Model save

README.md

@@ -32,7 +32,7 @@ More information needed
 ### Training hyperparameters
 
 The following hyperparameters were used during training:
-- learning_rate: 2e-06
+- learning_rate: 3e-06
 - train_batch_size: 2
 - eval_batch_size: 8
 - seed: 1

all_results.json

@@ -1,8 +1,8 @@
 {
     "epoch": 1.0,
-    "train_loss": 0.06185462889681322,
-    "train_runtime": 6360.6645,
-    "train_samples": 61134,
-    "train_samples_per_second": 9.611,
-    "train_steps_per_second": 0.075
+    "train_loss": 0.33912758298323187,
+    "train_runtime": 5289.8194,
+    "train_samples": 45548,
+    "train_samples_per_second": 8.611,
+    "train_steps_per_second": 0.067
 }

model-00001-of-00004.safetensors

@@ -1,3 +1,3 @@
 version https://git-lfs.github.com/spec/v1
-oid sha256:5d383b1d023d1ed1e3584695b798e77d278f800fc398598aa6e9c9e42e3e2ad1
+oid sha256:733a532a112774a114c1a3fbd7d3848a1f045cac921ba2700a6e624cf81570ce
 size 4976698672

model-00002-of-00004.safetensors

@@ -1,3 +1,3 @@
 version https://git-lfs.github.com/spec/v1
-oid sha256:cb98d9e0bb313a95a034b93ea04d8620ae709b5c4390e5f6ee5637ed55a68aca
+oid sha256:56a7c7182cf25359b8e1cc662cc8b0deff28eb8a0f8f32ef5dac2729a67f5103
 size 4999802720

model-00003-of-00004.safetensors

@@ -1,3 +1,3 @@
 version https://git-lfs.github.com/spec/v1
-oid sha256:2b65b68220db877b0e0b92a876b84cf38946255244aa69d62cb4ac1c1c8fd5cd
+oid sha256:a668f5ca1ceb705045c372af7bb4df0e367e21106b1b109f41eb0d5e49616574
 size 4915916176

model-00004-of-00004.safetensors

@@ -1,3 +1,3 @@
 version https://git-lfs.github.com/spec/v1
-oid sha256:641a1860fbabe30e2297682bc87a6ca0f8a090cfb74308285997d02428d923de
+oid sha256:d7ca431de24cc7f662911fbb7a1e63d9f4a271a428bf509023179893c75685ab
 size 1168138808

train_results.json

@@ -1,8 +1,8 @@
 {
     "epoch": 1.0,
-    "train_loss": 0.06185462889681322,
-    "train_runtime": 6360.6645,
-    "train_samples": 61134,
-    "train_samples_per_second": 9.611,
-    "train_steps_per_second": 0.075
+    "train_loss": 0.33912758298323187,
+    "train_runtime": 5289.8194,
+    "train_samples": 45548,
+    "train_samples_per_second": 8.611,
+    "train_steps_per_second": 0.067
 }

trainer_state.json

@@ -1,21 +1,21 @@
 {
   "best_metric": null,
   "best_model_checkpoint": null,
-  "epoch": 0.998691442030882,
+  "epoch": 0.9975412715138743,
   "eval_steps": 10000,
-  "global_step": 477,
+  "global_step": 355,
   "is_hyper_param_search": false,
   "is_local_process_zero": true,
   "is_world_process_zero": true,
   "log_history": [
     {
       "epoch": 0.0,
-      "learning_rate": 4.166666666666666e-08,
-      "logits/chosen": 0.21113795042037964,
-      "logits/rejected": 0.15092352032661438,
-      "logps/chosen": -422.9735412597656,
-      "logps/rejected": -444.97625732421875,
-      "loss": 0.1408,
+      "learning_rate": 8.333333333333333e-08,
+      "logits/chosen": -0.13174405694007874,
+      "logits/rejected": -0.027169257402420044,
+      "logps/chosen": -477.4691162109375,
+      "logps/rejected": -277.6482238769531,
+      "loss": 0.4106,
       "rewards/accuracies": 0.0,
       "rewards/chosen": 0.0,
       "rewards/margins": 0.0,
@@ -23,675 +23,507 @@
       "step": 1
     },
     {
-      "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,
+      "epoch": 0.03,
+      "learning_rate": 8.333333333333334e-07,
+      "logits/chosen": -0.08925160765647888,
+      "logits/rejected": -0.04610097035765648,
+      "logps/chosen": -334.2904052734375,
+      "logps/rejected": -264.34820556640625,
+      "loss": 0.4194,
+      "rewards/accuracies": 0.4305555522441864,
+      "rewards/chosen": -0.002801957307383418,
+      "rewards/margins": -0.0008487759623676538,
+      "rewards/rejected": -0.001953181577846408,
       "step": 10
     },
     {
-      "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,
+      "epoch": 0.06,
+      "learning_rate": 1.6666666666666669e-06,
+      "logits/chosen": -0.06430797278881073,
+      "logits/rejected": -0.02448318526148796,
+      "logps/chosen": -314.1076965332031,
+      "logps/rejected": -218.28536987304688,
+      "loss": 0.439,
+      "rewards/accuracies": 0.5687500238418579,
+      "rewards/chosen": -0.017305215820670128,
+      "rewards/margins": 0.050741471350193024,
+      "rewards/rejected": -0.0680466890335083,
       "step": 20
     },
     {
-      "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,
+      "epoch": 0.08,
+      "learning_rate": 2.5e-06,
+      "logits/chosen": -0.07502040266990662,
+      "logits/rejected": -0.030765822157263756,
+      "logps/chosen": -375.57470703125,
+      "logps/rejected": -248.95822143554688,
+      "loss": 0.4779,
+      "rewards/accuracies": 0.606249988079071,
+      "rewards/chosen": -0.04765231907367706,
+      "rewards/margins": 0.21400682628154755,
+      "rewards/rejected": -0.2616591453552246,
       "step": 30
     },
     {
-      "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,
+      "epoch": 0.11,
+      "learning_rate": 2.9988362934929793e-06,
+      "logits/chosen": -0.17460331320762634,
+      "logits/rejected": -0.11901924759149551,
+      "logps/chosen": -320.1922607421875,
+      "logps/rejected": -252.71444702148438,
+      "loss": 0.4782,
+      "rewards/accuracies": 0.59375,
+      "rewards/chosen": 0.04120566323399544,
+      "rewards/margins": 0.15689076483249664,
+      "rewards/rejected": -0.11568509042263031,
       "step": 40
     },
     {
-      "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,
+      "epoch": 0.14,
+      "learning_rate": 2.985765322825759e-06,
+      "logits/chosen": -0.0304028932005167,
+      "logits/rejected": 0.039170511066913605,
+      "logps/chosen": -332.5814208984375,
+      "logps/rejected": -268.250244140625,
+      "loss": 0.4316,
+      "rewards/accuracies": 0.581250011920929,
+      "rewards/chosen": 0.07894544303417206,
+      "rewards/margins": 0.14912132918834686,
+      "rewards/rejected": -0.07017587125301361,
       "step": 50
     },
     {
-      "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,
+      "epoch": 0.17,
+      "learning_rate": 2.9582958419982717e-06,
+      "logits/chosen": -0.22259502112865448,
+      "logits/rejected": -0.12899939715862274,
+      "logps/chosen": -398.3349609375,
+      "logps/rejected": -249.4995574951172,
+      "loss": 0.4568,
+      "rewards/accuracies": 0.71875,
+      "rewards/chosen": 0.09524339437484741,
+      "rewards/margins": 0.3634737432003021,
+      "rewards/rejected": -0.2682303786277771,
       "step": 60
     },
     {
-      "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,
+      "epoch": 0.2,
+      "learning_rate": 2.916694056980408e-06,
+      "logits/chosen": -0.20538434386253357,
+      "logits/rejected": -0.07854785025119781,
+      "logps/chosen": -396.68438720703125,
+      "logps/rejected": -272.218994140625,
+      "loss": 0.4356,
+      "rewards/accuracies": 0.643750011920929,
+      "rewards/chosen": -0.04246167838573456,
+      "rewards/margins": 0.27019184827804565,
+      "rewards/rejected": -0.3126535713672638,
       "step": 70
     },
     {
-      "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,
+      "epoch": 0.22,
+      "learning_rate": 2.8613631295064357e-06,
+      "logits/chosen": 0.12309768050909042,
+      "logits/rejected": 0.16693784296512604,
+      "logps/chosen": -316.2131652832031,
+      "logps/rejected": -260.7282409667969,
+      "loss": 0.3878,
+      "rewards/accuracies": 0.5874999761581421,
+      "rewards/chosen": -0.370283842086792,
+      "rewards/margins": 0.24990463256835938,
+      "rewards/rejected": -0.6201884150505066,
       "step": 80
     },
     {
-      "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,
+      "epoch": 0.25,
+      "learning_rate": 2.792839270045916e-06,
+      "logits/chosen": 0.007012078072875738,
+      "logits/rejected": 0.08523766696453094,
+      "logps/chosen": -414.2920837402344,
+      "logps/rejected": -269.9517517089844,
+      "loss": 0.3726,
+      "rewards/accuracies": 0.7124999761581421,
+      "rewards/chosen": -0.20565040409564972,
+      "rewards/margins": 0.39593735337257385,
+      "rewards/rejected": -0.6015877723693848,
       "step": 90
     },
     {
-      "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,
+      "epoch": 0.28,
+      "learning_rate": 2.711786541403051e-06,
+      "logits/chosen": -0.011275473982095718,
+      "logits/rejected": 0.01660158298909664,
+      "logps/chosen": -397.98016357421875,
+      "logps/rejected": -313.91571044921875,
+      "loss": 0.3614,
+      "rewards/accuracies": 0.625,
+      "rewards/chosen": -0.33979809284210205,
+      "rewards/margins": 0.1854400336742401,
+      "rewards/rejected": -0.5252381563186646,
       "step": 100
     },
     {
-      "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,
+      "epoch": 0.31,
+      "learning_rate": 2.6189904233026363e-06,
+      "logits/chosen": -0.12849292159080505,
+      "logits/rejected": -0.017519207671284676,
+      "logps/chosen": -357.66461181640625,
+      "logps/rejected": -311.74127197265625,
+      "loss": 0.351,
+      "rewards/accuracies": 0.59375,
+      "rewards/chosen": -0.4424869418144226,
+      "rewards/margins": 0.11273103952407837,
+      "rewards/rejected": -0.555217981338501,
       "step": 110
     },
     {
-      "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,
+      "epoch": 0.34,
+      "learning_rate": 2.515350200328027e-06,
+      "logits/chosen": 0.001210880232974887,
+      "logits/rejected": 0.054382193833589554,
+      "logps/chosen": -392.6571350097656,
+      "logps/rejected": -285.66156005859375,
+      "loss": 0.3253,
+      "rewards/accuracies": 0.699999988079071,
+      "rewards/chosen": -0.5204089283943176,
+      "rewards/margins": 0.2852626144886017,
+      "rewards/rejected": -0.8056715130805969,
       "step": 120
     },
     {
-      "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,
+      "epoch": 0.37,
+      "learning_rate": 2.401870246979413e-06,
+      "logits/chosen": -0.03578253090381622,
+      "logits/rejected": 0.09569860994815826,
+      "logps/chosen": -432.0030212402344,
+      "logps/rejected": -300.4866943359375,
+      "loss": 0.3486,
+      "rewards/accuracies": 0.668749988079071,
+      "rewards/chosen": -0.5831924080848694,
+      "rewards/margins": 0.3761444389820099,
+      "rewards/rejected": -0.9593368768692017,
       "step": 130
     },
     {
-      "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,
+      "epoch": 0.39,
+      "learning_rate": 2.279650294308645e-06,
+      "logits/chosen": 0.009150633588433266,
+      "logits/rejected": 0.04822584241628647,
+      "logps/chosen": -375.3031005859375,
+      "logps/rejected": -296.3115234375,
+      "loss": 0.339,
+      "rewards/accuracies": 0.581250011920929,
+      "rewards/chosen": -0.563973069190979,
+      "rewards/margins": 0.11898863315582275,
+      "rewards/rejected": -0.682961642742157,
       "step": 140
     },
     {
-      "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,
+      "epoch": 0.42,
+      "learning_rate": 2.1498747724563957e-06,
+      "logits/chosen": -0.01069603580981493,
+      "logits/rejected": 0.06919459253549576,
+      "logps/chosen": -346.59552001953125,
+      "logps/rejected": -311.71453857421875,
+      "loss": 0.3286,
+      "rewards/accuracies": 0.637499988079071,
+      "rewards/chosen": -0.5837494134902954,
+      "rewards/margins": 0.212588831782341,
+      "rewards/rejected": -0.7963383197784424,
       "step": 150
     },
     {
-      "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,
+      "epoch": 0.45,
+      "learning_rate": 2.0138013323728074e-06,
+      "logits/chosen": -0.10445091873407364,
+      "logits/rejected": -0.0012316614156588912,
+      "logps/chosen": -440.801025390625,
+      "logps/rejected": -332.5092468261719,
+      "loss": 0.3141,
+      "rewards/accuracies": 0.6000000238418579,
+      "rewards/chosen": -0.7070592641830444,
+      "rewards/margins": 0.26387742161750793,
+      "rewards/rejected": -0.9709367752075195,
       "step": 160
     },
     {
-      "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,
+      "epoch": 0.48,
+      "learning_rate": 1.8727486579573409e-06,
+      "logits/chosen": -0.013905433006584644,
+      "logits/rejected": 0.06436058133840561,
+      "logps/chosen": -396.68731689453125,
+      "logps/rejected": -354.0695495605469,
+      "loss": 0.2845,
+      "rewards/accuracies": 0.59375,
+      "rewards/chosen": -0.8083154559135437,
+      "rewards/margins": 0.2599693536758423,
+      "rewards/rejected": -1.0682848691940308,
       "step": 170
     },
     {
-      "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,
+      "epoch": 0.51,
+      "learning_rate": 1.7280836867300083e-06,
+      "logits/chosen": -0.11434350162744522,
+      "logits/rejected": 0.023699652403593063,
+      "logps/chosen": -409.00396728515625,
+      "logps/rejected": -343.5912170410156,
+      "loss": 0.2934,
+      "rewards/accuracies": 0.5687500238418579,
+      "rewards/chosen": -0.8608970642089844,
+      "rewards/margins": 0.23020252585411072,
+      "rewards/rejected": -1.0910996198654175,
       "step": 180
     },
     {
-      "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,
+      "epoch": 0.53,
+      "learning_rate": 1.5812083628781265e-06,
+      "logits/chosen": -0.12255267053842545,
+      "logits/rejected": -0.05903090164065361,
+      "logps/chosen": -413.7901916503906,
+      "logps/rejected": -309.20904541015625,
+      "loss": 0.2889,
+      "rewards/accuracies": 0.612500011920929,
+      "rewards/chosen": -0.7189455032348633,
+      "rewards/margins": 0.388428270816803,
+      "rewards/rejected": -1.107373595237732,
       "step": 190
     },
     {
-      "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,
+      "epoch": 0.56,
+      "learning_rate": 1.433546051054432e-06,
+      "logits/chosen": -0.26926541328430176,
+      "logits/rejected": -0.16134600341320038,
+      "logps/chosen": -413.9627990722656,
+      "logps/rejected": -353.7982177734375,
+      "loss": 0.2944,
+      "rewards/accuracies": 0.5687500238418579,
+      "rewards/chosen": -1.0403170585632324,
+      "rewards/margins": 0.2550913691520691,
+      "rewards/rejected": -1.2954083681106567,
       "step": 200
     },
     {
-      "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,
+      "epoch": 0.59,
+      "learning_rate": 1.2865277425900725e-06,
+      "logits/chosen": -0.1475798785686493,
+      "logits/rejected": -0.10237713158130646,
+      "logps/chosen": -472.66339111328125,
+      "logps/rejected": -371.1333923339844,
+      "loss": 0.2912,
+      "rewards/accuracies": 0.625,
+      "rewards/chosen": -0.9276388883590698,
+      "rewards/margins": 0.35251864790916443,
+      "rewards/rejected": -1.280157446861267,
       "step": 210
     },
     {
-      "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,
+      "epoch": 0.62,
+      "learning_rate": 1.141578187797663e-06,
+      "logits/chosen": -0.22542640566825867,
+      "logits/rejected": -0.14302000403404236,
+      "logps/chosen": -440.4434509277344,
+      "logps/rejected": -330.79278564453125,
+      "loss": 0.2937,
+      "rewards/accuracies": 0.643750011920929,
+      "rewards/chosen": -0.8558899164199829,
+      "rewards/margins": 0.2570609748363495,
+      "rewards/rejected": -1.1129509210586548,
       "step": 220
     },
     {
-      "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,
+      "epoch": 0.65,
+      "learning_rate": 1.0001020887558839e-06,
+      "logits/chosen": -0.1044299378991127,
+      "logits/rejected": 0.011820942163467407,
+      "logps/chosen": -398.46728515625,
+      "logps/rejected": -318.6748352050781,
+      "loss": 0.2992,
+      "rewards/accuracies": 0.643750011920929,
+      "rewards/chosen": -0.7419184446334839,
+      "rewards/margins": 0.29130107164382935,
+      "rewards/rejected": -1.033219575881958,
       "step": 230
     },
     {
-      "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,
+      "epoch": 0.67,
+      "learning_rate": 8.634704863809502e-07,
+      "logits/chosen": -0.11705616861581802,
+      "logits/rejected": 0.025935053825378418,
+      "logps/chosen": -457.593994140625,
+      "logps/rejected": -347.49322509765625,
+      "loss": 0.2807,
+      "rewards/accuracies": 0.643750011920929,
+      "rewards/chosen": -0.8912258148193359,
+      "rewards/margins": 0.24232976138591766,
+      "rewards/rejected": -1.1335556507110596,
       "step": 240
     },
     {
-      "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,
+      "epoch": 0.7,
+      "learning_rate": 7.330074737074666e-07,
+      "logits/chosen": -0.09622545540332794,
+      "logits/rejected": 0.04155969247221947,
+      "logps/chosen": -463.0812072753906,
+      "logps/rejected": -348.09552001953125,
+      "loss": 0.2864,
+      "rewards/accuracies": 0.65625,
+      "rewards/chosen": -0.8856151700019836,
+      "rewards/margins": 0.3970854878425598,
+      "rewards/rejected": -1.2827006578445435,
       "step": 250
     },
     {
-      "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,
+      "epoch": 0.73,
+      "learning_rate": 6.099773641398835e-07,
+      "logits/chosen": -0.11350131034851074,
+      "logits/rejected": -0.020252907648682594,
+      "logps/chosen": -455.5005798339844,
+      "logps/rejected": -365.80120849609375,
+      "loss": 0.2874,
+      "rewards/accuracies": 0.6000000238418579,
+      "rewards/chosen": -1.0101550817489624,
+      "rewards/margins": 0.3649025559425354,
+      "rewards/rejected": -1.375057578086853,
       "step": 260
     },
     {
-      "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,
+      "epoch": 0.76,
+      "learning_rate": 4.955724390266841e-07,
+      "logits/chosen": -0.08897687494754791,
+      "logits/rejected": -0.018877411261200905,
+      "logps/chosen": -417.6068420410156,
+      "logps/rejected": -356.24029541015625,
+      "loss": 0.2851,
+      "rewards/accuracies": 0.643750011920929,
+      "rewards/chosen": -0.9667322039604187,
+      "rewards/margins": 0.37253400683403015,
+      "rewards/rejected": -1.3392661809921265,
       "step": 270
     },
     {
-      "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,
+      "epoch": 0.79,
+      "learning_rate": 3.9090139329520333e-07,
+      "logits/chosen": -0.07754144817590714,
+      "logits/rejected": 0.0014429532457143068,
+      "logps/chosen": -463.06182861328125,
+      "logps/rejected": -349.83331298828125,
+      "loss": 0.306,
+      "rewards/accuracies": 0.6812499761581421,
+      "rewards/chosen": -0.8562005758285522,
+      "rewards/margins": 0.40673941373825073,
+      "rewards/rejected": -1.2629398107528687,
       "step": 280
     },
     {
-      "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,
+      "epoch": 0.81,
+      "learning_rate": 2.9697859112011724e-07,
+      "logits/chosen": -0.0834055170416832,
+      "logits/rejected": 0.01585310325026512,
+      "logps/chosen": -417.92547607421875,
+      "logps/rejected": -353.6193542480469,
+      "loss": 0.3084,
+      "rewards/accuracies": 0.6187499761581421,
+      "rewards/chosen": -0.8791738748550415,
+      "rewards/margins": 0.2505313456058502,
+      "rewards/rejected": -1.1297051906585693,
       "step": 290
     },
     {
-      "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,
+      "epoch": 0.84,
+      "learning_rate": 2.1471423574861643e-07,
+      "logits/chosen": -0.10339058935642242,
+      "logits/rejected": 0.019864924252033234,
+      "logps/chosen": -434.0230407714844,
+      "logps/rejected": -332.9052734375,
+      "loss": 0.2997,
+      "rewards/accuracies": 0.6187499761581421,
+      "rewards/chosen": -0.6771033406257629,
+      "rewards/margins": 0.413799524307251,
+      "rewards/rejected": -1.0909026861190796,
       "step": 300
     },
     {
-      "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,
+      "epoch": 0.87,
+      "learning_rate": 1.449055487462102e-07,
+      "logits/chosen": -0.09986544400453568,
+      "logits/rejected": 0.010388897731900215,
+      "logps/chosen": -410.6251525878906,
+      "logps/rejected": -353.5132751464844,
+      "loss": 0.2973,
+      "rewards/accuracies": 0.6499999761581421,
+      "rewards/chosen": -0.7978848814964294,
+      "rewards/margins": 0.3202103078365326,
+      "rewards/rejected": -1.1180951595306396,
       "step": 310
     },
     {
-      "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,
+      "epoch": 0.9,
+      "learning_rate": 8.822904414485194e-08,
+      "logits/chosen": -0.02717159129679203,
+      "logits/rejected": 0.01261619757860899,
+      "logps/chosen": -388.2625427246094,
+      "logps/rejected": -338.05938720703125,
+      "loss": 0.2954,
+      "rewards/accuracies": 0.6312500238418579,
+      "rewards/chosen": -0.7534447312355042,
+      "rewards/margins": 0.24238231778144836,
+      "rewards/rejected": -0.9958270788192749,
       "step": 320
     },
     {
-      "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": 0.71,
-      "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": 0.73,
-      "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,
+      "epoch": 0.93,
+      "learning_rate": 4.523397236438398e-08,
+      "logits/chosen": -0.1445978581905365,
+      "logits/rejected": -0.06250996887683868,
+      "logps/chosen": -485.3677673339844,
+      "logps/rejected": -386.36688232421875,
+      "loss": 0.2896,
       "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
+      "rewards/chosen": -0.7756272554397583,
+      "rewards/margins": 0.4196205139160156,
+      "rewards/rejected": -1.1952476501464844,
+      "step": 330
     },
     {
       "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
+      "learning_rate": 1.6336997442095825e-08,
+      "logits/chosen": -0.07295690476894379,
+      "logits/rejected": -0.002095238771289587,
+      "logps/chosen": -430.9432678222656,
+      "logps/rejected": -346.85638427734375,
+      "loss": 0.306,
+      "rewards/accuracies": 0.643750011920929,
+      "rewards/chosen": -0.7955079078674316,
+      "rewards/margins": 0.34720322489738464,
+      "rewards/rejected": -1.1427110433578491,
+      "step": 340
     },
     {
       "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
+      "learning_rate": 1.8181591531977737e-09,
+      "logits/chosen": -0.11711341142654419,
+      "logits/rejected": -0.08915948867797852,
+      "logps/chosen": -413.8916931152344,
+      "logps/rejected": -344.609130859375,
+      "loss": 0.3091,
+      "rewards/accuracies": 0.6812499761581421,
+      "rewards/chosen": -0.766974687576294,
+      "rewards/margins": 0.3948844075202942,
+      "rewards/rejected": -1.1618590354919434,
+      "step": 350
     },
     {
       "epoch": 1.0,
-      "step": 477,
+      "step": 355,
       "total_flos": 0.0,
-      "train_loss": 0.06185462889681322,
-      "train_runtime": 6360.6645,
-      "train_samples_per_second": 9.611,
-      "train_steps_per_second": 0.075
+      "train_loss": 0.33912758298323187,
+      "train_runtime": 5289.8194,
+      "train_samples_per_second": 8.611,
+      "train_steps_per_second": 0.067
     }
   ],
   "logging_steps": 10,
-  "max_steps": 477,
+  "max_steps": 355,
   "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 sha256:9df90afe6489dd288d2f7e8506e57b4ca17d4164a74df570e062390c04a57b25
+oid sha256:49e4ccaf4f269c36cbaf74181809222cfd8bd58f6db22e369ff0440368f2d18d
 size 6008