Model save

README.md

@@ -37,7 +37,7 @@ The following hyperparameters were used during training:
 - learning_rate: 5e-07
 - train_batch_size: 8
 - eval_batch_size: 8
-- seed: 4
+- seed: 5
 - distributed_type: multi-GPU
 - num_devices: 8
 - gradient_accumulation_steps: 2

all_results.json

@@ -1,8 +1,8 @@
 {
     "epoch": 1.0,
-    "train_loss": 0.08520494085035206,
-    "train_runtime": 3257.6401,
+    "train_loss": 0.08637741568647785,
+    "train_runtime": 3241.0704,
     "train_samples": 51894,
-    "train_samples_per_second": 15.93,
-    "train_steps_per_second": 0.124
+    "train_samples_per_second": 16.011,
+    "train_steps_per_second": 0.125
 }

model-00001-of-00003.safetensors

@@ -1,3 +1,3 @@
 version https://git-lfs.github.com/spec/v1
-oid sha256:8cc657a2ee48daf3c730e4aab1598173ac5f267b70b973a7eb21736bbef1e546
+oid sha256:95705cc51e5605b33982ea0df54777bcf1c925b1f1ddedae22001a4d729bf3cd
 size 4943162336

model-00002-of-00003.safetensors

@@ -1,3 +1,3 @@
 version https://git-lfs.github.com/spec/v1
-oid sha256:d97d51c686e2d180bd32387ba32d6b5c4c1187ea7765c11ec897d31c8c7030e5
+oid sha256:acbfb972feae803aa016bf84a68601e82ede3de4db685302a0702f4eed2610c8
 size 4999819336

model-00003-of-00003.safetensors

@@ -1,3 +1,3 @@
 version https://git-lfs.github.com/spec/v1
-oid sha256:afb80c7d1ddf42019d5701c07a7752e01c5b98b2bf02d3dc3764e39ca350bc77
+oid sha256:b859840e065ab8f09e5fb0a306f50c6cfc01bfee07ca846d4556b15e6766b7b6
 size 4540516344

train_results.json

@@ -1,8 +1,8 @@
 {
     "epoch": 1.0,
-    "train_loss": 0.08520494085035206,
-    "train_runtime": 3257.6401,
+    "train_loss": 0.08637741568647785,
+    "train_runtime": 3241.0704,
     "train_samples": 51894,
-    "train_samples_per_second": 15.93,
-    "train_steps_per_second": 0.124
+    "train_samples_per_second": 16.011,
+    "train_steps_per_second": 0.125
 }

trainer_state.json

@@ -11,11 +11,11 @@
     {
       "epoch": 0.0,
       "learning_rate": 1.2195121951219512e-08,
-      "logits/chosen": -2.8681135177612305,
-      "logits/rejected": -2.8858838081359863,
-      "logps/chosen": -518.1907958984375,
-      "logps/rejected": -109.31971740722656,
-      "loss": 0.3939,
+      "logits/chosen": -2.8695335388183594,
+      "logits/rejected": -2.8522377014160156,
+      "logps/chosen": -537.80126953125,
+      "logps/rejected": -108.91968536376953,
+      "loss": 0.3731,
       "rewards/accuracies": 0.0,
       "rewards/chosen": 0.0,
       "rewards/margins": 0.0,
@@ -25,571 +25,571 @@
     {
       "epoch": 0.02,
       "learning_rate": 1.219512195121951e-07,
-      "logits/chosen": -2.7987546920776367,
-      "logits/rejected": -2.7523815631866455,
-      "logps/chosen": -434.1848449707031,
-      "logps/rejected": -114.18251037597656,
-      "loss": 0.3844,
-      "rewards/accuracies": 0.4513888955116272,
-      "rewards/chosen": 0.0004987930878996849,
-      "rewards/margins": 0.001031861756928265,
-      "rewards/rejected": -0.000533068785443902,
+      "logits/chosen": -2.8006839752197266,
+      "logits/rejected": -2.7513468265533447,
+      "logps/chosen": -339.1521911621094,
+      "logps/rejected": -113.4110336303711,
+      "loss": 0.3869,
+      "rewards/accuracies": 0.5,
+      "rewards/chosen": 0.0008597287815064192,
+      "rewards/margins": 0.0015750755555927753,
+      "rewards/rejected": -0.0007153467740863562,
       "step": 10
     },
     {
       "epoch": 0.05,
       "learning_rate": 2.439024390243902e-07,
-      "logits/chosen": -2.8212170600891113,
-      "logits/rejected": -2.7980713844299316,
-      "logps/chosen": -417.22845458984375,
-      "logps/rejected": -118.038330078125,
-      "loss": 0.3804,
-      "rewards/accuracies": 0.768750011920929,
-      "rewards/chosen": 0.02029377594590187,
-      "rewards/margins": 0.036424122750759125,
-      "rewards/rejected": -0.0161303523927927,
+      "logits/chosen": -2.8162007331848145,
+      "logits/rejected": -2.8079230785369873,
+      "logps/chosen": -435.24627685546875,
+      "logps/rejected": -116.09075927734375,
+      "loss": 0.3762,
+      "rewards/accuracies": 0.7875000238418579,
+      "rewards/chosen": 0.021398674696683884,
+      "rewards/margins": 0.03878290206193924,
+      "rewards/rejected": -0.017384223639965057,
       "step": 20
     },
     {
       "epoch": 0.07,
       "learning_rate": 3.6585365853658536e-07,
-      "logits/chosen": -2.656278133392334,
-      "logits/rejected": -2.644028663635254,
-      "logps/chosen": -398.81500244140625,
-      "logps/rejected": -125.97016906738281,
-      "loss": 0.3437,
-      "rewards/accuracies": 0.7562500238418579,
-      "rewards/chosen": 0.0762760117650032,
-      "rewards/margins": 0.2021373212337494,
-      "rewards/rejected": -0.12586131691932678,
+      "logits/chosen": -2.7241621017456055,
+      "logits/rejected": -2.686893939971924,
+      "logps/chosen": -437.8358459472656,
+      "logps/rejected": -141.41867065429688,
+      "loss": 0.3485,
+      "rewards/accuracies": 0.84375,
+      "rewards/chosen": 0.08630482107400894,
+      "rewards/margins": 0.20522558689117432,
+      "rewards/rejected": -0.11892076581716537,
       "step": 30
     },
     {
       "epoch": 0.1,
       "learning_rate": 4.878048780487804e-07,
-      "logits/chosen": -2.5408267974853516,
-      "logits/rejected": -2.5265488624572754,
-      "logps/chosen": -384.7110595703125,
-      "logps/rejected": -169.5616912841797,
-      "loss": 0.2883,
-      "rewards/accuracies": 0.737500011920929,
-      "rewards/chosen": 0.022841650992631912,
-      "rewards/margins": 0.42677050828933716,
-      "rewards/rejected": -0.40392884612083435,
+      "logits/chosen": -2.5932071208953857,
+      "logits/rejected": -2.5736117362976074,
+      "logps/chosen": -414.60986328125,
+      "logps/rejected": -169.4441375732422,
+      "loss": 0.2897,
+      "rewards/accuracies": 0.7875000238418579,
+      "rewards/chosen": 0.01923329569399357,
+      "rewards/margins": 0.4725423753261566,
+      "rewards/rejected": -0.4533091187477112,
       "step": 40
     },
     {
       "epoch": 0.12,
       "learning_rate": 4.992461696250783e-07,
-      "logits/chosen": -2.42402982711792,
-      "logits/rejected": -2.390493154525757,
-      "logps/chosen": -437.21337890625,
-      "logps/rejected": -221.1737823486328,
-      "loss": 0.204,
-      "rewards/accuracies": 0.7749999761581421,
-      "rewards/chosen": -0.1747589260339737,
-      "rewards/margins": 0.8014341592788696,
-      "rewards/rejected": -0.9761930704116821,
+      "logits/chosen": -2.471208333969116,
+      "logits/rejected": -2.458268642425537,
+      "logps/chosen": -421.943359375,
+      "logps/rejected": -212.53466796875,
+      "loss": 0.2122,
+      "rewards/accuracies": 0.7250000238418579,
+      "rewards/chosen": -0.12829390168190002,
+      "rewards/margins": 0.7995172739028931,
+      "rewards/rejected": -0.9278111457824707,
       "step": 50
     },
     {
       "epoch": 0.15,
       "learning_rate": 4.966461721767899e-07,
-      "logits/chosen": -2.378632068634033,
-      "logits/rejected": -2.3337972164154053,
-      "logps/chosen": -439.27764892578125,
-      "logps/rejected": -243.908203125,
-      "loss": 0.1586,
-      "rewards/accuracies": 0.7749999761581421,
-      "rewards/chosen": -0.3428526818752289,
-      "rewards/margins": 0.9457007646560669,
-      "rewards/rejected": -1.2885534763336182,
+      "logits/chosen": -2.404550075531006,
+      "logits/rejected": -2.372710704803467,
+      "logps/chosen": -378.82940673828125,
+      "logps/rejected": -210.00704956054688,
+      "loss": 0.1372,
+      "rewards/accuracies": 0.7562500238418579,
+      "rewards/chosen": -0.3348023593425751,
+      "rewards/margins": 0.8655646443367004,
+      "rewards/rejected": -1.2003670930862427,
       "step": 60
     },
     {
       "epoch": 0.17,
       "learning_rate": 4.922100518015975e-07,
-      "logits/chosen": -2.37459659576416,
-      "logits/rejected": -2.327303409576416,
-      "logps/chosen": -420.84381103515625,
-      "logps/rejected": -267.9267578125,
-      "loss": 0.119,
+      "logits/chosen": -2.402493953704834,
+      "logits/rejected": -2.36720871925354,
+      "logps/chosen": -450.5000915527344,
+      "logps/rejected": -274.5812683105469,
+      "loss": 0.1023,
       "rewards/accuracies": 0.762499988079071,
-      "rewards/chosen": -0.40542110800743103,
-      "rewards/margins": 1.2117817401885986,
-      "rewards/rejected": -1.6172027587890625,
+      "rewards/chosen": -0.5065652132034302,
+      "rewards/margins": 1.22487473487854,
+      "rewards/rejected": -1.7314398288726807,
       "step": 70
     },
     {
       "epoch": 0.2,
       "learning_rate": 4.859708325770919e-07,
-      "logits/chosen": -2.428609609603882,
-      "logits/rejected": -2.370483875274658,
-      "logps/chosen": -474.28582763671875,
-      "logps/rejected": -337.1321716308594,
-      "loss": 0.0755,
-      "rewards/accuracies": 0.824999988079071,
-      "rewards/chosen": -0.5983784794807434,
-      "rewards/margins": 1.5884076356887817,
-      "rewards/rejected": -2.18678617477417,
+      "logits/chosen": -2.3470957279205322,
+      "logits/rejected": -2.2893269062042236,
+      "logps/chosen": -458.8729553222656,
+      "logps/rejected": -316.4507141113281,
+      "loss": 0.0833,
+      "rewards/accuracies": 0.8374999761581421,
+      "rewards/chosen": -0.47927385568618774,
+      "rewards/margins": 1.5929757356643677,
+      "rewards/rejected": -2.0722496509552,
       "step": 80
     },
     {
       "epoch": 0.22,
       "learning_rate": 4.779749614980225e-07,
-      "logits/chosen": -2.3890507221221924,
-      "logits/rejected": -2.3450489044189453,
-      "logps/chosen": -500.783203125,
-      "logps/rejected": -368.41094970703125,
-      "loss": 0.0584,
-      "rewards/accuracies": 0.793749988079071,
-      "rewards/chosen": -0.796062171459198,
-      "rewards/margins": 1.7808799743652344,
-      "rewards/rejected": -2.576941967010498,
+      "logits/chosen": -2.353701114654541,
+      "logits/rejected": -2.29787015914917,
+      "logps/chosen": -460.30462646484375,
+      "logps/rejected": -341.1560974121094,
+      "loss": 0.0731,
+      "rewards/accuracies": 0.7875000238418579,
+      "rewards/chosen": -0.7498773336410522,
+      "rewards/margins": 1.5740221738815308,
+      "rewards/rejected": -2.323899745941162,
       "step": 90
     },
     {
       "epoch": 0.25,
       "learning_rate": 4.682819627081427e-07,
-      "logits/chosen": -2.384782075881958,
-      "logits/rejected": -2.3418028354644775,
-      "logps/chosen": -502.3565368652344,
-      "logps/rejected": -362.3828125,
-      "loss": 0.0628,
-      "rewards/accuracies": 0.8500000238418579,
-      "rewards/chosen": -0.5387237071990967,
-      "rewards/margins": 1.9631805419921875,
-      "rewards/rejected": -2.501904249191284,
+      "logits/chosen": -2.3233580589294434,
+      "logits/rejected": -2.2358224391937256,
+      "logps/chosen": -494.71783447265625,
+      "logps/rejected": -365.74462890625,
+      "loss": 0.0787,
+      "rewards/accuracies": 0.737500011920929,
+      "rewards/chosen": -0.7132088541984558,
+      "rewards/margins": 1.8270628452301025,
+      "rewards/rejected": -2.540271520614624,
       "step": 100
     },
     {
       "epoch": 0.27,
       "learning_rate": 4.569639943810477e-07,
-      "logits/chosen": -2.3396503925323486,
-      "logits/rejected": -2.2695436477661133,
-      "logps/chosen": -466.52679443359375,
-      "logps/rejected": -395.2033996582031,
-      "loss": 0.0506,
-      "rewards/accuracies": 0.762499988079071,
-      "rewards/chosen": -1.0900439023971558,
-      "rewards/margins": 1.7583458423614502,
-      "rewards/rejected": -2.8483898639678955,
+      "logits/chosen": -2.3129024505615234,
+      "logits/rejected": -2.2176918983459473,
+      "logps/chosen": -534.7408447265625,
+      "logps/rejected": -389.4866638183594,
+      "loss": 0.0587,
+      "rewards/accuracies": 0.793749988079071,
+      "rewards/chosen": -0.9212908744812012,
+      "rewards/margins": 1.8763319253921509,
+      "rewards/rejected": -2.7976229190826416,
       "step": 110
     },
     {
       "epoch": 0.3,
       "learning_rate": 4.4410531154874543e-07,
-      "logits/chosen": -2.3576321601867676,
-      "logits/rejected": -2.3109076023101807,
-      "logps/chosen": -484.6878356933594,
-      "logps/rejected": -393.3995666503906,
-      "loss": 0.0561,
+      "logits/chosen": -2.316215991973877,
+      "logits/rejected": -2.223778486251831,
+      "logps/chosen": -538.7808837890625,
+      "logps/rejected": -417.7806701660156,
+      "loss": 0.0448,
       "rewards/accuracies": 0.8062499761581421,
-      "rewards/chosen": -0.9047477841377258,
-      "rewards/margins": 1.829101324081421,
-      "rewards/rejected": -2.733849048614502,
+      "rewards/chosen": -1.0477784872055054,
+      "rewards/margins": 1.9544658660888672,
+      "rewards/rejected": -3.002244234085083,
       "step": 120
     },
     {
       "epoch": 0.32,
       "learning_rate": 4.298016388768561e-07,
-      "logits/chosen": -2.3168373107910156,
-      "logits/rejected": -2.277609348297119,
-      "logps/chosen": -474.674560546875,
-      "logps/rejected": -389.1014709472656,
-      "loss": 0.0576,
-      "rewards/accuracies": 0.8062499761581421,
-      "rewards/chosen": -1.033732533454895,
-      "rewards/margins": 1.7740589380264282,
-      "rewards/rejected": -2.807791233062744,
+      "logits/chosen": -2.342869281768799,
+      "logits/rejected": -2.2497687339782715,
+      "logps/chosen": -501.41131591796875,
+      "logps/rejected": -390.2490234375,
+      "loss": 0.0676,
+      "rewards/accuracies": 0.8500000238418579,
+      "rewards/chosen": -0.7455350160598755,
+      "rewards/margins": 1.8945093154907227,
+      "rewards/rejected": -2.6400444507598877,
       "step": 130
     },
     {
       "epoch": 0.35,
       "learning_rate": 4.1415945805573005e-07,
-      "logits/chosen": -2.27695631980896,
-      "logits/rejected": -2.22804594039917,
-      "logps/chosen": -521.7109985351562,
-      "logps/rejected": -422.5042419433594,
-      "loss": 0.0485,
-      "rewards/accuracies": 0.862500011920929,
-      "rewards/chosen": -0.9181007146835327,
-      "rewards/margins": 2.139181613922119,
-      "rewards/rejected": -3.0572826862335205,
+      "logits/chosen": -2.343029499053955,
+      "logits/rejected": -2.2743804454803467,
+      "logps/chosen": -512.1886596679688,
+      "logps/rejected": -406.03045654296875,
+      "loss": 0.0574,
+      "rewards/accuracies": 0.856249988079071,
+      "rewards/chosen": -0.6720421314239502,
+      "rewards/margins": 2.2327818870544434,
+      "rewards/rejected": -2.9048237800598145,
       "step": 140
     },
     {
       "epoch": 0.37,
       "learning_rate": 3.972952151123984e-07,
-      "logits/chosen": -2.2969155311584473,
-      "logits/rejected": -2.2290713787078857,
-      "logps/chosen": -517.611328125,
-      "logps/rejected": -414.35491943359375,
-      "loss": 0.0446,
-      "rewards/accuracies": 0.824999988079071,
-      "rewards/chosen": -1.0711259841918945,
-      "rewards/margins": 1.9469735622406006,
-      "rewards/rejected": -3.018099546432495,
+      "logits/chosen": -2.246107816696167,
+      "logits/rejected": -2.1575684547424316,
+      "logps/chosen": -559.8818969726562,
+      "logps/rejected": -472.61920166015625,
+      "loss": 0.0424,
+      "rewards/accuracies": 0.8374999761581421,
+      "rewards/chosen": -1.3547112941741943,
+      "rewards/margins": 2.0968470573425293,
+      "rewards/rejected": -3.4515585899353027,
       "step": 150
     },
     {
       "epoch": 0.39,
       "learning_rate": 3.793344535444142e-07,
-      "logits/chosen": -2.294722080230713,
-      "logits/rejected": -2.2438769340515137,
-      "logps/chosen": -532.6485595703125,
-      "logps/rejected": -426.20513916015625,
-      "loss": 0.0502,
-      "rewards/accuracies": 0.768750011920929,
-      "rewards/chosen": -1.2362943887710571,
-      "rewards/margins": 1.7946383953094482,
-      "rewards/rejected": -3.030932903289795,
+      "logits/chosen": -2.282726526260376,
+      "logits/rejected": -2.1828060150146484,
+      "logps/chosen": -590.2351684570312,
+      "logps/rejected": -452.65087890625,
+      "loss": 0.0402,
+      "rewards/accuracies": 0.856249988079071,
+      "rewards/chosen": -1.2386449575424194,
+      "rewards/margins": 2.1290228366851807,
+      "rewards/rejected": -3.3676681518554688,
       "step": 160
     },
     {
       "epoch": 0.42,
       "learning_rate": 3.604108797288461e-07,
-      "logits/chosen": -2.2703588008880615,
-      "logits/rejected": -2.2441458702087402,
-      "logps/chosen": -449.19091796875,
-      "logps/rejected": -373.4166259765625,
-      "loss": 0.059,
-      "rewards/accuracies": 0.7875000238418579,
-      "rewards/chosen": -1.1028985977172852,
-      "rewards/margins": 1.593082070350647,
-      "rewards/rejected": -2.6959805488586426,
+      "logits/chosen": -2.2551448345184326,
+      "logits/rejected": -2.1805520057678223,
+      "logps/chosen": -529.7485961914062,
+      "logps/rejected": -446.873779296875,
+      "loss": 0.042,
+      "rewards/accuracies": 0.793749988079071,
+      "rewards/chosen": -1.2568132877349854,
+      "rewards/margins": 2.019052505493164,
+      "rewards/rejected": -3.2758660316467285,
       "step": 170
     },
     {
       "epoch": 0.44,
       "learning_rate": 3.40665367563858e-07,
-      "logits/chosen": -2.300586700439453,
-      "logits/rejected": -2.2104036808013916,
-      "logps/chosen": -538.3341064453125,
-      "logps/rejected": -435.9353942871094,
-      "loss": 0.0478,
-      "rewards/accuracies": 0.856249988079071,
-      "rewards/chosen": -0.782570481300354,
-      "rewards/margins": 2.3180627822875977,
-      "rewards/rejected": -3.100633144378662,
+      "logits/chosen": -2.2764365673065186,
+      "logits/rejected": -2.175426483154297,
+      "logps/chosen": -488.26385498046875,
+      "logps/rejected": -383.0807189941406,
+      "loss": 0.049,
+      "rewards/accuracies": 0.7749999761581421,
+      "rewards/chosen": -1.1468641757965088,
+      "rewards/margins": 1.7448303699493408,
+      "rewards/rejected": -2.8916945457458496,
       "step": 180
     },
     {
       "epoch": 0.47,
       "learning_rate": 3.202449097526798e-07,
-      "logits/chosen": -2.273599624633789,
-      "logits/rejected": -2.2272255420684814,
-      "logps/chosen": -511.9939880371094,
-      "logps/rejected": -422.14044189453125,
-      "loss": 0.0502,
-      "rewards/accuracies": 0.793749988079071,
-      "rewards/chosen": -1.0683417320251465,
-      "rewards/margins": 1.9933398962020874,
-      "rewards/rejected": -3.0616817474365234,
+      "logits/chosen": -2.3102352619171143,
+      "logits/rejected": -2.215864419937134,
+      "logps/chosen": -495.09588623046875,
+      "logps/rejected": -415.6620178222656,
+      "loss": 0.0605,
+      "rewards/accuracies": 0.831250011920929,
+      "rewards/chosen": -1.0821647644042969,
+      "rewards/margins": 1.9470455646514893,
+      "rewards/rejected": -3.0292105674743652,
       "step": 190
     },
     {
       "epoch": 0.49,
       "learning_rate": 2.993015235369905e-07,
-      "logits/chosen": -2.2274768352508545,
-      "logits/rejected": -2.178544282913208,
-      "logps/chosen": -531.9685668945312,
-      "logps/rejected": -466.0660095214844,
-      "loss": 0.0415,
-      "rewards/accuracies": 0.800000011920929,
-      "rewards/chosen": -1.5277751684188843,
-      "rewards/margins": 1.9938457012176514,
-      "rewards/rejected": -3.521620988845825,
+      "logits/chosen": -2.3042426109313965,
+      "logits/rejected": -2.210401773452759,
+      "logps/chosen": -522.3523559570312,
+      "logps/rejected": -429.0350036621094,
+      "loss": 0.0546,
+      "rewards/accuracies": 0.887499988079071,
+      "rewards/chosen": -0.8798307180404663,
+      "rewards/margins": 2.224181890487671,
+      "rewards/rejected": -3.1040127277374268,
       "step": 200
     },
     {
       "epoch": 0.52,
       "learning_rate": 2.7799111902582693e-07,
-      "logits/chosen": -2.239495038986206,
-      "logits/rejected": -2.14347505569458,
-      "logps/chosen": -574.8614501953125,
-      "logps/rejected": -501.8006896972656,
-      "loss": 0.0305,
-      "rewards/accuracies": 0.8187500238418579,
-      "rewards/chosen": -1.3575702905654907,
-      "rewards/margins": 2.561095714569092,
-      "rewards/rejected": -3.918666124343872,
+      "logits/chosen": -2.227726936340332,
+      "logits/rejected": -2.1312689781188965,
+      "logps/chosen": -507.91546630859375,
+      "logps/rejected": -440.345703125,
+      "loss": 0.0461,
+      "rewards/accuracies": 0.8374999761581421,
+      "rewards/chosen": -1.2374424934387207,
+      "rewards/margins": 2.0522685050964355,
+      "rewards/rejected": -3.2897109985351562,
       "step": 210
     },
     {
       "epoch": 0.54,
       "learning_rate": 2.564723385445869e-07,
-      "logits/chosen": -2.3055145740509033,
-      "logits/rejected": -2.2241768836975098,
-      "logps/chosen": -569.2613525390625,
-      "logps/rejected": -487.712158203125,
-      "loss": 0.0477,
-      "rewards/accuracies": 0.768750011920929,
-      "rewards/chosen": -1.3685243129730225,
-      "rewards/margins": 2.2271132469177246,
-      "rewards/rejected": -3.595637559890747,
+      "logits/chosen": -2.1844215393066406,
+      "logits/rejected": -2.0822980403900146,
+      "logps/chosen": -558.802978515625,
+      "logps/rejected": -482.9683532714844,
+      "loss": 0.0439,
+      "rewards/accuracies": 0.8187500238418579,
+      "rewards/chosen": -1.1958107948303223,
+      "rewards/margins": 2.303767204284668,
+      "rewards/rejected": -3.4995779991149902,
       "step": 220
     },
     {
       "epoch": 0.57,
       "learning_rate": 2.3490537564442845e-07,
-      "logits/chosen": -2.2792344093322754,
-      "logits/rejected": -2.2088494300842285,
-      "logps/chosen": -494.75286865234375,
-      "logps/rejected": -409.4736328125,
-      "loss": 0.0545,
-      "rewards/accuracies": 0.768750011920929,
-      "rewards/chosen": -1.0721429586410522,
-      "rewards/margins": 1.7748781442642212,
-      "rewards/rejected": -2.8470211029052734,
+      "logits/chosen": -2.213812828063965,
+      "logits/rejected": -2.1118857860565186,
+      "logps/chosen": -527.6865234375,
+      "logps/rejected": -451.37347412109375,
+      "loss": 0.0375,
+      "rewards/accuracies": 0.793749988079071,
+      "rewards/chosen": -1.212794303894043,
+      "rewards/margins": 2.118138074874878,
+      "rewards/rejected": -3.3309326171875,
       "step": 230
     },
     {
       "epoch": 0.59,
       "learning_rate": 2.1345078256378801e-07,
-      "logits/chosen": -2.3294692039489746,
-      "logits/rejected": -2.264793634414673,
-      "logps/chosen": -521.0982666015625,
-      "logps/rejected": -408.40289306640625,
-      "loss": 0.0513,
-      "rewards/accuracies": 0.8125,
-      "rewards/chosen": -1.0901975631713867,
-      "rewards/margins": 1.8291746377944946,
-      "rewards/rejected": -2.919372320175171,
+      "logits/chosen": -2.228127956390381,
+      "logits/rejected": -2.1170496940612793,
+      "logps/chosen": -518.7861938476562,
+      "logps/rejected": -471.34930419921875,
+      "loss": 0.0325,
+      "rewards/accuracies": 0.856249988079071,
+      "rewards/chosen": -1.140577793121338,
+      "rewards/margins": 2.372342348098755,
+      "rewards/rejected": -3.5129196643829346,
       "step": 240
     },
     {
       "epoch": 0.62,
       "learning_rate": 1.9226827501969865e-07,
-      "logits/chosen": -2.317063808441162,
-      "logits/rejected": -2.2598507404327393,
-      "logps/chosen": -524.32666015625,
-      "logps/rejected": -434.41925048828125,
-      "loss": 0.0528,
-      "rewards/accuracies": 0.8062499761581421,
-      "rewards/chosen": -1.0961580276489258,
-      "rewards/margins": 2.1600213050842285,
-      "rewards/rejected": -3.2561793327331543,
+      "logits/chosen": -2.22481107711792,
+      "logits/rejected": -2.097669839859009,
+      "logps/chosen": -545.9178466796875,
+      "logps/rejected": -495.44329833984375,
+      "loss": 0.0482,
+      "rewards/accuracies": 0.862500011920929,
+      "rewards/chosen": -1.1768572330474854,
+      "rewards/margins": 2.6003925800323486,
+      "rewards/rejected": -3.777249574661255,
       "step": 250
     },
     {
       "epoch": 0.64,
       "learning_rate": 1.715155432264775e-07,
-      "logits/chosen": -2.2932469844818115,
-      "logits/rejected": -2.188689708709717,
-      "logps/chosen": -496.59967041015625,
-      "logps/rejected": -417.621826171875,
-      "loss": 0.0481,
-      "rewards/accuracies": 0.793749988079071,
-      "rewards/chosen": -1.067463755607605,
-      "rewards/margins": 2.0409655570983887,
-      "rewards/rejected": -3.108428955078125,
+      "logits/chosen": -2.218748092651367,
+      "logits/rejected": -2.1094837188720703,
+      "logps/chosen": -496.226318359375,
+      "logps/rejected": -427.3712463378906,
+      "loss": 0.0615,
+      "rewards/accuracies": 0.84375,
+      "rewards/chosen": -1.1070716381072998,
+      "rewards/margins": 2.072988986968994,
+      "rewards/rejected": -3.180060386657715,
       "step": 260
     },
     {
       "epoch": 0.67,
       "learning_rate": 1.51347077992983e-07,
-      "logits/chosen": -2.3267765045166016,
-      "logits/rejected": -2.231224775314331,
-      "logps/chosen": -548.4693603515625,
-      "logps/rejected": -429.4169921875,
-      "loss": 0.0427,
-      "rewards/accuracies": 0.831250011920929,
-      "rewards/chosen": -1.0470257997512817,
-      "rewards/margins": 2.1003241539001465,
-      "rewards/rejected": -3.1473500728607178,
+      "logits/chosen": -2.2541496753692627,
+      "logits/rejected": -2.1432924270629883,
+      "logps/chosen": -493.8655700683594,
+      "logps/rejected": -428.19403076171875,
+      "loss": 0.0488,
+      "rewards/accuracies": 0.84375,
+      "rewards/chosen": -0.9823986291885376,
+      "rewards/margins": 2.0746474266052246,
+      "rewards/rejected": -3.057046413421631,
       "step": 270
     },
     {
       "epoch": 0.69,
       "learning_rate": 1.3191302063739906e-07,
-      "logits/chosen": -2.334428310394287,
-      "logits/rejected": -2.2525620460510254,
-      "logps/chosen": -515.8858642578125,
-      "logps/rejected": -440.9449768066406,
-      "loss": 0.0527,
-      "rewards/accuracies": 0.8187500238418579,
-      "rewards/chosen": -1.1162147521972656,
-      "rewards/margins": 2.022719383239746,
-      "rewards/rejected": -3.1389341354370117,
+      "logits/chosen": -2.2505111694335938,
+      "logits/rejected": -2.137561798095703,
+      "logps/chosen": -520.2976684570312,
+      "logps/rejected": -448.7735900878906,
+      "loss": 0.045,
+      "rewards/accuracies": 0.824999988079071,
+      "rewards/chosen": -1.226133108139038,
+      "rewards/margins": 2.1102890968322754,
+      "rewards/rejected": -3.3364219665527344,
       "step": 280
     },
     {
       "epoch": 0.72,
       "learning_rate": 1.1335804528119475e-07,
-      "logits/chosen": -2.344043016433716,
-      "logits/rejected": -2.2616305351257324,
-      "logps/chosen": -535.1497192382812,
-      "logps/rejected": -425.8858337402344,
-      "loss": 0.0575,
-      "rewards/accuracies": 0.8500000238418579,
-      "rewards/chosen": -0.9090016484260559,
-      "rewards/margins": 2.22568678855896,
-      "rewards/rejected": -3.13468861579895,
+      "logits/chosen": -2.232018232345581,
+      "logits/rejected": -2.0879502296447754,
+      "logps/chosen": -566.6788330078125,
+      "logps/rejected": -472.78228759765625,
+      "loss": 0.0366,
+      "rewards/accuracies": 0.862500011920929,
+      "rewards/chosen": -1.2060538530349731,
+      "rewards/margins": 2.433617353439331,
+      "rewards/rejected": -3.6396713256835938,
       "step": 290
     },
     {
       "epoch": 0.74,
       "learning_rate": 9.582028184286423e-08,
-      "logits/chosen": -2.3820149898529053,
-      "logits/rejected": -2.352027416229248,
-      "logps/chosen": -549.5676879882812,
-      "logps/rejected": -470.928466796875,
-      "loss": 0.0489,
-      "rewards/accuracies": 0.8062499761581421,
-      "rewards/chosen": -1.1177103519439697,
-      "rewards/margins": 2.212918758392334,
-      "rewards/rejected": -3.3306288719177246,
+      "logits/chosen": -2.226560354232788,
+      "logits/rejected": -2.1372122764587402,
+      "logps/chosen": -492.98162841796875,
+      "logps/rejected": -440.3003845214844,
+      "loss": 0.0416,
+      "rewards/accuracies": 0.762499988079071,
+      "rewards/chosen": -1.3394794464111328,
+      "rewards/margins": 1.9935524463653564,
+      "rewards/rejected": -3.3330321311950684,
       "step": 300
     },
     {
       "epoch": 0.76,
       "learning_rate": 7.943028774907065e-08,
-      "logits/chosen": -2.3420827388763428,
-      "logits/rejected": -2.2547786235809326,
-      "logps/chosen": -522.0978393554688,
-      "logps/rejected": -427.39801025390625,
-      "loss": 0.0367,
-      "rewards/accuracies": 0.78125,
-      "rewards/chosen": -1.0316474437713623,
-      "rewards/margins": 2.1541504859924316,
-      "rewards/rejected": -3.185797929763794,
+      "logits/chosen": -2.216187000274658,
+      "logits/rejected": -2.1210825443267822,
+      "logps/chosen": -514.5095825195312,
+      "logps/rejected": -437.02569580078125,
+      "loss": 0.0429,
+      "rewards/accuracies": 0.800000011920929,
+      "rewards/chosen": -1.0884000062942505,
+      "rewards/margins": 2.209381580352783,
+      "rewards/rejected": -3.2977817058563232,
       "step": 310
     },
     {
       "epoch": 0.79,
       "learning_rate": 6.431007601814637e-08,
-      "logits/chosen": -2.299654006958008,
-      "logits/rejected": -2.211099863052368,
-      "logps/chosen": -534.5721435546875,
-      "logps/rejected": -457.3140563964844,
-      "loss": 0.0373,
-      "rewards/accuracies": 0.768750011920929,
-      "rewards/chosen": -1.5419642925262451,
-      "rewards/margins": 1.9639537334442139,
-      "rewards/rejected": -3.50591778755188,
+      "logits/chosen": -2.319559335708618,
+      "logits/rejected": -2.206366777420044,
+      "logps/chosen": -559.8201904296875,
+      "logps/rejected": -476.9498596191406,
+      "loss": 0.0418,
+      "rewards/accuracies": 0.800000011920929,
+      "rewards/chosen": -1.3187508583068848,
+      "rewards/margins": 2.246459722518921,
+      "rewards/rejected": -3.5652108192443848,
       "step": 320
     },
     {
       "epoch": 0.81,
       "learning_rate": 5.0572206951246e-08,
-      "logits/chosen": -2.315218687057495,
-      "logits/rejected": -2.2397372722625732,
-      "logps/chosen": -560.6382446289062,
-      "logps/rejected": -482.7513732910156,
-      "loss": 0.0347,
-      "rewards/accuracies": 0.800000011920929,
-      "rewards/chosen": -1.36269211769104,
-      "rewards/margins": 2.1836559772491455,
-      "rewards/rejected": -3.5463478565216064,
+      "logits/chosen": -2.2584736347198486,
+      "logits/rejected": -2.124389410018921,
+      "logps/chosen": -562.3218994140625,
+      "logps/rejected": -474.72393798828125,
+      "loss": 0.0445,
+      "rewards/accuracies": 0.8812500238418579,
+      "rewards/chosen": -1.0979220867156982,
+      "rewards/margins": 2.4956517219543457,
+      "rewards/rejected": -3.593573808670044,
       "step": 330
     },
     {
       "epoch": 0.84,
       "learning_rate": 3.831895019292897e-08,
-      "logits/chosen": -2.3570194244384766,
-      "logits/rejected": -2.2624592781066895,
-      "logps/chosen": -610.1851806640625,
-      "logps/rejected": -510.604248046875,
-      "loss": 0.0387,
-      "rewards/accuracies": 0.7875000238418579,
-      "rewards/chosen": -1.4758695363998413,
-      "rewards/margins": 2.291604518890381,
-      "rewards/rejected": -3.767474412918091,
+      "logits/chosen": -2.296125888824463,
+      "logits/rejected": -2.1779675483703613,
+      "logps/chosen": -574.4061889648438,
+      "logps/rejected": -500.864501953125,
+      "loss": 0.0416,
+      "rewards/accuracies": 0.856249988079071,
+      "rewards/chosen": -1.0942927598953247,
+      "rewards/margins": 2.5931239128112793,
+      "rewards/rejected": -3.6874160766601562,
       "step": 340
     },
     {
       "epoch": 0.86,
       "learning_rate": 2.764152339909756e-08,
-      "logits/chosen": -2.341916561126709,
-      "logits/rejected": -2.257068157196045,
-      "logps/chosen": -559.6868286132812,
-      "logps/rejected": -464.418701171875,
-      "loss": 0.0324,
+      "logits/chosen": -2.2426624298095703,
+      "logits/rejected": -2.1238327026367188,
+      "logps/chosen": -519.2926025390625,
+      "logps/rejected": -443.22265625,
+      "loss": 0.0479,
       "rewards/accuracies": 0.8062499761581421,
-      "rewards/chosen": -1.2571017742156982,
-      "rewards/margins": 2.2919716835021973,
-      "rewards/rejected": -3.5490734577178955,
+      "rewards/chosen": -1.2578840255737305,
+      "rewards/margins": 2.1321566104888916,
+      "rewards/rejected": -3.390040636062622,
       "step": 350
     },
     {
       "epoch": 0.89,
       "learning_rate": 1.861941317991664e-08,
-      "logits/chosen": -2.354764699935913,
-      "logits/rejected": -2.2795515060424805,
-      "logps/chosen": -564.1604614257812,
-      "logps/rejected": -485.11944580078125,
-      "loss": 0.0352,
-      "rewards/accuracies": 0.8125,
-      "rewards/chosen": -1.2872124910354614,
-      "rewards/margins": 2.3846843242645264,
-      "rewards/rejected": -3.6718971729278564,
+      "logits/chosen": -2.227678060531616,
+      "logits/rejected": -2.120879888534546,
+      "logps/chosen": -526.3479614257812,
+      "logps/rejected": -452.90447998046875,
+      "loss": 0.0426,
+      "rewards/accuracies": 0.768750011920929,
+      "rewards/chosen": -1.417136788368225,
+      "rewards/margins": 1.8834213018417358,
+      "rewards/rejected": -3.3005576133728027,
       "step": 360
     },
     {
       "epoch": 0.91,
       "learning_rate": 1.13197833728636e-08,
-      "logits/chosen": -2.319873809814453,
-      "logits/rejected": -2.2500181198120117,
-      "logps/chosen": -572.5164794921875,
-      "logps/rejected": -500.48760986328125,
-      "loss": 0.0369,
-      "rewards/accuracies": 0.8062499761581421,
-      "rewards/chosen": -1.1619113683700562,
-      "rewards/margins": 2.508751392364502,
-      "rewards/rejected": -3.6706624031066895,
+      "logits/chosen": -2.2354464530944824,
+      "logits/rejected": -2.125568151473999,
+      "logps/chosen": -536.7144775390625,
+      "logps/rejected": -455.36993408203125,
+      "loss": 0.0411,
+      "rewards/accuracies": 0.800000011920929,
+      "rewards/chosen": -1.1943690776824951,
+      "rewards/margins": 2.2170112133026123,
+      "rewards/rejected": -3.4113802909851074,
       "step": 370
     },
     {
       "epoch": 0.94,
       "learning_rate": 5.79697505093521e-09,
-      "logits/chosen": -2.334282636642456,
-      "logits/rejected": -2.2317328453063965,
-      "logps/chosen": -550.3582153320312,
-      "logps/rejected": -479.588623046875,
-      "loss": 0.0448,
-      "rewards/accuracies": 0.856249988079071,
-      "rewards/chosen": -1.32636559009552,
-      "rewards/margins": 2.201442003250122,
-      "rewards/rejected": -3.5278077125549316,
+      "logits/chosen": -2.2122418880462646,
+      "logits/rejected": -2.113687515258789,
+      "logps/chosen": -501.590087890625,
+      "logps/rejected": -421.33984375,
+      "loss": 0.0386,
+      "rewards/accuracies": 0.78125,
+      "rewards/chosen": -1.1698087453842163,
+      "rewards/margins": 1.9851791858673096,
+      "rewards/rejected": -3.1549878120422363,
       "step": 380
     },
     {
       "epoch": 0.96,
       "learning_rate": 2.092101988131256e-09,
-      "logits/chosen": -2.3516108989715576,
-      "logits/rejected": -2.2636828422546387,
-      "logps/chosen": -587.2901611328125,
-      "logps/rejected": -482.85198974609375,
-      "loss": 0.0351,
-      "rewards/accuracies": 0.8062499761581421,
-      "rewards/chosen": -1.2399132251739502,
-      "rewards/margins": 2.379091739654541,
-      "rewards/rejected": -3.619004726409912,
+      "logits/chosen": -2.291910409927368,
+      "logits/rejected": -2.174203395843506,
+      "logps/chosen": -549.9415893554688,
+      "logps/rejected": -461.7567443847656,
+      "loss": 0.0376,
+      "rewards/accuracies": 0.800000011920929,
+      "rewards/chosen": -1.1343244314193726,
+      "rewards/margins": 2.2652435302734375,
+      "rewards/rejected": -3.3995678424835205,
       "step": 390
     },
     {
       "epoch": 0.99,
       "learning_rate": 2.327445937151673e-10,
-      "logits/chosen": -2.351771831512451,
-      "logits/rejected": -2.2641963958740234,
-      "logps/chosen": -545.67822265625,
-      "logps/rejected": -469.13568115234375,
-      "loss": 0.039,
-      "rewards/accuracies": 0.800000011920929,
-      "rewards/chosen": -1.5256686210632324,
-      "rewards/margins": 2.0123348236083984,
-      "rewards/rejected": -3.538003444671631,
+      "logits/chosen": -2.214627742767334,
+      "logits/rejected": -2.1191391944885254,
+      "logps/chosen": -537.9795532226562,
+      "logps/rejected": -451.3349609375,
+      "loss": 0.0457,
+      "rewards/accuracies": 0.84375,
+      "rewards/chosen": -1.1944307088851929,
+      "rewards/margins": 2.0842816829681396,
+      "rewards/rejected": -3.278712511062622,
       "step": 400
     },
     {
       "epoch": 1.0,
       "step": 405,
       "total_flos": 0.0,
-      "train_loss": 0.08520494085035206,
-      "train_runtime": 3257.6401,
-      "train_samples_per_second": 15.93,
-      "train_steps_per_second": 0.124
+      "train_loss": 0.08637741568647785,
+      "train_runtime": 3241.0704,
+      "train_samples_per_second": 16.011,
+      "train_steps_per_second": 0.125
     }
   ],
   "logging_steps": 10,

training_args.bin

@@ -1,3 +1,3 @@
 version https://git-lfs.github.com/spec/v1
-oid sha256:b7deba97c0b442a6672b07e7b7c225827b2e1a1e18ad8c3c0925d66cddd6281e
+oid sha256:d71b48e7a062b5b93db1d9a47f6551977b7c6cb8be396f9ad2015197910fd029
 size 5944