End of training

all_results.json

@@ -1,12 +1,12 @@
 {
     "epoch": 3.0,
-    "eval_loss": 0.9258912801742554,
-    "eval_runtime": 72.4588,
-    "eval_samples_per_second": 16.658,
-    "eval_steps_per_second": 2.084,
-    "total_flos": 5859024508620360.0,
-    "train_loss": 0.24355060232561188,
-    "train_runtime": 12410.1332,
-    "train_samples_per_second": 2.624,
-    "train_steps_per_second": 1.312
+    "eval_loss": 0.8672720789909363,
+    "eval_runtime": 60.0659,
+    "eval_samples_per_second": 16.432,
+    "eval_steps_per_second": 2.064,
+    "total_flos": 1.01200024209438e+16,
+    "train_loss": 0.22731135058321644,
+    "train_runtime": 20704.962,
+    "train_samples_per_second": 2.717,
+    "train_steps_per_second": 1.359
 }

eval_results.json

@@ -1,7 +1,7 @@
 {
     "epoch": 3.0,
-    "eval_loss": 0.9258912801742554,
-    "eval_runtime": 72.4588,
-    "eval_samples_per_second": 16.658,
-    "eval_steps_per_second": 2.084
+    "eval_loss": 0.8672720789909363,
+    "eval_runtime": 60.0659,
+    "eval_samples_per_second": 16.432,
+    "eval_steps_per_second": 2.064
 }

runs/Aug24_17-41-19_wolf-creek/events.out.tfevents.1724563837.wolf-creek.2226397.1

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:c1d8c117d26b95993ab4131df742f735d8179261ab37af4e7fae5e7cd92c4930
+size 316

train_results.json

@@ -1,8 +1,8 @@
 {
     "epoch": 3.0,
-    "total_flos": 5859024508620360.0,
-    "train_loss": 0.24355060232561188,
-    "train_runtime": 12410.1332,
-    "train_samples_per_second": 2.624,
-    "train_steps_per_second": 1.312
+    "total_flos": 1.01200024209438e+16,
+    "train_loss": 0.22731135058321644,
+    "train_runtime": 20704.962,
+    "train_samples_per_second": 2.717,
+    "train_steps_per_second": 1.359
 }

trainer_state.json

@@ -1,505 +1,865 @@
 {
-  "best_metric": 0.9258912801742554,
-  "best_model_checkpoint": "./output/clip-finetuned-csu-p14-336-e3l58-l/checkpoint-16000",
+  "best_metric": 0.8672720789909363,
+  "best_model_checkpoint": "./output/clip-finetuned-csu-p14-336-e3l58-l/checkpoint-28000",
   "epoch": 3.0,
   "eval_steps": 500,
-  "global_step": 16284,
+  "global_step": 28128,
   "is_hyper_param_search": false,
   "is_local_process_zero": true,
   "is_world_process_zero": true,
   "log_history": [
     {
-      "epoch": 0.09211495946941783,
-      "grad_norm": 19.808820724487305,
-      "learning_rate": 4.84647506755097e-08,
-      "loss": 0.3758,
+      "epoch": 0.05332764505119454,
+      "grad_norm": 217.40109252929688,
+      "learning_rate": 4.911120591581342e-08,
+      "loss": 0.4667,
       "step": 500
     },
     {
-      "epoch": 0.09211495946941783,
-      "eval_loss": 1.418845295906067,
-      "eval_runtime": 75.2377,
-      "eval_samples_per_second": 16.042,
-      "eval_steps_per_second": 2.007,
+      "epoch": 0.05332764505119454,
+      "eval_loss": 1.4427576065063477,
+      "eval_runtime": 60.328,
+      "eval_samples_per_second": 16.361,
+      "eval_steps_per_second": 2.055,
       "step": 500
     },
     {
-      "epoch": 0.18422991893883567,
-      "grad_norm": 76.37991333007812,
-      "learning_rate": 4.69295013510194e-08,
-      "loss": 0.4108,
+      "epoch": 0.10665529010238908,
+      "grad_norm": 70.38941955566406,
+      "learning_rate": 4.822241183162685e-08,
+      "loss": 0.4534,
       "step": 1000
     },
     {
-      "epoch": 0.18422991893883567,
-      "eval_loss": 1.3513683080673218,
-      "eval_runtime": 74.9686,
-      "eval_samples_per_second": 16.1,
-      "eval_steps_per_second": 2.014,
+      "epoch": 0.10665529010238908,
+      "eval_loss": 1.3823094367980957,
+      "eval_runtime": 61.4884,
+      "eval_samples_per_second": 16.052,
+      "eval_steps_per_second": 2.017,
       "step": 1000
     },
     {
-      "epoch": 0.2763448784082535,
-      "grad_norm": 0.5795987844467163,
-      "learning_rate": 4.539425202652911e-08,
-      "loss": 0.4335,
+      "epoch": 0.1599829351535836,
+      "grad_norm": 389.9967956542969,
+      "learning_rate": 4.733361774744027e-08,
+      "loss": 0.3753,
       "step": 1500
     },
     {
-      "epoch": 0.2763448784082535,
-      "eval_loss": 1.2895005941390991,
-      "eval_runtime": 76.713,
-      "eval_samples_per_second": 15.734,
-      "eval_steps_per_second": 1.968,
+      "epoch": 0.1599829351535836,
+      "eval_loss": 1.3328827619552612,
+      "eval_runtime": 63.0378,
+      "eval_samples_per_second": 15.657,
+      "eval_steps_per_second": 1.967,
       "step": 1500
     },
     {
-      "epoch": 0.36845983787767134,
-      "grad_norm": 0.1463615745306015,
-      "learning_rate": 4.385900270203881e-08,
-      "loss": 0.3436,
+      "epoch": 0.21331058020477817,
+      "grad_norm": 13.427445411682129,
+      "learning_rate": 4.6444823663253694e-08,
+      "loss": 0.3361,
       "step": 2000
     },
     {
-      "epoch": 0.36845983787767134,
-      "eval_loss": 1.2414860725402832,
-      "eval_runtime": 76.9749,
-      "eval_samples_per_second": 15.68,
+      "epoch": 0.21331058020477817,
+      "eval_loss": 1.2922736406326294,
+      "eval_runtime": 63.1958,
+      "eval_samples_per_second": 15.618,
       "eval_steps_per_second": 1.962,
       "step": 2000
     },
     {
-      "epoch": 0.46057479734708917,
-      "grad_norm": 323.8576354980469,
-      "learning_rate": 4.232375337754851e-08,
-      "loss": 0.366,
+      "epoch": 0.2666382252559727,
+      "grad_norm": 320.7197265625,
+      "learning_rate": 4.555602957906712e-08,
+      "loss": 0.3597,
       "step": 2500
     },
     {
-      "epoch": 0.46057479734708917,
-      "eval_loss": 1.1949334144592285,
-      "eval_runtime": 77.0381,
-      "eval_samples_per_second": 15.668,
-      "eval_steps_per_second": 1.96,
+      "epoch": 0.2666382252559727,
+      "eval_loss": 1.257211685180664,
+      "eval_runtime": 63.3241,
+      "eval_samples_per_second": 15.586,
+      "eval_steps_per_second": 1.958,
       "step": 2500
     },
     {
-      "epoch": 0.552689756816507,
-      "grad_norm": 339.28533935546875,
-      "learning_rate": 4.078850405305821e-08,
-      "loss": 0.3191,
+      "epoch": 0.3199658703071672,
+      "grad_norm": 0.0019184721168130636,
+      "learning_rate": 4.4667235494880546e-08,
+      "loss": 0.3042,
       "step": 3000
     },
     {
-      "epoch": 0.552689756816507,
-      "eval_loss": 1.1560280323028564,
-      "eval_runtime": 77.0026,
-      "eval_samples_per_second": 15.675,
-      "eval_steps_per_second": 1.961,
+      "epoch": 0.3199658703071672,
+      "eval_loss": 1.2240594625473022,
+      "eval_runtime": 63.5118,
+      "eval_samples_per_second": 15.54,
+      "eval_steps_per_second": 1.952,
       "step": 3000
     },
     {
-      "epoch": 0.6448047162859248,
-      "grad_norm": 60.27859878540039,
-      "learning_rate": 3.925325472856792e-08,
-      "loss": 0.2779,
+      "epoch": 0.37329351535836175,
+      "grad_norm": 11.738512992858887,
+      "learning_rate": 4.3778441410693965e-08,
+      "loss": 0.3252,
       "step": 3500
     },
     {
-      "epoch": 0.6448047162859248,
-      "eval_loss": 1.1233855485916138,
-      "eval_runtime": 76.9006,
-      "eval_samples_per_second": 15.696,
-      "eval_steps_per_second": 1.964,
+      "epoch": 0.37329351535836175,
+      "eval_loss": 1.1988952159881592,
+      "eval_runtime": 63.5384,
+      "eval_samples_per_second": 15.534,
+      "eval_steps_per_second": 1.952,
       "step": 3500
     },
     {
-      "epoch": 0.7369196757553427,
-      "grad_norm": 18.268455505371094,
-      "learning_rate": 3.7718005404077616e-08,
-      "loss": 0.2942,
+      "epoch": 0.42662116040955633,
+      "grad_norm": 5.839956283569336,
+      "learning_rate": 4.28896473265074e-08,
+      "loss": 0.2439,
       "step": 4000
     },
     {
-      "epoch": 0.7369196757553427,
-      "eval_loss": 1.0993207693099976,
-      "eval_runtime": 77.2834,
-      "eval_samples_per_second": 15.618,
-      "eval_steps_per_second": 1.954,
+      "epoch": 0.42662116040955633,
+      "eval_loss": 1.1730674505233765,
+      "eval_runtime": 62.3497,
+      "eval_samples_per_second": 15.83,
+      "eval_steps_per_second": 1.989,
       "step": 4000
     },
     {
-      "epoch": 0.8290346352247605,
-      "grad_norm": 505.22723388671875,
-      "learning_rate": 3.6182756079587324e-08,
-      "loss": 0.2615,
+      "epoch": 0.47994880546075086,
+      "grad_norm": 154.76502990722656,
+      "learning_rate": 4.2000853242320816e-08,
+      "loss": 0.3004,
       "step": 4500
     },
     {
-      "epoch": 0.8290346352247605,
-      "eval_loss": 1.080470085144043,
-      "eval_runtime": 77.1214,
-      "eval_samples_per_second": 15.651,
-      "eval_steps_per_second": 1.958,
+      "epoch": 0.47994880546075086,
+      "eval_loss": 1.1497682332992554,
+      "eval_runtime": 63.5491,
+      "eval_samples_per_second": 15.531,
+      "eval_steps_per_second": 1.951,
       "step": 4500
     },
     {
-      "epoch": 0.9211495946941783,
-      "grad_norm": 1.2971173524856567,
-      "learning_rate": 3.4647506755097025e-08,
-      "loss": 0.2715,
+      "epoch": 0.5332764505119454,
+      "grad_norm": 0.13150489330291748,
+      "learning_rate": 4.111205915813424e-08,
+      "loss": 0.3205,
       "step": 5000
     },
     {
-      "epoch": 0.9211495946941783,
-      "eval_loss": 1.0614728927612305,
-      "eval_runtime": 77.2982,
-      "eval_samples_per_second": 15.615,
-      "eval_steps_per_second": 1.953,
+      "epoch": 0.5332764505119454,
+      "eval_loss": 1.1258739233016968,
+      "eval_runtime": 62.1884,
+      "eval_samples_per_second": 15.871,
+      "eval_steps_per_second": 1.994,
       "step": 5000
     },
     {
-      "epoch": 1.013264554163596,
-      "grad_norm": 0.17611829936504364,
-      "learning_rate": 3.3112257430606727e-08,
-      "loss": 0.2509,
+      "epoch": 0.58660409556314,
+      "grad_norm": 0.3719326853752136,
+      "learning_rate": 4.022326507394767e-08,
+      "loss": 0.3091,
       "step": 5500
     },
     {
-      "epoch": 1.013264554163596,
-      "eval_loss": 1.0461839437484741,
-      "eval_runtime": 75.3494,
-      "eval_samples_per_second": 16.019,
-      "eval_steps_per_second": 2.004,
+      "epoch": 0.58660409556314,
+      "eval_loss": 1.1053005456924438,
+      "eval_runtime": 63.437,
+      "eval_samples_per_second": 15.559,
+      "eval_steps_per_second": 1.955,
       "step": 5500
     },
     {
-      "epoch": 1.105379513633014,
-      "grad_norm": 228.2562713623047,
-      "learning_rate": 3.1577008106116435e-08,
-      "loss": 0.2653,
+      "epoch": 0.6399317406143344,
+      "grad_norm": 170.89846801757812,
+      "learning_rate": 3.9334470989761093e-08,
+      "loss": 0.2657,
       "step": 6000
     },
     {
-      "epoch": 1.105379513633014,
-      "eval_loss": 1.0305675268173218,
-      "eval_runtime": 77.0542,
-      "eval_samples_per_second": 15.664,
-      "eval_steps_per_second": 1.96,
+      "epoch": 0.6399317406143344,
+      "eval_loss": 1.0857468843460083,
+      "eval_runtime": 62.4055,
+      "eval_samples_per_second": 15.816,
+      "eval_steps_per_second": 1.987,
       "step": 6000
     },
     {
-      "epoch": 1.1974944731024317,
-      "grad_norm": 432.4784240722656,
-      "learning_rate": 3.0041758781626136e-08,
-      "loss": 0.2199,
+      "epoch": 0.693259385665529,
+      "grad_norm": 323.0997619628906,
+      "learning_rate": 3.844567690557451e-08,
+      "loss": 0.2251,
       "step": 6500
     },
     {
-      "epoch": 1.1974944731024317,
-      "eval_loss": 1.0182703733444214,
-      "eval_runtime": 76.296,
-      "eval_samples_per_second": 15.82,
-      "eval_steps_per_second": 1.979,
+      "epoch": 0.693259385665529,
+      "eval_loss": 1.0695208311080933,
+      "eval_runtime": 63.1085,
+      "eval_samples_per_second": 15.64,
+      "eval_steps_per_second": 1.965,
       "step": 6500
     },
     {
-      "epoch": 1.2896094325718497,
-      "grad_norm": 0.6385967135429382,
-      "learning_rate": 2.8506509457135834e-08,
-      "loss": 0.1844,
+      "epoch": 0.7465870307167235,
+      "grad_norm": 0.13254186511039734,
+      "learning_rate": 3.7556882821387945e-08,
+      "loss": 0.248,
       "step": 7000
     },
     {
-      "epoch": 1.2896094325718497,
-      "eval_loss": 1.0078601837158203,
-      "eval_runtime": 76.5441,
-      "eval_samples_per_second": 15.769,
-      "eval_steps_per_second": 1.973,
+      "epoch": 0.7465870307167235,
+      "eval_loss": 1.0535719394683838,
+      "eval_runtime": 62.3504,
+      "eval_samples_per_second": 15.83,
+      "eval_steps_per_second": 1.989,
       "step": 7000
     },
     {
-      "epoch": 1.3817243920412676,
-      "grad_norm": 57.47743606567383,
-      "learning_rate": 2.697126013264554e-08,
-      "loss": 0.2143,
+      "epoch": 0.7999146757679181,
+      "grad_norm": 63.14700698852539,
+      "learning_rate": 3.6668088737201364e-08,
+      "loss": 0.2745,
       "step": 7500
     },
     {
-      "epoch": 1.3817243920412676,
-      "eval_loss": 0.9969784617424011,
-      "eval_runtime": 75.6186,
-      "eval_samples_per_second": 15.962,
-      "eval_steps_per_second": 1.997,
+      "epoch": 0.7999146757679181,
+      "eval_loss": 1.0388950109481812,
+      "eval_runtime": 62.2311,
+      "eval_samples_per_second": 15.86,
+      "eval_steps_per_second": 1.993,
       "step": 7500
     },
     {
-      "epoch": 1.4738393515106853,
-      "grad_norm": 0.11704438179731369,
-      "learning_rate": 2.5436010808155244e-08,
-      "loss": 0.2011,
+      "epoch": 0.8532423208191127,
+      "grad_norm": 623.6299438476562,
+      "learning_rate": 3.577929465301478e-08,
+      "loss": 0.2857,
       "step": 8000
     },
     {
-      "epoch": 1.4738393515106853,
-      "eval_loss": 0.9886976480484009,
-      "eval_runtime": 75.238,
-      "eval_samples_per_second": 16.042,
-      "eval_steps_per_second": 2.007,
+      "epoch": 0.8532423208191127,
+      "eval_loss": 1.0260424613952637,
+      "eval_runtime": 62.2792,
+      "eval_samples_per_second": 15.848,
+      "eval_steps_per_second": 1.991,
       "step": 8000
     },
     {
-      "epoch": 1.565954310980103,
-      "grad_norm": 39.71356964111328,
-      "learning_rate": 2.3900761483664948e-08,
-      "loss": 0.1995,
+      "epoch": 0.9065699658703071,
+      "grad_norm": 155.54579162597656,
+      "learning_rate": 3.4890500568828216e-08,
+      "loss": 0.2365,
       "step": 8500
     },
     {
-      "epoch": 1.565954310980103,
-      "eval_loss": 0.980699360370636,
-      "eval_runtime": 75.7533,
-      "eval_samples_per_second": 15.933,
-      "eval_steps_per_second": 1.993,
+      "epoch": 0.9065699658703071,
+      "eval_loss": 1.0136038064956665,
+      "eval_runtime": 62.583,
+      "eval_samples_per_second": 15.771,
+      "eval_steps_per_second": 1.981,
       "step": 8500
     },
     {
-      "epoch": 1.658069270449521,
-      "grad_norm": 0.09355564415454865,
-      "learning_rate": 2.236551215917465e-08,
-      "loss": 0.2493,
+      "epoch": 0.9598976109215017,
+      "grad_norm": 76.14079284667969,
+      "learning_rate": 3.4001706484641635e-08,
+      "loss": 0.2382,
       "step": 9000
     },
     {
-      "epoch": 1.658069270449521,
-      "eval_loss": 0.9733120203018188,
-      "eval_runtime": 75.1219,
-      "eval_samples_per_second": 16.067,
-      "eval_steps_per_second": 2.01,
+      "epoch": 0.9598976109215017,
+      "eval_loss": 1.0034172534942627,
+      "eval_runtime": 63.5592,
+      "eval_samples_per_second": 15.529,
+      "eval_steps_per_second": 1.951,
       "step": 9000
     },
     {
-      "epoch": 1.750184229918939,
-      "grad_norm": 0.12566468119621277,
-      "learning_rate": 2.083026283468435e-08,
-      "loss": 0.2192,
+      "epoch": 1.0132252559726962,
+      "grad_norm": 571.6434936523438,
+      "learning_rate": 3.311291240045506e-08,
+      "loss": 0.2419,
       "step": 9500
     },
     {
-      "epoch": 1.750184229918939,
-      "eval_loss": 0.966866672039032,
-      "eval_runtime": 76.5812,
-      "eval_samples_per_second": 15.761,
-      "eval_steps_per_second": 1.972,
+      "epoch": 1.0132252559726962,
+      "eval_loss": 0.994463324546814,
+      "eval_runtime": 61.8407,
+      "eval_samples_per_second": 15.96,
+      "eval_steps_per_second": 2.005,
       "step": 9500
     },
     {
-      "epoch": 1.8422991893883567,
-      "grad_norm": 1.9423623085021973,
-      "learning_rate": 1.9295013510194056e-08,
-      "loss": 0.1882,
+      "epoch": 1.0665529010238908,
+      "grad_norm": 540.1460571289062,
+      "learning_rate": 3.2224118316268486e-08,
+      "loss": 0.2047,
       "step": 10000
     },
     {
-      "epoch": 1.8422991893883567,
-      "eval_loss": 0.9616385698318481,
-      "eval_runtime": 75.5454,
-      "eval_samples_per_second": 15.977,
-      "eval_steps_per_second": 1.999,
+      "epoch": 1.0665529010238908,
+      "eval_loss": 0.9851205348968506,
+      "eval_runtime": 62.1686,
+      "eval_samples_per_second": 15.876,
+      "eval_steps_per_second": 1.995,
       "step": 10000
     },
     {
-      "epoch": 1.9344141488577744,
-      "grad_norm": 0.12048076093196869,
-      "learning_rate": 1.7759764185703757e-08,
-      "loss": 0.2193,
+      "epoch": 1.1198805460750854,
+      "grad_norm": 6.2389655113220215,
+      "learning_rate": 3.133532423208191e-08,
+      "loss": 0.2133,
       "step": 10500
     },
     {
-      "epoch": 1.9344141488577744,
-      "eval_loss": 0.9550592303276062,
-      "eval_runtime": 75.1134,
-      "eval_samples_per_second": 16.069,
-      "eval_steps_per_second": 2.01,
+      "epoch": 1.1198805460750854,
+      "eval_loss": 0.9759647250175476,
+      "eval_runtime": 62.8691,
+      "eval_samples_per_second": 15.699,
+      "eval_steps_per_second": 1.972,
       "step": 10500
     },
     {
-      "epoch": 2.026529108327192,
-      "grad_norm": 0.19933176040649414,
-      "learning_rate": 1.622451486121346e-08,
-      "loss": 0.2148,
+      "epoch": 1.17320819112628,
+      "grad_norm": 321.5069885253906,
+      "learning_rate": 3.044653014789533e-08,
+      "loss": 0.207,
       "step": 11000
     },
     {
-      "epoch": 2.026529108327192,
-      "eval_loss": 0.9494845271110535,
-      "eval_runtime": 76.2803,
-      "eval_samples_per_second": 15.823,
-      "eval_steps_per_second": 1.98,
+      "epoch": 1.17320819112628,
+      "eval_loss": 0.9683472514152527,
+      "eval_runtime": 61.915,
+      "eval_samples_per_second": 15.941,
+      "eval_steps_per_second": 2.003,
       "step": 11000
     },
     {
-      "epoch": 2.1186440677966103,
-      "grad_norm": 13.599303245544434,
-      "learning_rate": 1.4689265536723163e-08,
-      "loss": 0.1975,
+      "epoch": 1.2265358361774745,
+      "grad_norm": 0.0028632767498493195,
+      "learning_rate": 2.955773606370876e-08,
+      "loss": 0.2369,
       "step": 11500
     },
     {
-      "epoch": 2.1186440677966103,
-      "eval_loss": 0.9449298977851868,
-      "eval_runtime": 75.6422,
-      "eval_samples_per_second": 15.957,
-      "eval_steps_per_second": 1.996,
+      "epoch": 1.2265358361774745,
+      "eval_loss": 0.9597958922386169,
+      "eval_runtime": 62.4216,
+      "eval_samples_per_second": 15.812,
+      "eval_steps_per_second": 1.986,
       "step": 11500
     },
     {
-      "epoch": 2.210759027266028,
-      "grad_norm": 0.8174373507499695,
-      "learning_rate": 1.3154016212232866e-08,
-      "loss": 0.1791,
+      "epoch": 1.2798634812286689,
+      "grad_norm": 1.046075701713562,
+      "learning_rate": 2.8668941979522183e-08,
+      "loss": 0.1834,
       "step": 12000
     },
     {
-      "epoch": 2.210759027266028,
-      "eval_loss": 0.940927267074585,
-      "eval_runtime": 75.2379,
-      "eval_samples_per_second": 16.042,
-      "eval_steps_per_second": 2.007,
+      "epoch": 1.2798634812286689,
+      "eval_loss": 0.9528058767318726,
+      "eval_runtime": 62.9687,
+      "eval_samples_per_second": 15.674,
+      "eval_steps_per_second": 1.969,
       "step": 12000
     },
     {
-      "epoch": 2.3028739867354457,
-      "grad_norm": 705.062744140625,
-      "learning_rate": 1.161876688774257e-08,
-      "loss": 0.2057,
+      "epoch": 1.3331911262798635,
+      "grad_norm": 0.01093329768627882,
+      "learning_rate": 2.778014789533561e-08,
+      "loss": 0.2002,
       "step": 12500
     },
     {
-      "epoch": 2.3028739867354457,
-      "eval_loss": 0.9382032155990601,
-      "eval_runtime": 76.1075,
-      "eval_samples_per_second": 15.859,
-      "eval_steps_per_second": 1.984,
+      "epoch": 1.3331911262798635,
+      "eval_loss": 0.9465218782424927,
+      "eval_runtime": 62.0222,
+      "eval_samples_per_second": 15.914,
+      "eval_steps_per_second": 1.999,
       "step": 12500
     },
     {
-      "epoch": 2.3949889462048635,
-      "grad_norm": 0.7270774841308594,
-      "learning_rate": 1.0083517563252273e-08,
-      "loss": 0.2037,
+      "epoch": 1.386518771331058,
+      "grad_norm": 0.13472874462604523,
+      "learning_rate": 2.689135381114903e-08,
+      "loss": 0.1918,
       "step": 13000
     },
     {
-      "epoch": 2.3949889462048635,
-      "eval_loss": 0.9352426528930664,
-      "eval_runtime": 75.5821,
-      "eval_samples_per_second": 15.969,
-      "eval_steps_per_second": 1.998,
+      "epoch": 1.386518771331058,
+      "eval_loss": 0.9417626857757568,
+      "eval_runtime": 62.4603,
+      "eval_samples_per_second": 15.802,
+      "eval_steps_per_second": 1.985,
       "step": 13000
     },
     {
-      "epoch": 2.4871039056742816,
-      "grad_norm": 2.034397602081299,
-      "learning_rate": 8.548268238761974e-09,
-      "loss": 0.2011,
+      "epoch": 1.4398464163822526,
+      "grad_norm": 19.54829216003418,
+      "learning_rate": 2.600255972696246e-08,
+      "loss": 0.2264,
       "step": 13500
     },
     {
-      "epoch": 2.4871039056742816,
-      "eval_loss": 0.9317955374717712,
-      "eval_runtime": 75.3697,
-      "eval_samples_per_second": 16.014,
-      "eval_steps_per_second": 2.003,
+      "epoch": 1.4398464163822526,
+      "eval_loss": 0.9353039860725403,
+      "eval_runtime": 63.2324,
+      "eval_samples_per_second": 15.609,
+      "eval_steps_per_second": 1.961,
       "step": 13500
     },
     {
-      "epoch": 2.5792188651436994,
-      "grad_norm": 299.313232421875,
-      "learning_rate": 7.013018914271678e-09,
-      "loss": 0.1551,
+      "epoch": 1.493174061433447,
+      "grad_norm": 6.569502353668213,
+      "learning_rate": 2.511376564277588e-08,
+      "loss": 0.171,
       "step": 14000
     },
     {
-      "epoch": 2.5792188651436994,
-      "eval_loss": 0.9295068979263306,
-      "eval_runtime": 76.109,
-      "eval_samples_per_second": 15.859,
-      "eval_steps_per_second": 1.984,
+      "epoch": 1.493174061433447,
+      "eval_loss": 0.930813729763031,
+      "eval_runtime": 61.9703,
+      "eval_samples_per_second": 15.927,
+      "eval_steps_per_second": 2.001,
       "step": 14000
     },
     {
-      "epoch": 2.671333824613117,
-      "grad_norm": 0.04151364788413048,
-      "learning_rate": 5.47776958978138e-09,
-      "loss": 0.1565,
+      "epoch": 1.5465017064846416,
+      "grad_norm": 153.30908203125,
+      "learning_rate": 2.4224971558589305e-08,
+      "loss": 0.2029,
       "step": 14500
     },
     {
-      "epoch": 2.671333824613117,
-      "eval_loss": 0.9278931617736816,
-      "eval_runtime": 75.8114,
-      "eval_samples_per_second": 15.921,
-      "eval_steps_per_second": 1.992,
+      "epoch": 1.5465017064846416,
+      "eval_loss": 0.9262459874153137,
+      "eval_runtime": 62.3851,
+      "eval_samples_per_second": 15.821,
+      "eval_steps_per_second": 1.988,
       "step": 14500
     },
     {
-      "epoch": 2.7634487840825352,
-      "grad_norm": 4.239564895629883,
-      "learning_rate": 3.942520265291083e-09,
-      "loss": 0.1963,
+      "epoch": 1.5998293515358362,
+      "grad_norm": 578.005615234375,
+      "learning_rate": 2.333617747440273e-08,
+      "loss": 0.173,
       "step": 15000
     },
     {
-      "epoch": 2.7634487840825352,
-      "eval_loss": 0.9268263578414917,
-      "eval_runtime": 75.5159,
-      "eval_samples_per_second": 15.983,
-      "eval_steps_per_second": 2.0,
+      "epoch": 1.5998293515358362,
+      "eval_loss": 0.9224158525466919,
+      "eval_runtime": 63.0306,
+      "eval_samples_per_second": 15.659,
+      "eval_steps_per_second": 1.967,
       "step": 15000
     },
     {
-      "epoch": 2.855563743551953,
-      "grad_norm": 0.001815026975236833,
-      "learning_rate": 2.407270940800786e-09,
-      "loss": 0.1823,
+      "epoch": 1.6531569965870307,
+      "grad_norm": 38.18622970581055,
+      "learning_rate": 2.2447383390216153e-08,
+      "loss": 0.2126,
       "step": 15500
     },
     {
-      "epoch": 2.855563743551953,
-      "eval_loss": 0.9261995553970337,
-      "eval_runtime": 76.2027,
-      "eval_samples_per_second": 15.839,
-      "eval_steps_per_second": 1.982,
+      "epoch": 1.6531569965870307,
+      "eval_loss": 0.9170305132865906,
+      "eval_runtime": 61.9661,
+      "eval_samples_per_second": 15.928,
+      "eval_steps_per_second": 2.001,
       "step": 15500
     },
     {
-      "epoch": 2.9476787030213707,
-      "grad_norm": 221.31707763671875,
-      "learning_rate": 8.720216163104888e-10,
-      "loss": 0.1854,
+      "epoch": 1.7064846416382253,
+      "grad_norm": 279.6297302246094,
+      "learning_rate": 2.155858930602958e-08,
+      "loss": 0.1933,
       "step": 16000
     },
     {
-      "epoch": 2.9476787030213707,
-      "eval_loss": 0.9258912801742554,
-      "eval_runtime": 75.783,
-      "eval_samples_per_second": 15.927,
-      "eval_steps_per_second": 1.993,
+      "epoch": 1.7064846416382253,
+      "eval_loss": 0.9142164587974548,
+      "eval_runtime": 62.0524,
+      "eval_samples_per_second": 15.906,
+      "eval_steps_per_second": 1.998,
       "step": 16000
     },
+    {
+      "epoch": 1.75981228668942,
+      "grad_norm": 32.153900146484375,
+      "learning_rate": 2.0669795221843004e-08,
+      "loss": 0.1911,
+      "step": 16500
+    },
+    {
+      "epoch": 1.75981228668942,
+      "eval_loss": 0.9099463820457458,
+      "eval_runtime": 63.0816,
+      "eval_samples_per_second": 15.646,
+      "eval_steps_per_second": 1.966,
+      "step": 16500
+    },
+    {
+      "epoch": 1.8131399317406145,
+      "grad_norm": 1.433034062385559,
+      "learning_rate": 1.9781001137656427e-08,
+      "loss": 0.1809,
+      "step": 17000
+    },
+    {
+      "epoch": 1.8131399317406145,
+      "eval_loss": 0.9075033664703369,
+      "eval_runtime": 62.1223,
+      "eval_samples_per_second": 15.888,
+      "eval_steps_per_second": 1.996,
+      "step": 17000
+    },
+    {
+      "epoch": 1.8664675767918089,
+      "grad_norm": 211.28334045410156,
+      "learning_rate": 1.8892207053469853e-08,
+      "loss": 0.2699,
+      "step": 17500
+    },
+    {
+      "epoch": 1.8664675767918089,
+      "eval_loss": 0.9023045897483826,
+      "eval_runtime": 62.2896,
+      "eval_samples_per_second": 15.845,
+      "eval_steps_per_second": 1.991,
+      "step": 17500
+    },
+    {
+      "epoch": 1.9197952218430034,
+      "grad_norm": 0.22130072116851807,
+      "learning_rate": 1.800341296928328e-08,
+      "loss": 0.1682,
+      "step": 18000
+    },
+    {
+      "epoch": 1.9197952218430034,
+      "eval_loss": 0.8992673754692078,
+      "eval_runtime": 63.0217,
+      "eval_samples_per_second": 15.661,
+      "eval_steps_per_second": 1.968,
+      "step": 18000
+    },
+    {
+      "epoch": 1.9731228668941978,
+      "grad_norm": 463.95916748046875,
+      "learning_rate": 1.7114618885096697e-08,
+      "loss": 0.2248,
+      "step": 18500
+    },
+    {
+      "epoch": 1.9731228668941978,
+      "eval_loss": 0.8949461579322815,
+      "eval_runtime": 62.3865,
+      "eval_samples_per_second": 15.821,
+      "eval_steps_per_second": 1.988,
+      "step": 18500
+    },
+    {
+      "epoch": 2.0264505119453924,
+      "grad_norm": 257.9815979003906,
+      "learning_rate": 1.6225824800910123e-08,
+      "loss": 0.1892,
+      "step": 19000
+    },
+    {
+      "epoch": 2.0264505119453924,
+      "eval_loss": 0.892412543296814,
+      "eval_runtime": 62.3214,
+      "eval_samples_per_second": 15.837,
+      "eval_steps_per_second": 1.99,
+      "step": 19000
+    },
+    {
+      "epoch": 2.079778156996587,
+      "grad_norm": 50.01844787597656,
+      "learning_rate": 1.533703071672355e-08,
+      "loss": 0.1907,
+      "step": 19500
+    },
+    {
+      "epoch": 2.079778156996587,
+      "eval_loss": 0.8893100619316101,
+      "eval_runtime": 62.9983,
+      "eval_samples_per_second": 15.667,
+      "eval_steps_per_second": 1.968,
+      "step": 19500
+    },
+    {
+      "epoch": 2.1331058020477816,
+      "grad_norm": 1.0831713676452637,
+      "learning_rate": 1.4448236632536973e-08,
+      "loss": 0.1661,
+      "step": 20000
+    },
+    {
+      "epoch": 2.1331058020477816,
+      "eval_loss": 0.887154757976532,
+      "eval_runtime": 62.2417,
+      "eval_samples_per_second": 15.858,
+      "eval_steps_per_second": 1.992,
+      "step": 20000
+    },
+    {
+      "epoch": 2.186433447098976,
+      "grad_norm": 0.08456582576036453,
+      "learning_rate": 1.3559442548350397e-08,
+      "loss": 0.1852,
+      "step": 20500
+    },
+    {
+      "epoch": 2.186433447098976,
+      "eval_loss": 0.8855507969856262,
+      "eval_runtime": 61.7252,
+      "eval_samples_per_second": 15.99,
+      "eval_steps_per_second": 2.009,
+      "step": 20500
+    },
+    {
+      "epoch": 2.2397610921501707,
+      "grad_norm": 19.829164505004883,
+      "learning_rate": 1.2670648464163821e-08,
+      "loss": 0.1561,
+      "step": 21000
+    },
+    {
+      "epoch": 2.2397610921501707,
+      "eval_loss": 0.8829911947250366,
+      "eval_runtime": 62.7443,
+      "eval_samples_per_second": 15.731,
+      "eval_steps_per_second": 1.976,
+      "step": 21000
+    },
+    {
+      "epoch": 2.2930887372013653,
+      "grad_norm": 0.12170098721981049,
+      "learning_rate": 1.1781854379977247e-08,
+      "loss": 0.2146,
+      "step": 21500
+    },
+    {
+      "epoch": 2.2930887372013653,
+      "eval_loss": 0.8807759881019592,
+      "eval_runtime": 62.1264,
+      "eval_samples_per_second": 15.887,
+      "eval_steps_per_second": 1.996,
+      "step": 21500
+    },
+    {
+      "epoch": 2.34641638225256,
+      "grad_norm": 0.0019679605029523373,
+      "learning_rate": 1.0893060295790671e-08,
+      "loss": 0.16,
+      "step": 22000
+    },
+    {
+      "epoch": 2.34641638225256,
+      "eval_loss": 0.8785866498947144,
+      "eval_runtime": 62.1778,
+      "eval_samples_per_second": 15.874,
+      "eval_steps_per_second": 1.994,
+      "step": 22000
+    },
+    {
+      "epoch": 2.399744027303754,
+      "grad_norm": 0.0032679687719792128,
+      "learning_rate": 1.0004266211604095e-08,
+      "loss": 0.1904,
+      "step": 22500
+    },
+    {
+      "epoch": 2.399744027303754,
+      "eval_loss": 0.8765972852706909,
+      "eval_runtime": 63.05,
+      "eval_samples_per_second": 15.654,
+      "eval_steps_per_second": 1.967,
+      "step": 22500
+    },
+    {
+      "epoch": 2.453071672354949,
+      "grad_norm": 0.3566090762615204,
+      "learning_rate": 9.11547212741752e-09,
+      "loss": 0.1746,
+      "step": 23000
+    },
+    {
+      "epoch": 2.453071672354949,
+      "eval_loss": 0.874836266040802,
+      "eval_runtime": 62.6532,
+      "eval_samples_per_second": 15.753,
+      "eval_steps_per_second": 1.979,
+      "step": 23000
+    },
+    {
+      "epoch": 2.506399317406143,
+      "grad_norm": 14.910808563232422,
+      "learning_rate": 8.226678043230943e-09,
+      "loss": 0.1767,
+      "step": 23500
+    },
+    {
+      "epoch": 2.506399317406143,
+      "eval_loss": 0.8737635612487793,
+      "eval_runtime": 62.3335,
+      "eval_samples_per_second": 15.834,
+      "eval_steps_per_second": 1.989,
+      "step": 23500
+    },
+    {
+      "epoch": 2.5597269624573378,
+      "grad_norm": 199.9237518310547,
+      "learning_rate": 7.3378839590443675e-09,
+      "loss": 0.1922,
+      "step": 24000
+    },
+    {
+      "epoch": 2.5597269624573378,
+      "eval_loss": 0.8720763921737671,
+      "eval_runtime": 62.5247,
+      "eval_samples_per_second": 15.786,
+      "eval_steps_per_second": 1.983,
+      "step": 24000
+    },
+    {
+      "epoch": 2.6130546075085324,
+      "grad_norm": 3.2796316146850586,
+      "learning_rate": 6.4490898748577924e-09,
+      "loss": 0.1556,
+      "step": 24500
+    },
+    {
+      "epoch": 2.6130546075085324,
+      "eval_loss": 0.8707790970802307,
+      "eval_runtime": 61.7705,
+      "eval_samples_per_second": 15.978,
+      "eval_steps_per_second": 2.007,
+      "step": 24500
+    },
+    {
+      "epoch": 2.666382252559727,
+      "grad_norm": 124.43791961669922,
+      "learning_rate": 5.560295790671217e-09,
+      "loss": 0.1388,
+      "step": 25000
+    },
+    {
+      "epoch": 2.666382252559727,
+      "eval_loss": 0.8700253963470459,
+      "eval_runtime": 62.0486,
+      "eval_samples_per_second": 15.907,
+      "eval_steps_per_second": 1.998,
+      "step": 25000
+    },
+    {
+      "epoch": 2.7197098976109215,
+      "grad_norm": 0.05777175351977348,
+      "learning_rate": 4.6715017064846414e-09,
+      "loss": 0.1766,
+      "step": 25500
+    },
+    {
+      "epoch": 2.7197098976109215,
+      "eval_loss": 0.8693736791610718,
+      "eval_runtime": 63.5834,
+      "eval_samples_per_second": 15.523,
+      "eval_steps_per_second": 1.95,
+      "step": 25500
+    },
+    {
+      "epoch": 2.773037542662116,
+      "grad_norm": 110.81836700439453,
+      "learning_rate": 3.782707622298066e-09,
+      "loss": 0.1615,
+      "step": 26000
+    },
+    {
+      "epoch": 2.773037542662116,
+      "eval_loss": 0.8683955669403076,
+      "eval_runtime": 62.1251,
+      "eval_samples_per_second": 15.887,
+      "eval_steps_per_second": 1.996,
+      "step": 26000
+    },
+    {
+      "epoch": 2.8263651877133107,
+      "grad_norm": 3.5993499755859375,
+      "learning_rate": 2.8939135381114905e-09,
+      "loss": 0.1591,
+      "step": 26500
+    },
+    {
+      "epoch": 2.8263651877133107,
+      "eval_loss": 0.8679539561271667,
+      "eval_runtime": 62.0886,
+      "eval_samples_per_second": 15.897,
+      "eval_steps_per_second": 1.997,
+      "step": 26500
+    },
+    {
+      "epoch": 2.8796928327645053,
+      "grad_norm": 29.50309944152832,
+      "learning_rate": 2.0051194539249145e-09,
+      "loss": 0.1812,
+      "step": 27000
+    },
+    {
+      "epoch": 2.8796928327645053,
+      "eval_loss": 0.8675134181976318,
+      "eval_runtime": 63.3395,
+      "eval_samples_per_second": 15.583,
+      "eval_steps_per_second": 1.958,
+      "step": 27000
+    },
+    {
+      "epoch": 2.9330204778157,
+      "grad_norm": 1.3827078342437744,
+      "learning_rate": 1.116325369738339e-09,
+      "loss": 0.1408,
+      "step": 27500
+    },
+    {
+      "epoch": 2.9330204778157,
+      "eval_loss": 0.8673344254493713,
+      "eval_runtime": 62.2027,
+      "eval_samples_per_second": 15.867,
+      "eval_steps_per_second": 1.993,
+      "step": 27500
+    },
+    {
+      "epoch": 2.986348122866894,
+      "grad_norm": 399.4076843261719,
+      "learning_rate": 2.2753128555176336e-10,
+      "loss": 0.191,
+      "step": 28000
+    },
+    {
+      "epoch": 2.986348122866894,
+      "eval_loss": 0.8672720789909363,
+      "eval_runtime": 62.0226,
+      "eval_samples_per_second": 15.914,
+      "eval_steps_per_second": 1.999,
+      "step": 28000
+    },
     {
       "epoch": 3.0,
-      "step": 16284,
-      "total_flos": 5859024508620360.0,
-      "train_loss": 0.24355060232561188,
-      "train_runtime": 12410.1332,
-      "train_samples_per_second": 2.624,
-      "train_steps_per_second": 1.312
+      "step": 28128,
+      "total_flos": 1.01200024209438e+16,
+      "train_loss": 0.22731135058321644,
+      "train_runtime": 20704.962,
+      "train_samples_per_second": 2.717,
+      "train_steps_per_second": 1.359
     }
   ],
   "logging_steps": 500,
-  "max_steps": 16284,
+  "max_steps": 28128,
   "num_input_tokens_seen": 0,
   "num_train_epochs": 3,
   "save_steps": 500,
@@ -515,7 +875,7 @@
       "attributes": {}
     }
   },
-  "total_flos": 5859024508620360.0,
+  "total_flos": 1.01200024209438e+16,
   "train_batch_size": 2,
   "trial_name": null,
   "trial_params": null