End of training

README.md

@@ -15,7 +15,7 @@ should probably proofread and complete it, then remove this comment. -->
 
 This model is a fine-tuned version of [openai/clip-vit-large-patch14-336](https://huggingface.co/openai/clip-vit-large-patch14-336) on an unknown dataset.
 It achieves the following results on the evaluation set:
-- Loss: 2.0670
+- Loss: 1.9574
 
 ## Model description
 

all_results.json

@@ -1,12 +1,12 @@
 {
     "epoch": 3.0,
-    "eval_loss": 1.9056034088134766,
-    "eval_runtime": 75.5584,
-    "eval_samples_per_second": 15.961,
-    "eval_steps_per_second": 1.998,
-    "total_flos": 5859564146147250.0,
-    "train_loss": 0.010180538147276201,
-    "train_runtime": 230.6503,
-    "train_samples_per_second": 141.214,
-    "train_steps_per_second": 70.613
+    "eval_loss": 1.9574228525161743,
+    "eval_runtime": 71.0993,
+    "eval_samples_per_second": 16.976,
+    "eval_steps_per_second": 2.124,
+    "total_flos": 5859024508620360.0,
+    "train_loss": 0.6559209774413702,
+    "train_runtime": 12369.044,
+    "train_samples_per_second": 2.633,
+    "train_steps_per_second": 1.317
 }

eval_results.json

@@ -1,7 +1,7 @@
 {
     "epoch": 3.0,
-    "eval_loss": 1.9056034088134766,
-    "eval_runtime": 75.5584,
-    "eval_samples_per_second": 15.961,
-    "eval_steps_per_second": 1.998
+    "eval_loss": 1.9574228525161743,
+    "eval_runtime": 71.0993,
+    "eval_samples_per_second": 16.976,
+    "eval_steps_per_second": 2.124
 }

runs/Aug22_19-57-06_winter-park/events.out.tfevents.1724390757.winter-park.3947668.1

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:3a253c86a2fc241e99683992877eff99d62b84178e495245a25738ad7c6f7f4f
+size 311

train_results.json

@@ -1,8 +1,8 @@
 {
     "epoch": 3.0,
-    "total_flos": 5859564146147250.0,
-    "train_loss": 0.010180538147276201,
-    "train_runtime": 230.6503,
-    "train_samples_per_second": 141.214,
-    "train_steps_per_second": 70.613
+    "total_flos": 5859024508620360.0,
+    "train_loss": 0.6559209774413702,
+    "train_runtime": 12369.044,
+    "train_samples_per_second": 2.633,
+    "train_steps_per_second": 1.317
 }

trainer_state.json

@@ -1,505 +1,505 @@
 {
-  "best_metric": 1.9056034088134766,
-  "best_model_checkpoint": "./output/clip-finetuned-csu-p14-336-e3l55-l/checkpoint-13000",
+  "best_metric": 1.9574228525161743,
+  "best_model_checkpoint": "./output/clip-finetuned-csu-p14-336-e3l55-l/checkpoint-9000",
   "epoch": 3.0,
   "eval_steps": 500,
-  "global_step": 16287,
+  "global_step": 16284,
   "is_hyper_param_search": false,
   "is_local_process_zero": true,
   "is_world_process_zero": true,
   "log_history": [
     {
-      "epoch": 0.09209799226376865,
-      "grad_norm": 0.009597355499863625,
-      "learning_rate": 4.846503346227052e-05,
-      "loss": 0.7336,
+      "epoch": 0.09211495946941783,
+      "grad_norm": 0.030529862269759178,
+      "learning_rate": 4.8464750675509704e-05,
+      "loss": 0.7076,
       "step": 500
     },
     {
-      "epoch": 0.09209799226376865,
-      "eval_loss": 2.0780556201934814,
-      "eval_runtime": 77.1404,
-      "eval_samples_per_second": 15.634,
-      "eval_steps_per_second": 1.957,
+      "epoch": 0.09211495946941783,
+      "eval_loss": 2.07871675491333,
+      "eval_runtime": 73.4415,
+      "eval_samples_per_second": 16.435,
+      "eval_steps_per_second": 2.056,
       "step": 500
     },
     {
-      "epoch": 0.1841959845275373,
-      "grad_norm": 0.10049372166395187,
-      "learning_rate": 4.693006692454105e-05,
-      "loss": 0.6953,
+      "epoch": 0.18422991893883567,
+      "grad_norm": 0.02040918357670307,
+      "learning_rate": 4.6929501351019406e-05,
+      "loss": 0.7045,
       "step": 1000
     },
     {
-      "epoch": 0.1841959845275373,
-      "eval_loss": 2.0782744884490967,
-      "eval_runtime": 77.8965,
-      "eval_samples_per_second": 15.482,
-      "eval_steps_per_second": 1.938,
+      "epoch": 0.18422991893883567,
+      "eval_loss": 2.0783257484436035,
+      "eval_runtime": 74.6654,
+      "eval_samples_per_second": 16.165,
+      "eval_steps_per_second": 2.022,
       "step": 1000
     },
     {
-      "epoch": 0.27629397679130596,
-      "grad_norm": 0.0041180383414030075,
-      "learning_rate": 4.539510038681157e-05,
-      "loss": 0.6953,
+      "epoch": 0.2763448784082535,
+      "grad_norm": 0.8010361194610596,
+      "learning_rate": 4.539425202652911e-05,
+      "loss": 0.7106,
       "step": 1500
     },
     {
-      "epoch": 0.27629397679130596,
-      "eval_loss": 2.077699899673462,
-      "eval_runtime": 78.9586,
-      "eval_samples_per_second": 15.274,
-      "eval_steps_per_second": 1.912,
+      "epoch": 0.2763448784082535,
+      "eval_loss": 2.0473792552948,
+      "eval_runtime": 75.4141,
+      "eval_samples_per_second": 16.005,
+      "eval_steps_per_second": 2.002,
       "step": 1500
     },
     {
-      "epoch": 0.3683919690550746,
-      "grad_norm": 0.29397013783454895,
-      "learning_rate": 4.386013384908209e-05,
-      "loss": 0.6932,
+      "epoch": 0.36845983787767134,
+      "grad_norm": 1.1971429586410522,
+      "learning_rate": 4.3859002702038817e-05,
+      "loss": 0.704,
       "step": 2000
     },
     {
-      "epoch": 0.3683919690550746,
-      "eval_loss": 2.0721774101257324,
-      "eval_runtime": 77.349,
-      "eval_samples_per_second": 15.592,
-      "eval_steps_per_second": 1.952,
+      "epoch": 0.36845983787767134,
+      "eval_loss": 2.281594753265381,
+      "eval_runtime": 76.3969,
+      "eval_samples_per_second": 15.799,
+      "eval_steps_per_second": 1.977,
       "step": 2000
     },
     {
-      "epoch": 0.46048996131884323,
-      "grad_norm": 0.9136261343955994,
-      "learning_rate": 4.232516731135262e-05,
-      "loss": 0.6907,
+      "epoch": 0.46057479734708917,
+      "grad_norm": 1.5381091833114624,
+      "learning_rate": 4.232375337754851e-05,
+      "loss": 0.6996,
       "step": 2500
     },
     {
-      "epoch": 0.46048996131884323,
-      "eval_loss": 2.079211711883545,
-      "eval_runtime": 79.3204,
-      "eval_samples_per_second": 15.204,
-      "eval_steps_per_second": 1.904,
+      "epoch": 0.46057479734708917,
+      "eval_loss": 2.045022964477539,
+      "eval_runtime": 76.3329,
+      "eval_samples_per_second": 15.812,
+      "eval_steps_per_second": 1.978,
       "step": 2500
     },
     {
-      "epoch": 0.5525879535826119,
-      "grad_norm": 4.484213829040527,
-      "learning_rate": 4.079020077362314e-05,
-      "loss": 0.7043,
+      "epoch": 0.552689756816507,
+      "grad_norm": 5.1257524490356445,
+      "learning_rate": 4.078850405305822e-05,
+      "loss": 0.6739,
       "step": 3000
     },
     {
-      "epoch": 0.5525879535826119,
-      "eval_loss": 2.0204577445983887,
-      "eval_runtime": 78.612,
-      "eval_samples_per_second": 15.341,
-      "eval_steps_per_second": 1.921,
+      "epoch": 0.552689756816507,
+      "eval_loss": 2.0969045162200928,
+      "eval_runtime": 76.2547,
+      "eval_samples_per_second": 15.829,
+      "eval_steps_per_second": 1.98,
       "step": 3000
     },
     {
-      "epoch": 0.6446859458463805,
-      "grad_norm": 1.1021015644073486,
-      "learning_rate": 3.9255234235893664e-05,
-      "loss": 0.7078,
+      "epoch": 0.6448047162859248,
+      "grad_norm": 6.156252861022949,
+      "learning_rate": 3.925325472856792e-05,
+      "loss": 0.6687,
       "step": 3500
     },
     {
-      "epoch": 0.6446859458463805,
-      "eval_loss": 2.1256344318389893,
-      "eval_runtime": 78.63,
-      "eval_samples_per_second": 15.338,
-      "eval_steps_per_second": 1.92,
+      "epoch": 0.6448047162859248,
+      "eval_loss": 2.0315287113189697,
+      "eval_runtime": 76.4108,
+      "eval_samples_per_second": 15.796,
+      "eval_steps_per_second": 1.976,
       "step": 3500
     },
     {
-      "epoch": 0.7367839381101492,
-      "grad_norm": 16.864044189453125,
-      "learning_rate": 3.772026769816418e-05,
-      "loss": 0.7034,
+      "epoch": 0.7369196757553427,
+      "grad_norm": 0.053296659141778946,
+      "learning_rate": 3.7718005404077624e-05,
+      "loss": 0.8387,
       "step": 4000
     },
     {
-      "epoch": 0.7367839381101492,
-      "eval_loss": 2.0000314712524414,
-      "eval_runtime": 78.6543,
-      "eval_samples_per_second": 15.333,
-      "eval_steps_per_second": 1.92,
+      "epoch": 0.7369196757553427,
+      "eval_loss": 2.3810112476348877,
+      "eval_runtime": 74.9737,
+      "eval_samples_per_second": 16.099,
+      "eval_steps_per_second": 2.014,
       "step": 4000
     },
     {
-      "epoch": 0.8288819303739179,
-      "grad_norm": 20.071943283081055,
-      "learning_rate": 3.6185301160434705e-05,
-      "loss": 0.6671,
+      "epoch": 0.8290346352247605,
+      "grad_norm": 0.1077703982591629,
+      "learning_rate": 3.6182756079587326e-05,
+      "loss": 0.6988,
       "step": 4500
     },
     {
-      "epoch": 0.8288819303739179,
-      "eval_loss": 2.1607539653778076,
-      "eval_runtime": 78.6404,
-      "eval_samples_per_second": 15.336,
-      "eval_steps_per_second": 1.92,
+      "epoch": 0.8290346352247605,
+      "eval_loss": 2.0409226417541504,
+      "eval_runtime": 76.7206,
+      "eval_samples_per_second": 15.732,
+      "eval_steps_per_second": 1.968,
       "step": 4500
     },
     {
-      "epoch": 0.9209799226376865,
-      "grad_norm": 17.817607879638672,
-      "learning_rate": 3.4650334622705225e-05,
-      "loss": 0.6899,
+      "epoch": 0.9211495946941783,
+      "grad_norm": 4.246299743652344,
+      "learning_rate": 3.4647506755097034e-05,
+      "loss": 0.6808,
       "step": 5000
     },
     {
-      "epoch": 0.9209799226376865,
-      "eval_loss": 2.093959331512451,
-      "eval_runtime": 78.539,
-      "eval_samples_per_second": 15.355,
-      "eval_steps_per_second": 1.923,
+      "epoch": 0.9211495946941783,
+      "eval_loss": 1.9745088815689087,
+      "eval_runtime": 74.928,
+      "eval_samples_per_second": 16.109,
+      "eval_steps_per_second": 2.015,
       "step": 5000
     },
     {
-      "epoch": 1.013077914901455,
-      "grad_norm": 41.0746955871582,
-      "learning_rate": 3.311536808497575e-05,
-      "loss": 0.6649,
+      "epoch": 1.013264554163596,
+      "grad_norm": 2.2185893058776855,
+      "learning_rate": 3.311225743060673e-05,
+      "loss": 0.6584,
       "step": 5500
     },
     {
-      "epoch": 1.013077914901455,
-      "eval_loss": 1.9776026010513306,
-      "eval_runtime": 78.5095,
-      "eval_samples_per_second": 15.361,
-      "eval_steps_per_second": 1.923,
+      "epoch": 1.013264554163596,
+      "eval_loss": 1.9648858308792114,
+      "eval_runtime": 76.0743,
+      "eval_samples_per_second": 15.866,
+      "eval_steps_per_second": 1.985,
       "step": 5500
     },
     {
-      "epoch": 1.1051759071652238,
-      "grad_norm": 35.426021575927734,
-      "learning_rate": 3.158040154724627e-05,
-      "loss": 0.6748,
+      "epoch": 1.105379513633014,
+      "grad_norm": 6.992915153503418,
+      "learning_rate": 3.157700810611644e-05,
+      "loss": 0.6567,
       "step": 6000
     },
     {
-      "epoch": 1.1051759071652238,
-      "eval_loss": 1.9851746559143066,
-      "eval_runtime": 78.5944,
-      "eval_samples_per_second": 15.345,
-      "eval_steps_per_second": 1.921,
+      "epoch": 1.105379513633014,
+      "eval_loss": 2.0568628311157227,
+      "eval_runtime": 76.0502,
+      "eval_samples_per_second": 15.871,
+      "eval_steps_per_second": 1.986,
       "step": 6000
     },
     {
-      "epoch": 1.1972738994289924,
-      "grad_norm": 0.8773899078369141,
-      "learning_rate": 3.0045435009516792e-05,
-      "loss": 0.6456,
+      "epoch": 1.1974944731024317,
+      "grad_norm": 13.946557998657227,
+      "learning_rate": 3.004175878162614e-05,
+      "loss": 0.6542,
       "step": 6500
     },
     {
-      "epoch": 1.1972738994289924,
-      "eval_loss": 1.9981426000595093,
-      "eval_runtime": 78.4741,
-      "eval_samples_per_second": 15.368,
-      "eval_steps_per_second": 1.924,
+      "epoch": 1.1974944731024317,
+      "eval_loss": 1.9788861274719238,
+      "eval_runtime": 75.9196,
+      "eval_samples_per_second": 15.898,
+      "eval_steps_per_second": 1.989,
       "step": 6500
     },
     {
-      "epoch": 1.289371891692761,
-      "grad_norm": 19.7215518951416,
-      "learning_rate": 2.8510468471787316e-05,
-      "loss": 0.6514,
+      "epoch": 1.2896094325718497,
+      "grad_norm": 2.426790237426758,
+      "learning_rate": 2.8506509457135838e-05,
+      "loss": 0.6536,
       "step": 7000
     },
     {
-      "epoch": 1.289371891692761,
-      "eval_loss": 1.9636509418487549,
-      "eval_runtime": 78.4704,
-      "eval_samples_per_second": 15.369,
-      "eval_steps_per_second": 1.924,
+      "epoch": 1.2896094325718497,
+      "eval_loss": 1.9676496982574463,
+      "eval_runtime": 75.155,
+      "eval_samples_per_second": 16.06,
+      "eval_steps_per_second": 2.009,
       "step": 7000
     },
     {
-      "epoch": 1.3814698839565298,
-      "grad_norm": 5.96924352645874,
-      "learning_rate": 2.697550193405784e-05,
-      "loss": 0.6179,
+      "epoch": 1.3817243920412676,
+      "grad_norm": 1.7985050678253174,
+      "learning_rate": 2.6971260132645544e-05,
+      "loss": 0.637,
       "step": 7500
     },
     {
-      "epoch": 1.3814698839565298,
-      "eval_loss": 2.0034029483795166,
-      "eval_runtime": 78.4771,
-      "eval_samples_per_second": 15.368,
-      "eval_steps_per_second": 1.924,
+      "epoch": 1.3817243920412676,
+      "eval_loss": 2.003986358642578,
+      "eval_runtime": 74.9913,
+      "eval_samples_per_second": 16.095,
+      "eval_steps_per_second": 2.014,
       "step": 7500
     },
     {
-      "epoch": 1.4735678762202984,
-      "grad_norm": 1.2145048379898071,
-      "learning_rate": 2.5440535396328363e-05,
-      "loss": 0.6561,
+      "epoch": 1.4738393515106853,
+      "grad_norm": 22.398109436035156,
+      "learning_rate": 2.543601080815525e-05,
+      "loss": 0.6409,
       "step": 8000
     },
     {
-      "epoch": 1.4735678762202984,
-      "eval_loss": 2.2864511013031006,
-      "eval_runtime": 78.3606,
-      "eval_samples_per_second": 15.39,
-      "eval_steps_per_second": 1.927,
+      "epoch": 1.4738393515106853,
+      "eval_loss": 1.9904581308364868,
+      "eval_runtime": 74.9317,
+      "eval_samples_per_second": 16.108,
+      "eval_steps_per_second": 2.015,
       "step": 8000
     },
     {
-      "epoch": 1.565665868484067,
-      "grad_norm": 15.816754341125488,
-      "learning_rate": 2.3905568858598883e-05,
-      "loss": 0.6328,
+      "epoch": 1.565954310980103,
+      "grad_norm": 4.255030155181885,
+      "learning_rate": 2.390076148366495e-05,
+      "loss": 0.6127,
       "step": 8500
     },
     {
-      "epoch": 1.565665868484067,
-      "eval_loss": 2.4807703495025635,
-      "eval_runtime": 77.0504,
-      "eval_samples_per_second": 15.652,
-      "eval_steps_per_second": 1.96,
+      "epoch": 1.565954310980103,
+      "eval_loss": 2.2638871669769287,
+      "eval_runtime": 74.9634,
+      "eval_samples_per_second": 16.101,
+      "eval_steps_per_second": 2.014,
       "step": 8500
     },
     {
-      "epoch": 1.6577638607478358,
-      "grad_norm": 18.753801345825195,
-      "learning_rate": 2.2370602320869404e-05,
-      "loss": 0.7053,
+      "epoch": 1.658069270449521,
+      "grad_norm": 11.839683532714844,
+      "learning_rate": 2.2365512159174652e-05,
+      "loss": 0.6564,
       "step": 9000
     },
     {
-      "epoch": 1.6577638607478358,
-      "eval_loss": 2.1609902381896973,
-      "eval_runtime": 76.5781,
-      "eval_samples_per_second": 15.749,
-      "eval_steps_per_second": 1.972,
+      "epoch": 1.658069270449521,
+      "eval_loss": 1.9574228525161743,
+      "eval_runtime": 74.8273,
+      "eval_samples_per_second": 16.13,
+      "eval_steps_per_second": 2.018,
       "step": 9000
     },
     {
-      "epoch": 1.7498618530116044,
-      "grad_norm": 3.3683111667633057,
-      "learning_rate": 2.0835635783139927e-05,
-      "loss": 0.6533,
+      "epoch": 1.750184229918939,
+      "grad_norm": 7.12531042098999,
+      "learning_rate": 2.0830262834684354e-05,
+      "loss": 0.6813,
       "step": 9500
     },
     {
-      "epoch": 1.7498618530116044,
-      "eval_loss": 1.9845130443572998,
-      "eval_runtime": 78.8415,
-      "eval_samples_per_second": 15.297,
-      "eval_steps_per_second": 1.915,
+      "epoch": 1.750184229918939,
+      "eval_loss": 2.036475896835327,
+      "eval_runtime": 75.9457,
+      "eval_samples_per_second": 15.893,
+      "eval_steps_per_second": 1.988,
       "step": 9500
     },
     {
-      "epoch": 1.841959845275373,
-      "grad_norm": 5.0674967765808105,
-      "learning_rate": 1.930066924541045e-05,
-      "loss": 0.6594,
+      "epoch": 1.8422991893883567,
+      "grad_norm": 2.8198225498199463,
+      "learning_rate": 1.9295013510194056e-05,
+      "loss": 0.6421,
       "step": 10000
     },
     {
-      "epoch": 1.841959845275373,
-      "eval_loss": 1.9894555807113647,
-      "eval_runtime": 77.1087,
-      "eval_samples_per_second": 15.64,
-      "eval_steps_per_second": 1.958,
+      "epoch": 1.8422991893883567,
+      "eval_loss": 2.2431833744049072,
+      "eval_runtime": 75.2519,
+      "eval_samples_per_second": 16.039,
+      "eval_steps_per_second": 2.007,
       "step": 10000
     },
     {
-      "epoch": 1.9340578375391417,
-      "grad_norm": 45.641239166259766,
-      "learning_rate": 1.7765702707680974e-05,
-      "loss": 0.6597,
+      "epoch": 1.9344141488577744,
+      "grad_norm": 23.107166290283203,
+      "learning_rate": 1.7759764185703758e-05,
+      "loss": 0.6531,
       "step": 10500
     },
     {
-      "epoch": 1.9340578375391417,
-      "eval_loss": 2.0103421211242676,
-      "eval_runtime": 77.0894,
-      "eval_samples_per_second": 15.644,
-      "eval_steps_per_second": 1.959,
+      "epoch": 1.9344141488577744,
+      "eval_loss": 1.9985452890396118,
+      "eval_runtime": 74.9704,
+      "eval_samples_per_second": 16.1,
+      "eval_steps_per_second": 2.014,
       "step": 10500
     },
     {
-      "epoch": 2.02615582980291,
-      "grad_norm": 7.026269435882568,
-      "learning_rate": 1.6230736169951498e-05,
-      "loss": 0.6435,
+      "epoch": 2.026529108327192,
+      "grad_norm": 4.0393805503845215,
+      "learning_rate": 1.622451486121346e-05,
+      "loss": 0.6388,
       "step": 11000
     },
     {
-      "epoch": 2.02615582980291,
-      "eval_loss": 2.058932304382324,
-      "eval_runtime": 78.5759,
-      "eval_samples_per_second": 15.348,
-      "eval_steps_per_second": 1.922,
+      "epoch": 2.026529108327192,
+      "eval_loss": 1.9725395441055298,
+      "eval_runtime": 74.9174,
+      "eval_samples_per_second": 16.111,
+      "eval_steps_per_second": 2.016,
       "step": 11000
     },
     {
-      "epoch": 2.118253822066679,
-      "grad_norm": 11.815095901489258,
-      "learning_rate": 1.4695769632222017e-05,
-      "loss": 0.6404,
+      "epoch": 2.1186440677966103,
+      "grad_norm": 21.68863868713379,
+      "learning_rate": 1.4689265536723165e-05,
+      "loss": 0.6394,
       "step": 11500
     },
     {
-      "epoch": 2.118253822066679,
-      "eval_loss": 2.0517008304595947,
-      "eval_runtime": 76.9172,
-      "eval_samples_per_second": 15.679,
-      "eval_steps_per_second": 1.963,
+      "epoch": 2.1186440677966103,
+      "eval_loss": 2.0170812606811523,
+      "eval_runtime": 75.0053,
+      "eval_samples_per_second": 16.092,
+      "eval_steps_per_second": 2.013,
       "step": 11500
     },
     {
-      "epoch": 2.2103518143304477,
-      "grad_norm": 3.2536027431488037,
-      "learning_rate": 1.316080309449254e-05,
-      "loss": 0.6348,
+      "epoch": 2.210759027266028,
+      "grad_norm": 1.8485018014907837,
+      "learning_rate": 1.3154016212232867e-05,
+      "loss": 0.6151,
       "step": 12000
     },
     {
-      "epoch": 2.2103518143304477,
-      "eval_loss": 1.9749338626861572,
-      "eval_runtime": 78.0449,
-      "eval_samples_per_second": 15.453,
-      "eval_steps_per_second": 1.935,
+      "epoch": 2.210759027266028,
+      "eval_loss": 1.9586845636367798,
+      "eval_runtime": 74.7694,
+      "eval_samples_per_second": 16.143,
+      "eval_steps_per_second": 2.02,
       "step": 12000
     },
     {
-      "epoch": 2.302449806594216,
-      "grad_norm": 12.233312606811523,
-      "learning_rate": 1.1625836556763064e-05,
-      "loss": 0.6545,
+      "epoch": 2.3028739867354457,
+      "grad_norm": 14.168105125427246,
+      "learning_rate": 1.161876688774257e-05,
+      "loss": 0.5947,
       "step": 12500
     },
     {
-      "epoch": 2.302449806594216,
-      "eval_loss": 1.9680836200714111,
-      "eval_runtime": 76.9869,
-      "eval_samples_per_second": 15.665,
-      "eval_steps_per_second": 1.961,
+      "epoch": 2.3028739867354457,
+      "eval_loss": 2.0683774948120117,
+      "eval_runtime": 74.9327,
+      "eval_samples_per_second": 16.108,
+      "eval_steps_per_second": 2.015,
       "step": 12500
     },
     {
-      "epoch": 2.394547798857985,
-      "grad_norm": 16.69487953186035,
-      "learning_rate": 1.0090870019033586e-05,
-      "loss": 0.6323,
+      "epoch": 2.3949889462048635,
+      "grad_norm": 1.8051279783248901,
+      "learning_rate": 1.0083517563252274e-05,
+      "loss": 0.6166,
       "step": 13000
     },
     {
-      "epoch": 2.394547798857985,
-      "eval_loss": 1.9056034088134766,
-      "eval_runtime": 78.3412,
-      "eval_samples_per_second": 15.394,
-      "eval_steps_per_second": 1.927,
+      "epoch": 2.3949889462048635,
+      "eval_loss": 2.05798602104187,
+      "eval_runtime": 74.9282,
+      "eval_samples_per_second": 16.109,
+      "eval_steps_per_second": 2.015,
       "step": 13000
     },
     {
-      "epoch": 2.4866457911217537,
-      "grad_norm": 11.504305839538574,
-      "learning_rate": 8.555903481304108e-06,
-      "loss": 0.6045,
+      "epoch": 2.4871039056742816,
+      "grad_norm": 23.26294708251953,
+      "learning_rate": 8.548268238761976e-06,
+      "loss": 0.6382,
       "step": 13500
     },
     {
-      "epoch": 2.4866457911217537,
-      "eval_loss": 1.9470211267471313,
-      "eval_runtime": 77.1033,
-      "eval_samples_per_second": 15.641,
-      "eval_steps_per_second": 1.958,
+      "epoch": 2.4871039056742816,
+      "eval_loss": 2.020942211151123,
+      "eval_runtime": 74.9346,
+      "eval_samples_per_second": 16.107,
+      "eval_steps_per_second": 2.015,
       "step": 13500
     },
     {
-      "epoch": 2.578743783385522,
-      "grad_norm": 188.8672332763672,
-      "learning_rate": 7.0209369435746295e-06,
-      "loss": 0.6492,
+      "epoch": 2.5792188651436994,
+      "grad_norm": 14.594226837158203,
+      "learning_rate": 7.0130189142716784e-06,
+      "loss": 0.5949,
       "step": 14000
     },
     {
-      "epoch": 2.578743783385522,
-      "eval_loss": 2.017714023590088,
-      "eval_runtime": 77.6236,
-      "eval_samples_per_second": 15.537,
-      "eval_steps_per_second": 1.945,
+      "epoch": 2.5792188651436994,
+      "eval_loss": 2.0775811672210693,
+      "eval_runtime": 74.871,
+      "eval_samples_per_second": 16.121,
+      "eval_steps_per_second": 2.017,
       "step": 14000
     },
     {
-      "epoch": 2.670841775649291,
-      "grad_norm": 10.520608901977539,
-      "learning_rate": 5.485970405845152e-06,
-      "loss": 0.6512,
+      "epoch": 2.671333824613117,
+      "grad_norm": 6.04954719543457,
+      "learning_rate": 5.47776958978138e-06,
+      "loss": 0.6282,
       "step": 14500
     },
     {
-      "epoch": 2.670841775649291,
-      "eval_loss": 1.948345422744751,
-      "eval_runtime": 77.8588,
-      "eval_samples_per_second": 15.49,
-      "eval_steps_per_second": 1.939,
+      "epoch": 2.671333824613117,
+      "eval_loss": 2.072066307067871,
+      "eval_runtime": 74.9104,
+      "eval_samples_per_second": 16.113,
+      "eval_steps_per_second": 2.016,
       "step": 14500
     },
     {
-      "epoch": 2.7629397679130596,
-      "grad_norm": 13.767060279846191,
-      "learning_rate": 3.951003868115676e-06,
-      "loss": 0.6382,
+      "epoch": 2.7634487840825352,
+      "grad_norm": 1.6737792491912842,
+      "learning_rate": 3.942520265291084e-06,
+      "loss": 0.6057,
       "step": 15000
     },
     {
-      "epoch": 2.7629397679130596,
-      "eval_loss": 1.9499462842941284,
-      "eval_runtime": 77.4986,
-      "eval_samples_per_second": 15.562,
-      "eval_steps_per_second": 1.948,
+      "epoch": 2.7634487840825352,
+      "eval_loss": 2.0812695026397705,
+      "eval_runtime": 74.9464,
+      "eval_samples_per_second": 16.105,
+      "eval_steps_per_second": 2.015,
       "step": 15000
     },
     {
-      "epoch": 2.8550377601768284,
-      "grad_norm": 0.9124027490615845,
-      "learning_rate": 2.416037330386198e-06,
-      "loss": 0.6203,
+      "epoch": 2.855563743551953,
+      "grad_norm": 13.746774673461914,
+      "learning_rate": 2.407270940800786e-06,
+      "loss": 0.6229,
       "step": 15500
     },
     {
-      "epoch": 2.8550377601768284,
-      "eval_loss": 1.9528756141662598,
-      "eval_runtime": 77.9576,
-      "eval_samples_per_second": 15.47,
-      "eval_steps_per_second": 1.937,
+      "epoch": 2.855563743551953,
+      "eval_loss": 2.0467605590820312,
+      "eval_runtime": 74.8136,
+      "eval_samples_per_second": 16.133,
+      "eval_steps_per_second": 2.018,
       "step": 15500
     },
     {
-      "epoch": 2.947135752440597,
-      "grad_norm": 8.062443733215332,
-      "learning_rate": 8.810707926567202e-07,
-      "loss": 0.6241,
+      "epoch": 2.9476787030213707,
+      "grad_norm": 11.62869644165039,
+      "learning_rate": 8.720216163104889e-07,
+      "loss": 0.612,
       "step": 16000
     },
     {
-      "epoch": 2.947135752440597,
-      "eval_loss": 1.981693148612976,
-      "eval_runtime": 77.6267,
-      "eval_samples_per_second": 15.536,
-      "eval_steps_per_second": 1.945,
+      "epoch": 2.9476787030213707,
+      "eval_loss": 2.0670080184936523,
+      "eval_runtime": 74.9975,
+      "eval_samples_per_second": 16.094,
+      "eval_steps_per_second": 2.013,
       "step": 16000
     },
     {
       "epoch": 3.0,
-      "step": 16287,
-      "total_flos": 5859564146147250.0,
-      "train_loss": 0.010180538147276201,
-      "train_runtime": 230.6503,
-      "train_samples_per_second": 141.214,
-      "train_steps_per_second": 70.613
+      "step": 16284,
+      "total_flos": 5859024508620360.0,
+      "train_loss": 0.6559209774413702,
+      "train_runtime": 12369.044,
+      "train_samples_per_second": 2.633,
+      "train_steps_per_second": 1.317
     }
   ],
   "logging_steps": 500,
-  "max_steps": 16287,
+  "max_steps": 16284,
   "num_input_tokens_seen": 0,
   "num_train_epochs": 3,
   "save_steps": 500,
@@ -515,7 +515,7 @@
       "attributes": {}
     }
   },
-  "total_flos": 5859564146147250.0,
+  "total_flos": 5859024508620360.0,
   "train_batch_size": 2,
   "trial_name": null,
   "trial_params": null