Model save

README.md

@@ -13,7 +13,7 @@ should probably proofread and complete it, then remove this comment. -->
 
 # zephyr-7b-dpo-full
 
-This model was trained from scratch on an unknown dataset.
+This model was trained from scratch on the None dataset.
 
 ## Model description
 
@@ -54,5 +54,5 @@ The following hyperparameters were used during training:
 
 - Transformers 4.39.3
 - Pytorch 2.1.2+cu118
-- Datasets 2.19.1
+- Datasets 2.16.1
 - Tokenizers 0.15.2

all_results.json

@@ -1,8 +1,8 @@
 {
     "epoch": 1.0,
-    "train_loss": 0.22171047773116676,
-    "train_runtime": 5890.091,
-    "train_samples": 50000,
-    "train_samples_per_second": 8.489,
-    "train_steps_per_second": 0.066
+    "train_loss": 0.4806597235705027,
+    "train_runtime": 6367.8161,
+    "train_samples": 52922,
+    "train_samples_per_second": 8.311,
+    "train_steps_per_second": 0.065
 }

model-00001-of-00003.safetensors

@@ -1,3 +1,3 @@
 version https://git-lfs.github.com/spec/v1
-oid sha256:d0abc7bcb57ebd4c85bca100446495fa1f35516ea4513f630bb3534e01a3e139
+oid sha256:57f19e858eea12b0670cd9a0f2bfae2bc9fe01a746d7440b448096c2ce2adfb5
 size 4943162336

model-00002-of-00003.safetensors

@@ -1,3 +1,3 @@
 version https://git-lfs.github.com/spec/v1
-oid sha256:9901d8efdf08230c45749f927c9bcca0fc53e1bb330c70a7989c85951effcb3e
+oid sha256:2aa4b6e10a72b39cce164d59d13823a231c012529340b9cc611ac0ed6ca31663
 size 4999819336

model-00003-of-00003.safetensors

@@ -1,3 +1,3 @@
 version https://git-lfs.github.com/spec/v1
-oid sha256:07758d9f276a4287d05416d1f89f26070bd0cd4f6fe426596e860df642b6c6e8
+oid sha256:1f15a0897c91764d009463e408e4e41110f373d906a7b49f0c5a7f33d46a1e44
 size 4540516344

runs/Jun30_02-30-12_n136-129-074/events.out.tfevents.1719685832.n136-129-074.957046.0

@@ -1,3 +1,3 @@
 version https://git-lfs.github.com/spec/v1
-oid sha256:9d1d36dc58ab04f95beb21e0dfbaead3e72c3ffbfad82ed28f170f9be1b5d0ed
-size 32947
+oid sha256:075b2ad2d9cd39a7ce491a270755179912bd03617c64afe13203471444b32bd2
+size 33989

train_results.json

@@ -1,8 +1,8 @@
 {
     "epoch": 1.0,
-    "train_loss": 0.22171047773116676,
-    "train_runtime": 5890.091,
-    "train_samples": 50000,
-    "train_samples_per_second": 8.489,
-    "train_steps_per_second": 0.066
+    "train_loss": 0.4806597235705027,
+    "train_runtime": 6367.8161,
+    "train_samples": 52922,
+    "train_samples_per_second": 8.311,
+    "train_steps_per_second": 0.065
 }

trainer_state.json

@@ -1,21 +1,21 @@
 {
   "best_metric": null,
   "best_model_checkpoint": null,
-  "epoch": 0.9980806142034548,
-  "eval_steps": 1000000,
-  "global_step": 390,
+  "epoch": 0.9987908101571947,
+  "eval_steps": 10000000,
+  "global_step": 413,
   "is_hyper_param_search": false,
   "is_local_process_zero": true,
   "is_world_process_zero": true,
   "log_history": [
     {
       "epoch": 0.0,
-      "grad_norm": 746.9239950470109,
-      "learning_rate": 1.0256410256410256e-08,
-      "logits/chosen": -2.5617921352386475,
-      "logits/rejected": -2.415619373321533,
-      "logps/chosen": -258.1644592285156,
-      "logps/rejected": -191.65736389160156,
+      "grad_norm": 26.17978267050873,
+      "learning_rate": 9.523809523809522e-09,
+      "logits/chosen": -2.7005977630615234,
+      "logits/rejected": -2.6288318634033203,
+      "logps/chosen": -1.1158788204193115,
+      "logps/rejected": -1.1333446502685547,
       "loss": 0.6931,
       "rewards/accuracies": 0.0,
       "rewards/chosen": 0.0,
@@ -24,602 +24,632 @@
       "step": 1
     },
     {
-      "epoch": 0.03,
-      "grad_norm": 712.2130131356953,
-      "learning_rate": 1.0256410256410255e-07,
-      "logits/chosen": -2.611403465270996,
-      "logits/rejected": -2.524329900741577,
-      "logps/chosen": -267.30902099609375,
-      "logps/rejected": -198.22096252441406,
-      "loss": 0.6986,
-      "rewards/accuracies": 0.4652777910232544,
-      "rewards/chosen": -0.010162555612623692,
-      "rewards/margins": 0.005689369514584541,
-      "rewards/rejected": -0.015851924195885658,
+      "epoch": 0.02,
+      "grad_norm": 20.16477598811574,
+      "learning_rate": 9.523809523809523e-08,
+      "logits/chosen": -2.762394666671753,
+      "logits/rejected": -2.697201728820801,
+      "logps/chosen": -0.8377183079719543,
+      "logps/rejected": -0.8181390762329102,
+      "loss": 0.693,
+      "rewards/accuracies": 0.4513888955116272,
+      "rewards/chosen": -0.00019922231149394065,
+      "rewards/margins": 0.0010956709738820791,
+      "rewards/rejected": -0.0012948933290317655,
       "step": 10
     },
     {
       "epoch": 0.05,
-      "grad_norm": 343.2600415646243,
-      "learning_rate": 2.051282051282051e-07,
-      "logits/chosen": -2.631699800491333,
-      "logits/rejected": -2.5292506217956543,
-      "logps/chosen": -260.6354064941406,
-      "logps/rejected": -198.5509796142578,
-      "loss": 0.5075,
-      "rewards/accuracies": 0.793749988079071,
-      "rewards/chosen": 0.3450052738189697,
-      "rewards/margins": 0.5796785354614258,
-      "rewards/rejected": -0.23467323184013367,
+      "grad_norm": 23.11513047113543,
+      "learning_rate": 1.9047619047619045e-07,
+      "logits/chosen": -2.6904549598693848,
+      "logits/rejected": -2.6507537364959717,
+      "logps/chosen": -1.0053201913833618,
+      "logps/rejected": -1.0543372631072998,
+      "loss": 0.6916,
+      "rewards/accuracies": 0.6312500238418579,
+      "rewards/chosen": -0.02715405263006687,
+      "rewards/margins": 0.008816033601760864,
+      "rewards/rejected": -0.03597008436918259,
       "step": 20
     },
     {
-      "epoch": 0.08,
-      "grad_norm": 460.51807031992456,
-      "learning_rate": 3.076923076923077e-07,
-      "logits/chosen": -2.6552329063415527,
-      "logits/rejected": -2.5521411895751953,
-      "logps/chosen": -251.4678497314453,
-      "logps/rejected": -195.8263702392578,
-      "loss": 0.2598,
-      "rewards/accuracies": 0.918749988079071,
-      "rewards/chosen": 1.8457540273666382,
-      "rewards/margins": 2.781405210494995,
-      "rewards/rejected": -0.9356509447097778,
+      "epoch": 0.07,
+      "grad_norm": 32.75609621630006,
+      "learning_rate": 2.857142857142857e-07,
+      "logits/chosen": -2.7400002479553223,
+      "logits/rejected": -2.6848459243774414,
+      "logps/chosen": -1.0142773389816284,
+      "logps/rejected": -1.0437568426132202,
+      "loss": 0.6808,
+      "rewards/accuracies": 0.7124999761581421,
+      "rewards/chosen": -0.09552808105945587,
+      "rewards/margins": 0.04477732628583908,
+      "rewards/rejected": -0.14030542969703674,
       "step": 30
     },
     {
       "epoch": 0.1,
-      "grad_norm": 386.50821557830585,
-      "learning_rate": 3.999919890759704e-07,
-      "logits/chosen": -2.6647181510925293,
-      "logits/rejected": -2.5375823974609375,
-      "logps/chosen": -243.2847442626953,
-      "logps/rejected": -199.77682495117188,
-      "loss": 0.1941,
-      "rewards/accuracies": 0.90625,
-      "rewards/chosen": 3.5360469818115234,
-      "rewards/margins": 4.843508243560791,
-      "rewards/rejected": -1.3074613809585571,
+      "grad_norm": 33.37233941529707,
+      "learning_rate": 3.809523809523809e-07,
+      "logits/chosen": -2.705505847930908,
+      "logits/rejected": -2.6594300270080566,
+      "logps/chosen": -1.0757502317428589,
+      "logps/rejected": -1.0365426540374756,
+      "loss": 0.6798,
+      "rewards/accuracies": 0.7437499761581421,
+      "rewards/chosen": -0.17448952794075012,
+      "rewards/margins": 0.07824571430683136,
+      "rewards/rejected": -0.2527352273464203,
       "step": 40
     },
     {
-      "epoch": 0.13,
-      "grad_norm": 288.39653383863197,
-      "learning_rate": 3.9903145446619833e-07,
-      "logits/chosen": -2.6978812217712402,
-      "logits/rejected": -2.5775580406188965,
-      "logps/chosen": -256.521240234375,
-      "logps/rejected": -205.7974853515625,
-      "loss": 0.2479,
-      "rewards/accuracies": 0.8999999761581421,
-      "rewards/chosen": 5.355835914611816,
-      "rewards/margins": 6.771894931793213,
-      "rewards/rejected": -1.416058897972107,
+      "epoch": 0.12,
+      "grad_norm": 13.882814308284438,
+      "learning_rate": 3.995412608484087e-07,
+      "logits/chosen": -2.8172850608825684,
+      "logits/rejected": -2.7554986476898193,
+      "logps/chosen": -1.0229161977767944,
+      "logps/rejected": -1.0335018634796143,
+      "loss": 0.6582,
+      "rewards/accuracies": 0.8125,
+      "rewards/chosen": -0.1122991219162941,
+      "rewards/margins": 0.14092899858951569,
+      "rewards/rejected": -0.2532281279563904,
       "step": 50
     },
     {
       "epoch": 0.15,
-      "grad_norm": 288.38654209122495,
-      "learning_rate": 3.9647754748313294e-07,
-      "logits/chosen": -2.67028546333313,
-      "logits/rejected": -2.5481607913970947,
-      "logps/chosen": -258.8858337402344,
-      "logps/rejected": -193.81582641601562,
-      "loss": 0.2103,
-      "rewards/accuracies": 0.925000011920929,
-      "rewards/chosen": 4.908812522888184,
-      "rewards/margins": 7.172672271728516,
-      "rewards/rejected": -2.263859748840332,
+      "grad_norm": 17.509586275945495,
+      "learning_rate": 3.976812391485896e-07,
+      "logits/chosen": -2.867166757583618,
+      "logits/rejected": -2.789944887161255,
+      "logps/chosen": -1.0946844816207886,
+      "logps/rejected": -1.1673001050949097,
+      "loss": 0.6324,
+      "rewards/accuracies": 0.800000011920929,
+      "rewards/chosen": -0.35667115449905396,
+      "rewards/margins": 0.16513890027999878,
+      "rewards/rejected": -0.5218100547790527,
       "step": 60
     },
     {
-      "epoch": 0.18,
-      "grad_norm": 271.6769920625721,
-      "learning_rate": 3.9235071376357917e-07,
-      "logits/chosen": -2.653860569000244,
-      "logits/rejected": -2.506782293319702,
-      "logps/chosen": -288.6905517578125,
-      "logps/rejected": -225.65853881835938,
-      "loss": 0.1949,
-      "rewards/accuracies": 0.9312499761581421,
-      "rewards/chosen": 4.613911151885986,
-      "rewards/margins": 8.780401229858398,
-      "rewards/rejected": -4.166489601135254,
+      "epoch": 0.17,
+      "grad_norm": 14.561097424837552,
+      "learning_rate": 3.9440458281608213e-07,
+      "logits/chosen": -2.9396042823791504,
+      "logits/rejected": -2.9019951820373535,
+      "logps/chosen": -1.211810827255249,
+      "logps/rejected": -1.2822917699813843,
+      "loss": 0.6054,
+      "rewards/accuracies": 0.793749988079071,
+      "rewards/chosen": -0.5700233578681946,
+      "rewards/margins": 0.2668416500091553,
+      "rewards/rejected": -0.8368650674819946,
       "step": 70
     },
     {
-      "epoch": 0.2,
-      "grad_norm": 381.16800166622966,
-      "learning_rate": 3.8668399121531677e-07,
-      "logits/chosen": -2.605896472930908,
-      "logits/rejected": -2.4767518043518066,
-      "logps/chosen": -260.63116455078125,
-      "logps/rejected": -215.01742553710938,
-      "loss": 0.183,
-      "rewards/accuracies": 0.925000011920929,
-      "rewards/chosen": 3.8556149005889893,
-      "rewards/margins": 8.602324485778809,
-      "rewards/rejected": -4.74670934677124,
+      "epoch": 0.19,
+      "grad_norm": 25.626763106792357,
+      "learning_rate": 3.897347732134074e-07,
+      "logits/chosen": -2.943155288696289,
+      "logits/rejected": -2.8804593086242676,
+      "logps/chosen": -1.3322759866714478,
+      "logps/rejected": -1.5549732446670532,
+      "loss": 0.5868,
+      "rewards/accuracies": 0.737500011920929,
+      "rewards/chosen": -0.8756327629089355,
+      "rewards/margins": 0.26477211713790894,
+      "rewards/rejected": -1.1404049396514893,
       "step": 80
     },
     {
-      "epoch": 0.23,
-      "grad_norm": 609.4338656498198,
-      "learning_rate": 3.795227455278028e-07,
-      "logits/chosen": -2.646833896636963,
-      "logits/rejected": -2.5277552604675293,
-      "logps/chosen": -252.8790740966797,
-      "logps/rejected": -193.67758178710938,
-      "loss": 0.2079,
-      "rewards/accuracies": 0.893750011920929,
-      "rewards/chosen": 3.7215206623077393,
-      "rewards/margins": 7.463395118713379,
-      "rewards/rejected": -3.7418742179870605,
+      "epoch": 0.22,
+      "grad_norm": 21.077234131324868,
+      "learning_rate": 3.8370527539794614e-07,
+      "logits/chosen": -2.992217540740967,
+      "logits/rejected": -2.941014051437378,
+      "logps/chosen": -1.633958101272583,
+      "logps/rejected": -1.875692367553711,
+      "loss": 0.5636,
+      "rewards/accuracies": 0.7124999761581421,
+      "rewards/chosen": -1.240767478942871,
+      "rewards/margins": 0.4708392024040222,
+      "rewards/rejected": -1.7116069793701172,
       "step": 90
     },
     {
-      "epoch": 0.26,
-      "grad_norm": 259.36924157384516,
-      "learning_rate": 3.7092430699120403e-07,
-      "logits/chosen": -2.680565357208252,
-      "logits/rejected": -2.561267375946045,
-      "logps/chosen": -233.5777587890625,
-      "logps/rejected": -198.25820922851562,
-      "loss": 0.1876,
-      "rewards/accuracies": 0.918749988079071,
-      "rewards/chosen": 4.603877067565918,
-      "rewards/margins": 7.88943338394165,
-      "rewards/rejected": -3.285557270050049,
+      "epoch": 0.24,
+      "grad_norm": 20.60031539796354,
+      "learning_rate": 3.763592983027255e-07,
+      "logits/chosen": -3.0840325355529785,
+      "logits/rejected": -3.028627872467041,
+      "logps/chosen": -1.6897590160369873,
+      "logps/rejected": -1.8915116786956787,
+      "loss": 0.5526,
+      "rewards/accuracies": 0.731249988079071,
+      "rewards/chosen": -1.596776008605957,
+      "rewards/margins": 0.387612909078598,
+      "rewards/rejected": -1.9843889474868774,
       "step": 100
     },
     {
-      "epoch": 0.28,
-      "grad_norm": 344.20413939741326,
-      "learning_rate": 3.6095751153125107e-07,
-      "logits/chosen": -2.6716485023498535,
-      "logits/rejected": -2.5542380809783936,
-      "logps/chosen": -252.8291778564453,
-      "logps/rejected": -193.85076904296875,
-      "loss": 0.1768,
-      "rewards/accuracies": 0.9437500238418579,
-      "rewards/chosen": 5.924564361572266,
-      "rewards/margins": 8.784211158752441,
-      "rewards/rejected": -2.8596463203430176,
+      "epoch": 0.27,
+      "grad_norm": 27.384903203298304,
+      "learning_rate": 3.6774948509008527e-07,
+      "logits/chosen": -3.1617236137390137,
+      "logits/rejected": -3.1181647777557373,
+      "logps/chosen": -2.0058510303497314,
+      "logps/rejected": -2.295015335083008,
+      "loss": 0.5213,
+      "rewards/accuracies": 0.71875,
+      "rewards/chosen": -2.055260181427002,
+      "rewards/margins": 0.6770964860916138,
+      "rewards/rejected": -2.732356548309326,
       "step": 110
     },
     {
-      "epoch": 0.31,
-      "grad_norm": 298.19415193137047,
-      "learning_rate": 3.497021496342202e-07,
-      "logits/chosen": -2.648298501968384,
-      "logits/rejected": -2.5618278980255127,
-      "logps/chosen": -257.97100830078125,
-      "logps/rejected": -214.90670776367188,
-      "loss": 0.2066,
-      "rewards/accuracies": 0.925000011920929,
-      "rewards/chosen": 4.884228229522705,
-      "rewards/margins": 9.177003860473633,
-      "rewards/rejected": -4.292778015136719,
+      "epoch": 0.29,
+      "grad_norm": 26.797786638298437,
+      "learning_rate": 3.579375358972288e-07,
+      "logits/chosen": -3.113551616668701,
+      "logits/rejected": -3.081834316253662,
+      "logps/chosen": -1.9728782176971436,
+      "logps/rejected": -2.4605767726898193,
+      "loss": 0.4952,
+      "rewards/accuracies": 0.75,
+      "rewards/chosen": -2.184112548828125,
+      "rewards/margins": 0.8185291290283203,
+      "rewards/rejected": -3.0026419162750244,
       "step": 120
     },
     {
-      "epoch": 0.33,
-      "grad_norm": 143.10334660096893,
-      "learning_rate": 3.372483275737467e-07,
-      "logits/chosen": -2.6251893043518066,
-      "logits/rejected": -2.5034987926483154,
-      "logps/chosen": -250.27548217773438,
-      "logps/rejected": -191.64712524414062,
-      "loss": 0.1938,
-      "rewards/accuracies": 0.9125000238418579,
-      "rewards/chosen": 3.4552054405212402,
-      "rewards/margins": 9.01502799987793,
-      "rewards/rejected": -5.559823513031006,
+      "epoch": 0.31,
+      "grad_norm": 26.030117791456227,
+      "learning_rate": 3.4699376567716156e-07,
+      "logits/chosen": -3.1078505516052246,
+      "logits/rejected": -3.083446979522705,
+      "logps/chosen": -2.0211844444274902,
+      "logps/rejected": -2.5097596645355225,
+      "loss": 0.5018,
+      "rewards/accuracies": 0.737500011920929,
+      "rewards/chosen": -2.3030853271484375,
+      "rewards/margins": 1.0088802576065063,
+      "rewards/rejected": -3.3119654655456543,
       "step": 130
     },
     {
-      "epoch": 0.36,
-      "grad_norm": 396.95461669932075,
-      "learning_rate": 3.2369574605325094e-07,
-      "logits/chosen": -2.6585159301757812,
-      "logits/rejected": -2.5590972900390625,
-      "logps/chosen": -270.13775634765625,
-      "logps/rejected": -210.39706420898438,
-      "loss": 0.2585,
-      "rewards/accuracies": 0.9312499761581421,
-      "rewards/chosen": 3.534144639968872,
-      "rewards/margins": 8.91390609741211,
-      "rewards/rejected": -5.379761695861816,
+      "epoch": 0.34,
+      "grad_norm": 29.499216323929453,
+      "learning_rate": 3.349966003036421e-07,
+      "logits/chosen": -3.024627208709717,
+      "logits/rejected": -3.009237766265869,
+      "logps/chosen": -1.9538856744766235,
+      "logps/rejected": -2.388256311416626,
+      "loss": 0.4969,
+      "rewards/accuracies": 0.71875,
+      "rewards/chosen": -2.0561270713806152,
+      "rewards/margins": 0.8914540410041809,
+      "rewards/rejected": -2.9475810527801514,
       "step": 140
     },
     {
-      "epoch": 0.38,
-      "grad_norm": 248.0981758080623,
-      "learning_rate": 3.091529020389009e-07,
-      "logits/chosen": -2.6475913524627686,
-      "logits/rejected": -2.5541744232177734,
-      "logps/chosen": -262.02142333984375,
-      "logps/rejected": -214.84573364257812,
-      "loss": 0.2768,
-      "rewards/accuracies": 0.918749988079071,
-      "rewards/chosen": 4.376931190490723,
-      "rewards/margins": 9.941673278808594,
-      "rewards/rejected": -5.564742088317871,
+      "epoch": 0.36,
+      "grad_norm": 27.290843756174784,
+      "learning_rate": 3.220320145511884e-07,
+      "logits/chosen": -3.063863754272461,
+      "logits/rejected": -3.0345683097839355,
+      "logps/chosen": -2.092048168182373,
+      "logps/rejected": -2.6104140281677246,
+      "loss": 0.4448,
+      "rewards/accuracies": 0.768750011920929,
+      "rewards/chosen": -2.272033929824829,
+      "rewards/margins": 0.9999080896377563,
+      "rewards/rejected": -3.271941661834717,
       "step": 150
     },
     {
-      "epoch": 0.41,
-      "grad_norm": 531.5498022382194,
-      "learning_rate": 2.937362201729407e-07,
-      "logits/chosen": -2.6315155029296875,
-      "logits/rejected": -2.5306897163391113,
-      "logps/chosen": -240.9332275390625,
-      "logps/rejected": -186.1042022705078,
-      "loss": 0.2307,
-      "rewards/accuracies": 0.9437500238418579,
-      "rewards/chosen": 4.558727741241455,
-      "rewards/margins": 8.91313362121582,
-      "rewards/rejected": -4.354406356811523,
+      "epoch": 0.39,
+      "grad_norm": 58.56715174010242,
+      "learning_rate": 3.0819291597771795e-07,
+      "logits/chosen": -3.0004026889801025,
+      "logits/rejected": -2.9992613792419434,
+      "logps/chosen": -2.5250463485717773,
+      "logps/rejected": -3.2703194618225098,
+      "loss": 0.4704,
+      "rewards/accuracies": 0.7749999761581421,
+      "rewards/chosen": -3.229747772216797,
+      "rewards/margins": 1.4721379280090332,
+      "rewards/rejected": -4.701886177062988,
       "step": 160
     },
     {
-      "epoch": 0.44,
-      "grad_norm": 204.85526693169226,
-      "learning_rate": 2.77569120720971e-07,
-      "logits/chosen": -2.6248703002929688,
-      "logits/rejected": -2.520388126373291,
-      "logps/chosen": -259.86029052734375,
-      "logps/rejected": -202.1363067626953,
-      "loss": 0.1656,
-      "rewards/accuracies": 0.956250011920929,
-      "rewards/chosen": 4.666792869567871,
-      "rewards/margins": 9.15965747833252,
-      "rewards/rejected": -4.492865085601807,
+      "epoch": 0.41,
+      "grad_norm": 25.42012705976763,
+      "learning_rate": 2.9357847912507786e-07,
+      "logits/chosen": -2.9128940105438232,
+      "logits/rejected": -2.865865468978882,
+      "logps/chosen": -2.3983154296875,
+      "logps/rejected": -3.011444568634033,
+      "loss": 0.4553,
+      "rewards/accuracies": 0.7875000238418579,
+      "rewards/chosen": -3.0126404762268066,
+      "rewards/margins": 1.2925684452056885,
+      "rewards/rejected": -4.305209159851074,
       "step": 170
     },
     {
-      "epoch": 0.46,
-      "grad_norm": 346.922389452496,
-      "learning_rate": 2.6078103151484935e-07,
-      "logits/chosen": -2.6694693565368652,
-      "logits/rejected": -2.5681331157684326,
-      "logps/chosen": -235.467041015625,
-      "logps/rejected": -193.00096130371094,
-      "loss": 0.2092,
-      "rewards/accuracies": 0.9312499761581421,
-      "rewards/chosen": 3.1894354820251465,
-      "rewards/margins": 8.24107551574707,
-      "rewards/rejected": -5.051640033721924,
+      "epoch": 0.44,
+      "grad_norm": 38.47933377900249,
+      "learning_rate": 2.7829343480875617e-07,
+      "logits/chosen": -2.856142520904541,
+      "logits/rejected": -2.8031108379364014,
+      "logps/chosen": -2.4536030292510986,
+      "logps/rejected": -3.0898194313049316,
+      "loss": 0.4496,
+      "rewards/accuracies": 0.7562500238418579,
+      "rewards/chosen": -3.0219366550445557,
+      "rewards/margins": 1.2693630456924438,
+      "rewards/rejected": -4.291299343109131,
       "step": 180
     },
     {
-      "epoch": 0.49,
-      "grad_norm": 469.7545752914211,
-      "learning_rate": 2.435063518012335e-07,
-      "logits/chosen": -2.627861499786377,
-      "logits/rejected": -2.537619113922119,
-      "logps/chosen": -257.1391906738281,
-      "logps/rejected": -208.8784637451172,
-      "loss": 0.2104,
-      "rewards/accuracies": 0.918749988079071,
-      "rewards/chosen": 3.447594404220581,
-      "rewards/margins": 8.870077133178711,
-      "rewards/rejected": -5.422482013702393,
+      "epoch": 0.46,
+      "grad_norm": 32.18961859239255,
+      "learning_rate": 2.624473195899052e-07,
+      "logits/chosen": -2.8849470615386963,
+      "logits/rejected": -2.8798046112060547,
+      "logps/chosen": -2.532538890838623,
+      "logps/rejected": -3.2981109619140625,
+      "loss": 0.4317,
+      "rewards/accuracies": 0.75,
+      "rewards/chosen": -3.1526973247528076,
+      "rewards/margins": 1.4344627857208252,
+      "rewards/rejected": -4.587159633636475,
       "step": 190
     },
     {
-      "epoch": 0.51,
-      "grad_norm": 532.7452067278298,
-      "learning_rate": 2.2588337629081105e-07,
-      "logits/chosen": -2.669545888900757,
-      "logits/rejected": -2.5410807132720947,
-      "logps/chosen": -270.0424499511719,
-      "logps/rejected": -204.46751403808594,
-      "loss": 0.1802,
-      "rewards/accuracies": 0.925000011920929,
-      "rewards/chosen": 5.255965709686279,
-      "rewards/margins": 10.433439254760742,
-      "rewards/rejected": -5.177473068237305,
+      "epoch": 0.48,
+      "grad_norm": 27.9570447315806,
+      "learning_rate": 2.4615369080815547e-07,
+      "logits/chosen": -2.8126208782196045,
+      "logits/rejected": -2.783909797668457,
+      "logps/chosen": -2.4390881061553955,
+      "logps/rejected": -3.303173780441284,
+      "loss": 0.415,
+      "rewards/accuracies": 0.7875000238418579,
+      "rewards/chosen": -3.163785219192505,
+      "rewards/margins": 1.6035875082015991,
+      "rewards/rejected": -4.7673726081848145,
       "step": 200
     },
     {
-      "epoch": 0.54,
-      "grad_norm": 489.8003677533757,
-      "learning_rate": 2.0805318802188303e-07,
-      "logits/chosen": -2.656721353530884,
-      "logits/rejected": -2.548938035964966,
-      "logps/chosen": -262.646728515625,
-      "logps/rejected": -213.17138671875,
-      "loss": 0.1767,
-      "rewards/accuracies": 0.925000011920929,
-      "rewards/chosen": 4.205015659332275,
-      "rewards/margins": 9.936458587646484,
-      "rewards/rejected": -5.731442451477051,
+      "epoch": 0.51,
+      "grad_norm": 27.22162141586115,
+      "learning_rate": 2.2952931280049625e-07,
+      "logits/chosen": -2.817441940307617,
+      "logits/rejected": -2.767930507659912,
+      "logps/chosen": -2.603651762008667,
+      "logps/rejected": -3.2926082611083984,
+      "loss": 0.4406,
+      "rewards/accuracies": 0.7124999761581421,
+      "rewards/chosen": -3.163153886795044,
+      "rewards/margins": 1.5288965702056885,
+      "rewards/rejected": -4.692050933837891,
       "step": 210
     },
     {
-      "epoch": 0.56,
-      "grad_norm": 398.2303862264868,
-      "learning_rate": 1.9015852890162436e-07,
-      "logits/chosen": -2.65899658203125,
-      "logits/rejected": -2.564746856689453,
-      "logps/chosen": -255.7983856201172,
-      "logps/rejected": -209.6800079345703,
-      "loss": 0.2859,
-      "rewards/accuracies": 0.918749988079071,
-      "rewards/chosen": 3.3606581687927246,
-      "rewards/margins": 9.43185806274414,
-      "rewards/rejected": -6.071199417114258,
+      "epoch": 0.53,
+      "grad_norm": 29.81747102837208,
+      "learning_rate": 2.1269332013798747e-07,
+      "logits/chosen": -2.772239923477173,
+      "logits/rejected": -2.751819133758545,
+      "logps/chosen": -2.564336061477661,
+      "logps/rejected": -3.258159637451172,
+      "loss": 0.4602,
+      "rewards/accuracies": 0.7437499761581421,
+      "rewards/chosen": -3.343303680419922,
+      "rewards/margins": 1.45319402217865,
+      "rewards/rejected": -4.7964982986450195,
       "step": 220
     },
     {
-      "epoch": 0.59,
-      "grad_norm": 345.5121551134584,
-      "learning_rate": 1.723426569670534e-07,
-      "logits/chosen": -2.685997247695923,
-      "logits/rejected": -2.5591225624084473,
-      "logps/chosen": -264.55902099609375,
-      "logps/rejected": -209.5423126220703,
-      "loss": 0.1839,
-      "rewards/accuracies": 0.893750011920929,
-      "rewards/chosen": 4.420236587524414,
-      "rewards/margins": 9.53612232208252,
-      "rewards/rejected": -5.1158857345581055,
+      "epoch": 0.56,
+      "grad_norm": 26.477303615737316,
+      "learning_rate": 1.9576636387676436e-07,
+      "logits/chosen": -2.7296347618103027,
+      "logits/rejected": -2.693775177001953,
+      "logps/chosen": -1.9129469394683838,
+      "logps/rejected": -2.428316593170166,
+      "loss": 0.4332,
+      "rewards/accuracies": 0.762499988079071,
+      "rewards/chosen": -1.9380298852920532,
+      "rewards/margins": 1.0814852714538574,
+      "rewards/rejected": -3.019515037536621,
       "step": 230
     },
     {
-      "epoch": 0.61,
-      "grad_norm": 366.69298678156184,
-      "learning_rate": 1.547481995140556e-07,
-      "logits/chosen": -2.6587154865264893,
-      "logits/rejected": -2.5532352924346924,
-      "logps/chosen": -258.416259765625,
-      "logps/rejected": -213.5823974609375,
-      "loss": 0.1755,
-      "rewards/accuracies": 0.9312499761581421,
-      "rewards/chosen": 4.108377456665039,
-      "rewards/margins": 10.194208145141602,
-      "rewards/rejected": -6.085830211639404,
+      "epoch": 0.58,
+      "grad_norm": 27.65148512042932,
+      "learning_rate": 1.7886974694151976e-07,
+      "logits/chosen": -2.672154188156128,
+      "logits/rejected": -2.644366502761841,
+      "logps/chosen": -2.446423053741455,
+      "logps/rejected": -3.194565534591675,
+      "loss": 0.4135,
+      "rewards/accuracies": 0.8062499761581421,
+      "rewards/chosen": -2.9146509170532227,
+      "rewards/margins": 1.5267184972763062,
+      "rewards/rejected": -4.44136905670166,
       "step": 240
     },
     {
-      "epoch": 0.64,
-      "grad_norm": 355.57895419726833,
-      "learning_rate": 1.375160112758885e-07,
-      "logits/chosen": -2.6688976287841797,
-      "logits/rejected": -2.559256076812744,
-      "logps/chosen": -264.78363037109375,
-      "logps/rejected": -219.9192352294922,
-      "loss": 0.1867,
-      "rewards/accuracies": 0.956250011920929,
-      "rewards/chosen": 3.6294102668762207,
-      "rewards/margins": 10.07470417022705,
-      "rewards/rejected": -6.445294380187988,
+      "epoch": 0.6,
+      "grad_norm": 34.59064472191795,
+      "learning_rate": 1.6212455483752895e-07,
+      "logits/chosen": -2.6410295963287354,
+      "logits/rejected": -2.525394916534424,
+      "logps/chosen": -2.9665780067443848,
+      "logps/rejected": -4.038729667663574,
+      "loss": 0.4021,
+      "rewards/accuracies": 0.78125,
+      "rewards/chosen": -4.13940954208374,
+      "rewards/margins": 2.119044065475464,
+      "rewards/rejected": -6.258453845977783,
       "step": 250
     },
     {
-      "epoch": 0.67,
-      "grad_norm": 284.33502896066835,
-      "learning_rate": 1.2078404679216862e-07,
-      "logits/chosen": -2.6680378913879395,
-      "logits/rejected": -2.5718884468078613,
-      "logps/chosen": -258.0495910644531,
-      "logps/rejected": -212.5093231201172,
-      "loss": 0.1465,
-      "rewards/accuracies": 0.949999988079071,
-      "rewards/chosen": 3.9134533405303955,
-      "rewards/margins": 9.667370796203613,
-      "rewards/rejected": -5.753917694091797,
+      "epoch": 0.63,
+      "grad_norm": 42.39824760145083,
+      "learning_rate": 1.4565078792075733e-07,
+      "logits/chosen": -2.5688576698303223,
+      "logits/rejected": -2.4721179008483887,
+      "logps/chosen": -2.9468441009521484,
+      "logps/rejected": -3.990417003631592,
+      "loss": 0.4211,
+      "rewards/accuracies": 0.8187500238418579,
+      "rewards/chosen": -3.876873016357422,
+      "rewards/margins": 2.1601932048797607,
+      "rewards/rejected": -6.037066459655762,
       "step": 260
     },
     {
-      "epoch": 0.69,
-      "grad_norm": 281.9194802132945,
-      "learning_rate": 1.0468625599573841e-07,
-      "logits/chosen": -2.6398463249206543,
-      "logits/rejected": -2.5165765285491943,
-      "logps/chosen": -268.0271911621094,
-      "logps/rejected": -219.8000946044922,
-      "loss": 0.1901,
-      "rewards/accuracies": 0.9437500238418579,
-      "rewards/chosen": 4.4066267013549805,
-      "rewards/margins": 10.626806259155273,
-      "rewards/rejected": -6.220180511474609,
+      "epoch": 0.65,
+      "grad_norm": 34.18349239080256,
+      "learning_rate": 1.295665014444281e-07,
+      "logits/chosen": -2.6383066177368164,
+      "logits/rejected": -2.543112277984619,
+      "logps/chosen": -2.683907985687256,
+      "logps/rejected": -3.600834369659424,
+      "loss": 0.4175,
+      "rewards/accuracies": 0.831250011920929,
+      "rewards/chosen": -3.4698543548583984,
+      "rewards/margins": 1.888292908668518,
+      "rewards/rejected": -5.358147621154785,
       "step": 270
     },
     {
-      "epoch": 0.72,
-      "grad_norm": 262.1937325706126,
-      "learning_rate": 8.935151185893727e-08,
-      "logits/chosen": -2.6787869930267334,
-      "logits/rejected": -2.56150484085083,
-      "logps/chosen": -266.0087585449219,
-      "logps/rejected": -210.2588348388672,
-      "loss": 0.1777,
-      "rewards/accuracies": 0.918749988079071,
-      "rewards/chosen": 3.8459548950195312,
-      "rewards/margins": 10.500468254089355,
-      "rewards/rejected": -6.654513359069824,
+      "epoch": 0.68,
+      "grad_norm": 32.21777836724207,
+      "learning_rate": 1.1398695954469597e-07,
+      "logits/chosen": -2.6166112422943115,
+      "logits/rejected": -2.5435919761657715,
+      "logps/chosen": -2.4326510429382324,
+      "logps/rejected": -3.2459170818328857,
+      "loss": 0.4301,
+      "rewards/accuracies": 0.7562500238418579,
+      "rewards/chosen": -3.0369038581848145,
+      "rewards/margins": 1.6798131465911865,
+      "rewards/rejected": -4.716716766357422,
       "step": 280
     },
     {
-      "epoch": 0.74,
-      "grad_norm": 267.1170614305438,
-      "learning_rate": 7.490257868414449e-08,
-      "logits/chosen": -2.6337618827819824,
-      "logits/rejected": -2.5607006549835205,
-      "logps/chosen": -266.64654541015625,
-      "logps/rejected": -227.7572479248047,
-      "loss": 0.1881,
-      "rewards/accuracies": 0.9375,
-      "rewards/chosen": 3.585491180419922,
-      "rewards/margins": 9.039448738098145,
-      "rewards/rejected": -5.453957557678223,
+      "epoch": 0.7,
+      "grad_norm": 34.76609006558408,
+      "learning_rate": 9.902380922818425e-08,
+      "logits/chosen": -2.639005184173584,
+      "logits/rejected": -2.5904557704925537,
+      "logps/chosen": -2.9825644493103027,
+      "logps/rejected": -3.8297367095947266,
+      "loss": 0.3838,
+      "rewards/accuracies": 0.800000011920929,
+      "rewards/chosen": -3.9611072540283203,
+      "rewards/margins": 1.7571494579315186,
+      "rewards/rejected": -5.718256950378418,
       "step": 290
     },
     {
-      "epoch": 0.77,
-      "grad_norm": 337.4409958446907,
-      "learning_rate": 6.145512929808013e-08,
-      "logits/chosen": -2.6632189750671387,
-      "logits/rejected": -2.548957109451294,
-      "logps/chosen": -270.69012451171875,
-      "logps/rejected": -219.18807983398438,
-      "loss": 0.1755,
-      "rewards/accuracies": 0.956250011920929,
-      "rewards/chosen": 3.8013317584991455,
-      "rewards/margins": 10.349567413330078,
-      "rewards/rejected": -6.5482354164123535,
+      "epoch": 0.73,
+      "grad_norm": 31.40350929501295,
+      "learning_rate": 8.478428028080398e-08,
+      "logits/chosen": -2.602800130844116,
+      "logits/rejected": -2.540292263031006,
+      "logps/chosen": -2.592560052871704,
+      "logps/rejected": -3.4401028156280518,
+      "loss": 0.4113,
+      "rewards/accuracies": 0.793749988079071,
+      "rewards/chosen": -3.383810520172119,
+      "rewards/margins": 1.6390845775604248,
+      "rewards/rejected": -5.022894859313965,
       "step": 300
     },
     {
-      "epoch": 0.79,
-      "grad_norm": 161.29540411078162,
-      "learning_rate": 4.911681901784197e-08,
-      "logits/chosen": -2.620042324066162,
-      "logits/rejected": -2.5278143882751465,
-      "logps/chosen": -275.6562805175781,
-      "logps/rejected": -213.7178192138672,
-      "loss": 0.188,
-      "rewards/accuracies": 0.9312499761581421,
-      "rewards/chosen": 3.946884870529175,
-      "rewards/margins": 10.046659469604492,
-      "rewards/rejected": -6.0997748374938965,
+      "epoch": 0.75,
+      "grad_norm": 36.190730721174226,
+      "learning_rate": 7.137041683151202e-08,
+      "logits/chosen": -2.5839016437530518,
+      "logits/rejected": -2.506042003631592,
+      "logps/chosen": -2.746112585067749,
+      "logps/rejected": -3.739596128463745,
+      "loss": 0.4097,
+      "rewards/accuracies": 0.793749988079071,
+      "rewards/chosen": -3.345579147338867,
+      "rewards/margins": 2.1395156383514404,
+      "rewards/rejected": -5.485095024108887,
       "step": 310
     },
     {
-      "epoch": 0.82,
-      "grad_norm": 394.51415258074036,
-      "learning_rate": 3.79864238021667e-08,
-      "logits/chosen": -2.6458163261413574,
-      "logits/rejected": -2.5520365238189697,
-      "logps/chosen": -258.85113525390625,
-      "logps/rejected": -212.78414916992188,
-      "loss": 0.2281,
-      "rewards/accuracies": 0.887499988079071,
-      "rewards/chosen": 3.033263683319092,
-      "rewards/margins": 9.521443367004395,
-      "rewards/rejected": -6.4881792068481445,
+      "epoch": 0.77,
+      "grad_norm": 38.28420667279529,
+      "learning_rate": 5.8878346077822135e-08,
+      "logits/chosen": -2.605834484100342,
+      "logits/rejected": -2.5087852478027344,
+      "logps/chosen": -2.6911044120788574,
+      "logps/rejected": -3.7380852699279785,
+      "loss": 0.3982,
+      "rewards/accuracies": 0.8187500238418579,
+      "rewards/chosen": -3.573197841644287,
+      "rewards/margins": 2.0680930614471436,
+      "rewards/rejected": -5.641291618347168,
       "step": 320
     },
     {
-      "epoch": 0.84,
-      "grad_norm": 235.27856577918993,
-      "learning_rate": 2.8153049487556634e-08,
-      "logits/chosen": -2.6645383834838867,
-      "logits/rejected": -2.5732433795928955,
-      "logps/chosen": -260.6364440917969,
-      "logps/rejected": -212.11898803710938,
-      "loss": 0.1828,
-      "rewards/accuracies": 0.9750000238418579,
-      "rewards/chosen": 4.755698204040527,
-      "rewards/margins": 10.375238418579102,
-      "rewards/rejected": -5.619540214538574,
+      "epoch": 0.8,
+      "grad_norm": 27.520252680470655,
+      "learning_rate": 4.73975894135696e-08,
+      "logits/chosen": -2.558149814605713,
+      "logits/rejected": -2.4873461723327637,
+      "logps/chosen": -2.4535582065582275,
+      "logps/rejected": -3.269908905029297,
+      "loss": 0.4337,
+      "rewards/accuracies": 0.8062499761581421,
+      "rewards/chosen": -3.0535290241241455,
+      "rewards/margins": 1.616397500038147,
+      "rewards/rejected": -4.669926643371582,
       "step": 330
     },
     {
-      "epoch": 0.87,
-      "grad_norm": 326.32410341900527,
-      "learning_rate": 1.9695418439836796e-08,
-      "logits/chosen": -2.6518936157226562,
-      "logits/rejected": -2.555978298187256,
-      "logps/chosen": -256.3904724121094,
-      "logps/rejected": -200.47169494628906,
-      "loss": 0.2192,
-      "rewards/accuracies": 0.9437500238418579,
-      "rewards/chosen": 4.210465908050537,
-      "rewards/margins": 10.564642906188965,
-      "rewards/rejected": -6.3541765213012695,
+      "epoch": 0.82,
+      "grad_norm": 40.737401644115494,
+      "learning_rate": 3.701042089556483e-08,
+      "logits/chosen": -2.635023593902588,
+      "logits/rejected": -2.5549774169921875,
+      "logps/chosen": -2.5029196739196777,
+      "logps/rejected": -3.4127399921417236,
+      "loss": 0.3918,
+      "rewards/accuracies": 0.824999988079071,
+      "rewards/chosen": -3.2250847816467285,
+      "rewards/margins": 1.7689367532730103,
+      "rewards/rejected": -4.994021415710449,
       "step": 340
     },
     {
-      "epoch": 0.9,
-      "grad_norm": 356.14147259437647,
-      "learning_rate": 1.2681239331945692e-08,
-      "logits/chosen": -2.666987895965576,
-      "logits/rejected": -2.55633544921875,
-      "logps/chosen": -265.26556396484375,
-      "logps/rejected": -217.6700897216797,
-      "loss": 0.2137,
-      "rewards/accuracies": 0.918749988079071,
-      "rewards/chosen": 3.833866596221924,
-      "rewards/margins": 9.782751083374023,
-      "rewards/rejected": -5.9488844871521,
+      "epoch": 0.85,
+      "grad_norm": 27.371576928481083,
+      "learning_rate": 2.779127764652889e-08,
+      "logits/chosen": -2.5536205768585205,
+      "logits/rejected": -2.4854509830474854,
+      "logps/chosen": -2.5617735385894775,
+      "logps/rejected": -3.4692695140838623,
+      "loss": 0.415,
+      "rewards/accuracies": 0.793749988079071,
+      "rewards/chosen": -3.1769766807556152,
+      "rewards/margins": 1.873204231262207,
+      "rewards/rejected": -5.0501813888549805,
       "step": 350
     },
     {
-      "epoch": 0.92,
-      "grad_norm": 330.15606711737206,
-      "learning_rate": 7.166665093287538e-09,
-      "logits/chosen": -2.6728262901306152,
-      "logits/rejected": -2.573294162750244,
-      "logps/chosen": -271.0104675292969,
-      "logps/rejected": -214.70834350585938,
-      "loss": 0.1861,
-      "rewards/accuracies": 0.949999988079071,
-      "rewards/chosen": 4.187949180603027,
-      "rewards/margins": 10.108842849731445,
-      "rewards/rejected": -5.920893669128418,
+      "epoch": 0.87,
+      "grad_norm": 27.636977967012264,
+      "learning_rate": 1.9806226419516193e-08,
+      "logits/chosen": -2.5588951110839844,
+      "logits/rejected": -2.494626998901367,
+      "logps/chosen": -2.6627163887023926,
+      "logps/rejected": -3.467442750930786,
+      "loss": 0.4046,
+      "rewards/accuracies": 0.7437499761581421,
+      "rewards/chosen": -3.390608310699463,
+      "rewards/margins": 1.5652096271514893,
+      "rewards/rejected": -4.955817699432373,
       "step": 360
     },
     {
-      "epoch": 0.95,
-      "grad_norm": 244.55642297419294,
-      "learning_rate": 3.1958433701019694e-09,
-      "logits/chosen": -2.6283528804779053,
-      "logits/rejected": -2.5266776084899902,
-      "logps/chosen": -269.41925048828125,
-      "logps/rejected": -217.0562744140625,
-      "loss": 0.1859,
-      "rewards/accuracies": 0.9125000238418579,
-      "rewards/chosen": 4.10398006439209,
-      "rewards/margins": 9.084988594055176,
-      "rewards/rejected": -4.981008052825928,
+      "epoch": 0.89,
+      "grad_norm": 32.06703942545683,
+      "learning_rate": 1.3112490146559552e-08,
+      "logits/chosen": -2.5901262760162354,
+      "logits/rejected": -2.505229949951172,
+      "logps/chosen": -2.562215805053711,
+      "logps/rejected": -3.4887568950653076,
+      "loss": 0.3925,
+      "rewards/accuracies": 0.8125,
+      "rewards/chosen": -3.3314082622528076,
+      "rewards/margins": 1.8955705165863037,
+      "rewards/rejected": -5.2269792556762695,
       "step": 370
     },
     {
-      "epoch": 0.97,
-      "grad_norm": 197.28407297465554,
-      "learning_rate": 8.005630957010012e-10,
-      "logits/chosen": -2.682586193084717,
-      "logits/rejected": -2.5935988426208496,
-      "logps/chosen": -260.6213073730469,
-      "logps/rejected": -221.34423828125,
-      "loss": 0.1775,
-      "rewards/accuracies": 0.918749988079071,
-      "rewards/chosen": 4.256922245025635,
-      "rewards/margins": 9.350181579589844,
-      "rewards/rejected": -5.093258857727051,
+      "epoch": 0.92,
+      "grad_norm": 38.64904344644471,
+      "learning_rate": 7.758037864413247e-09,
+      "logits/chosen": -2.5616049766540527,
+      "logits/rejected": -2.516554117202759,
+      "logps/chosen": -2.6926965713500977,
+      "logps/rejected": -3.6225249767303467,
+      "loss": 0.4086,
+      "rewards/accuracies": 0.8187500238418579,
+      "rewards/chosen": -3.5749175548553467,
+      "rewards/margins": 1.7677252292633057,
+      "rewards/rejected": -5.342642784118652,
       "step": 380
     },
     {
-      "epoch": 1.0,
-      "grad_norm": 504.7364263980203,
-      "learning_rate": 0.0,
-      "logits/chosen": -2.686406135559082,
-      "logits/rejected": -2.582782745361328,
-      "logps/chosen": -248.7636260986328,
-      "logps/rejected": -209.47628784179688,
-      "loss": 0.1791,
-      "rewards/accuracies": 0.9312499761581421,
-      "rewards/chosen": 3.9559149742126465,
-      "rewards/margins": 10.19153118133545,
-      "rewards/rejected": -6.235617160797119,
+      "epoch": 0.94,
+      "grad_norm": 31.576303147697956,
+      "learning_rate": 3.78124095609087e-09,
+      "logits/chosen": -2.529954671859741,
+      "logits/rejected": -2.4696826934814453,
+      "logps/chosen": -2.5880253314971924,
+      "logps/rejected": -3.5366291999816895,
+      "loss": 0.3992,
+      "rewards/accuracies": 0.8187500238418579,
+      "rewards/chosen": -3.3176732063293457,
+      "rewards/margins": 1.8059123754501343,
+      "rewards/rejected": -5.1235857009887695,
       "step": 390
     },
+    {
+      "epoch": 0.97,
+      "grad_norm": 31.75576068376339,
+      "learning_rate": 1.2105981716597603e-09,
+      "logits/chosen": -2.5555756092071533,
+      "logits/rejected": -2.4442620277404785,
+      "logps/chosen": -2.6657116413116455,
+      "logps/rejected": -3.635850429534912,
+      "loss": 0.4179,
+      "rewards/accuracies": 0.78125,
+      "rewards/chosen": -3.39148211479187,
+      "rewards/margins": 2.059187412261963,
+      "rewards/rejected": -5.450669288635254,
+      "step": 400
+    },
+    {
+      "epoch": 0.99,
+      "grad_norm": 31.71069378165557,
+      "learning_rate": 6.453139886395398e-11,
+      "logits/chosen": -2.5864222049713135,
+      "logits/rejected": -2.5135445594787598,
+      "logps/chosen": -2.647493362426758,
+      "logps/rejected": -3.583590030670166,
+      "loss": 0.414,
+      "rewards/accuracies": 0.7875000238418579,
+      "rewards/chosen": -3.436647891998291,
+      "rewards/margins": 1.8721681833267212,
+      "rewards/rejected": -5.308815956115723,
+      "step": 410
+    },
     {
       "epoch": 1.0,
-      "step": 390,
+      "step": 413,
       "total_flos": 0.0,
-      "train_loss": 0.22171047773116676,
-      "train_runtime": 5890.091,
-      "train_samples_per_second": 8.489,
-      "train_steps_per_second": 0.066
+      "train_loss": 0.4806597235705027,
+      "train_runtime": 6367.8161,
+      "train_samples_per_second": 8.311,
+      "train_steps_per_second": 0.065
     }
   ],
   "logging_steps": 10,
-  "max_steps": 390,
+  "max_steps": 413,
   "num_input_tokens_seen": 0,
   "num_train_epochs": 1,
   "save_steps": 100,