Model save

README.md

@@ -32,7 +32,7 @@ More information needed
 ### Training hyperparameters
 
 The following hyperparameters were used during training:
-- learning_rate: 5e-08
+- learning_rate: 5e-07
 - train_batch_size: 8
 - eval_batch_size: 8
 - seed: 42

all_results.json

@@ -1,8 +1,8 @@
 {
     "epoch": 1.0,
-    "train_loss": 0.5437000210468586,
-    "train_runtime": 4622.4032,
-    "train_samples": 39942,
-    "train_samples_per_second": 8.641,
+    "train_loss": 0.5533875935638006,
+    "train_runtime": 12702.5109,
+    "train_samples": 111132,
+    "train_samples_per_second": 8.749,
     "train_steps_per_second": 0.034
 }

model-00001-of-00003.safetensors

@@ -1,3 +1,3 @@
 version https://git-lfs.github.com/spec/v1
-oid sha256:0f1145c8df98da5d886c710139d12133f030eeb6eac516ce616aa261c1a4af95
+oid sha256:bcec15c11061f01145410326e83e22e9e119675ae40a0022f9d01119851e5b43
 size 4943178720

model-00002-of-00003.safetensors

@@ -1,3 +1,3 @@
 version https://git-lfs.github.com/spec/v1
-oid sha256:dadcd0d623ac649e14077272450402fe741a935010b44760a1cf23f14bf4046a
+oid sha256:3388da565e73d0a53d42775f84d6f31e380786fc86693f449f99a6a5c5ee5e87
 size 4999819336

model-00003-of-00003.safetensors

@@ -1,3 +1,3 @@
 version https://git-lfs.github.com/spec/v1
-oid sha256:731c8da9d0797c68151c2f5e91ca88680335e53ef3600a69e45056d95a511ebc
+oid sha256:6b5c26a3b832e0dc2dd5eb31988b1522bdb63734b704f3b28700dc261401dfa0
 size 4540532728

runs/Jun06_13-29-12_n136-129-074/events.out.tfevents.1717652376.n136-129-074.1035360.0

@@ -1,3 +1,3 @@
 version https://git-lfs.github.com/spec/v1
-oid sha256:5873eeb4e9ce1908b23071b4817dca301c089338fd4b3e1ad524b739fe07813b
-size 32918
+oid sha256:93fed13b8d0407afd27cc592dea0936b76b9d64f86cb9d3e4b7bbc08da040aa7
+size 35336

train_results.json

@@ -1,8 +1,8 @@
 {
     "epoch": 1.0,
-    "train_loss": 0.5437000210468586,
-    "train_runtime": 4622.4032,
-    "train_samples": 39942,
-    "train_samples_per_second": 8.641,
+    "train_loss": 0.5533875935638006,
+    "train_runtime": 12702.5109,
+    "train_samples": 111132,
+    "train_samples_per_second": 8.749,
     "train_steps_per_second": 0.034
 }

trainer_state.json

@@ -1,22 +1,22 @@
 {
   "best_metric": null,
   "best_model_checkpoint": null,
-  "epoch": 0.9984,
+  "epoch": 0.9994242947610823,
   "eval_steps": 500,
-  "global_step": 156,
+  "global_step": 434,
   "is_hyper_param_search": false,
   "is_local_process_zero": true,
   "is_world_process_zero": true,
   "log_history": [
     {
-      "epoch": 0.01,
-      "grad_norm": 1061.4292758130512,
-      "learning_rate": 3.125e-09,
-      "logits/chosen": -3.9499800205230713,
-      "logits/rejected": -4.237819194793701,
-      "logps/chosen": -300.693115234375,
-      "logps/rejected": -249.96307373046875,
-      "loss": 0.6941,
+      "epoch": 0.0,
+      "grad_norm": 41.760020423030454,
+      "learning_rate": 1.1363636363636363e-08,
+      "logits/chosen": -4.852883815765381,
+      "logits/rejected": -5.038845539093018,
+      "logps/chosen": -266.87384033203125,
+      "logps/rejected": -341.158935546875,
+      "loss": 0.6931,
       "rewards/accuracies": 0.0,
       "rewards/chosen": 0.0,
       "rewards/margins": 0.0,
@@ -24,242 +24,662 @@
       "step": 1
     },
     {
-      "epoch": 0.06,
-      "grad_norm": 1100.0352396547798,
-      "learning_rate": 3.125e-08,
-      "logits/chosen": -4.129236221313477,
-      "logits/rejected": -4.351477146148682,
-      "logps/chosen": -351.48150634765625,
-      "logps/rejected": -308.8130187988281,
-      "loss": 0.7282,
-      "rewards/accuracies": 0.3611111044883728,
-      "rewards/chosen": -0.027542730793356895,
-      "rewards/margins": -0.10096076130867004,
-      "rewards/rejected": 0.073418028652668,
+      "epoch": 0.02,
+      "grad_norm": 35.2404459189821,
+      "learning_rate": 1.1363636363636363e-07,
+      "logits/chosen": -4.649970054626465,
+      "logits/rejected": -4.861534118652344,
+      "logps/chosen": -329.62591552734375,
+      "logps/rejected": -358.3674621582031,
+      "loss": 0.6926,
+      "rewards/accuracies": 0.4236111044883728,
+      "rewards/chosen": -0.003552212379872799,
+      "rewards/margins": 0.00014615118561778218,
+      "rewards/rejected": -0.0036983639001846313,
       "step": 10
     },
     {
-      "epoch": 0.13,
-      "grad_norm": 1023.2979189429582,
-      "learning_rate": 4.9899357349880975e-08,
-      "logits/chosen": -4.195385932922363,
-      "logits/rejected": -4.382458686828613,
-      "logps/chosen": -334.9761962890625,
-      "logps/rejected": -293.8690185546875,
-      "loss": 0.6994,
-      "rewards/accuracies": 0.5625,
-      "rewards/chosen": 0.18794658780097961,
-      "rewards/margins": 0.09464438259601593,
-      "rewards/rejected": 0.09330219030380249,
+      "epoch": 0.05,
+      "grad_norm": 38.80276500420471,
+      "learning_rate": 2.2727272727272726e-07,
+      "logits/chosen": -4.702324867248535,
+      "logits/rejected": -4.93120813369751,
+      "logps/chosen": -323.15277099609375,
+      "logps/rejected": -350.1238708496094,
+      "loss": 0.6758,
+      "rewards/accuracies": 0.609375,
+      "rewards/chosen": -0.09294315427541733,
+      "rewards/margins": 0.03752884641289711,
+      "rewards/rejected": -0.13047200441360474,
       "step": 20
     },
     {
-      "epoch": 0.19,
-      "grad_norm": 716.4790257172086,
-      "learning_rate": 4.877641290737884e-08,
-      "logits/chosen": -4.231568813323975,
-      "logits/rejected": -4.364924430847168,
-      "logps/chosen": -327.92413330078125,
-      "logps/rejected": -295.37176513671875,
-      "loss": 0.6055,
-      "rewards/accuracies": 0.684374988079071,
-      "rewards/chosen": 0.8660896420478821,
-      "rewards/margins": 0.425116628408432,
-      "rewards/rejected": 0.44097304344177246,
+      "epoch": 0.07,
+      "grad_norm": 44.150425429765605,
+      "learning_rate": 3.4090909090909085e-07,
+      "logits/chosen": -4.943687915802002,
+      "logits/rejected": -5.059121608734131,
+      "logps/chosen": -358.85186767578125,
+      "logps/rejected": -428.28997802734375,
+      "loss": 0.6356,
+      "rewards/accuracies": 0.6312500238418579,
+      "rewards/chosen": -0.5626996159553528,
+      "rewards/margins": 0.20829734206199646,
+      "rewards/rejected": -0.7709969282150269,
       "step": 30
     },
     {
-      "epoch": 0.26,
-      "grad_norm": 820.682467230606,
-      "learning_rate": 4.646121984004665e-08,
-      "logits/chosen": -4.1439104080200195,
-      "logits/rejected": -4.345718860626221,
-      "logps/chosen": -330.7722473144531,
-      "logps/rejected": -288.60272216796875,
-      "loss": 0.5304,
-      "rewards/accuracies": 0.831250011920929,
-      "rewards/chosen": 1.6021575927734375,
-      "rewards/margins": 0.956769585609436,
-      "rewards/rejected": 0.6453880071640015,
+      "epoch": 0.09,
+      "grad_norm": 40.317859636548626,
+      "learning_rate": 4.545454545454545e-07,
+      "logits/chosen": -4.930764198303223,
+      "logits/rejected": -5.059735298156738,
+      "logps/chosen": -406.4983825683594,
+      "logps/rejected": -480.08984375,
+      "loss": 0.6181,
+      "rewards/accuracies": 0.640625,
+      "rewards/chosen": -0.804176926612854,
+      "rewards/margins": 0.33279144763946533,
+      "rewards/rejected": -1.1369682550430298,
       "step": 40
     },
     {
-      "epoch": 0.32,
-      "grad_norm": 587.6344375884266,
-      "learning_rate": 4.3069871595684784e-08,
-      "logits/chosen": -4.235655784606934,
-      "logits/rejected": -4.4143476486206055,
-      "logps/chosen": -330.8318786621094,
-      "logps/rejected": -291.75067138671875,
-      "loss": 0.525,
-      "rewards/accuracies": 0.8125,
-      "rewards/chosen": 2.010468006134033,
-      "rewards/margins": 1.2047243118286133,
-      "rewards/rejected": 0.8057435750961304,
+      "epoch": 0.12,
+      "grad_norm": 41.480329345920765,
+      "learning_rate": 4.997080567080816e-07,
+      "logits/chosen": -5.149637222290039,
+      "logits/rejected": -5.36204719543457,
+      "logps/chosen": -328.46356201171875,
+      "logps/rejected": -432.31854248046875,
+      "loss": 0.5989,
+      "rewards/accuracies": 0.7093750238418579,
+      "rewards/chosen": -0.586052417755127,
+      "rewards/margins": 0.4240691661834717,
+      "rewards/rejected": -1.0101215839385986,
       "step": 50
     },
     {
-      "epoch": 0.38,
-      "grad_norm": 586.1895231882175,
-      "learning_rate": 3.8772424536302564e-08,
-      "logits/chosen": -4.247714519500732,
-      "logits/rejected": -4.417937278747559,
-      "logps/chosen": -322.380126953125,
-      "logps/rejected": -291.76654052734375,
-      "loss": 0.517,
-      "rewards/accuracies": 0.8218749761581421,
-      "rewards/chosen": 2.19193696975708,
-      "rewards/margins": 1.311265230178833,
-      "rewards/rejected": 0.8806716203689575,
+      "epoch": 0.14,
+      "grad_norm": 45.66439552100321,
+      "learning_rate": 4.979264274553905e-07,
+      "logits/chosen": -5.387632846832275,
+      "logits/rejected": -5.600171089172363,
+      "logps/chosen": -370.166259765625,
+      "logps/rejected": -468.5503845214844,
+      "loss": 0.5899,
+      "rewards/accuracies": 0.7124999761581421,
+      "rewards/chosen": -0.6436477899551392,
+      "rewards/margins": 0.5441282987594604,
+      "rewards/rejected": -1.1877760887145996,
       "step": 60
     },
     {
-      "epoch": 0.45,
-      "grad_norm": 528.8947001003231,
-      "learning_rate": 3.378437060203357e-08,
-      "logits/chosen": -4.170197486877441,
-      "logits/rejected": -4.359009742736816,
-      "logps/chosen": -322.4336242675781,
-      "logps/rejected": -289.197998046875,
-      "loss": 0.5096,
-      "rewards/accuracies": 0.815625011920929,
-      "rewards/chosen": 2.203657865524292,
-      "rewards/margins": 1.3729288578033447,
-      "rewards/rejected": 0.8307291269302368,
+      "epoch": 0.16,
+      "grad_norm": 58.36091247583923,
+      "learning_rate": 4.945369001834514e-07,
+      "logits/chosen": -5.68746280670166,
+      "logits/rejected": -5.95988655090332,
+      "logps/chosen": -392.7463684082031,
+      "logps/rejected": -469.6172790527344,
+      "loss": 0.582,
+      "rewards/accuracies": 0.6656249761581421,
+      "rewards/chosen": -0.8184071779251099,
+      "rewards/margins": 0.44322291016578674,
+      "rewards/rejected": -1.2616300582885742,
       "step": 70
     },
     {
-      "epoch": 0.51,
-      "grad_norm": 544.8408089721645,
-      "learning_rate": 2.8355831645441387e-08,
-      "logits/chosen": -4.038945198059082,
-      "logits/rejected": -4.326292991638184,
-      "logps/chosen": -348.65472412109375,
-      "logps/rejected": -308.1936950683594,
-      "loss": 0.5115,
-      "rewards/accuracies": 0.875,
-      "rewards/chosen": 2.3145015239715576,
-      "rewards/margins": 1.5040271282196045,
-      "rewards/rejected": 0.8104745149612427,
+      "epoch": 0.18,
+      "grad_norm": 44.87515888678066,
+      "learning_rate": 4.895614572772916e-07,
+      "logits/chosen": -5.767647743225098,
+      "logits/rejected": -6.016678810119629,
+      "logps/chosen": -372.30438232421875,
+      "logps/rejected": -485.02850341796875,
+      "loss": 0.5658,
+      "rewards/accuracies": 0.6656249761581421,
+      "rewards/chosen": -0.6735275983810425,
+      "rewards/margins": 0.523722767829895,
+      "rewards/rejected": -1.1972506046295166,
       "step": 80
     },
     {
-      "epoch": 0.58,
-      "grad_norm": 526.2097671297138,
-      "learning_rate": 2.2759017277414164e-08,
-      "logits/chosen": -4.170694351196289,
-      "logits/rejected": -4.379502296447754,
-      "logps/chosen": -335.43402099609375,
-      "logps/rejected": -295.8087463378906,
-      "loss": 0.5066,
-      "rewards/accuracies": 0.859375,
-      "rewards/chosen": 1.9496158361434937,
-      "rewards/margins": 1.3254307508468628,
-      "rewards/rejected": 0.6241849660873413,
+      "epoch": 0.21,
+      "grad_norm": 52.56006964617041,
+      "learning_rate": 4.830323663933919e-07,
+      "logits/chosen": -5.834498405456543,
+      "logits/rejected": -6.01763916015625,
+      "logps/chosen": -381.7452392578125,
+      "logps/rejected": -476.9891662597656,
+      "loss": 0.5659,
+      "rewards/accuracies": 0.6656249761581421,
+      "rewards/chosen": -0.829800009727478,
+      "rewards/margins": 0.5006512403488159,
+      "rewards/rejected": -1.3304513692855835,
       "step": 90
     },
     {
-      "epoch": 0.64,
-      "grad_norm": 595.4206000749634,
-      "learning_rate": 1.7274575140626317e-08,
-      "logits/chosen": -4.159605979919434,
-      "logits/rejected": -4.377494812011719,
-      "logps/chosen": -333.1622314453125,
-      "logps/rejected": -286.3451232910156,
-      "loss": 0.5142,
-      "rewards/accuracies": 0.871874988079071,
-      "rewards/chosen": 2.0654988288879395,
-      "rewards/margins": 1.4395055770874023,
-      "rewards/rejected": 0.6259931921958923,
+      "epoch": 0.23,
+      "grad_norm": 55.69644328341761,
+      "learning_rate": 4.74991971191553e-07,
+      "logits/chosen": -5.832143306732178,
+      "logits/rejected": -6.132119655609131,
+      "logps/chosen": -373.9161682128906,
+      "logps/rejected": -479.5479431152344,
+      "loss": 0.5632,
+      "rewards/accuracies": 0.6781250238418579,
+      "rewards/chosen": -0.693705677986145,
+      "rewards/margins": 0.6272264719009399,
+      "rewards/rejected": -1.320932149887085,
       "step": 100
     },
     {
-      "epoch": 0.7,
-      "grad_norm": 528.6243658840883,
-      "learning_rate": 1.217751806485235e-08,
-      "logits/chosen": -4.138543128967285,
-      "logits/rejected": -4.379323959350586,
-      "logps/chosen": -314.89556884765625,
-      "logps/rejected": -277.0162658691406,
-      "loss": 0.5071,
-      "rewards/accuracies": 0.840624988079071,
-      "rewards/chosen": 2.0163025856018066,
-      "rewards/margins": 1.3896347284317017,
-      "rewards/rejected": 0.6266676783561707,
+      "epoch": 0.25,
+      "grad_norm": 55.95067399887189,
+      "learning_rate": 4.6549241672001225e-07,
+      "logits/chosen": -6.0772786140441895,
+      "logits/rejected": -6.3634233474731445,
+      "logps/chosen": -431.4512634277344,
+      "logps/rejected": -542.3984985351562,
+      "loss": 0.5495,
+      "rewards/accuracies": 0.706250011920929,
+      "rewards/chosen": -1.0600697994232178,
+      "rewards/margins": 0.6067067384719849,
+      "rewards/rejected": -1.6667766571044922,
       "step": 110
     },
     {
-      "epoch": 0.77,
-      "grad_norm": 557.9046270518194,
-      "learning_rate": 7.723433775328384e-09,
-      "logits/chosen": -4.136676788330078,
-      "logits/rejected": -4.345128536224365,
-      "logps/chosen": -328.94482421875,
-      "logps/rejected": -281.4302673339844,
-      "loss": 0.5052,
-      "rewards/accuracies": 0.831250011920929,
-      "rewards/chosen": 1.9894015789031982,
-      "rewards/margins": 1.363441824913025,
-      "rewards/rejected": 0.625959575176239,
+      "epoch": 0.28,
+      "grad_norm": 48.05186252180738,
+      "learning_rate": 4.5459531123479673e-07,
+      "logits/chosen": -6.061857223510742,
+      "logits/rejected": -6.299743175506592,
+      "logps/chosen": -365.68341064453125,
+      "logps/rejected": -485.478271484375,
+      "loss": 0.5609,
+      "rewards/accuracies": 0.6968749761581421,
+      "rewards/chosen": -0.7879248261451721,
+      "rewards/margins": 0.5600650906562805,
+      "rewards/rejected": -1.347989797592163,
       "step": 120
     },
     {
-      "epoch": 0.83,
-      "grad_norm": 523.8022952711281,
-      "learning_rate": 4.135668656967433e-09,
-      "logits/chosen": -4.222664833068848,
-      "logits/rejected": -4.383507251739502,
-      "logps/chosen": -334.4266357421875,
-      "logps/rejected": -287.63067626953125,
-      "loss": 0.5059,
-      "rewards/accuracies": 0.8687499761581421,
-      "rewards/chosen": 2.026325225830078,
-      "rewards/margins": 1.3963022232055664,
-      "rewards/rejected": 0.6300228834152222,
+      "epoch": 0.3,
+      "grad_norm": 56.849323822551575,
+      "learning_rate": 4.4237132664654147e-07,
+      "logits/chosen": -6.179199695587158,
+      "logits/rejected": -6.447500705718994,
+      "logps/chosen": -373.60626220703125,
+      "logps/rejected": -507.93450927734375,
+      "loss": 0.552,
+      "rewards/accuracies": 0.675000011920929,
+      "rewards/chosen": -0.806027889251709,
+      "rewards/margins": 0.6611197590827942,
+      "rewards/rejected": -1.467147707939148,
       "step": 130
     },
     {
-      "epoch": 0.9,
-      "grad_norm": 601.0204054485596,
-      "learning_rate": 1.5941282340065698e-09,
-      "logits/chosen": -4.176892280578613,
-      "logits/rejected": -4.3910746574401855,
-      "logps/chosen": -335.8815612792969,
-      "logps/rejected": -304.35723876953125,
-      "loss": 0.5101,
-      "rewards/accuracies": 0.831250011920929,
-      "rewards/chosen": 1.9034252166748047,
-      "rewards/margins": 1.3076425790786743,
-      "rewards/rejected": 0.5957827568054199,
+      "epoch": 0.32,
+      "grad_norm": 51.72316001946749,
+      "learning_rate": 4.2889974018603024e-07,
+      "logits/chosen": -6.394374847412109,
+      "logits/rejected": -6.6608686447143555,
+      "logps/chosen": -405.6766662597656,
+      "logps/rejected": -509.36065673828125,
+      "loss": 0.5535,
+      "rewards/accuracies": 0.6937500238418579,
+      "rewards/chosen": -0.810355544090271,
+      "rewards/margins": 0.6747696399688721,
+      "rewards/rejected": -1.4851253032684326,
       "step": 140
     },
     {
-      "epoch": 0.96,
-      "grad_norm": 496.1092972793449,
-      "learning_rate": 2.262559558016325e-10,
-      "logits/chosen": -4.114329814910889,
-      "logits/rejected": -4.342306137084961,
-      "logps/chosen": -342.4503479003906,
-      "logps/rejected": -296.07086181640625,
-      "loss": 0.506,
-      "rewards/accuracies": 0.840624988079071,
-      "rewards/chosen": 2.0278801918029785,
-      "rewards/margins": 1.378021001815796,
-      "rewards/rejected": 0.6498591303825378,
+      "epoch": 0.35,
+      "grad_norm": 56.109386147317295,
+      "learning_rate": 4.142679202609327e-07,
+      "logits/chosen": -6.469355583190918,
+      "logits/rejected": -6.71613073348999,
+      "logps/chosen": -389.1719665527344,
+      "logps/rejected": -522.1373291015625,
+      "loss": 0.5577,
+      "rewards/accuracies": 0.706250011920929,
+      "rewards/chosen": -0.880142867565155,
+      "rewards/margins": 0.8348616361618042,
+      "rewards/rejected": -1.715004563331604,
       "step": 150
     },
+    {
+      "epoch": 0.37,
+      "grad_norm": 52.97530937607907,
+      "learning_rate": 3.9857075983815435e-07,
+      "logits/chosen": -6.588133811950684,
+      "logits/rejected": -6.864637851715088,
+      "logps/chosen": -383.6964416503906,
+      "logps/rejected": -510.33111572265625,
+      "loss": 0.5416,
+      "rewards/accuracies": 0.746874988079071,
+      "rewards/chosen": -0.7866679430007935,
+      "rewards/margins": 0.8415088653564453,
+      "rewards/rejected": -1.6281766891479492,
+      "step": 160
+    },
+    {
+      "epoch": 0.39,
+      "grad_norm": 55.77171290205848,
+      "learning_rate": 3.8191006102653317e-07,
+      "logits/chosen": -6.785226345062256,
+      "logits/rejected": -7.160543918609619,
+      "logps/chosen": -397.39068603515625,
+      "logps/rejected": -506.77398681640625,
+      "loss": 0.5291,
+      "rewards/accuracies": 0.703125,
+      "rewards/chosen": -0.8594799041748047,
+      "rewards/margins": 0.8123966455459595,
+      "rewards/rejected": -1.6718765497207642,
+      "step": 170
+    },
+    {
+      "epoch": 0.41,
+      "grad_norm": 54.05797462315199,
+      "learning_rate": 3.6439387485109883e-07,
+      "logits/chosen": -6.769280433654785,
+      "logits/rejected": -7.096776008605957,
+      "logps/chosen": -387.23101806640625,
+      "logps/rejected": -501.2037048339844,
+      "loss": 0.5379,
+      "rewards/accuracies": 0.7124999761581421,
+      "rewards/chosen": -0.8674057126045227,
+      "rewards/margins": 0.7561392784118652,
+      "rewards/rejected": -1.6235449314117432,
+      "step": 180
+    },
+    {
+      "epoch": 0.44,
+      "grad_norm": 56.88992267517494,
+      "learning_rate": 3.4613580050071274e-07,
+      "logits/chosen": -6.776444435119629,
+      "logits/rejected": -6.957192897796631,
+      "logps/chosen": -407.02349853515625,
+      "logps/rejected": -540.9013671875,
+      "loss": 0.5546,
+      "rewards/accuracies": 0.6875,
+      "rewards/chosen": -1.035926103591919,
+      "rewards/margins": 0.7458062171936035,
+      "rewards/rejected": -1.7817323207855225,
+      "step": 190
+    },
+    {
+      "epoch": 0.46,
+      "grad_norm": 48.953131469141674,
+      "learning_rate": 3.272542485937368e-07,
+      "logits/chosen": -6.675101280212402,
+      "logits/rejected": -6.999788761138916,
+      "logps/chosen": -379.14483642578125,
+      "logps/rejected": -494.6246032714844,
+      "loss": 0.532,
+      "rewards/accuracies": 0.703125,
+      "rewards/chosen": -0.6772204637527466,
+      "rewards/margins": 0.6743000745773315,
+      "rewards/rejected": -1.3515205383300781,
+      "step": 200
+    },
+    {
+      "epoch": 0.48,
+      "grad_norm": 59.16183169308755,
+      "learning_rate": 3.078716732397358e-07,
+      "logits/chosen": -6.790053367614746,
+      "logits/rejected": -7.169698238372803,
+      "logps/chosen": -404.49749755859375,
+      "logps/rejected": -520.8421020507812,
+      "loss": 0.5446,
+      "rewards/accuracies": 0.6937500238418579,
+      "rewards/chosen": -0.8419032096862793,
+      "rewards/margins": 0.7671428918838501,
+      "rewards/rejected": -1.609046220779419,
+      "step": 210
+    },
+    {
+      "epoch": 0.51,
+      "grad_norm": 51.81706264683318,
+      "learning_rate": 2.881137778775863e-07,
+      "logits/chosen": -6.8976874351501465,
+      "logits/rejected": -7.1951165199279785,
+      "logps/chosen": -386.2132873535156,
+      "logps/rejected": -497.3343811035156,
+      "loss": 0.5343,
+      "rewards/accuracies": 0.7250000238418579,
+      "rewards/chosen": -0.8116914629936218,
+      "rewards/margins": 0.7524455785751343,
+      "rewards/rejected": -1.5641369819641113,
+      "step": 220
+    },
+    {
+      "epoch": 0.53,
+      "grad_norm": 55.97935905060418,
+      "learning_rate": 2.681087000404406e-07,
+      "logits/chosen": -6.740978240966797,
+      "logits/rejected": -7.123350620269775,
+      "logps/chosen": -392.87066650390625,
+      "logps/rejected": -503.4215393066406,
+      "loss": 0.5453,
+      "rewards/accuracies": 0.6812499761581421,
+      "rewards/chosen": -0.8021535873413086,
+      "rewards/margins": 0.7195227742195129,
+      "rewards/rejected": -1.5216763019561768,
+      "step": 230
+    },
+    {
+      "epoch": 0.55,
+      "grad_norm": 49.37528181856528,
+      "learning_rate": 2.4798618033465255e-07,
+      "logits/chosen": -6.945050239562988,
+      "logits/rejected": -7.23672342300415,
+      "logps/chosen": -388.6627502441406,
+      "logps/rejected": -537.4052124023438,
+      "loss": 0.5321,
+      "rewards/accuracies": 0.7406250238418579,
+      "rewards/chosen": -0.9606342315673828,
+      "rewards/margins": 0.8123178482055664,
+      "rewards/rejected": -1.7729520797729492,
+      "step": 240
+    },
+    {
+      "epoch": 0.58,
+      "grad_norm": 54.105423705553996,
+      "learning_rate": 2.278767210221604e-07,
+      "logits/chosen": -6.868166923522949,
+      "logits/rejected": -7.291008949279785,
+      "logps/chosen": -392.64984130859375,
+      "logps/rejected": -484.0719299316406,
+      "loss": 0.5417,
+      "rewards/accuracies": 0.731249988079071,
+      "rewards/chosen": -0.779350996017456,
+      "rewards/margins": 0.814343273639679,
+      "rewards/rejected": -1.5936942100524902,
+      "step": 250
+    },
+    {
+      "epoch": 0.6,
+      "grad_norm": 56.43268807738197,
+      "learning_rate": 2.0791073966324034e-07,
+      "logits/chosen": -7.0641374588012695,
+      "logits/rejected": -7.506702423095703,
+      "logps/chosen": -392.3079528808594,
+      "logps/rejected": -523.6732788085938,
+      "loss": 0.5417,
+      "rewards/accuracies": 0.7281249761581421,
+      "rewards/chosen": -0.9223122596740723,
+      "rewards/margins": 0.6978656053543091,
+      "rewards/rejected": -1.6201778650283813,
+      "step": 260
+    },
+    {
+      "epoch": 0.62,
+      "grad_norm": 57.791839351610484,
+      "learning_rate": 1.8821772330858257e-07,
+      "logits/chosen": -7.262925624847412,
+      "logits/rejected": -7.752607822418213,
+      "logps/chosen": -377.09088134765625,
+      "logps/rejected": -489.00811767578125,
+      "loss": 0.5303,
+      "rewards/accuracies": 0.737500011920929,
+      "rewards/chosen": -0.7971266508102417,
+      "rewards/margins": 0.7223898768424988,
+      "rewards/rejected": -1.5195165872573853,
+      "step": 270
+    },
+    {
+      "epoch": 0.64,
+      "grad_norm": 52.90664589103769,
+      "learning_rate": 1.6892538872607933e-07,
+      "logits/chosen": -7.113929748535156,
+      "logits/rejected": -7.584033012390137,
+      "logps/chosen": -380.8609619140625,
+      "logps/rejected": -483.22265625,
+      "loss": 0.5418,
+      "rewards/accuracies": 0.699999988079071,
+      "rewards/chosen": -0.7374104261398315,
+      "rewards/margins": 0.6971267461776733,
+      "rewards/rejected": -1.4345371723175049,
+      "step": 280
+    },
+    {
+      "epoch": 0.67,
+      "grad_norm": 52.54626658594162,
+      "learning_rate": 1.5015885410857614e-07,
+      "logits/chosen": -7.044780731201172,
+      "logits/rejected": -7.396302700042725,
+      "logps/chosen": -402.6374206542969,
+      "logps/rejected": -516.0399169921875,
+      "loss": 0.5395,
+      "rewards/accuracies": 0.7250000238418579,
+      "rewards/chosen": -0.8658789396286011,
+      "rewards/margins": 0.7155196666717529,
+      "rewards/rejected": -1.581398606300354,
+      "step": 290
+    },
+    {
+      "epoch": 0.69,
+      "grad_norm": 49.517270927729605,
+      "learning_rate": 1.320398276343795e-07,
+      "logits/chosen": -7.218907356262207,
+      "logits/rejected": -7.622152805328369,
+      "logps/chosen": -381.68310546875,
+      "logps/rejected": -497.5284118652344,
+      "loss": 0.5304,
+      "rewards/accuracies": 0.71875,
+      "rewards/chosen": -0.8602964282035828,
+      "rewards/margins": 0.7747164368629456,
+      "rewards/rejected": -1.6350128650665283,
+      "step": 300
+    },
+    {
+      "epoch": 0.71,
+      "grad_norm": 56.86193381024086,
+      "learning_rate": 1.1468581814301717e-07,
+      "logits/chosen": -7.034183502197266,
+      "logits/rejected": -7.478652000427246,
+      "logps/chosen": -426.84246826171875,
+      "logps/rejected": -530.8802490234375,
+      "loss": 0.5331,
+      "rewards/accuracies": 0.715624988079071,
+      "rewards/chosen": -0.9773043394088745,
+      "rewards/margins": 0.6900089979171753,
+      "rewards/rejected": -1.6673132181167603,
+      "step": 310
+    },
+    {
+      "epoch": 0.74,
+      "grad_norm": 54.684846687132904,
+      "learning_rate": 9.82093730453222e-08,
+      "logits/chosen": -7.0423383712768555,
+      "logits/rejected": -7.299139499664307,
+      "logps/chosen": -368.8961181640625,
+      "logps/rejected": -539.9181518554688,
+      "loss": 0.5349,
+      "rewards/accuracies": 0.7281249761581421,
+      "rewards/chosen": -0.8125068545341492,
+      "rewards/margins": 0.9283336400985718,
+      "rewards/rejected": -1.7408406734466553,
+      "step": 320
+    },
+    {
+      "epoch": 0.76,
+      "grad_norm": 50.829923121131,
+      "learning_rate": 8.271734841028552e-08,
+      "logits/chosen": -7.086381435394287,
+      "logits/rejected": -7.435737609863281,
+      "logps/chosen": -373.8357849121094,
+      "logps/rejected": -515.7593994140625,
+      "loss": 0.5378,
+      "rewards/accuracies": 0.7281249761581421,
+      "rewards/chosen": -0.8114317655563354,
+      "rewards/margins": 0.869731068611145,
+      "rewards/rejected": -1.6811625957489014,
+      "step": 330
+    },
+    {
+      "epoch": 0.78,
+      "grad_norm": 55.18994010345068,
+      "learning_rate": 6.831021596244424e-08,
+      "logits/chosen": -7.145020484924316,
+      "logits/rejected": -7.365900993347168,
+      "logps/chosen": -399.44464111328125,
+      "logps/rejected": -547.0018920898438,
+      "loss": 0.5091,
+      "rewards/accuracies": 0.75,
+      "rewards/chosen": -0.8901287317276001,
+      "rewards/margins": 0.9701870083808899,
+      "rewards/rejected": -1.8603156805038452,
+      "step": 340
+    },
+    {
+      "epoch": 0.81,
+      "grad_norm": 54.117959716474104,
+      "learning_rate": 5.508141148419443e-08,
+      "logits/chosen": -7.050479888916016,
+      "logits/rejected": -7.3631181716918945,
+      "logps/chosen": -395.1244201660156,
+      "logps/rejected": -546.9639282226562,
+      "loss": 0.5287,
+      "rewards/accuracies": 0.746874988079071,
+      "rewards/chosen": -0.9601947069168091,
+      "rewards/margins": 0.8646580576896667,
+      "rewards/rejected": -1.8248529434204102,
+      "step": 350
+    },
+    {
+      "epoch": 0.83,
+      "grad_norm": 59.05434076734903,
+      "learning_rate": 4.311672884888756e-08,
+      "logits/chosen": -7.259505271911621,
+      "logits/rejected": -7.578073024749756,
+      "logps/chosen": -371.8418884277344,
+      "logps/rejected": -483.1830139160156,
+      "loss": 0.5215,
+      "rewards/accuracies": 0.6812499761581421,
+      "rewards/chosen": -0.8195570707321167,
+      "rewards/margins": 0.8197137117385864,
+      "rewards/rejected": -1.6392707824707031,
+      "step": 360
+    },
+    {
+      "epoch": 0.85,
+      "grad_norm": 54.176289282529886,
+      "learning_rate": 3.249376361464021e-08,
+      "logits/chosen": -7.098161220550537,
+      "logits/rejected": -7.438362121582031,
+      "logps/chosen": -408.6588439941406,
+      "logps/rejected": -542.7982788085938,
+      "loss": 0.5208,
+      "rewards/accuracies": 0.75,
+      "rewards/chosen": -0.9817103147506714,
+      "rewards/margins": 0.7650116682052612,
+      "rewards/rejected": -1.746721863746643,
+      "step": 370
+    },
+    {
+      "epoch": 0.88,
+      "grad_norm": 60.447603354663286,
+      "learning_rate": 2.3281409787363648e-08,
+      "logits/chosen": -7.142265319824219,
+      "logits/rejected": -7.481961727142334,
+      "logps/chosen": -387.88934326171875,
+      "logps/rejected": -509.31011962890625,
+      "loss": 0.5432,
+      "rewards/accuracies": 0.6937500238418579,
+      "rewards/chosen": -0.8285897374153137,
+      "rewards/margins": 0.6989187598228455,
+      "rewards/rejected": -1.5275084972381592,
+      "step": 380
+    },
+    {
+      "epoch": 0.9,
+      "grad_norm": 53.49129578625785,
+      "learning_rate": 1.553941301669892e-08,
+      "logits/chosen": -7.165548801422119,
+      "logits/rejected": -7.568087577819824,
+      "logps/chosen": -370.10491943359375,
+      "logps/rejected": -490.369873046875,
+      "loss": 0.5315,
+      "rewards/accuracies": 0.706250011920929,
+      "rewards/chosen": -0.814159095287323,
+      "rewards/margins": 0.7283869981765747,
+      "rewards/rejected": -1.542546033859253,
+      "step": 390
+    },
+    {
+      "epoch": 0.92,
+      "grad_norm": 52.37813865787704,
+      "learning_rate": 9.31798312255233e-09,
+      "logits/chosen": -7.114771842956543,
+      "logits/rejected": -7.5348968505859375,
+      "logps/chosen": -390.9321594238281,
+      "logps/rejected": -521.2406005859375,
+      "loss": 0.5151,
+      "rewards/accuracies": 0.699999988079071,
+      "rewards/chosen": -0.8433186411857605,
+      "rewards/margins": 0.7988412976264954,
+      "rewards/rejected": -1.6421600580215454,
+      "step": 400
+    },
+    {
+      "epoch": 0.94,
+      "grad_norm": 53.77720647192081,
+      "learning_rate": 4.657468465146641e-09,
+      "logits/chosen": -7.192900657653809,
+      "logits/rejected": -7.582442283630371,
+      "logps/chosen": -385.35809326171875,
+      "logps/rejected": -510.85919189453125,
+      "loss": 0.5351,
+      "rewards/accuracies": 0.7250000238418579,
+      "rewards/chosen": -0.8629342913627625,
+      "rewards/margins": 0.8567607998847961,
+      "rewards/rejected": -1.7196948528289795,
+      "step": 410
+    },
+    {
+      "epoch": 0.97,
+      "grad_norm": 59.87565888131157,
+      "learning_rate": 1.5880942704217526e-09,
+      "logits/chosen": -6.9229278564453125,
+      "logits/rejected": -7.265672206878662,
+      "logps/chosen": -432.6807556152344,
+      "logps/rejected": -540.7777099609375,
+      "loss": 0.5192,
+      "rewards/accuracies": 0.715624988079071,
+      "rewards/chosen": -0.9274924993515015,
+      "rewards/margins": 0.6948927640914917,
+      "rewards/rejected": -1.6223852634429932,
+      "step": 420
+    },
+    {
+      "epoch": 0.99,
+      "grad_norm": 54.84275580772891,
+      "learning_rate": 1.297666078462767e-10,
+      "logits/chosen": -7.139874458312988,
+      "logits/rejected": -7.5435380935668945,
+      "logps/chosen": -396.08514404296875,
+      "logps/rejected": -513.01806640625,
+      "loss": 0.5269,
+      "rewards/accuracies": 0.7593749761581421,
+      "rewards/chosen": -0.8496484756469727,
+      "rewards/margins": 0.8382279276847839,
+      "rewards/rejected": -1.6878764629364014,
+      "step": 430
+    },
     {
       "epoch": 1.0,
-      "step": 156,
+      "step": 434,
       "total_flos": 0.0,
-      "train_loss": 0.5437000210468586,
-      "train_runtime": 4622.4032,
-      "train_samples_per_second": 8.641,
+      "train_loss": 0.5533875935638006,
+      "train_runtime": 12702.5109,
+      "train_samples_per_second": 8.749,
       "train_steps_per_second": 0.034
     }
   ],
   "logging_steps": 10,
-  "max_steps": 156,
+  "max_steps": 434,
   "num_input_tokens_seen": 0,
   "num_train_epochs": 1,
   "save_steps": 100,