Model save

all_results.json

@@ -1,8 +1,8 @@
 {
     "epoch": 1.0,
-    "train_loss": 0.37752047600880473,
-    "train_runtime": 5207.0243,
-    "train_samples": 45548,
-    "train_samples_per_second": 8.747,
-    "train_steps_per_second": 0.068
+    "train_loss": 0.06185462889681322,
+    "train_runtime": 6360.6645,
+    "train_samples": 61134,
+    "train_samples_per_second": 9.611,
+    "train_steps_per_second": 0.075
 }

model-00001-of-00004.safetensors

@@ -1,3 +1,3 @@
 version https://git-lfs.github.com/spec/v1
-oid sha256:41f84531bc53c6aaed5ceb01e417e1c7621240bbab9fd02352fb518cb487cc5d
+oid sha256:5d383b1d023d1ed1e3584695b798e77d278f800fc398598aa6e9c9e42e3e2ad1
 size 4976698672

model-00002-of-00004.safetensors

@@ -1,3 +1,3 @@
 version https://git-lfs.github.com/spec/v1
-oid sha256:c3273f73bf9bf355f91a630daea63952a3b7a04db5efe9bebc0fa5ff1ececb25
+oid sha256:cb98d9e0bb313a95a034b93ea04d8620ae709b5c4390e5f6ee5637ed55a68aca
 size 4999802720

model-00003-of-00004.safetensors

@@ -1,3 +1,3 @@
 version https://git-lfs.github.com/spec/v1
-oid sha256:506a25cb979eb73d988e7126c35a3e373e38d0458ca3b9a253297a093b9235b6
+oid sha256:2b65b68220db877b0e0b92a876b84cf38946255244aa69d62cb4ac1c1c8fd5cd
 size 4915916176

model-00004-of-00004.safetensors

@@ -1,3 +1,3 @@
 version https://git-lfs.github.com/spec/v1
-oid sha256:9fa4998481e763210d32919a0fc1bfbd47699ac3ff332337103a8a7cd091a081
+oid sha256:641a1860fbabe30e2297682bc87a6ca0f8a090cfb74308285997d02428d923de
 size 1168138808

train_results.json

@@ -1,8 +1,8 @@
 {
     "epoch": 1.0,
-    "train_loss": 0.37752047600880473,
-    "train_runtime": 5207.0243,
-    "train_samples": 45548,
-    "train_samples_per_second": 8.747,
-    "train_steps_per_second": 0.068
+    "train_loss": 0.06185462889681322,
+    "train_runtime": 6360.6645,
+    "train_samples": 61134,
+    "train_samples_per_second": 9.611,
+    "train_steps_per_second": 0.075
 }

trainer_state.json

@@ -1,21 +1,21 @@
 {
   "best_metric": null,
   "best_model_checkpoint": null,
-  "epoch": 0.9975412715138743,
+  "epoch": 0.998691442030882,
   "eval_steps": 10000,
-  "global_step": 355,
+  "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": 5.555555555555555e-08,
-      "logits/chosen": -0.13174405694007874,
-      "logits/rejected": -0.027169257402420044,
-      "logps/chosen": -477.4691162109375,
-      "logps/rejected": -277.6482238769531,
-      "loss": 0.4106,
+      "learning_rate": 4.166666666666666e-08,
+      "logits/chosen": 0.21113795042037964,
+      "logits/rejected": 0.15092352032661438,
+      "logps/chosen": -422.9735412597656,
+      "logps/rejected": -444.97625732421875,
+      "loss": 0.1408,
       "rewards/accuracies": 0.0,
       "rewards/chosen": 0.0,
       "rewards/margins": 0.0,
@@ -23,507 +23,675 @@
       "step": 1
     },
     {
-      "epoch": 0.03,
-      "learning_rate": 5.555555555555555e-07,
-      "logits/chosen": -0.08890426158905029,
-      "logits/rejected": -0.04657414183020592,
-      "logps/chosen": -334.1969909667969,
-      "logps/rejected": -264.19439697265625,
-      "loss": 0.419,
-      "rewards/accuracies": 0.4166666567325592,
-      "rewards/chosen": -0.0018680854700505733,
-      "rewards/margins": -0.0014530097832903266,
-      "rewards/rejected": -0.0004150756576564163,
+      "epoch": 0.02,
+      "learning_rate": 4.1666666666666667e-07,
+      "logits/chosen": 0.11635181307792664,
+      "logits/rejected": 0.20632369816303253,
+      "logps/chosen": -370.0391845703125,
+      "logps/rejected": -310.927490234375,
+      "loss": 0.1373,
+      "rewards/accuracies": 0.4513888955116272,
+      "rewards/chosen": 0.00027829466853290796,
+      "rewards/margins": 0.000870375195518136,
+      "rewards/rejected": -0.0005920806434005499,
       "step": 10
     },
     {
-      "epoch": 0.06,
-      "learning_rate": 1.111111111111111e-06,
-      "logits/chosen": -0.06340397149324417,
-      "logits/rejected": -0.02432728186249733,
-      "logps/chosen": -312.5043029785156,
-      "logps/rejected": -214.32138061523438,
-      "loss": 0.429,
-      "rewards/accuracies": 0.5375000238418579,
-      "rewards/chosen": -0.001270765089429915,
-      "rewards/margins": 0.027136042714118958,
-      "rewards/rejected": -0.02840680442750454,
+      "epoch": 0.04,
+      "learning_rate": 8.333333333333333e-07,
+      "logits/chosen": 0.030925219878554344,
+      "logits/rejected": 0.16167232394218445,
+      "logps/chosen": -337.4822082519531,
+      "logps/rejected": -295.37750244140625,
+      "loss": 0.1407,
+      "rewards/accuracies": 0.6625000238418579,
+      "rewards/chosen": 0.0036961566656827927,
+      "rewards/margins": 0.007187467068433762,
+      "rewards/rejected": -0.0034913099370896816,
       "step": 20
     },
     {
-      "epoch": 0.08,
-      "learning_rate": 1.6666666666666667e-06,
-      "logits/chosen": -0.044113289564847946,
-      "logits/rejected": 0.002516481326892972,
-      "logps/chosen": -372.5130310058594,
-      "logps/rejected": -238.03060913085938,
-      "loss": 0.4539,
-      "rewards/accuracies": 0.5874999761581421,
-      "rewards/chosen": -0.01703575626015663,
-      "rewards/margins": 0.13534700870513916,
-      "rewards/rejected": -0.1523827612400055,
+      "epoch": 0.06,
+      "learning_rate": 1.2499999999999999e-06,
+      "logits/chosen": 0.15862567722797394,
+      "logits/rejected": 0.2354920357465744,
+      "logps/chosen": -314.4154052734375,
+      "logps/rejected": -320.0315856933594,
+      "loss": 0.1444,
+      "rewards/accuracies": 0.6875,
+      "rewards/chosen": 0.009034913964569569,
+      "rewards/margins": 0.03341209888458252,
+      "rewards/rejected": -0.024377182126045227,
       "step": 30
     },
     {
-      "epoch": 0.11,
-      "learning_rate": 1.999224195661986e-06,
-      "logits/chosen": -0.11524984985589981,
-      "logits/rejected": -0.06750848144292831,
-      "logps/chosen": -330.25140380859375,
-      "logps/rejected": -260.4798278808594,
-      "loss": 0.4867,
-      "rewards/accuracies": 0.606249988079071,
-      "rewards/chosen": -0.05938565731048584,
-      "rewards/margins": 0.13395321369171143,
-      "rewards/rejected": -0.19333890080451965,
+      "epoch": 0.08,
+      "learning_rate": 1.6666666666666667e-06,
+      "logits/chosen": 0.18349352478981018,
+      "logits/rejected": 0.31235384941101074,
+      "logps/chosen": -364.74676513671875,
+      "logps/rejected": -345.8722229003906,
+      "loss": 0.1495,
+      "rewards/accuracies": 0.675000011920929,
+      "rewards/chosen": 0.02523394301533699,
+      "rewards/margins": 0.08020723611116409,
+      "rewards/rejected": -0.054973285645246506,
       "step": 40
     },
     {
-      "epoch": 0.14,
-      "learning_rate": 1.9905102152171726e-06,
-      "logits/chosen": -0.15043430030345917,
-      "logits/rejected": -0.09099297970533371,
-      "logps/chosen": -333.2086181640625,
-      "logps/rejected": -272.1454772949219,
-      "loss": 0.4736,
-      "rewards/accuracies": 0.59375,
-      "rewards/chosen": 0.07267357409000397,
-      "rewards/margins": 0.18180176615715027,
-      "rewards/rejected": -0.10912821441888809,
+      "epoch": 0.1,
+      "learning_rate": 1.9998927475076105e-06,
+      "logits/chosen": 0.4573761820793152,
+      "logits/rejected": 0.5666500926017761,
+      "logps/chosen": -346.28973388671875,
+      "logps/rejected": -342.87890625,
+      "loss": 0.1114,
+      "rewards/accuracies": 0.7124999761581421,
+      "rewards/chosen": -0.16050797700881958,
+      "rewards/margins": 0.23386959731578827,
+      "rewards/rejected": -0.39437755942344666,
       "step": 50
     },
     {
-      "epoch": 0.17,
-      "learning_rate": 1.9721972279988477e-06,
-      "logits/chosen": -0.3500102758407593,
-      "logits/rejected": -0.28142938017845154,
-      "logps/chosen": -389.78326416015625,
-      "logps/rejected": -238.11001586914062,
-      "loss": 0.4607,
-      "rewards/accuracies": 0.7124999761581421,
-      "rewards/chosen": 0.18076026439666748,
-      "rewards/margins": 0.3350951075553894,
-      "rewards/rejected": -0.15433481335639954,
+      "epoch": 0.13,
+      "learning_rate": 1.9961413253717214e-06,
+      "logits/chosen": 0.5153708457946777,
+      "logits/rejected": 0.790372371673584,
+      "logps/chosen": -362.19476318359375,
+      "logps/rejected": -374.3284606933594,
+      "loss": 0.0785,
+      "rewards/accuracies": 0.6625000238418579,
+      "rewards/chosen": -0.48446550965309143,
+      "rewards/margins": 0.3205246031284332,
+      "rewards/rejected": -0.8049899935722351,
       "step": 60
     },
     {
-      "epoch": 0.2,
-      "learning_rate": 1.9444627046536053e-06,
-      "logits/chosen": -0.18015244603157043,
-      "logits/rejected": -0.05930241197347641,
-      "logps/chosen": -385.66790771484375,
-      "logps/rejected": -259.8401794433594,
-      "loss": 0.4287,
-      "rewards/accuracies": 0.6312500238418579,
-      "rewards/chosen": 0.06770263612270355,
-      "rewards/margins": 0.25656843185424805,
-      "rewards/rejected": -0.1888658106327057,
+      "epoch": 0.15,
+      "learning_rate": 1.9870502626379126e-06,
+      "logits/chosen": 0.31514567136764526,
+      "logits/rejected": 0.530523419380188,
+      "logps/chosen": -391.986328125,
+      "logps/rejected": -424.04156494140625,
+      "loss": 0.0807,
+      "rewards/accuracies": 0.625,
+      "rewards/chosen": -0.4213319420814514,
+      "rewards/margins": 0.43596163392066956,
+      "rewards/rejected": -0.8572934865951538,
       "step": 70
     },
     {
-      "epoch": 0.22,
-      "learning_rate": 1.907575419670957e-06,
-      "logits/chosen": -0.02038487419486046,
-      "logits/rejected": -0.006824444979429245,
-      "logps/chosen": -291.40362548828125,
-      "logps/rejected": -235.5683135986328,
-      "loss": 0.4392,
-      "rewards/accuracies": 0.5874999761581421,
-      "rewards/chosen": -0.12218828499317169,
-      "rewards/margins": 0.24640071392059326,
-      "rewards/rejected": -0.36858901381492615,
+      "epoch": 0.17,
+      "learning_rate": 1.972668290351084e-06,
+      "logits/chosen": 0.4700261652469635,
+      "logits/rejected": 0.7672463655471802,
+      "logps/chosen": -379.77764892578125,
+      "logps/rejected": -388.63079833984375,
+      "loss": 0.072,
+      "rewards/accuracies": 0.75,
+      "rewards/chosen": -0.40196624398231506,
+      "rewards/margins": 0.43673914670944214,
+      "rewards/rejected": -0.8387054204940796,
       "step": 80
     },
     {
-      "epoch": 0.25,
-      "learning_rate": 1.861892846697277e-06,
-      "logits/chosen": -0.07922788709402084,
-      "logits/rejected": 0.0028565465472638607,
-      "logps/chosen": -392.5978088378906,
-      "logps/rejected": -249.13351440429688,
-      "loss": 0.4232,
-      "rewards/accuracies": 0.737500011920929,
-      "rewards/chosen": 0.011292616836726665,
-      "rewards/margins": 0.4046979546546936,
-      "rewards/rejected": -0.39340537786483765,
+      "epoch": 0.19,
+      "learning_rate": 1.9530725005474194e-06,
+      "logits/chosen": -0.013982963748276234,
+      "logits/rejected": 0.312116801738739,
+      "logps/chosen": -401.41705322265625,
+      "logps/rejected": -368.51715087890625,
+      "loss": 0.0776,
+      "rewards/accuracies": 0.643750011920929,
+      "rewards/chosen": -0.44402894377708435,
+      "rewards/margins": 0.389095664024353,
+      "rewards/rejected": -0.8331245183944702,
       "step": 90
     },
     {
-      "epoch": 0.28,
-      "learning_rate": 1.8078576942687008e-06,
-      "logits/chosen": -0.0182100310921669,
-      "logits/rejected": 0.01910480670630932,
-      "logps/chosen": -365.0164489746094,
-      "logps/rejected": -280.1756286621094,
-      "loss": 0.4191,
-      "rewards/accuracies": 0.6312500238418579,
-      "rewards/chosen": -0.010161404497921467,
-      "rewards/margins": 0.17767608165740967,
-      "rewards/rejected": -0.18783748149871826,
+      "epoch": 0.21,
+      "learning_rate": 1.9283679330160725e-06,
+      "logits/chosen": 0.3012472689151764,
+      "logits/rejected": 0.662305474281311,
+      "logps/chosen": -419.91937255859375,
+      "logps/rejected": -426.59710693359375,
+      "loss": 0.0654,
+      "rewards/accuracies": 0.706250011920929,
+      "rewards/chosen": -0.6331688165664673,
+      "rewards/margins": 0.5270191431045532,
+      "rewards/rejected": -1.160187840461731,
       "step": 100
     },
     {
-      "epoch": 0.31,
-      "learning_rate": 1.7459936155350907e-06,
-      "logits/chosen": -0.06832268089056015,
-      "logits/rejected": 0.0665212869644165,
-      "logps/chosen": -330.8415832519531,
-      "logps/rejected": -285.3780822753906,
-      "loss": 0.3916,
-      "rewards/accuracies": 0.59375,
-      "rewards/chosen": -0.17425654828548431,
-      "rewards/margins": 0.1173292025923729,
-      "rewards/rejected": -0.2915857434272766,
+      "epoch": 0.23,
+      "learning_rate": 1.8986870122518259e-06,
+      "logits/chosen": 0.4358280599117279,
+      "logits/rejected": 0.5628947019577026,
+      "logps/chosen": -437.00555419921875,
+      "logps/rejected": -478.0025329589844,
+      "loss": 0.0524,
+      "rewards/accuracies": 0.65625,
+      "rewards/chosen": -1.0241305828094482,
+      "rewards/margins": 0.5563775897026062,
+      "rewards/rejected": -1.5805081129074097,
       "step": 110
     },
     {
-      "epoch": 0.34,
-      "learning_rate": 1.6769001335520179e-06,
-      "logits/chosen": 0.10953323543071747,
-      "logits/rejected": 0.1718703806400299,
-      "logps/chosen": -365.9538269042969,
-      "logps/rejected": -254.223388671875,
-      "loss": 0.3635,
-      "rewards/accuracies": 0.6937500238418579,
-      "rewards/chosen": -0.2533760368824005,
-      "rewards/margins": 0.23791372776031494,
-      "rewards/rejected": -0.49128976464271545,
+      "epoch": 0.25,
+      "learning_rate": 1.8641888376168483e-06,
+      "logits/chosen": 0.4048207700252533,
+      "logits/rejected": 0.601729154586792,
+      "logps/chosen": -457.0703125,
+      "logps/rejected": -516.7349853515625,
+      "loss": 0.0385,
+      "rewards/accuracies": 0.6625000238418579,
+      "rewards/chosen": -1.1407548189163208,
+      "rewards/margins": 0.6236680150032043,
+      "rewards/rejected": -1.7644230127334595,
       "step": 120
     },
     {
-      "epoch": 0.37,
-      "learning_rate": 1.6012468313196084e-06,
-      "logits/chosen": 0.11821316182613373,
-      "logits/rejected": 0.24304287135601044,
-      "logps/chosen": -399.622802734375,
-      "logps/rejected": -265.00897216796875,
-      "loss": 0.3923,
-      "rewards/accuracies": 0.643750011920929,
-      "rewards/chosen": -0.25938984751701355,
-      "rewards/margins": 0.34516972303390503,
-      "rewards/rejected": -0.6045595407485962,
+      "epoch": 0.27,
+      "learning_rate": 1.8250583305165094e-06,
+      "logits/chosen": 0.3097571134567261,
+      "logits/rejected": 0.6428357362747192,
+      "logps/chosen": -466.8719177246094,
+      "logps/rejected": -459.4197692871094,
+      "loss": 0.0465,
+      "rewards/accuracies": 0.6312500238418579,
+      "rewards/chosen": -1.0680186748504639,
+      "rewards/margins": 0.4782249927520752,
+      "rewards/rejected": -1.5462435483932495,
       "step": 130
     },
     {
-      "epoch": 0.39,
-      "learning_rate": 1.51976686287243e-06,
-      "logits/chosen": 0.17939393222332,
-      "logits/rejected": 0.2452642023563385,
-      "logps/chosen": -350.52655029296875,
-      "logps/rejected": -276.53729248046875,
-      "loss": 0.3797,
-      "rewards/accuracies": 0.6312500238418579,
-      "rewards/chosen": -0.316207617521286,
-      "rewards/margins": 0.16901156306266785,
-      "rewards/rejected": -0.48521918058395386,
+      "epoch": 0.29,
+      "learning_rate": 1.78150524316067e-06,
+      "logits/chosen": 0.26445186138153076,
+      "logits/rejected": 0.6199995279312134,
+      "logps/chosen": -435.9942321777344,
+      "logps/rejected": -411.48284912109375,
+      "loss": 0.0546,
+      "rewards/accuracies": 0.668749988079071,
+      "rewards/chosen": -0.8276292085647583,
+      "rewards/margins": 0.39598146080970764,
+      "rewards/rejected": -1.2236106395721436,
       "step": 140
     },
     {
-      "epoch": 0.42,
-      "learning_rate": 1.4332498483042636e-06,
-      "logits/chosen": 0.07941923290491104,
-      "logits/rejected": 0.1618000715970993,
-      "logps/chosen": -321.47528076171875,
-      "logps/rejected": -287.1752014160156,
-      "loss": 0.3807,
-      "rewards/accuracies": 0.65625,
-      "rewards/chosen": -0.33254697918891907,
-      "rewards/margins": 0.21839794516563416,
-      "rewards/rejected": -0.5509449243545532,
+      "epoch": 0.31,
+      "learning_rate": 1.7337630342238039e-06,
+      "logits/chosen": 0.4062308371067047,
+      "logits/rejected": 0.5630704164505005,
+      "logps/chosen": -418.47186279296875,
+      "logps/rejected": -450.25897216796875,
+      "loss": 0.0745,
+      "rewards/accuracies": 0.731249988079071,
+      "rewards/chosen": -0.6129023432731628,
+      "rewards/margins": 0.5388969779014587,
+      "rewards/rejected": -1.1517994403839111,
       "step": 150
     },
     {
-      "epoch": 0.45,
-      "learning_rate": 1.3425342215818716e-06,
-      "logits/chosen": 0.05771520733833313,
-      "logits/rejected": 0.1935223489999771,
-      "logps/chosen": -428.2752380371094,
-      "logps/rejected": -319.6703186035156,
-      "loss": 0.3346,
-      "rewards/accuracies": 0.6187499761581421,
-      "rewards/chosen": -0.5818016529083252,
-      "rewards/margins": 0.2607456147670746,
-      "rewards/rejected": -0.8425472378730774,
+      "epoch": 0.33,
+      "learning_rate": 1.6820876174307821e-06,
+      "logits/chosen": 0.2119346559047699,
+      "logits/rejected": 0.3295299708843231,
+      "logps/chosen": -374.5950622558594,
+      "logps/rejected": -423.4131774902344,
+      "loss": 0.0626,
+      "rewards/accuracies": 0.737500011920929,
+      "rewards/chosen": -0.5443166494369507,
+      "rewards/margins": 0.6955867409706116,
+      "rewards/rejected": -1.2399033308029175,
       "step": 160
     },
     {
-      "epoch": 0.48,
-      "learning_rate": 1.248499105304894e-06,
-      "logits/chosen": 0.04840540513396263,
-      "logits/rejected": 0.1421603113412857,
-      "logps/chosen": -381.43975830078125,
-      "logps/rejected": -337.63140869140625,
-      "loss": 0.3159,
-      "rewards/accuracies": 0.574999988079071,
-      "rewards/chosen": -0.6558400392532349,
-      "rewards/margins": 0.2480638474225998,
-      "rewards/rejected": -0.9039037823677063,
+      "epoch": 0.36,
+      "learning_rate": 1.6267559897763027e-06,
+      "logits/chosen": 0.12537510693073273,
+      "logits/rejected": 0.33151909708976746,
+      "logps/chosen": -444.45074462890625,
+      "logps/rejected": -490.6385803222656,
+      "loss": 0.0543,
+      "rewards/accuracies": 0.7437499761581421,
+      "rewards/chosen": -0.8747140169143677,
+      "rewards/margins": 0.7744320631027222,
+      "rewards/rejected": -1.6491460800170898,
       "step": 170
     },
     {
-      "epoch": 0.51,
-      "learning_rate": 1.1520557911533388e-06,
-      "logits/chosen": -0.2732602655887604,
-      "logits/rejected": -0.20177264511585236,
-      "logps/chosen": -389.3359680175781,
-      "logps/rejected": -316.3470764160156,
-      "loss": 0.3382,
-      "rewards/accuracies": 0.5249999761581421,
-      "rewards/chosen": -0.6642172336578369,
-      "rewards/margins": 0.1544409543275833,
-      "rewards/rejected": -0.8186582326889038,
+      "epoch": 0.38,
+      "learning_rate": 1.5680647467311555e-06,
+      "logits/chosen": 0.12561924755573273,
+      "logits/rejected": 0.4202500283718109,
+      "logps/chosen": -442.4149475097656,
+      "logps/rejected": -493.0331115722656,
+      "loss": 0.0438,
+      "rewards/accuracies": 0.762499988079071,
+      "rewards/chosen": -0.9780979156494141,
+      "rewards/margins": 0.7627999782562256,
+      "rewards/rejected": -1.74089777469635,
       "step": 180
     },
     {
-      "epoch": 0.53,
-      "learning_rate": 1.0541389085854176e-06,
-      "logits/chosen": -0.3075205385684967,
-      "logits/rejected": -0.2349202185869217,
-      "logps/chosen": -391.30889892578125,
-      "logps/rejected": -284.8778381347656,
-      "loss": 0.3435,
-      "rewards/accuracies": 0.6312500238418579,
-      "rewards/chosen": -0.49413275718688965,
-      "rewards/margins": 0.36992865800857544,
-      "rewards/rejected": -0.8640613555908203,
+      "epoch": 0.4,
+      "learning_rate": 1.5063284923943028e-06,
+      "logits/chosen": 0.280507355928421,
+      "logits/rejected": 0.47573453187942505,
+      "logps/chosen": -495.60870361328125,
+      "logps/rejected": -534.2664794921875,
+      "loss": 0.0388,
+      "rewards/accuracies": 0.668749988079071,
+      "rewards/chosen": -1.390926480293274,
+      "rewards/margins": 0.6419951319694519,
+      "rewards/rejected": -2.032921552658081,
       "step": 190
     },
     {
-      "epoch": 0.56,
-      "learning_rate": 9.556973673696213e-07,
-      "logits/chosen": -0.22508034110069275,
-      "logits/rejected": -0.11411086469888687,
-      "logps/chosen": -370.9524230957031,
-      "logps/rejected": -309.0107727050781,
-      "loss": 0.3695,
-      "rewards/accuracies": 0.5687500238418579,
-      "rewards/chosen": -0.6102132797241211,
-      "rewards/margins": 0.2373207062482834,
-      "rewards/rejected": -0.8475340008735657,
+      "epoch": 0.42,
+      "learning_rate": 1.4418781531128635e-06,
+      "logits/chosen": 0.23784343898296356,
+      "logits/rejected": 0.4942964017391205,
+      "logps/chosen": -456.57403564453125,
+      "logps/rejected": -482.3223571777344,
+      "loss": 0.0436,
+      "rewards/accuracies": 0.706250011920929,
+      "rewards/chosen": -0.9536474943161011,
+      "rewards/margins": 0.7362873554229736,
+      "rewards/rejected": -1.6899349689483643,
       "step": 200
     },
     {
-      "epoch": 0.59,
-      "learning_rate": 8.576851617267149e-07,
-      "logits/chosen": -0.06804222613573074,
-      "logits/rejected": 0.026474561542272568,
-      "logps/chosen": -429.52972412109375,
-      "logps/rejected": -319.5240783691406,
-      "loss": 0.3553,
-      "rewards/accuracies": 0.606249988079071,
-      "rewards/chosen": -0.4963027536869049,
-      "rewards/margins": 0.2677620053291321,
-      "rewards/rejected": -0.7640646696090698,
+      "epoch": 0.44,
+      "learning_rate": 1.3750592036095619e-06,
+      "logits/chosen": 0.3661119043827057,
+      "logits/rejected": 0.6860957145690918,
+      "logps/chosen": -410.147705078125,
+      "logps/rejected": -429.26617431640625,
+      "loss": 0.0509,
+      "rewards/accuracies": 0.7437499761581421,
+      "rewards/chosen": -0.7587505578994751,
+      "rewards/margins": 0.646103024482727,
+      "rewards/rejected": -1.4048535823822021,
       "step": 210
     },
     {
-      "epoch": 0.62,
-      "learning_rate": 7.610521251984419e-07,
-      "logits/chosen": -0.1526344269514084,
-      "logits/rejected": -0.09809421747922897,
-      "logps/chosen": -410.1642150878906,
-      "logps/rejected": -300.3708190917969,
-      "loss": 0.3397,
-      "rewards/accuracies": 0.643750011920929,
-      "rewards/chosen": -0.5530972480773926,
-      "rewards/margins": 0.2556341588497162,
-      "rewards/rejected": -0.8087313771247864,
+      "epoch": 0.46,
+      "learning_rate": 1.3062298151261591e-06,
+      "logits/chosen": 0.18747654557228088,
+      "logits/rejected": 0.5437577366828918,
+      "logps/chosen": -402.08258056640625,
+      "logps/rejected": -397.69793701171875,
+      "loss": 0.0654,
+      "rewards/accuracies": 0.699999988079071,
+      "rewards/chosen": -0.6500775218009949,
+      "rewards/margins": 0.5026715993881226,
+      "rewards/rejected": -1.1527493000030518,
       "step": 220
     },
     {
-      "epoch": 0.65,
-      "learning_rate": 6.667347258372558e-07,
-      "logits/chosen": -0.022926704958081245,
-      "logits/rejected": 0.09270887076854706,
-      "logps/chosen": -381.1971435546875,
-      "logps/rejected": -297.2896728515625,
-      "loss": 0.3384,
-      "rewards/accuracies": 0.606249988079071,
-      "rewards/chosen": -0.569217324256897,
-      "rewards/margins": 0.2501508295536041,
-      "rewards/rejected": -0.8193681836128235,
+      "epoch": 0.48,
+      "learning_rate": 1.2357589355094273e-06,
+      "logits/chosen": 0.2568260729312897,
+      "logits/rejected": 0.43051618337631226,
+      "logps/chosen": -420.9578552246094,
+      "logps/rejected": -445.6470642089844,
+      "loss": 0.0653,
+      "rewards/accuracies": 0.6812499761581421,
+      "rewards/chosen": -0.7387503385543823,
+      "rewards/margins": 0.6945470571517944,
+      "rewards/rejected": -1.4332973957061768,
       "step": 230
     },
     {
-      "epoch": 0.67,
-      "learning_rate": 5.756469909206334e-07,
-      "logits/chosen": -0.06401239335536957,
-      "logits/rejected": 0.07748265564441681,
-      "logps/chosen": -423.3412170410156,
-      "logps/rejected": -315.5213928222656,
-      "loss": 0.3281,
-      "rewards/accuracies": 0.65625,
-      "rewards/chosen": -0.5486977696418762,
-      "rewards/margins": 0.2651389539241791,
-      "rewards/rejected": -0.8138366937637329,
+      "epoch": 0.5,
+      "learning_rate": 1.1640243115310217e-06,
+      "logits/chosen": 0.18585065007209778,
+      "logits/rejected": 0.31355416774749756,
+      "logps/chosen": -416.3084411621094,
+      "logps/rejected": -444.62371826171875,
+      "loss": 0.0616,
+      "rewards/accuracies": 0.6499999761581421,
+      "rewards/chosen": -0.8420212864875793,
+      "rewards/margins": 0.4050542712211609,
+      "rewards/rejected": -1.2470756769180298,
       "step": 240
     },
     {
-      "epoch": 0.7,
-      "learning_rate": 4.88671649138311e-07,
-      "logits/chosen": -0.05412503331899643,
-      "logits/rejected": 0.08518598228693008,
-      "logps/chosen": -420.9993591308594,
-      "logps/rejected": -304.08599853515625,
-      "loss": 0.3417,
-      "rewards/accuracies": 0.6625000238418579,
-      "rewards/chosen": -0.46479707956314087,
-      "rewards/margins": 0.37780848145484924,
-      "rewards/rejected": -0.8426055908203125,
+      "epoch": 0.52,
+      "learning_rate": 1.0914104640422679e-06,
+      "logits/chosen": 0.13035783171653748,
+      "logits/rejected": 0.3304864764213562,
+      "logps/chosen": -409.50335693359375,
+      "logps/rejected": -436.58203125,
+      "loss": 0.0568,
+      "rewards/accuracies": 0.7250000238418579,
+      "rewards/chosen": -0.7695704102516174,
+      "rewards/margins": 0.576435923576355,
+      "rewards/rejected": -1.3460063934326172,
       "step": 250
     },
     {
-      "epoch": 0.73,
-      "learning_rate": 4.0665157609325563e-07,
-      "logits/chosen": -0.0160801000893116,
-      "logits/rejected": 0.07172103226184845,
-      "logps/chosen": -409.5190734863281,
-      "logps/rejected": -316.1880798339844,
-      "loss": 0.3521,
-      "rewards/accuracies": 0.6000000238418579,
-      "rewards/chosen": -0.5503396987915039,
-      "rewards/margins": 0.3285866379737854,
-      "rewards/rejected": -0.8789263963699341,
+      "epoch": 0.54,
+      "learning_rate": 1.0183066268176775e-06,
+      "logits/chosen": 0.01489760261029005,
+      "logits/rejected": 0.22449907660484314,
+      "logps/chosen": -483.8492126464844,
+      "logps/rejected": -479.94439697265625,
+      "loss": 0.049,
+      "rewards/accuracies": 0.6625000238418579,
+      "rewards/chosen": -1.0832984447479248,
+      "rewards/margins": 0.5218774080276489,
+      "rewards/rejected": -1.6051762104034424,
       "step": 260
     },
     {
-      "epoch": 0.76,
-      "learning_rate": 3.303816260177894e-07,
-      "logits/chosen": 0.03344980627298355,
-      "logits/rejected": 0.11460290104150772,
-      "logps/chosen": -383.3497619628906,
-      "logps/rejected": -318.8149108886719,
-      "loss": 0.3367,
-      "rewards/accuracies": 0.643750011920929,
-      "rewards/chosen": -0.6241618394851685,
-      "rewards/margins": 0.340850293636322,
-      "rewards/rejected": -0.9650121927261353,
+      "epoch": 0.57,
+      "learning_rate": 9.451046601356725e-07,
+      "logits/chosen": 0.16816282272338867,
+      "logits/rejected": 0.3088468611240387,
+      "logps/chosen": -456.3360290527344,
+      "logps/rejected": -502.9310607910156,
+      "loss": 0.0476,
+      "rewards/accuracies": 0.6875,
+      "rewards/chosen": -1.0681060552597046,
+      "rewards/margins": 0.6512666940689087,
+      "rewards/rejected": -1.7193725109100342,
       "step": 270
     },
     {
-      "epoch": 0.79,
-      "learning_rate": 2.6060092886346885e-07,
-      "logits/chosen": 0.06961329281330109,
-      "logits/rejected": 0.1703241467475891,
-      "logps/chosen": -435.59130859375,
-      "logps/rejected": -319.23138427734375,
-      "loss": 0.3508,
-      "rewards/accuracies": 0.6812499761581421,
-      "rewards/chosen": -0.5814953446388245,
-      "rewards/margins": 0.3754250407218933,
-      "rewards/rejected": -0.9569204449653625,
+      "epoch": 0.59,
+      "learning_rate": 8.721969502803953e-07,
+      "logits/chosen": 0.15908583998680115,
+      "logits/rejected": 0.19241809844970703,
+      "logps/chosen": -424.8711853027344,
+      "logps/rejected": -473.80364990234375,
+      "loss": 0.0513,
+      "rewards/accuracies": 0.737500011920929,
+      "rewards/chosen": -1.0908271074295044,
+      "rewards/margins": 0.5522018671035767,
+      "rewards/rejected": -1.643028974533081,
       "step": 280
     },
     {
-      "epoch": 0.81,
-      "learning_rate": 1.9798572741341148e-07,
-      "logits/chosen": 0.13535265624523163,
-      "logits/rejected": 0.270918071269989,
-      "logps/chosen": -399.9456787109375,
-      "logps/rejected": -334.4576416015625,
-      "loss": 0.3462,
-      "rewards/accuracies": 0.574999988079071,
-      "rewards/chosen": -0.6993755102157593,
-      "rewards/margins": 0.2387124001979828,
-      "rewards/rejected": -0.9380879402160645,
+      "epoch": 0.61,
+      "learning_rate": 7.999743062239557e-07,
+      "logits/chosen": 0.1455659717321396,
+      "logits/rejected": 0.3546694219112396,
+      "logps/chosen": -463.476318359375,
+      "logps/rejected": -514.99072265625,
+      "loss": 0.0507,
+      "rewards/accuracies": 0.793749988079071,
+      "rewards/chosen": -0.9700366258621216,
+      "rewards/margins": 0.7599349021911621,
+      "rewards/rejected": -1.7299716472625732,
       "step": 290
     },
     {
-      "epoch": 0.84,
-      "learning_rate": 1.4314282383241095e-07,
-      "logits/chosen": 0.036216866225004196,
-      "logits/rejected": 0.1895948201417923,
-      "logps/chosen": -417.3045349121094,
-      "logps/rejected": -311.4248046875,
-      "loss": 0.3407,
-      "rewards/accuracies": 0.6187499761581421,
-      "rewards/chosen": -0.5099183320999146,
-      "rewards/margins": 0.3661797046661377,
-      "rewards/rejected": -0.8760979771614075,
+      "epoch": 0.63,
+      "learning_rate": 7.288238647635829e-07,
+      "logits/chosen": 0.201935812830925,
+      "logits/rejected": 0.29931244254112244,
+      "logps/chosen": -406.6135559082031,
+      "logps/rejected": -511.62677001953125,
+      "loss": 0.0474,
+      "rewards/accuracies": 0.7437499761581421,
+      "rewards/chosen": -0.953034520149231,
+      "rewards/margins": 0.72877037525177,
+      "rewards/rejected": -1.681804895401001,
       "step": 300
     },
     {
-      "epoch": 0.87,
-      "learning_rate": 9.660369916414013e-08,
-      "logits/chosen": -0.024627825245261192,
-      "logits/rejected": 0.15113016963005066,
-      "logps/chosen": -380.7222900390625,
-      "logps/rejected": -324.25286865234375,
-      "loss": 0.3436,
-      "rewards/accuracies": 0.6312500238418579,
-      "rewards/chosen": -0.4988563060760498,
-      "rewards/margins": 0.3266349136829376,
-      "rewards/rejected": -0.8254912495613098,
+      "epoch": 0.65,
+      "learning_rate": 6.591270153428288e-07,
+      "logits/chosen": 0.18981854617595673,
+      "logits/rejected": 0.2584979236125946,
+      "logps/chosen": -407.3790283203125,
+      "logps/rejected": -500.0130310058594,
+      "loss": 0.0495,
+      "rewards/accuracies": 0.71875,
+      "rewards/chosen": -0.9219211339950562,
+      "rewards/margins": 0.7611614465713501,
+      "rewards/rejected": -1.6830825805664062,
       "step": 310
     },
     {
-      "epoch": 0.9,
-      "learning_rate": 5.881936276323462e-08,
-      "logits/chosen": 0.07192158699035645,
-      "logits/rejected": 0.13689716160297394,
-      "logps/chosen": -364.3758239746094,
-      "logps/rejected": -310.78253173828125,
-      "loss": 0.3421,
-      "rewards/accuracies": 0.6000000238418579,
-      "rewards/chosen": -0.5145777463912964,
-      "rewards/margins": 0.20848092436790466,
-      "rewards/rejected": -0.7230587005615234,
+      "epoch": 0.67,
+      "learning_rate": 5.912573556804452e-07,
+      "logits/chosen": 0.19075055420398712,
+      "logits/rejected": 0.42188405990600586,
+      "logps/chosen": -411.3404235839844,
+      "logps/rejected": -462.23516845703125,
+      "loss": 0.0518,
+      "rewards/accuracies": 0.75,
+      "rewards/chosen": -0.7666682004928589,
+      "rewards/margins": 0.7605694532394409,
+      "rewards/rejected": -1.5272376537322998,
       "step": 320
     },
     {
-      "epoch": 0.93,
-      "learning_rate": 3.015598157625598e-08,
-      "logits/chosen": -0.08804676681756973,
-      "logits/rejected": 0.00635142857208848,
-      "logps/chosen": -460.70257568359375,
-      "logps/rejected": -359.46258544921875,
-      "loss": 0.3382,
-      "rewards/accuracies": 0.6499999761581421,
-      "rewards/chosen": -0.5289751291275024,
-      "rewards/margins": 0.39722973108291626,
-      "rewards/rejected": -0.9262048006057739,
+      "epoch": 0.69,
+      "learning_rate": 5.255786891654399e-07,
+      "logits/chosen": 0.21234121918678284,
+      "logits/rejected": 0.3896685242652893,
+      "logps/chosen": -402.15509033203125,
+      "logps/rejected": -492.631103515625,
+      "loss": 0.0489,
+      "rewards/accuracies": 0.737500011920929,
+      "rewards/chosen": -0.8326274752616882,
+      "rewards/margins": 0.7936538457870483,
+      "rewards/rejected": -1.6262813806533813,
       "step": 330
     },
     {
-      "epoch": 0.96,
-      "learning_rate": 1.0891331628063882e-08,
-      "logits/chosen": -0.03883730247616768,
-      "logits/rejected": 0.06328520178794861,
-      "logps/chosen": -404.8522033691406,
-      "logps/rejected": -318.14471435546875,
-      "loss": 0.3546,
-      "rewards/accuracies": 0.668749988079071,
-      "rewards/chosen": -0.5345975160598755,
-      "rewards/margins": 0.32099664211273193,
-      "rewards/rejected": -0.8555941581726074,
+      "epoch": 0.71,
+      "learning_rate": 4.624430747529102e-07,
+      "logits/chosen": 0.15655192732810974,
+      "logits/rejected": 0.374599426984787,
+      "logps/chosen": -382.6220703125,
+      "logps/rejected": -439.4056091308594,
+      "loss": 0.0521,
+      "rewards/accuracies": 0.762499988079071,
+      "rewards/chosen": -0.7596348524093628,
+      "rewards/margins": 0.8444586992263794,
+      "rewards/rejected": -1.6040935516357422,
       "step": 340
     },
     {
-      "epoch": 0.98,
-      "learning_rate": 1.212106102131849e-09,
-      "logits/chosen": -0.05511404201388359,
-      "logits/rejected": -0.015045121312141418,
-      "logps/chosen": -391.4950256347656,
-      "logps/rejected": -317.26922607421875,
-      "loss": 0.3526,
-      "rewards/accuracies": 0.65625,
-      "rewards/chosen": -0.5430077314376831,
-      "rewards/margins": 0.3454528748989105,
-      "rewards/rejected": -0.888460636138916,
+      "epoch": 0.73,
+      "learning_rate": 4.0218893981385927e-07,
+      "logits/chosen": 0.11385490745306015,
+      "logits/rejected": 0.16609953343868256,
+      "logps/chosen": -397.93603515625,
+      "logps/rejected": -468.45440673828125,
+      "loss": 0.0507,
+      "rewards/accuracies": 0.699999988079071,
+      "rewards/chosen": -0.7674127221107483,
+      "rewards/margins": 0.6324188113212585,
+      "rewards/rejected": -1.3998315334320068,
       "step": 350
     },
+    {
+      "epoch": 0.75,
+      "learning_rate": 3.45139266054715e-07,
+      "logits/chosen": 0.038220081478357315,
+      "logits/rejected": 0.3043181896209717,
+      "logps/chosen": -426.51116943359375,
+      "logps/rejected": -444.027587890625,
+      "loss": 0.0509,
+      "rewards/accuracies": 0.7124999761581421,
+      "rewards/chosen": -0.7794097661972046,
+      "rewards/margins": 0.6512076258659363,
+      "rewards/rejected": -1.430617332458496,
+      "step": 360
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 2.915998582306299e-07,
+      "logits/chosen": 0.07361505925655365,
+      "logits/rejected": 0.3171204924583435,
+      "logps/chosen": -450.21435546875,
+      "logps/rejected": -491.0059509277344,
+      "loss": 0.0499,
+      "rewards/accuracies": 0.706250011920929,
+      "rewards/chosen": -0.8033027648925781,
+      "rewards/margins": 0.8250747919082642,
+      "rewards/rejected": -1.6283775568008423,
+      "step": 370
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 2.418577049328058e-07,
+      "logits/chosen": 0.16773228347301483,
+      "logits/rejected": 0.34962594509124756,
+      "logps/chosen": -406.15374755859375,
+      "logps/rejected": -453.6668395996094,
+      "loss": 0.0498,
+      "rewards/accuracies": 0.768750011920929,
+      "rewards/chosen": -0.9306628108024597,
+      "rewards/margins": 0.7057619094848633,
+      "rewards/rejected": -1.6364247798919678,
+      "step": 380
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 1.961794402365611e-07,
+      "logits/chosen": -0.025767967104911804,
+      "logits/rejected": 0.22987493872642517,
+      "logps/chosen": -450.62615966796875,
+      "logps/rejected": -491.8794860839844,
+      "loss": 0.0451,
+      "rewards/accuracies": 0.731249988079071,
+      "rewards/chosen": -0.8757637143135071,
+      "rewards/margins": 0.85161954164505,
+      "rewards/rejected": -1.7273832559585571,
+      "step": 390
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 1.5480991445620538e-07,
+      "logits/chosen": 0.07942807674407959,
+      "logits/rejected": 0.27572622895240784,
+      "logps/chosen": -420.186767578125,
+      "logps/rejected": -467.331298828125,
+      "loss": 0.0515,
+      "rewards/accuracies": 0.731249988079071,
+      "rewards/chosen": -0.8469425439834595,
+      "rewards/margins": 0.6868287324905396,
+      "rewards/rejected": -1.533771276473999,
+      "step": 400
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 1.1797088166794e-07,
+      "logits/chosen": 0.011249428614974022,
+      "logits/rejected": 0.1347026228904724,
+      "logps/chosen": -407.4740905761719,
+      "logps/rejected": -475.7933654785156,
+      "loss": 0.0427,
+      "rewards/accuracies": 0.6812499761581421,
+      "rewards/chosen": -0.9821971654891968,
+      "rewards/margins": 0.6814206838607788,
+      "rewards/rejected": -1.663617730140686,
+      "step": 410
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 8.585981103608342e-08,
+      "logits/chosen": 0.10574499517679214,
+      "logits/rejected": 0.29398587346076965,
+      "logps/chosen": -404.7536926269531,
+      "logps/rejected": -476.5546875,
+      "loss": 0.0425,
+      "rewards/accuracies": 0.71875,
+      "rewards/chosen": -0.9181003570556641,
+      "rewards/margins": 0.8052154779434204,
+      "rewards/rejected": -1.7233158349990845,
+      "step": 420
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 5.8648828314302735e-08,
+      "logits/chosen": 0.1234573945403099,
+      "logits/rejected": 0.21596229076385498,
+      "logps/chosen": -461.3655700683594,
+      "logps/rejected": -526.1497802734375,
+      "loss": 0.0487,
+      "rewards/accuracies": 0.7124999761581421,
+      "rewards/chosen": -1.061097264289856,
+      "rewards/margins": 0.7796509861946106,
+      "rewards/rejected": -1.8407481908798218,
+      "step": 430
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 3.648379319574568e-08,
+      "logits/chosen": 0.06800473481416702,
+      "logits/rejected": 0.1670626848936081,
+      "logps/chosen": -400.8985595703125,
+      "logps/rejected": -467.785400390625,
+      "loss": 0.0458,
+      "rewards/accuracies": 0.706250011920929,
+      "rewards/chosen": -0.9769018888473511,
+      "rewards/margins": 0.7274185419082642,
+      "rewards/rejected": -1.7043205499649048,
+      "step": 440
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 1.9483517457776434e-08,
+      "logits/chosen": 0.04293660447001457,
+      "logits/rejected": 0.3749822676181793,
+      "logps/chosen": -479.16107177734375,
+      "logps/rejected": -492.6188049316406,
+      "loss": 0.044,
+      "rewards/accuracies": 0.731249988079071,
+      "rewards/chosen": -1.1043040752410889,
+      "rewards/margins": 0.7897499203681946,
+      "rewards/rejected": -1.8940540552139282,
+      "step": 450
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 7.739128092312918e-09,
+      "logits/chosen": 0.08021829277276993,
+      "logits/rejected": 0.4172864854335785,
+      "logps/chosen": -407.2559509277344,
+      "logps/rejected": -451.20977783203125,
+      "loss": 0.0416,
+      "rewards/accuracies": 0.7437499761581421,
+      "rewards/chosen": -0.9067844152450562,
+      "rewards/margins": 0.7688730359077454,
+      "rewards/rejected": -1.6756572723388672,
+      "step": 460
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 1.313578835593465e-09,
+      "logits/chosen": 0.05767583101987839,
+      "logits/rejected": 0.32164663076400757,
+      "logps/chosen": -457.6153869628906,
+      "logps/rejected": -505.43402099609375,
+      "loss": 0.0456,
+      "rewards/accuracies": 0.762499988079071,
+      "rewards/chosen": -0.9809616804122925,
+      "rewards/margins": 0.9043967127799988,
+      "rewards/rejected": -1.885358452796936,
+      "step": 470
+    },
     {
       "epoch": 1.0,
-      "step": 355,
+      "step": 477,
       "total_flos": 0.0,
-      "train_loss": 0.37752047600880473,
-      "train_runtime": 5207.0243,
-      "train_samples_per_second": 8.747,
-      "train_steps_per_second": 0.068
+      "train_loss": 0.06185462889681322,
+      "train_runtime": 6360.6645,
+      "train_samples_per_second": 9.611,
+      "train_steps_per_second": 0.075
     }
   ],
   "logging_steps": 10,
-  "max_steps": 355,
+  "max_steps": 477,
   "num_train_epochs": 1,
   "save_steps": 10000,
   "total_flos": 0.0,

training_args.bin

@@ -1,3 +1,3 @@
 version https://git-lfs.github.com/spec/v1
-oid sha256:73a6913970550a2b439b746fec76691b22e0913489fbe7db2b2f3f15d5f5ff7c
+oid sha256:9df90afe6489dd288d2f7e8506e57b4ca17d4164a74df570e062390c04a57b25
 size 6008