e4a857628175385db7f617e54451cd7ffb62c1f9590ee9f30268f8c5b
+oid sha256:319be2a769ef7117b6826908976f080f9082f159f4f0ad75da1a151ad29d9bff
 size 4847