Training in progress, step 21000, checkpoint

checkpoint-21000/model.safetensors

@@ -1,3 +1,3 @@
 version https://git-lfs.github.com/spec/v1
-oid sha256:7858eebeb670a9ec334c334aa9bc793727b2021612fbcf88d7bd7aab517be05b
+oid sha256:79ae9b8c94942647fcee697da92825abb00613da96dd7c4d07ba61c3f4b51832
 size 1711848436

checkpoint-21000/optimizer.pt

@@ -1,3 +1,3 @@
 version https://git-lfs.github.com/spec/v1
-oid sha256:f3bbdb31e723e55da317ed3ae59750b882278d9fce407ffd858b972681c4c362
+oid sha256:4d1344a9063e4e273b35484041e5561a572f19dacb6d5e3d5272364fdc6143de
 size 3424043887

checkpoint-21000/trainer_state.json

@@ -1,5 +1,5 @@
 {
-  "best_metric": 0.7744118571281433,
+  "best_metric": 0.8657393455505371,
   "best_model_checkpoint": "./output/clip-finetuned-csu-p14-336-e4l58-l/checkpoint-21000",
   "epoch": 3.868828297715549,
   "eval_steps": 500,
@@ -9,633 +9,633 @@
   "is_world_process_zero": true,
   "log_history": [
     {
-      "epoch": 0.09209799226376865,
-      "grad_norm": 0.0014697719598188996,
-      "learning_rate": 4.884877509670289e-08,
-      "loss": 0.4588,
+      "epoch": 0.09211495946941783,
+      "grad_norm": 20.04043960571289,
+      "learning_rate": 4.8848563006632274e-08,
+      "loss": 0.3758,
       "step": 500
     },
     {
-      "epoch": 0.09209799226376865,
-      "eval_loss": 1.4194883108139038,
-      "eval_runtime": 78.7703,
-      "eval_samples_per_second": 15.31,
-      "eval_steps_per_second": 1.917,
+      "epoch": 0.09211495946941783,
+      "eval_loss": 1.4184927940368652,
+      "eval_runtime": 74.3562,
+      "eval_samples_per_second": 16.233,
+      "eval_steps_per_second": 2.031,
       "step": 500
     },
     {
-      "epoch": 0.1841959845275373,
-      "grad_norm": 301.76043701171875,
-      "learning_rate": 4.769755019340578e-08,
-      "loss": 0.4255,
+      "epoch": 0.18422991893883567,
+      "grad_norm": 74.25025939941406,
+      "learning_rate": 4.7697126013264556e-08,
+      "loss": 0.4103,
       "step": 1000
     },
     {
-      "epoch": 0.1841959845275373,
-      "eval_loss": 1.3416553735733032,
-      "eval_runtime": 77.0051,
-      "eval_samples_per_second": 15.661,
-      "eval_steps_per_second": 1.961,
+      "epoch": 0.18422991893883567,
+      "eval_loss": 1.3500770330429077,
+      "eval_runtime": 75.7013,
+      "eval_samples_per_second": 15.944,
+      "eval_steps_per_second": 1.995,
       "step": 1000
     },
     {
-      "epoch": 0.27629397679130596,
-      "grad_norm": 35.28853225708008,
-      "learning_rate": 4.654632529010867e-08,
-      "loss": 0.3724,
+      "epoch": 0.2763448784082535,
+      "grad_norm": 0.5102410912513733,
+      "learning_rate": 4.6545689019896826e-08,
+      "loss": 0.433,
       "step": 1500
     },
     {
-      "epoch": 0.27629397679130596,
-      "eval_loss": 1.2872973680496216,
-      "eval_runtime": 77.5439,
-      "eval_samples_per_second": 15.552,
-      "eval_steps_per_second": 1.947,
+      "epoch": 0.2763448784082535,
+      "eval_loss": 1.2885024547576904,
+      "eval_runtime": 75.197,
+      "eval_samples_per_second": 16.051,
+      "eval_steps_per_second": 2.008,
       "step": 1500
     },
     {
-      "epoch": 0.3683919690550746,
-      "grad_norm": 338.9349060058594,
-      "learning_rate": 4.539510038681156e-08,
-      "loss": 0.3251,
+      "epoch": 0.36845983787767134,
+      "grad_norm": 0.1667678952217102,
+      "learning_rate": 4.539425202652911e-08,
+      "loss": 0.3424,
       "step": 2000
     },
     {
-      "epoch": 0.3683919690550746,
-      "eval_loss": 1.234910249710083,
-      "eval_runtime": 78.7715,
-      "eval_samples_per_second": 15.31,
-      "eval_steps_per_second": 1.917,
+      "epoch": 0.36845983787767134,
+      "eval_loss": 1.239119052886963,
+      "eval_runtime": 75.387,
+      "eval_samples_per_second": 16.011,
+      "eval_steps_per_second": 2.003,
       "step": 2000
     },
     {
-      "epoch": 0.46048996131884323,
-      "grad_norm": 325.85101318359375,
-      "learning_rate": 4.4243875483514457e-08,
-      "loss": 0.3308,
+      "epoch": 0.46057479734708917,
+      "grad_norm": 343.8175048828125,
+      "learning_rate": 4.4242815033161385e-08,
+      "loss": 0.3645,
       "step": 2500
     },
     {
-      "epoch": 0.46048996131884323,
-      "eval_loss": 1.1944907903671265,
-      "eval_runtime": 79.0397,
-      "eval_samples_per_second": 15.258,
-      "eval_steps_per_second": 1.91,
+      "epoch": 0.46057479734708917,
+      "eval_loss": 1.1902339458465576,
+      "eval_runtime": 75.9899,
+      "eval_samples_per_second": 15.884,
+      "eval_steps_per_second": 1.987,
       "step": 2500
     },
     {
-      "epoch": 0.5525879535826119,
-      "grad_norm": 100.85855102539062,
-      "learning_rate": 4.309265058021735e-08,
-      "loss": 0.3017,
+      "epoch": 0.552689756816507,
+      "grad_norm": 305.5626525878906,
+      "learning_rate": 4.309137803979366e-08,
+      "loss": 0.3172,
       "step": 3000
     },
     {
-      "epoch": 0.5525879535826119,
-      "eval_loss": 1.1593009233474731,
-      "eval_runtime": 77.6354,
-      "eval_samples_per_second": 15.534,
-      "eval_steps_per_second": 1.945,
+      "epoch": 0.552689756816507,
+      "eval_loss": 1.1506118774414062,
+      "eval_runtime": 74.8655,
+      "eval_samples_per_second": 16.122,
+      "eval_steps_per_second": 2.017,
       "step": 3000
     },
     {
-      "epoch": 0.6446859458463805,
-      "grad_norm": 8.865598678588867,
-      "learning_rate": 4.194142567692024e-08,
-      "loss": 0.2962,
+      "epoch": 0.6448047162859248,
+      "grad_norm": 65.59048461914062,
+      "learning_rate": 4.193994104642594e-08,
+      "loss": 0.2751,
       "step": 3500
     },
     {
-      "epoch": 0.6446859458463805,
-      "eval_loss": 1.1259427070617676,
-      "eval_runtime": 78.2715,
-      "eval_samples_per_second": 15.408,
-      "eval_steps_per_second": 1.929,
+      "epoch": 0.6448047162859248,
+      "eval_loss": 1.1169357299804688,
+      "eval_runtime": 76.2936,
+      "eval_samples_per_second": 15.82,
+      "eval_steps_per_second": 1.979,
       "step": 3500
     },
     {
-      "epoch": 0.7367839381101492,
-      "grad_norm": 4.393447399139404,
-      "learning_rate": 4.079020077362314e-08,
+      "epoch": 0.7369196757553427,
+      "grad_norm": 18.106201171875,
+      "learning_rate": 4.078850405305821e-08,
       "loss": 0.2919,
       "step": 4000
     },
     {
-      "epoch": 0.7367839381101492,
-      "eval_loss": 1.0954023599624634,
-      "eval_runtime": 78.5366,
-      "eval_samples_per_second": 15.356,
-      "eval_steps_per_second": 1.923,
+      "epoch": 0.7369196757553427,
+      "eval_loss": 1.0920671224594116,
+      "eval_runtime": 74.9101,
+      "eval_samples_per_second": 16.113,
+      "eval_steps_per_second": 2.016,
       "step": 4000
     },
     {
-      "epoch": 0.8288819303739179,
-      "grad_norm": 1.6036031246185303,
-      "learning_rate": 3.9638975870326026e-08,
-      "loss": 0.307,
+      "epoch": 0.8290346352247605,
+      "grad_norm": 506.40570068359375,
+      "learning_rate": 3.9637067059690496e-08,
+      "loss": 0.2583,
       "step": 4500
     },
     {
-      "epoch": 0.8288819303739179,
-      "eval_loss": 1.0728861093521118,
-      "eval_runtime": 78.4032,
-      "eval_samples_per_second": 15.382,
-      "eval_steps_per_second": 1.926,
+      "epoch": 0.8290346352247605,
+      "eval_loss": 1.0721209049224854,
+      "eval_runtime": 75.8053,
+      "eval_samples_per_second": 15.922,
+      "eval_steps_per_second": 1.992,
       "step": 4500
     },
     {
-      "epoch": 0.9209799226376865,
-      "grad_norm": 11.990758895874023,
-      "learning_rate": 3.8487750967028915e-08,
-      "loss": 0.2764,
+      "epoch": 0.9211495946941783,
+      "grad_norm": 1.1508910655975342,
+      "learning_rate": 3.848563006632277e-08,
+      "loss": 0.2679,
       "step": 5000
     },
     {
-      "epoch": 0.9209799226376865,
-      "eval_loss": 1.0523593425750732,
-      "eval_runtime": 78.3726,
-      "eval_samples_per_second": 15.388,
-      "eval_steps_per_second": 1.927,
+      "epoch": 0.9211495946941783,
+      "eval_loss": 1.0519349575042725,
+      "eval_runtime": 75.226,
+      "eval_samples_per_second": 16.045,
+      "eval_steps_per_second": 2.007,
       "step": 5000
     },
     {
-      "epoch": 1.013077914901455,
-      "grad_norm": 43.865562438964844,
-      "learning_rate": 3.733652606373181e-08,
-      "loss": 0.2456,
+      "epoch": 1.013264554163596,
+      "grad_norm": 0.1393290013074875,
+      "learning_rate": 3.733419307295505e-08,
+      "loss": 0.2472,
       "step": 5500
     },
     {
-      "epoch": 1.013077914901455,
-      "eval_loss": 1.0375442504882812,
-      "eval_runtime": 78.2473,
-      "eval_samples_per_second": 15.413,
-      "eval_steps_per_second": 1.93,
+      "epoch": 1.013264554163596,
+      "eval_loss": 1.0355629920959473,
+      "eval_runtime": 74.6535,
+      "eval_samples_per_second": 16.168,
+      "eval_steps_per_second": 2.023,
       "step": 5500
     },
     {
-      "epoch": 1.1051759071652238,
-      "grad_norm": 105.91036987304688,
-      "learning_rate": 3.61853011604347e-08,
-      "loss": 0.2642,
+      "epoch": 1.105379513633014,
+      "grad_norm": 200.34664916992188,
+      "learning_rate": 3.6182756079587324e-08,
+      "loss": 0.26,
       "step": 6000
     },
     {
-      "epoch": 1.1051759071652238,
-      "eval_loss": 1.0233356952667236,
-      "eval_runtime": 78.3135,
-      "eval_samples_per_second": 15.4,
-      "eval_steps_per_second": 1.928,
+      "epoch": 1.105379513633014,
+      "eval_loss": 1.0176944732666016,
+      "eval_runtime": 74.4374,
+      "eval_samples_per_second": 16.215,
+      "eval_steps_per_second": 2.029,
       "step": 6000
     },
     {
-      "epoch": 1.1972738994289924,
-      "grad_norm": 221.4896240234375,
-      "learning_rate": 3.503407625713759e-08,
-      "loss": 0.2066,
+      "epoch": 1.1974944731024317,
+      "grad_norm": 431.1804504394531,
+      "learning_rate": 3.50313190862196e-08,
+      "loss": 0.2153,
       "step": 6500
     },
     {
-      "epoch": 1.1972738994289924,
-      "eval_loss": 1.0104238986968994,
-      "eval_runtime": 78.6485,
-      "eval_samples_per_second": 15.334,
-      "eval_steps_per_second": 1.92,
+      "epoch": 1.1974944731024317,
+      "eval_loss": 1.0044586658477783,
+      "eval_runtime": 74.3247,
+      "eval_samples_per_second": 16.24,
+      "eval_steps_per_second": 2.032,
       "step": 6500
     },
     {
-      "epoch": 1.289371891692761,
-      "grad_norm": 270.53253173828125,
-      "learning_rate": 3.3882851353840485e-08,
-      "loss": 0.2376,
+      "epoch": 1.2896094325718497,
+      "grad_norm": 0.6204941272735596,
+      "learning_rate": 3.387988209285188e-08,
+      "loss": 0.1791,
       "step": 7000
     },
     {
-      "epoch": 1.289371891692761,
-      "eval_loss": 0.998353898525238,
-      "eval_runtime": 78.5502,
-      "eval_samples_per_second": 15.353,
-      "eval_steps_per_second": 1.922,
+      "epoch": 1.2896094325718497,
+      "eval_loss": 0.9926707148551941,
+      "eval_runtime": 75.7317,
+      "eval_samples_per_second": 15.938,
+      "eval_steps_per_second": 1.994,
       "step": 7000
     },
     {
-      "epoch": 1.3814698839565298,
-      "grad_norm": 402.9339904785156,
-      "learning_rate": 3.2731626450543374e-08,
-      "loss": 0.1931,
+      "epoch": 1.3817243920412676,
+      "grad_norm": 52.27273178100586,
+      "learning_rate": 3.272844509948415e-08,
+      "loss": 0.2082,
       "step": 7500
     },
     {
-      "epoch": 1.3814698839565298,
-      "eval_loss": 0.9887453317642212,
-      "eval_runtime": 78.4833,
-      "eval_samples_per_second": 15.366,
-      "eval_steps_per_second": 1.924,
+      "epoch": 1.3817243920412676,
+      "eval_loss": 0.9803809523582458,
+      "eval_runtime": 74.9936,
+      "eval_samples_per_second": 16.095,
+      "eval_steps_per_second": 2.014,
       "step": 7500
     },
     {
-      "epoch": 1.4735678762202984,
-      "grad_norm": 0.006355441175401211,
-      "learning_rate": 3.158040154724626e-08,
-      "loss": 0.2163,
+      "epoch": 1.4738393515106853,
+      "grad_norm": 0.0976635068655014,
+      "learning_rate": 3.1577008106116435e-08,
+      "loss": 0.196,
       "step": 8000
     },
     {
-      "epoch": 1.4735678762202984,
-      "eval_loss": 0.9767189621925354,
-      "eval_runtime": 78.5499,
-      "eval_samples_per_second": 15.353,
-      "eval_steps_per_second": 1.922,
+      "epoch": 1.4738393515106853,
+      "eval_loss": 0.9711655378341675,
+      "eval_runtime": 74.3537,
+      "eval_samples_per_second": 16.233,
+      "eval_steps_per_second": 2.031,
       "step": 8000
     },
     {
-      "epoch": 1.565665868484067,
-      "grad_norm": 0.12982100248336792,
-      "learning_rate": 3.042917664394916e-08,
-      "loss": 0.1903,
+      "epoch": 1.565954310980103,
+      "grad_norm": 46.64162826538086,
+      "learning_rate": 3.042557111274871e-08,
+      "loss": 0.1946,
       "step": 8500
     },
     {
-      "epoch": 1.565665868484067,
-      "eval_loss": 0.9664921760559082,
-      "eval_runtime": 78.6832,
-      "eval_samples_per_second": 15.327,
-      "eval_steps_per_second": 1.919,
+      "epoch": 1.565954310980103,
+      "eval_loss": 0.9621181488037109,
+      "eval_runtime": 75.7205,
+      "eval_samples_per_second": 15.94,
+      "eval_steps_per_second": 1.994,
       "step": 8500
     },
     {
-      "epoch": 1.6577638607478358,
-      "grad_norm": 173.31788635253906,
-      "learning_rate": 2.927795174065205e-08,
-      "loss": 0.2069,
+      "epoch": 1.658069270449521,
+      "grad_norm": 0.05690698325634003,
+      "learning_rate": 2.927413411938099e-08,
+      "loss": 0.2422,
       "step": 9000
     },
     {
-      "epoch": 1.6577638607478358,
-      "eval_loss": 0.9571623206138611,
-      "eval_runtime": 78.544,
-      "eval_samples_per_second": 15.354,
-      "eval_steps_per_second": 1.922,
+      "epoch": 1.658069270449521,
+      "eval_loss": 0.9536784291267395,
+      "eval_runtime": 75.1745,
+      "eval_samples_per_second": 16.056,
+      "eval_steps_per_second": 2.009,
       "step": 9000
     },
     {
-      "epoch": 1.7498618530116044,
-      "grad_norm": 358.7403259277344,
-      "learning_rate": 2.8126726837354947e-08,
-      "loss": 0.2093,
+      "epoch": 1.750184229918939,
+      "grad_norm": 0.08470064401626587,
+      "learning_rate": 2.8122697126013263e-08,
+      "loss": 0.2106,
       "step": 9500
     },
     {
-      "epoch": 1.7498618530116044,
-      "eval_loss": 0.9496769905090332,
-      "eval_runtime": 78.3124,
-      "eval_samples_per_second": 15.4,
-      "eval_steps_per_second": 1.928,
+      "epoch": 1.750184229918939,
+      "eval_loss": 0.9457550644874573,
+      "eval_runtime": 74.5062,
+      "eval_samples_per_second": 16.2,
+      "eval_steps_per_second": 2.027,
       "step": 9500
     },
     {
-      "epoch": 1.841959845275373,
-      "grad_norm": 435.9449768066406,
-      "learning_rate": 2.6975501934057836e-08,
-      "loss": 0.2523,
+      "epoch": 1.8422991893883567,
+      "grad_norm": 1.5745967626571655,
+      "learning_rate": 2.697126013264554e-08,
+      "loss": 0.1801,
       "step": 10000
     },
     {
-      "epoch": 1.841959845275373,
-      "eval_loss": 0.9420022368431091,
-      "eval_runtime": 78.3702,
-      "eval_samples_per_second": 15.388,
-      "eval_steps_per_second": 1.927,
+      "epoch": 1.8422991893883567,
+      "eval_loss": 0.9393072724342346,
+      "eval_runtime": 75.9432,
+      "eval_samples_per_second": 15.893,
+      "eval_steps_per_second": 1.988,
       "step": 10000
     },
     {
-      "epoch": 1.9340578375391417,
-      "grad_norm": 23.49118423461914,
-      "learning_rate": 2.582427703076073e-08,
-      "loss": 0.2127,
+      "epoch": 1.9344141488577744,
+      "grad_norm": 0.06661002337932587,
+      "learning_rate": 2.5819823139277818e-08,
+      "loss": 0.2117,
       "step": 10500
     },
     {
-      "epoch": 1.9340578375391417,
-      "eval_loss": 0.9328628182411194,
-      "eval_runtime": 76.6904,
-      "eval_samples_per_second": 15.726,
-      "eval_steps_per_second": 1.969,
+      "epoch": 1.9344141488577744,
+      "eval_loss": 0.9308408498764038,
+      "eval_runtime": 75.0811,
+      "eval_samples_per_second": 16.076,
+      "eval_steps_per_second": 2.011,
       "step": 10500
     },
     {
-      "epoch": 2.02615582980291,
-      "grad_norm": 0.4029064476490021,
-      "learning_rate": 2.467305212746362e-08,
-      "loss": 0.1968,
+      "epoch": 2.026529108327192,
+      "grad_norm": 0.236691415309906,
+      "learning_rate": 2.4668386145910094e-08,
+      "loss": 0.2061,
       "step": 11000
     },
     {
-      "epoch": 2.02615582980291,
-      "eval_loss": 0.926975429058075,
-      "eval_runtime": 76.7161,
-      "eval_samples_per_second": 15.72,
-      "eval_steps_per_second": 1.968,
+      "epoch": 2.026529108327192,
+      "eval_loss": 0.9237484335899353,
+      "eval_runtime": 74.6134,
+      "eval_samples_per_second": 16.177,
+      "eval_steps_per_second": 2.024,
       "step": 11000
     },
     {
-      "epoch": 2.118253822066679,
-      "grad_norm": 274.0943603515625,
-      "learning_rate": 2.3521827224166513e-08,
-      "loss": 0.1879,
+      "epoch": 2.1186440677966103,
+      "grad_norm": 8.831416130065918,
+      "learning_rate": 2.3516949152542374e-08,
+      "loss": 0.1878,
       "step": 11500
     },
     {
-      "epoch": 2.118253822066679,
-      "eval_loss": 0.9231382012367249,
-      "eval_runtime": 78.6224,
-      "eval_samples_per_second": 15.339,
-      "eval_steps_per_second": 1.921,
+      "epoch": 2.1186440677966103,
+      "eval_loss": 0.9166750311851501,
+      "eval_runtime": 75.6886,
+      "eval_samples_per_second": 15.947,
+      "eval_steps_per_second": 1.995,
       "step": 11500
     },
     {
-      "epoch": 2.2103518143304477,
-      "grad_norm": 0.000765918695833534,
-      "learning_rate": 2.2370602320869402e-08,
-      "loss": 0.1981,
+      "epoch": 2.210759027266028,
+      "grad_norm": 0.787421464920044,
+      "learning_rate": 2.236551215917465e-08,
+      "loss": 0.1655,
       "step": 12000
     },
     {
-      "epoch": 2.2103518143304477,
-      "eval_loss": 0.9183884263038635,
-      "eval_runtime": 76.7707,
-      "eval_samples_per_second": 15.709,
-      "eval_steps_per_second": 1.967,
+      "epoch": 2.210759027266028,
+      "eval_loss": 0.9109458327293396,
+      "eval_runtime": 74.9866,
+      "eval_samples_per_second": 16.096,
+      "eval_steps_per_second": 2.014,
       "step": 12000
     },
     {
-      "epoch": 2.302449806594216,
-      "grad_norm": 0.6250036358833313,
-      "learning_rate": 2.1219377417572295e-08,
-      "loss": 0.1964,
+      "epoch": 2.3028739867354457,
+      "grad_norm": 535.4310913085938,
+      "learning_rate": 2.1214075165806926e-08,
+      "loss": 0.1946,
       "step": 12500
     },
     {
-      "epoch": 2.302449806594216,
-      "eval_loss": 0.9134928584098816,
-      "eval_runtime": 78.5128,
-      "eval_samples_per_second": 15.361,
-      "eval_steps_per_second": 1.923,
+      "epoch": 2.3028739867354457,
+      "eval_loss": 0.9070548415184021,
+      "eval_runtime": 74.3796,
+      "eval_samples_per_second": 16.228,
+      "eval_steps_per_second": 2.03,
       "step": 12500
     },
     {
-      "epoch": 2.394547798857985,
-      "grad_norm": 158.6990203857422,
-      "learning_rate": 2.0068152514275187e-08,
-      "loss": 0.1697,
+      "epoch": 2.3949889462048635,
+      "grad_norm": 0.8420014381408691,
+      "learning_rate": 2.0062638172439202e-08,
+      "loss": 0.1882,
       "step": 13000
     },
     {
-      "epoch": 2.394547798857985,
-      "eval_loss": 0.910048246383667,
-      "eval_runtime": 76.7476,
-      "eval_samples_per_second": 15.714,
-      "eval_steps_per_second": 1.967,
+      "epoch": 2.3949889462048635,
+      "eval_loss": 0.9021016359329224,
+      "eval_runtime": 75.8812,
+      "eval_samples_per_second": 15.906,
+      "eval_steps_per_second": 1.99,
       "step": 13000
     },
     {
-      "epoch": 2.4866457911217537,
-      "grad_norm": 1.288072943687439,
-      "learning_rate": 1.891692761097808e-08,
-      "loss": 0.2015,
+      "epoch": 2.4871039056742816,
+      "grad_norm": 1.4041686058044434,
+      "learning_rate": 1.891120117907148e-08,
+      "loss": 0.1871,
       "step": 13500
     },
     {
-      "epoch": 2.4866457911217537,
-      "eval_loss": 0.9051916599273682,
-      "eval_runtime": 78.3909,
-      "eval_samples_per_second": 15.384,
-      "eval_steps_per_second": 1.926,
+      "epoch": 2.4871039056742816,
+      "eval_loss": 0.8960274457931519,
+      "eval_runtime": 75.2283,
+      "eval_samples_per_second": 16.044,
+      "eval_steps_per_second": 2.007,
       "step": 13500
     },
     {
-      "epoch": 2.578743783385522,
-      "grad_norm": 5.835799311171286e-05,
-      "learning_rate": 1.7765702707680972e-08,
-      "loss": 0.1827,
+      "epoch": 2.5792188651436994,
+      "grad_norm": 327.2938537597656,
+      "learning_rate": 1.7759764185703757e-08,
+      "loss": 0.1419,
       "step": 14000
     },
     {
-      "epoch": 2.578743783385522,
-      "eval_loss": 0.9026296734809875,
-      "eval_runtime": 76.6591,
-      "eval_samples_per_second": 15.732,
-      "eval_steps_per_second": 1.97,
+      "epoch": 2.5792188651436994,
+      "eval_loss": 0.8913142085075378,
+      "eval_runtime": 74.4503,
+      "eval_samples_per_second": 16.212,
+      "eval_steps_per_second": 2.028,
       "step": 14000
     },
     {
-      "epoch": 2.670841775649291,
-      "grad_norm": 30.217784881591797,
-      "learning_rate": 1.6614477804383865e-08,
-      "loss": 0.1435,
+      "epoch": 2.671333824613117,
+      "grad_norm": 0.01779676228761673,
+      "learning_rate": 1.6608327192336037e-08,
+      "loss": 0.1431,
       "step": 14500
     },
     {
-      "epoch": 2.670841775649291,
-      "eval_loss": 0.899847686290741,
-      "eval_runtime": 76.742,
-      "eval_samples_per_second": 15.715,
-      "eval_steps_per_second": 1.968,
+      "epoch": 2.671333824613117,
+      "eval_loss": 0.8879275321960449,
+      "eval_runtime": 75.8636,
+      "eval_samples_per_second": 15.91,
+      "eval_steps_per_second": 1.99,
       "step": 14500
     },
     {
-      "epoch": 2.7629397679130596,
-      "grad_norm": 202.30496215820312,
-      "learning_rate": 1.5463252901086757e-08,
-      "loss": 0.1541,
+      "epoch": 2.7634487840825352,
+      "grad_norm": 1.6428073644638062,
+      "learning_rate": 1.5456890198968313e-08,
+      "loss": 0.1811,
       "step": 15000
     },
     {
-      "epoch": 2.7629397679130596,
-      "eval_loss": 0.896297812461853,
-      "eval_runtime": 78.5539,
-      "eval_samples_per_second": 15.353,
-      "eval_steps_per_second": 1.922,
+      "epoch": 2.7634487840825352,
+      "eval_loss": 0.8848333954811096,
+      "eval_runtime": 75.1911,
+      "eval_samples_per_second": 16.052,
+      "eval_steps_per_second": 2.008,
       "step": 15000
     },
     {
-      "epoch": 2.8550377601768284,
-      "grad_norm": 241.57069396972656,
-      "learning_rate": 1.4312027997789646e-08,
-      "loss": 0.1716,
+      "epoch": 2.855563743551953,
+      "grad_norm": 0.0006026208866387606,
+      "learning_rate": 1.430545320560059e-08,
+      "loss": 0.1694,
       "step": 15500
     },
     {
-      "epoch": 2.8550377601768284,
-      "eval_loss": 0.8935310244560242,
-      "eval_runtime": 76.9195,
-      "eval_samples_per_second": 15.679,
-      "eval_steps_per_second": 1.963,
+      "epoch": 2.855563743551953,
+      "eval_loss": 0.8826532959938049,
+      "eval_runtime": 74.6044,
+      "eval_samples_per_second": 16.179,
+      "eval_steps_per_second": 2.024,
       "step": 15500
     },
     {
-      "epoch": 2.947135752440597,
-      "grad_norm": 0.25739413499832153,
-      "learning_rate": 1.316080309449254e-08,
-      "loss": 0.2056,
+      "epoch": 2.9476787030213707,
+      "grad_norm": 226.09982299804688,
+      "learning_rate": 1.3154016212232866e-08,
+      "loss": 0.1718,
       "step": 16000
     },
     {
-      "epoch": 2.947135752440597,
-      "eval_loss": 0.890542209148407,
-      "eval_runtime": 76.7268,
-      "eval_samples_per_second": 15.718,
-      "eval_steps_per_second": 1.968,
+      "epoch": 2.9476787030213707,
+      "eval_loss": 0.8798421621322632,
+      "eval_runtime": 76.0825,
+      "eval_samples_per_second": 15.864,
+      "eval_steps_per_second": 1.985,
       "step": 16000
     },
     {
-      "epoch": 3.0392337447043656,
-      "grad_norm": 86.5535659790039,
-      "learning_rate": 1.2009578191195431e-08,
-      "loss": 0.1843,
+      "epoch": 3.0397936624907884,
+      "grad_norm": 5.1134706154698506e-05,
+      "learning_rate": 1.2002579218865142e-08,
+      "loss": 0.153,
       "step": 16500
     },
     {
-      "epoch": 3.0392337447043656,
-      "eval_loss": 0.8875145316123962,
-      "eval_runtime": 76.4556,
-      "eval_samples_per_second": 15.774,
-      "eval_steps_per_second": 1.975,
+      "epoch": 3.0397936624907884,
+      "eval_loss": 0.8776962757110596,
+      "eval_runtime": 75.3205,
+      "eval_samples_per_second": 16.025,
+      "eval_steps_per_second": 2.005,
       "step": 16500
     },
     {
-      "epoch": 3.131331736968134,
-      "grad_norm": 0.05445564165711403,
-      "learning_rate": 1.0858353287898323e-08,
-      "loss": 0.1611,
+      "epoch": 3.131908621960206,
+      "grad_norm": 0.017252299934625626,
+      "learning_rate": 1.085114222549742e-08,
+      "loss": 0.1715,
       "step": 17000
     },
     {
-      "epoch": 3.131331736968134,
-      "eval_loss": 0.8858152627944946,
-      "eval_runtime": 77.4814,
-      "eval_samples_per_second": 15.565,
-      "eval_steps_per_second": 1.949,
+      "epoch": 3.131908621960206,
+      "eval_loss": 0.8758523464202881,
+      "eval_runtime": 74.4544,
+      "eval_samples_per_second": 16.211,
+      "eval_steps_per_second": 2.028,
       "step": 17000
     },
     {
       "epoch": 3.2240235814296243,
-      "grad_norm": 44.673702239990234,
+      "grad_norm": 68.8683853149414,
       "learning_rate": 9.699705232129698e-09,
-      "loss": 0.1568,
+      "loss": 0.1558,
       "step": 17500
     },
     {
       "epoch": 3.2240235814296243,
-      "eval_loss": 0.7821062207221985,
-      "eval_runtime": 75.6502,
-      "eval_samples_per_second": 15.955,
-      "eval_steps_per_second": 1.996,
+      "eval_loss": 0.874191403388977,
+      "eval_runtime": 75.7598,
+      "eval_samples_per_second": 15.932,
+      "eval_steps_per_second": 1.993,
       "step": 17500
     },
     {
       "epoch": 3.316138540899042,
-      "grad_norm": 0.5617901086807251,
+      "grad_norm": 0.791266918182373,
       "learning_rate": 8.548268238761974e-09,
-      "loss": 0.1395,
+      "loss": 0.1384,
       "step": 18000
     },
     {
       "epoch": 3.316138540899042,
-      "eval_loss": 0.7793659567832947,
-      "eval_runtime": 76.0283,
-      "eval_samples_per_second": 15.876,
-      "eval_steps_per_second": 1.986,
+      "eval_loss": 0.8715030550956726,
+      "eval_runtime": 74.7336,
+      "eval_samples_per_second": 16.151,
+      "eval_steps_per_second": 2.021,
       "step": 18000
     },
     {
       "epoch": 3.4082535003684598,
-      "grad_norm": 6.4552321434021,
+      "grad_norm": 6.072669982910156,
       "learning_rate": 7.396831245394252e-09,
-      "loss": 0.1804,
+      "loss": 0.1788,
       "step": 18500
     },
     {
       "epoch": 3.4082535003684598,
-      "eval_loss": 0.7778191566467285,
-      "eval_runtime": 76.0536,
-      "eval_samples_per_second": 15.87,
-      "eval_steps_per_second": 1.985,
+      "eval_loss": 0.8695403933525085,
+      "eval_runtime": 75.2877,
+      "eval_samples_per_second": 16.032,
+      "eval_steps_per_second": 2.006,
       "step": 18500
     },
     {
       "epoch": 3.500368459837878,
-      "grad_norm": 0.31470850110054016,
+      "grad_norm": 0.43182799220085144,
       "learning_rate": 6.2453942520265294e-09,
-      "loss": 0.1728,
+      "loss": 0.1668,
       "step": 19000
     },
     {
       "epoch": 3.500368459837878,
-      "eval_loss": 0.7768829464912415,
-      "eval_runtime": 76.116,
-      "eval_samples_per_second": 15.857,
-      "eval_steps_per_second": 1.984,
+      "eval_loss": 0.8684694766998291,
+      "eval_runtime": 75.7597,
+      "eval_samples_per_second": 15.932,
+      "eval_steps_per_second": 1.993,
       "step": 19000
     },
     {
       "epoch": 3.5924834193072956,
-      "grad_norm": 0.029128307476639748,
+      "grad_norm": 0.01970759592950344,
       "learning_rate": 5.093957258658806e-09,
-      "loss": 0.179,
+      "loss": 0.1697,
       "step": 19500
     },
     {
       "epoch": 3.5924834193072956,
-      "eval_loss": 0.7758385539054871,
-      "eval_runtime": 76.1889,
-      "eval_samples_per_second": 15.842,
-      "eval_steps_per_second": 1.982,
+      "eval_loss": 0.8674086928367615,
+      "eval_runtime": 74.5942,
+      "eval_samples_per_second": 16.181,
+      "eval_steps_per_second": 2.024,
       "step": 19500
     },
     {
       "epoch": 3.6845983787767134,
-      "grad_norm": 0.0002703067730180919,
+      "grad_norm": 0.0002275915176142007,
       "learning_rate": 3.942520265291083e-09,
-      "loss": 0.179,
+      "loss": 0.1764,
       "step": 20000
     },
     {
       "epoch": 3.6845983787767134,
-      "eval_loss": 0.7751660943031311,
-      "eval_runtime": 76.1566,
-      "eval_samples_per_second": 15.849,
-      "eval_steps_per_second": 1.983,
+      "eval_loss": 0.8666148781776428,
+      "eval_runtime": 74.3994,
+      "eval_samples_per_second": 16.223,
+      "eval_steps_per_second": 2.03,
       "step": 20000
     },
     {
       "epoch": 3.776713338246131,
-      "grad_norm": 259.0263366699219,
+      "grad_norm": 199.484375,
       "learning_rate": 2.79108327192336e-09,
-      "loss": 0.1454,
+      "loss": 0.1417,
       "step": 20500
     },
     {
       "epoch": 3.776713338246131,
-      "eval_loss": 0.7746543884277344,
-      "eval_runtime": 76.3059,
-      "eval_samples_per_second": 15.818,
-      "eval_steps_per_second": 1.979,
+      "eval_loss": 0.8660197854042053,
+      "eval_runtime": 75.6487,
+      "eval_samples_per_second": 15.955,
+      "eval_steps_per_second": 1.996,
       "step": 20500
     },
     {
       "epoch": 3.868828297715549,
-      "grad_norm": 0.0011778937187045813,
+      "grad_norm": 0.0007738114218227565,
       "learning_rate": 1.6396462785556374e-09,
-      "loss": 0.1568,
+      "loss": 0.1556,
       "step": 21000
     },
     {
       "epoch": 3.868828297715549,
-      "eval_loss": 0.7744118571281433,
-      "eval_runtime": 76.0825,
-      "eval_samples_per_second": 15.864,
-      "eval_steps_per_second": 1.985,
+      "eval_loss": 0.8657393455505371,
+      "eval_runtime": 74.9437,
+      "eval_samples_per_second": 16.105,
+      "eval_steps_per_second": 2.015,
       "step": 21000
     }
   ],
@@ -656,7 +656,7 @@
       "attributes": {}
     }
   },
-  "total_flos": 7555105255635630.0,
+  "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:af19727b1180ae73a7b4839d8539085faed51bc7e6cd0858188fac5c6ef07636
+oid sha256:9a29c2374e59ce3f57b34db9cb20d2d1bdc870c3dad128e9a4d462636c37e8b0
 size 4847