Model save

README.md

@@ -32,7 +32,7 @@ More information needed
 ### Training hyperparameters
 
 The following hyperparameters were used during training:
-- learning_rate: 5e-07
+- learning_rate: 3e-07
 - train_batch_size: 2
 - eval_batch_size: 8
 - seed: 1

all_results.json

@@ -1,8 +1,8 @@
 {
     "epoch": 1.0,
-    "train_loss": 0.663885760638449,
-    "train_runtime": 4268.7101,
-    "train_samples": 36897,
-    "train_samples_per_second": 8.644,
-    "train_steps_per_second": 0.067
+    "train_loss": 0.08108650413449206,
+    "train_runtime": 6869.1394,
+    "train_samples": 61134,
+    "train_samples_per_second": 8.9,
+    "train_steps_per_second": 0.069
 }

model-00001-of-00004.safetensors

@@ -1,3 +1,3 @@
 version https://git-lfs.github.com/spec/v1
-oid sha256:fd7540df4012cff72a7893bd996c2beda58421b6b2abf08da45fa0ee92680f45
+oid sha256:9e8fb30bfef374b2fb1936f5d61ecea9d8e4fb4d59f500aa087b6a42e641a8d2
 size 4976698672

model-00002-of-00004.safetensors

@@ -1,3 +1,3 @@
 version https://git-lfs.github.com/spec/v1
-oid sha256:3127229c4d6d16a97b9549926251f1607cbb7f90f95f3d164dec7a5b8f768fbb
+oid sha256:a83b987988a667bda420f68e1ec877a38db6e62191cd7fa0c04f9ae9fb533b13
 size 4999802720

model-00003-of-00004.safetensors

@@ -1,3 +1,3 @@
 version https://git-lfs.github.com/spec/v1
-oid sha256:c6731926ade047710eb1af86f6ac4fd269f1520887cba3438793bdda2447f1d1
+oid sha256:93811f6606f9d46e4be9e0e0be5754c75e8114961527ec5216c305097fc311cb
 size 4915916176

model-00004-of-00004.safetensors

@@ -1,3 +1,3 @@
 version https://git-lfs.github.com/spec/v1
-oid sha256:1a9274e29e7ea070f23e5c9c22241259dd91c76a2c2bff8eead27cf379bb5160
+oid sha256:a42e8fa97ce4d98209b7ccd68bfb0c129df90ecddebd1b578637ea3d64d33fa7
 size 1168138808

train_results.json

@@ -1,8 +1,8 @@
 {
     "epoch": 1.0,
-    "train_loss": 0.663885760638449,
-    "train_runtime": 4268.7101,
-    "train_samples": 36897,
-    "train_samples_per_second": 8.644,
-    "train_steps_per_second": 0.067
+    "train_loss": 0.08108650413449206,
+    "train_runtime": 6869.1394,
+    "train_samples": 61134,
+    "train_samples_per_second": 8.9,
+    "train_steps_per_second": 0.069
 }

trainer_state.json

@@ -1,21 +1,21 @@
 {
   "best_metric": null,
   "best_model_checkpoint": null,
-  "epoch": 0.9986996098829649,
-  "eval_steps": 1000,
-  "global_step": 288,
+  "epoch": 0.998691442030882,
+  "eval_steps": 10000,
+  "global_step": 477,
   "is_hyper_param_search": false,
   "is_local_process_zero": true,
   "is_world_process_zero": true,
   "log_history": [
     {
       "epoch": 0.0,
-      "learning_rate": 1.7241379310344825e-08,
-      "logits/chosen": 0.07189789414405823,
-      "logits/rejected": 0.02341889962553978,
-      "logps/chosen": -340.49188232421875,
-      "logps/rejected": -364.30450439453125,
-      "loss": 0.6931,
+      "learning_rate": 6.25e-09,
+      "logits/chosen": 0.23429542779922485,
+      "logits/rejected": 0.15932609140872955,
+      "logps/chosen": -461.0442199707031,
+      "logps/rejected": -483.8976745605469,
+      "loss": 0.1091,
       "rewards/accuracies": 0.0,
       "rewards/chosen": 0.0,
       "rewards/margins": 0.0,
@@ -23,411 +23,677 @@
       "step": 1
     },
     {
-      "epoch": 0.03,
-      "learning_rate": 1.7241379310344828e-07,
-      "logits/chosen": -0.09336617588996887,
-      "logits/rejected": -0.07832535356283188,
-      "logps/chosen": -238.44610595703125,
-      "logps/rejected": -242.68234252929688,
-      "loss": 0.6933,
+      "epoch": 0.02,
+      "learning_rate": 6.25e-08,
+      "logits/chosen": 0.10776016861200333,
+      "logits/rejected": 0.20574699342250824,
+      "logps/chosen": -404.8640441894531,
+      "logps/rejected": -347.9327392578125,
+      "loss": 0.1039,
       "rewards/accuracies": 0.4027777910232544,
-      "rewards/chosen": -0.0006303265690803528,
-      "rewards/margins": -0.0006773172062821686,
-      "rewards/rejected": 4.699068085756153e-05,
+      "rewards/chosen": 0.00012690963922068477,
+      "rewards/margins": -0.00029104368877597153,
+      "rewards/rejected": 0.00041795324068516493,
       "step": 10
     },
     {
-      "epoch": 0.07,
-      "learning_rate": 3.4482758620689656e-07,
-      "logits/chosen": -0.0782625749707222,
-      "logits/rejected": -0.08936318010091782,
-      "logps/chosen": -234.64584350585938,
-      "logps/rejected": -245.7167510986328,
-      "loss": 0.6931,
-      "rewards/accuracies": 0.4625000059604645,
-      "rewards/chosen": -0.002323926892131567,
-      "rewards/margins": 3.3687818358885124e-05,
-      "rewards/rejected": -0.002357614692300558,
+      "epoch": 0.04,
+      "learning_rate": 1.25e-07,
+      "logits/chosen": 0.018923137336969376,
+      "logits/rejected": 0.14658060669898987,
+      "logps/chosen": -374.73785400390625,
+      "logps/rejected": -331.5550842285156,
+      "loss": 0.1075,
+      "rewards/accuracies": 0.53125,
+      "rewards/chosen": 0.0007796655409038067,
+      "rewards/margins": 0.0010388566879555583,
+      "rewards/rejected": -0.0002591912052594125,
       "step": 20
     },
     {
-      "epoch": 0.1,
-      "learning_rate": 4.999816090185887e-07,
-      "logits/chosen": -0.0121148731559515,
-      "logits/rejected": -0.024416491389274597,
-      "logps/chosen": -217.1211700439453,
-      "logps/rejected": -226.70932006835938,
-      "loss": 0.6921,
-      "rewards/accuracies": 0.59375,
-      "rewards/chosen": -0.0074616484344005585,
-      "rewards/margins": 0.003082945942878723,
-      "rewards/rejected": -0.010544594377279282,
+      "epoch": 0.06,
+      "learning_rate": 1.875e-07,
+      "logits/chosen": 0.11497511714696884,
+      "logits/rejected": 0.18564197421073914,
+      "logps/chosen": -350.32647705078125,
+      "logps/rejected": -353.6821594238281,
+      "loss": 0.1093,
+      "rewards/accuracies": 0.4937500059604645,
+      "rewards/chosen": -0.0003703935071825981,
+      "rewards/margins": 0.0006811345228925347,
+      "rewards/rejected": -0.0010515280300751328,
       "step": 30
     },
     {
-      "epoch": 0.14,
-      "learning_rate": 4.977779634200946e-07,
-      "logits/chosen": -0.024179769679903984,
-      "logits/rejected": -0.007906658574938774,
-      "logps/chosen": -255.3112335205078,
-      "logps/rejected": -269.2708740234375,
-      "loss": 0.6907,
-      "rewards/accuracies": 0.612500011920929,
-      "rewards/chosen": -0.039654143154621124,
-      "rewards/margins": 0.006476263049989939,
-      "rewards/rejected": -0.046130403876304626,
+      "epoch": 0.08,
+      "learning_rate": 2.5e-07,
+      "logits/chosen": 0.04484361782670021,
+      "logits/rejected": 0.16449061036109924,
+      "logps/chosen": -402.95098876953125,
+      "logps/rejected": -375.9701843261719,
+      "loss": 0.1087,
+      "rewards/accuracies": 0.606249988079071,
+      "rewards/chosen": 0.000664212042465806,
+      "rewards/margins": 0.0027815275825560093,
+      "rewards/rejected": -0.002117315772920847,
       "step": 40
     },
     {
-      "epoch": 0.17,
-      "learning_rate": 4.919332367333748e-07,
-      "logits/chosen": -0.03670406714081764,
-      "logits/rejected": -0.016428029164671898,
-      "logps/chosen": -265.08740234375,
-      "logps/rejected": -260.3492736816406,
-      "loss": 0.6888,
-      "rewards/accuracies": 0.5874999761581421,
-      "rewards/chosen": -0.07955841720104218,
-      "rewards/margins": 0.008933933451771736,
-      "rewards/rejected": -0.08849234133958817,
+      "epoch": 0.1,
+      "learning_rate": 2.9998391212614154e-07,
+      "logits/chosen": 0.09781336784362793,
+      "logits/rejected": 0.17049923539161682,
+      "logps/chosen": -366.68865966796875,
+      "logps/rejected": -339.86761474609375,
+      "loss": 0.1046,
+      "rewards/accuracies": 0.6812499761581421,
+      "rewards/chosen": -0.001394082559272647,
+      "rewards/margins": 0.00876125879585743,
+      "rewards/rejected": -0.010155340656638145,
       "step": 50
     },
     {
-      "epoch": 0.21,
-      "learning_rate": 4.825333168376863e-07,
-      "logits/chosen": -0.0657847449183464,
-      "logits/rejected": -0.06315360963344574,
-      "logps/chosen": -231.0303192138672,
-      "logps/rejected": -239.0838623046875,
-      "loss": 0.6821,
-      "rewards/accuracies": 0.6625000238418579,
-      "rewards/chosen": -0.09296369552612305,
-      "rewards/margins": 0.021337883546948433,
-      "rewards/rejected": -0.11430158466100693,
+      "epoch": 0.13,
+      "learning_rate": 2.9942119880575817e-07,
+      "logits/chosen": 0.07368180900812149,
+      "logits/rejected": 0.22295121848583221,
+      "logps/chosen": -350.83502197265625,
+      "logps/rejected": -331.54931640625,
+      "loss": 0.1047,
+      "rewards/accuracies": 0.668749988079071,
+      "rewards/chosen": -0.004109091125428677,
+      "rewards/margins": 0.01390580553561449,
+      "rewards/rejected": -0.018014894798398018,
       "step": 60
     },
     {
-      "epoch": 0.24,
-      "learning_rate": 4.697163349457114e-07,
-      "logits/chosen": -0.06655992567539215,
-      "logits/rejected": -0.047813381999731064,
-      "logps/chosen": -284.04376220703125,
-      "logps/rejected": -287.72479248046875,
-      "loss": 0.6795,
-      "rewards/accuracies": 0.625,
-      "rewards/chosen": -0.2086440771818161,
-      "rewards/margins": 0.0424337163567543,
-      "rewards/rejected": -0.2510778307914734,
+      "epoch": 0.15,
+      "learning_rate": 2.980575393956869e-07,
+      "logits/chosen": 0.08261877298355103,
+      "logits/rejected": 0.16399553418159485,
+      "logps/chosen": -386.51324462890625,
+      "logps/rejected": -377.33319091796875,
+      "loss": 0.1079,
+      "rewards/accuracies": 0.606249988079071,
+      "rewards/chosen": -0.0024258007761090994,
+      "rewards/margins": 0.0236683189868927,
+      "rewards/rejected": -0.026094119995832443,
       "step": 70
     },
     {
-      "epoch": 0.28,
-      "learning_rate": 4.5367063577435216e-07,
-      "logits/chosen": -0.0791543573141098,
-      "logits/rejected": -0.07205502688884735,
-      "logps/chosen": -270.3536071777344,
-      "logps/rejected": -282.2701416015625,
-      "loss": 0.6759,
-      "rewards/accuracies": 0.5625,
-      "rewards/chosen": -0.28178995847702026,
-      "rewards/margins": 0.035790979862213135,
-      "rewards/rejected": -0.3175809383392334,
+      "epoch": 0.17,
+      "learning_rate": 2.9590024355266256e-07,
+      "logits/chosen": 0.135382741689682,
+      "logits/rejected": 0.22823040187358856,
+      "logps/chosen": -374.9875793457031,
+      "logps/rejected": -345.3629150390625,
+      "loss": 0.104,
+      "rewards/accuracies": 0.7250000238418579,
+      "rewards/chosen": -0.0036841861438006163,
+      "rewards/margins": 0.03979547321796417,
+      "rewards/rejected": -0.043479662388563156,
       "step": 80
     },
     {
-      "epoch": 0.31,
-      "learning_rate": 4.3463200983134283e-07,
-      "logits/chosen": -0.10235365480184555,
-      "logits/rejected": -0.07280973345041275,
-      "logps/chosen": -290.5710754394531,
-      "logps/rejected": -296.89007568359375,
-      "loss": 0.6788,
+      "epoch": 0.19,
+      "learning_rate": 2.929608750821129e-07,
+      "logits/chosen": 0.038639623671770096,
+      "logits/rejected": 0.19400770962238312,
+      "logps/chosen": -394.331787109375,
+      "logps/rejected": -327.37615966796875,
+      "loss": 0.1078,
       "rewards/accuracies": 0.6187499761581421,
-      "rewards/chosen": -0.3969407081604004,
-      "rewards/margins": 0.024875154718756676,
-      "rewards/rejected": -0.4218158721923828,
+      "rewards/chosen": -0.011807149276137352,
+      "rewards/margins": 0.04398493841290474,
+      "rewards/rejected": -0.05579209327697754,
       "step": 90
     },
     {
-      "epoch": 0.35,
-      "learning_rate": 4.128802284890448e-07,
-      "logits/chosen": -0.0756329670548439,
-      "logits/rejected": -0.060209862887859344,
-      "logps/chosen": -280.96728515625,
-      "logps/rejected": -284.9563903808594,
-      "loss": 0.6687,
-      "rewards/accuracies": 0.637499988079071,
-      "rewards/chosen": -0.37193718552589417,
-      "rewards/margins": 0.04876362159848213,
-      "rewards/rejected": -0.4207007884979248,
+      "epoch": 0.21,
+      "learning_rate": 2.892551899524109e-07,
+      "logits/chosen": 0.14743080735206604,
+      "logits/rejected": 0.3162938952445984,
+      "logps/chosen": -392.5331726074219,
+      "logps/rejected": -354.99078369140625,
+      "loss": 0.11,
+      "rewards/accuracies": 0.6875,
+      "rewards/chosen": -0.0037485682405531406,
+      "rewards/margins": 0.06982797384262085,
+      "rewards/rejected": -0.07357654720544815,
       "step": 100
     },
     {
-      "epoch": 0.38,
-      "learning_rate": 3.8873493276229006e-07,
-      "logits/chosen": -0.08315610140562057,
-      "logits/rejected": -0.04690064117312431,
-      "logps/chosen": -294.016845703125,
-      "logps/rejected": -299.04913330078125,
-      "loss": 0.6664,
-      "rewards/accuracies": 0.637499988079071,
-      "rewards/chosen": -0.30899888277053833,
-      "rewards/margins": 0.0818217471241951,
-      "rewards/rejected": -0.3908206820487976,
+      "epoch": 0.23,
+      "learning_rate": 2.848030518377739e-07,
+      "logits/chosen": 0.19383244216442108,
+      "logits/rejected": 0.2316429167985916,
+      "logps/chosen": -372.3023681640625,
+      "logps/rejected": -364.59344482421875,
+      "loss": 0.1093,
+      "rewards/accuracies": 0.6187499761581421,
+      "rewards/chosen": -0.01244374644011259,
+      "rewards/margins": 0.06392152607440948,
+      "rewards/rejected": -0.07636527717113495,
       "step": 110
     },
     {
-      "epoch": 0.42,
-      "learning_rate": 3.625509362044183e-07,
-      "logits/chosen": -0.110531747341156,
-      "logits/rejected": -0.11378206312656403,
-      "logps/chosen": -281.16558837890625,
-      "logps/rejected": -315.90399169921875,
-      "loss": 0.6619,
-      "rewards/accuracies": 0.65625,
-      "rewards/chosen": -0.36906224489212036,
-      "rewards/margins": 0.1327691376209259,
-      "rewards/rejected": -0.5018314123153687,
+      "epoch": 0.25,
+      "learning_rate": 2.7962832564252725e-07,
+      "logits/chosen": 0.17423847317695618,
+      "logits/rejected": 0.2554207742214203,
+      "logps/chosen": -380.5479736328125,
+      "logps/rejected": -387.89166259765625,
+      "loss": 0.1018,
+      "rewards/accuracies": 0.612500011920929,
+      "rewards/chosen": -0.017248453572392464,
+      "rewards/margins": 0.0897396057844162,
+      "rewards/rejected": -0.10698805749416351,
       "step": 120
     },
     {
-      "epoch": 0.45,
-      "learning_rate": 3.347130109451422e-07,
-      "logits/chosen": -0.13699400424957275,
-      "logits/rejected": -0.14944712817668915,
-      "logps/chosen": -271.65692138671875,
-      "logps/rejected": -296.51812744140625,
-      "loss": 0.6656,
-      "rewards/accuracies": 0.643750011920929,
-      "rewards/chosen": -0.36346161365509033,
-      "rewards/margins": 0.11088894307613373,
-      "rewards/rejected": -0.47435054183006287,
+      "epoch": 0.27,
+      "learning_rate": 2.737587495774764e-07,
+      "logits/chosen": 0.1611114740371704,
+      "logits/rejected": 0.3463302254676819,
+      "logps/chosen": -399.5164794921875,
+      "logps/rejected": -352.5214538574219,
+      "loss": 0.1062,
+      "rewards/accuracies": 0.6312500238418579,
+      "rewards/chosen": -0.03395389765501022,
+      "rewards/margins": 0.08507218956947327,
+      "rewards/rejected": -0.11902608722448349,
       "step": 130
     },
     {
-      "epoch": 0.49,
-      "learning_rate": 3.056302334890786e-07,
-      "logits/chosen": -0.07364296168088913,
-      "logits/rejected": -0.05889255553483963,
-      "logps/chosen": -255.48770141601562,
-      "logps/rejected": -275.1783142089844,
-      "loss": 0.6548,
-      "rewards/accuracies": 0.625,
-      "rewards/chosen": -0.28288859128952026,
-      "rewards/margins": 0.10128460824489594,
-      "rewards/rejected": -0.3841732144355774,
+      "epoch": 0.29,
+      "learning_rate": 2.672257864741005e-07,
+      "logits/chosen": 0.16581089794635773,
+      "logits/rejected": 0.3703039288520813,
+      "logps/chosen": -394.5074157714844,
+      "logps/rejected": -340.9477844238281,
+      "loss": 0.097,
+      "rewards/accuracies": 0.643750011920929,
+      "rewards/chosen": -0.056848086416721344,
+      "rewards/margins": 0.10620398819446564,
+      "rewards/rejected": -0.16305206716060638,
       "step": 140
     },
     {
-      "epoch": 0.52,
-      "learning_rate": 2.757299733630681e-07,
-      "logits/chosen": -0.1250859946012497,
-      "logits/rejected": -0.08450163900852203,
-      "logps/chosen": -283.5560607910156,
-      "logps/rejected": -298.326416015625,
-      "loss": 0.66,
-      "rewards/accuracies": 0.612500011920929,
-      "rewards/chosen": -0.39188164472579956,
-      "rewards/margins": 0.11459298431873322,
-      "rewards/rejected": -0.5064746737480164,
+      "epoch": 0.31,
+      "learning_rate": 2.6006445513357056e-07,
+      "logits/chosen": 0.29590731859207153,
+      "logits/rejected": 0.3648499846458435,
+      "logps/chosen": -397.1217346191406,
+      "logps/rejected": -389.2296142578125,
+      "loss": 0.0999,
+      "rewards/accuracies": 0.6812499761581421,
+      "rewards/chosen": -0.046468690037727356,
+      "rewards/margins": 0.14215725660324097,
+      "rewards/rejected": -0.18862596154212952,
       "step": 150
     },
     {
-      "epoch": 0.55,
-      "learning_rate": 2.454516129487201e-07,
-      "logits/chosen": -0.06747808307409286,
-      "logits/rejected": -0.04179771617054939,
-      "logps/chosen": -280.21014404296875,
-      "logps/rejected": -300.32843017578125,
-      "loss": 0.6516,
-      "rewards/accuracies": 0.6625000238418579,
-      "rewards/chosen": -0.40715375542640686,
-      "rewards/margins": 0.13689930737018585,
-      "rewards/rejected": -0.5440530776977539,
+      "epoch": 0.33,
+      "learning_rate": 2.523131426146173e-07,
+      "logits/chosen": 0.2799706757068634,
+      "logits/rejected": 0.3632858991622925,
+      "logps/chosen": -362.9356384277344,
+      "logps/rejected": -355.08538818359375,
+      "loss": 0.0958,
+      "rewards/accuracies": 0.71875,
+      "rewards/chosen": -0.049587804824113846,
+      "rewards/margins": 0.15337970852851868,
+      "rewards/rejected": -0.20296749472618103,
       "step": 160
     },
     {
-      "epoch": 0.59,
-      "learning_rate": 2.1524009078683714e-07,
-      "logits/chosen": -0.08217141777276993,
-      "logits/rejected": -0.09105993807315826,
-      "logps/chosen": -314.1512756347656,
-      "logps/rejected": -320.2287902832031,
-      "loss": 0.6533,
-      "rewards/accuracies": 0.65625,
-      "rewards/chosen": -0.5139371156692505,
-      "rewards/margins": 0.09399164468050003,
-      "rewards/rejected": -0.6079288125038147,
+      "epoch": 0.36,
+      "learning_rate": 2.440133984664454e-07,
+      "logits/chosen": 0.2953569293022156,
+      "logits/rejected": 0.3893248438835144,
+      "logps/chosen": -401.3814392089844,
+      "logps/rejected": -389.5867004394531,
+      "loss": 0.0914,
+      "rewards/accuracies": 0.6875,
+      "rewards/chosen": -0.0808958113193512,
+      "rewards/margins": 0.1940537691116333,
+      "rewards/rejected": -0.2749496102333069,
       "step": 170
     },
     {
-      "epoch": 0.62,
-      "learning_rate": 1.8553936323444056e-07,
-      "logits/chosen": -0.0693628191947937,
-      "logits/rejected": -0.06529299169778824,
-      "logps/chosen": -270.3302917480469,
-      "logps/rejected": -287.81878662109375,
-      "loss": 0.656,
-      "rewards/accuracies": 0.6875,
-      "rewards/chosen": -0.4169047772884369,
-      "rewards/margins": 0.16736415028572083,
-      "rewards/rejected": -0.5842689275741577,
+      "epoch": 0.38,
+      "learning_rate": 2.3520971200967334e-07,
+      "logits/chosen": 0.2695739269256592,
+      "logits/rejected": 0.4669845700263977,
+      "logps/chosen": -395.2998046875,
+      "logps/rejected": -387.2867736816406,
+      "loss": 0.0854,
+      "rewards/accuracies": 0.6937500238418579,
+      "rewards/chosen": -0.13700233399868011,
+      "rewards/margins": 0.18338099122047424,
+      "rewards/rejected": -0.32038331031799316,
       "step": 180
     },
     {
-      "epoch": 0.66,
-      "learning_rate": 1.5678588055492286e-07,
-      "logits/chosen": -0.06512834131717682,
-      "logits/rejected": -0.03425800800323486,
-      "logps/chosen": -281.94659423828125,
-      "logps/rejected": -297.2424621582031,
-      "loss": 0.6547,
-      "rewards/accuracies": 0.5687500238418579,
-      "rewards/chosen": -0.43236011266708374,
-      "rewards/margins": 0.12117085605859756,
-      "rewards/rejected": -0.5535309910774231,
+      "epoch": 0.4,
+      "learning_rate": 2.2594927385914543e-07,
+      "logits/chosen": 0.37522774934768677,
+      "logits/rejected": 0.5091076493263245,
+      "logps/chosen": -410.2521057128906,
+      "logps/rejected": -402.70196533203125,
+      "loss": 0.0892,
+      "rewards/accuracies": 0.6875,
+      "rewards/chosen": -0.16558507084846497,
+      "rewards/margins": 0.17272022366523743,
+      "rewards/rejected": -0.3383052945137024,
       "step": 190
     },
     {
-      "epoch": 0.69,
-      "learning_rate": 1.294021733097591e-07,
-      "logits/chosen": -0.019724350422620773,
-      "logits/rejected": 0.012522307224571705,
-      "logps/chosen": -256.83782958984375,
-      "logps/rejected": -270.844970703125,
-      "loss": 0.6614,
-      "rewards/accuracies": 0.6187499761581421,
-      "rewards/chosen": -0.3552347719669342,
-      "rewards/margins": 0.08165047317743301,
-      "rewards/rejected": -0.43688520789146423,
+      "epoch": 0.42,
+      "learning_rate": 2.162817229669295e-07,
+      "logits/chosen": 0.33533427119255066,
+      "logits/rejected": 0.5285639762878418,
+      "logps/chosen": -413.41278076171875,
+      "logps/rejected": -386.916748046875,
+      "loss": 0.0809,
+      "rewards/accuracies": 0.643750011920929,
+      "rewards/chosen": -0.1629498302936554,
+      "rewards/margins": 0.21286790072917938,
+      "rewards/rejected": -0.3758177161216736,
       "step": 200
     },
     {
-      "epoch": 0.73,
-      "learning_rate": 1.0379064329933384e-07,
-      "logits/chosen": -0.063674196600914,
-      "logits/rejected": -0.05707539990544319,
-      "logps/chosen": -282.6817626953125,
-      "logps/rejected": -298.39422607421875,
-      "loss": 0.6505,
-      "rewards/accuracies": 0.6937500238418579,
-      "rewards/chosen": -0.3917466402053833,
-      "rewards/margins": 0.16749300062656403,
-      "rewards/rejected": -0.5592396855354309,
+      "epoch": 0.44,
+      "learning_rate": 2.0625888054143427e-07,
+      "logits/chosen": 0.4836158752441406,
+      "logits/rejected": 0.7261173129081726,
+      "logps/chosen": -387.7673034667969,
+      "logps/rejected": -366.74945068359375,
+      "loss": 0.0766,
+      "rewards/accuracies": 0.6625000238418579,
+      "rewards/chosen": -0.17518684267997742,
+      "rewards/margins": 0.2312859296798706,
+      "rewards/rejected": -0.40647274255752563,
       "step": 210
     },
     {
-      "epoch": 0.76,
-      "learning_rate": 8.032765029460626e-08,
-      "logits/chosen": -0.07142199575901031,
-      "logits/rejected": -0.08730930835008621,
-      "logps/chosen": -300.83966064453125,
-      "logps/rejected": -317.92083740234375,
-      "loss": 0.6442,
+      "epoch": 0.46,
+      "learning_rate": 1.9593447226892386e-07,
+      "logits/chosen": 0.30768758058547974,
+      "logits/rejected": 0.5904788970947266,
+      "logps/chosen": -398.9160461425781,
+      "logps/rejected": -361.37255859375,
+      "loss": 0.078,
       "rewards/accuracies": 0.668749988079071,
-      "rewards/chosen": -0.453564316034317,
-      "rewards/margins": 0.15061169862747192,
-      "rewards/rejected": -0.6041759252548218,
+      "rewards/chosen": -0.24573536217212677,
+      "rewards/margins": 0.17728066444396973,
+      "rewards/rejected": -0.4230160117149353,
       "step": 220
     },
     {
-      "epoch": 0.8,
-      "learning_rate": 5.935798145470638e-08,
-      "logits/chosen": 0.0016148403519764543,
-      "logits/rejected": -0.010762521997094154,
-      "logps/chosen": -303.57305908203125,
-      "logps/rejected": -325.7623291015625,
-      "loss": 0.6602,
-      "rewards/accuracies": 0.675000011920929,
-      "rewards/chosen": -0.45530611276626587,
-      "rewards/margins": 0.17097248136997223,
-      "rewards/rejected": -0.6262785792350769,
+      "epoch": 0.48,
+      "learning_rate": 1.853638403264141e-07,
+      "logits/chosen": 0.4358161389827728,
+      "logits/rejected": 0.5574854016304016,
+      "logps/chosen": -410.8336486816406,
+      "logps/rejected": -385.32843017578125,
+      "loss": 0.0775,
+      "rewards/accuracies": 0.6312500238418579,
+      "rewards/chosen": -0.26229292154312134,
+      "rewards/margins": 0.1846790909767151,
+      "rewards/rejected": -0.44697198271751404,
       "step": 230
     },
     {
-      "epoch": 0.83,
-      "learning_rate": 4.1189784702013784e-08,
-      "logits/chosen": -0.04069576784968376,
-      "logits/rejected": -0.03334859013557434,
-      "logps/chosen": -321.64837646484375,
-      "logps/rejected": -341.418701171875,
-      "loss": 0.6401,
-      "rewards/accuracies": 0.7124999761581421,
-      "rewards/chosen": -0.5165733098983765,
-      "rewards/margins": 0.20258226990699768,
-      "rewards/rejected": -0.7191556692123413,
+      "epoch": 0.5,
+      "learning_rate": 1.7460364672965327e-07,
+      "logits/chosen": 0.4172348380088806,
+      "logits/rejected": 0.5195719003677368,
+      "logps/chosen": -400.6038513183594,
+      "logps/rejected": -395.9486389160156,
+      "loss": 0.073,
+      "rewards/accuracies": 0.581250011920929,
+      "rewards/chosen": -0.31180769205093384,
+      "rewards/margins": 0.09180188179016113,
+      "rewards/rejected": -0.40360957384109497,
       "step": 240
     },
     {
-      "epoch": 0.87,
-      "learning_rate": 2.6090040508440854e-08,
-      "logits/chosen": 0.03200738504528999,
-      "logits/rejected": 0.04725097864866257,
-      "logps/chosen": -326.45184326171875,
-      "logps/rejected": -352.00164794921875,
-      "loss": 0.6469,
+      "epoch": 0.52,
+      "learning_rate": 1.6371156960634017e-07,
+      "logits/chosen": 0.3972249925136566,
+      "logits/rejected": 0.584726095199585,
+      "logps/chosen": -395.7649230957031,
+      "logps/rejected": -386.7807312011719,
+      "loss": 0.0692,
       "rewards/accuracies": 0.6499999761581421,
-      "rewards/chosen": -0.5509322285652161,
-      "rewards/margins": 0.19423078000545502,
-      "rewards/rejected": -0.7451630234718323,
+      "rewards/chosen": -0.27468302845954895,
+      "rewards/margins": 0.22034874558448792,
+      "rewards/rejected": -0.49503177404403687,
       "step": 250
     },
     {
-      "epoch": 0.9,
-      "learning_rate": 1.4280638634728948e-08,
-      "logits/chosen": -0.08587409555912018,
-      "logits/rejected": -0.07037679105997086,
-      "logps/chosen": -310.7172546386719,
-      "logps/rejected": -318.9738464355469,
-      "loss": 0.6455,
-      "rewards/accuracies": 0.5874999761581421,
-      "rewards/chosen": -0.5080641508102417,
-      "rewards/margins": 0.10236090421676636,
-      "rewards/rejected": -0.6104251146316528,
+      "epoch": 0.54,
+      "learning_rate": 1.527459940226516e-07,
+      "logits/chosen": 0.3482324182987213,
+      "logits/rejected": 0.5496904253959656,
+      "logps/chosen": -445.3330993652344,
+      "logps/rejected": -406.5578918457031,
+      "loss": 0.0739,
+      "rewards/accuracies": 0.6187499761581421,
+      "rewards/chosen": -0.34145236015319824,
+      "rewards/margins": 0.16801035404205322,
+      "rewards/rejected": -0.5094627141952515,
       "step": 260
     },
     {
-      "epoch": 0.94,
-      "learning_rate": 5.935117474818036e-09,
-      "logits/chosen": 0.01848868653178215,
-      "logits/rejected": 0.031166959553956985,
-      "logps/chosen": -294.63177490234375,
-      "logps/rejected": -318.97650146484375,
-      "loss": 0.6431,
-      "rewards/accuracies": 0.643750011920929,
-      "rewards/chosen": -0.4989480972290039,
-      "rewards/margins": 0.12776677310466766,
-      "rewards/rejected": -0.6267148852348328,
+      "epoch": 0.57,
+      "learning_rate": 1.4176569902035086e-07,
+      "logits/chosen": 0.4846631586551666,
+      "logits/rejected": 0.6110225915908813,
+      "logps/chosen": -413.7703552246094,
+      "logps/rejected": -426.44818115234375,
+      "loss": 0.0703,
+      "rewards/accuracies": 0.675000011920929,
+      "rewards/chosen": -0.27306729555130005,
+      "rewards/margins": 0.3064786195755005,
+      "rewards/rejected": -0.5795458555221558,
       "step": 270
     },
     {
-      "epoch": 0.97,
-      "learning_rate": 1.1761139204126968e-09,
-      "logits/chosen": 0.020739596337080002,
-      "logits/rejected": 0.025563310831785202,
-      "logps/chosen": -305.96453857421875,
-      "logps/rejected": -330.4898376464844,
-      "loss": 0.6389,
-      "rewards/accuracies": 0.6187499761581421,
-      "rewards/chosen": -0.4924357533454895,
-      "rewards/margins": 0.1837814599275589,
-      "rewards/rejected": -0.6762171387672424,
+      "epoch": 0.59,
+      "learning_rate": 1.3082954254205928e-07,
+      "logits/chosen": 0.45010700821876526,
+      "logits/rejected": 0.5123178362846375,
+      "logps/chosen": -379.3990783691406,
+      "logps/rejected": -397.47491455078125,
+      "loss": 0.07,
+      "rewards/accuracies": 0.706250011920929,
+      "rewards/chosen": -0.27282288670539856,
+      "rewards/margins": 0.2435886114835739,
+      "rewards/rejected": -0.5164114832878113,
       "step": 280
     },
+    {
+      "epoch": 0.61,
+      "learning_rate": 1.1999614593359334e-07,
+      "logits/chosen": 0.4030938148498535,
+      "logits/rejected": 0.6075268983840942,
+      "logps/chosen": -435.68695068359375,
+      "logps/rejected": -439.18341064453125,
+      "loss": 0.0662,
+      "rewards/accuracies": 0.6937500238418579,
+      "rewards/chosen": -0.3328811526298523,
+      "rewards/margins": 0.27995553612709045,
+      "rewards/rejected": -0.6128367185592651,
+      "step": 290
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 1.0932357971453743e-07,
+      "logits/chosen": 0.45960482954978943,
+      "logits/rejected": 0.5529996752738953,
+      "logps/chosen": -379.45037841796875,
+      "logps/rejected": -435.02520751953125,
+      "loss": 0.0671,
+      "rewards/accuracies": 0.6812499761581421,
+      "rewards/chosen": -0.31138747930526733,
+      "rewards/margins": 0.24528288841247559,
+      "rewards/rejected": -0.5566703081130981,
+      "step": 300
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 9.886905230142431e-08,
+      "logits/chosen": 0.46065980195999146,
+      "logits/rejected": 0.5405734777450562,
+      "logps/chosen": -383.778076171875,
+      "logps/rejected": -427.1759338378906,
+      "loss": 0.0707,
+      "rewards/accuracies": 0.706250011920929,
+      "rewards/chosen": -0.3262946903705597,
+      "rewards/margins": 0.2724382281303406,
+      "rewards/rejected": -0.5987329483032227,
+      "step": 310
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 8.868860335206678e-08,
+      "logits/chosen": 0.4287898540496826,
+      "logits/rejected": 0.6984342336654663,
+      "logps/chosen": -401.4708557128906,
+      "logps/rejected": -404.81793212890625,
+      "loss": 0.0667,
+      "rewards/accuracies": 0.6875,
+      "rewards/chosen": -0.30442744493484497,
+      "rewards/margins": 0.30125004053115845,
+      "rewards/rejected": -0.6056774854660034,
+      "step": 320
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 7.883680337481599e-08,
+      "logits/chosen": 0.48913830518722534,
+      "logits/rejected": 0.6557382941246033,
+      "logps/chosen": -392.2198791503906,
+      "logps/rejected": -431.372802734375,
+      "loss": 0.062,
+      "rewards/accuracies": 0.6875,
+      "rewards/chosen": -0.36093270778656006,
+      "rewards/margins": 0.2822378873825073,
+      "rewards/rejected": -0.6431705951690674,
+      "step": 330
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 6.936646121293653e-08,
+      "logits/chosen": 0.43700629472732544,
+      "logits/rejected": 0.635035514831543,
+      "logps/chosen": -373.20635986328125,
+      "logps/rejected": -380.046142578125,
+      "loss": 0.0678,
+      "rewards/accuracies": 0.768750011920929,
+      "rewards/chosen": -0.313152015209198,
+      "rewards/margins": 0.34167423844337463,
+      "rewards/rejected": -0.654826283454895,
+      "step": 340
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 6.032834097207888e-08,
+      "logits/chosen": 0.4287714958190918,
+      "logits/rejected": 0.4792352318763733,
+      "logps/chosen": -394.83306884765625,
+      "logps/rejected": -425.98895263671875,
+      "loss": 0.06,
+      "rewards/accuracies": 0.606249988079071,
+      "rewards/chosen": -0.36768466234207153,
+      "rewards/margins": 0.24378108978271484,
+      "rewards/rejected": -0.6114657521247864,
+      "step": 350
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 5.177088990820725e-08,
+      "logits/chosen": 0.36616459488868713,
+      "logits/rejected": 0.5968407988548279,
+      "logps/chosen": -416.37451171875,
+      "logps/rejected": -392.8791198730469,
+      "loss": 0.0622,
+      "rewards/accuracies": 0.71875,
+      "rewards/chosen": -0.3135652542114258,
+      "rewards/margins": 0.253720223903656,
+      "rewards/rejected": -0.567285418510437,
+      "step": 360
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 4.373997873459449e-08,
+      "logits/chosen": 0.39764589071273804,
+      "logits/rejected": 0.5931984186172485,
+      "logps/chosen": -437.79150390625,
+      "logps/rejected": -433.67828369140625,
+      "loss": 0.0616,
+      "rewards/accuracies": 0.6812499761581421,
+      "rewards/chosen": -0.32311198115348816,
+      "rewards/margins": 0.3664843440055847,
+      "rewards/rejected": -0.6895962953567505,
+      "step": 370
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 3.627865573992087e-08,
+      "logits/chosen": 0.4783608317375183,
+      "logits/rejected": 0.6697598695755005,
+      "logps/chosen": -391.0821228027344,
+      "logps/rejected": -391.55908203125,
+      "loss": 0.0638,
+      "rewards/accuracies": 0.6937500238418579,
+      "rewards/chosen": -0.4051932692527771,
+      "rewards/margins": 0.24185864627361298,
+      "rewards/rejected": -0.6470519304275513,
+      "step": 380
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 2.942691603548416e-08,
+      "logits/chosen": 0.32613420486450195,
+      "logits/rejected": 0.5659363865852356,
+      "logps/chosen": -433.2505798339844,
+      "logps/rejected": -424.80670166015625,
+      "loss": 0.0617,
+      "rewards/accuracies": 0.71875,
+      "rewards/chosen": -0.3367374539375305,
+      "rewards/margins": 0.34916380047798157,
+      "rewards/rejected": -0.6859012842178345,
+      "step": 390
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 2.322148716843081e-08,
+      "logits/chosen": 0.4518936574459076,
+      "logits/rejected": 0.6020244359970093,
+      "logps/chosen": -401.5884094238281,
+      "logps/rejected": -410.43115234375,
+      "loss": 0.0665,
+      "rewards/accuracies": 0.71875,
+      "rewards/chosen": -0.2888852655887604,
+      "rewards/margins": 0.31569939851760864,
+      "rewards/rejected": -0.6045846939086914,
+      "step": 400
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 1.7695632250191e-08,
+      "logits/chosen": 0.37810713052749634,
+      "logits/rejected": 0.48380547761917114,
+      "logps/chosen": -383.5816955566406,
+      "logps/rejected": -410.0541076660156,
+      "loss": 0.0612,
+      "rewards/accuracies": 0.6312500238418579,
+      "rewards/chosen": -0.37703007459640503,
+      "rewards/margins": 0.26473695039749146,
+      "rewards/rejected": -0.6417670249938965,
+      "step": 410
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 1.2878971655412513e-08,
+      "logits/chosen": 0.4940028190612793,
+      "logits/rejected": 0.6443029046058655,
+      "logps/chosen": -383.65594482421875,
+      "logps/rejected": -405.99658203125,
+      "loss": 0.0611,
+      "rewards/accuracies": 0.6812499761581421,
+      "rewards/chosen": -0.33365705609321594,
+      "rewards/margins": 0.31004443764686584,
+      "rewards/rejected": -0.643701434135437,
+      "step": 420
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 8.79732424714541e-09,
+      "logits/chosen": 0.4920971989631653,
+      "logits/rejected": 0.5803946256637573,
+      "logps/chosen": -431.6775817871094,
+      "logps/rejected": -450.4654846191406,
+      "loss": 0.0669,
+      "rewards/accuracies": 0.7250000238418579,
+      "rewards/chosen": -0.4081823229789734,
+      "rewards/margins": 0.3052729368209839,
+      "rewards/rejected": -0.713455319404602,
+      "step": 430
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 5.472568979361852e-09,
+      "logits/chosen": 0.45466095209121704,
+      "logits/rejected": 0.5505176782608032,
+      "logps/chosen": -375.49822998046875,
+      "logps/rejected": -397.58526611328125,
+      "loss": 0.063,
+      "rewards/accuracies": 0.65625,
+      "rewards/chosen": -0.3610956072807312,
+      "rewards/margins": 0.26686498522758484,
+      "rewards/rejected": -0.6279605627059937,
+      "step": 440
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 2.922527618666465e-09,
+      "logits/chosen": 0.40739598870277405,
+      "logits/rejected": 0.7280329465866089,
+      "logps/chosen": -447.08428955078125,
+      "logps/rejected": -407.3627014160156,
+      "loss": 0.0607,
+      "rewards/accuracies": 0.668749988079071,
+      "rewards/chosen": -0.4314998686313629,
+      "rewards/margins": 0.25003308057785034,
+      "rewards/rejected": -0.6815329194068909,
+      "step": 450
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 1.1608692138469378e-09,
+      "logits/chosen": 0.45475348830223083,
+      "logits/rejected": 0.7208662033081055,
+      "logps/chosen": -389.51318359375,
+      "logps/rejected": -385.88629150390625,
+      "loss": 0.0584,
+      "rewards/accuracies": 0.7124999761581421,
+      "rewards/chosen": -0.36205732822418213,
+      "rewards/margins": 0.2959592938423157,
+      "rewards/rejected": -0.6580166220664978,
+      "step": 460
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 1.9703682533901976e-10,
+      "logits/chosen": 0.40651655197143555,
+      "logits/rejected": 0.6511154770851135,
+      "logps/chosen": -431.1214904785156,
+      "logps/rejected": -419.0870666503906,
+      "loss": 0.0629,
+      "rewards/accuracies": 0.668749988079071,
+      "rewards/chosen": -0.3773760199546814,
+      "rewards/margins": 0.311960905790329,
+      "rewards/rejected": -0.689336895942688,
+      "step": 470
+    },
     {
       "epoch": 1.0,
-      "step": 288,
+      "step": 477,
       "total_flos": 0.0,
-      "train_loss": 0.663885760638449,
-      "train_runtime": 4268.7101,
-      "train_samples_per_second": 8.644,
-      "train_steps_per_second": 0.067
+      "train_loss": 0.08108650413449206,
+      "train_runtime": 6869.1394,
+      "train_samples_per_second": 8.9,
+      "train_steps_per_second": 0.069
     }
   ],
   "logging_steps": 10,
-  "max_steps": 288,
+  "max_steps": 477,
   "num_train_epochs": 1,
-  "save_steps": 1000,
+  "save_steps": 10000,
   "total_flos": 0.0,
   "trial_name": null,
   "trial_params": null

training_args.bin

@@ -1,3 +1,3 @@
 version https://git-lfs.github.com/spec/v1
-oid sha256:a470dca6ff665650d1f336d25b4b572bd0354fa1af0ddd3648c4975f15d0c118
+oid sha256:4f02cd55a559c2c467d5a2e292b10d09d7ecc109a155111881d3477d63372b8f
 size 5944