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
 
@@ -32,7 +32,7 @@ More information needed
 ### Training hyperparameters
 
 The following hyperparameters were used during training:
-- learning_rate: 5e-09
+- learning_rate: 1e-08
 - train_batch_size: 8
 - eval_batch_size: 8
 - seed: 42
@@ -44,7 +44,7 @@ The following hyperparameters were used during training:
 - optimizer: Adam with betas=(0.9,0.999) and epsilon=1e-08
 - lr_scheduler_type: cosine
 - lr_scheduler_warmup_ratio: 0.1
-- num_epochs: 1
+- num_epochs: 2
 
 ### Training results
 
@@ -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": 0.7162073644002279,
-    "train_runtime": 4512.4874,
+    "epoch": 2.0,
+    "train_loss": 0.6411590957641602,
+    "train_runtime": 8942.072,
     "train_samples": 38445,
-    "train_samples_per_second": 8.52,
-    "train_steps_per_second": 0.033
+    "train_samples_per_second": 8.599,
+    "train_steps_per_second": 0.034
 }

model-00001-of-00003.safetensors

@@ -1,3 +1,3 @@
 version https://git-lfs.github.com/spec/v1
-oid sha256:c198eb9df08ec1da45189177d9165731677bcf8abcb6d6938bc58bf75d03977d
+oid sha256:e95be6f8c0aa9a00e2dec7409b07bb3971685d185f076a462b8f78d03a1941fd
 size 4943178720

model-00002-of-00003.safetensors

@@ -1,3 +1,3 @@
 version https://git-lfs.github.com/spec/v1
-oid sha256:57087d3701236ddd033204862662fca1d328a1d40449da21fe52584fb50fefff
+oid sha256:65d756c6023aac10ba06a637406ed4d2e2f4752c360d7d38f9d57ab38052466e
 size 4999819336

model-00003-of-00003.safetensors

@@ -1,3 +1,3 @@
 version https://git-lfs.github.com/spec/v1
-oid sha256:d22d542aa9e2e706d1339613cd7b76b3452024d5fd48f29b7543bbc9e61fc003
+oid sha256:a4dee0a340ae6ecce1b95adc398b7196eb2db45a4ad1ef0dc042944445a81bb6
 size 4540532728

runs/Jun17_01-59-50_n136-129-074/events.out.tfevents.1718560924.n136-129-074.1300229.0

@@ -1,3 +1,3 @@
 version https://git-lfs.github.com/spec/v1
-oid sha256:aa590a072804e4e12986ff8579b5512a4d087ed195a8a7fdab8596761c3285b4
-size 26092
+oid sha256:3d38c6a12ddad6fd5a37a8c85d17efcf257f35b1cf09dbb5272269fa21ba4220
+size 26446

train_results.json

@@ -1,8 +1,8 @@
 {
-    "epoch": 1.0,
-    "train_loss": 0.7162073644002279,
-    "train_runtime": 4512.4874,
+    "epoch": 2.0,
+    "train_loss": 0.6411590957641602,
+    "train_runtime": 8942.072,
     "train_samples": 38445,
-    "train_samples_per_second": 8.52,
-    "train_steps_per_second": 0.033
+    "train_samples_per_second": 8.599,
+    "train_steps_per_second": 0.034
 }

trainer_state.json

@@ -1,16 +1,16 @@
 {
   "best_metric": null,
   "best_model_checkpoint": null,
-  "epoch": 0.9983361064891847,
+  "epoch": 1.9966722129783694,
   "eval_steps": 500,
-  "global_step": 150,
+  "global_step": 300,
   "is_hyper_param_search": false,
   "is_local_process_zero": true,
   "is_world_process_zero": true,
   "log_history": [
     {
       "epoch": 0.01,
-      "grad_norm": 1151.6129333090275,
+      "grad_norm": 1151.6184415193795,
       "learning_rate": 3.333333333333333e-10,
       "logits/chosen": -4.106247425079346,
       "logits/rejected": -4.200438499450684,
@@ -25,243 +25,468 @@
     },
     {
       "epoch": 0.07,
-      "grad_norm": 1175.0279383615512,
+      "grad_norm": 1153.8530059959064,
       "learning_rate": 3.3333333333333334e-09,
-      "logits/chosen": -4.217203617095947,
-      "logits/rejected": -4.32081413269043,
-      "logps/chosen": -334.6714172363281,
-      "logps/rejected": -313.4322509765625,
-      "loss": 0.7317,
-      "rewards/accuracies": 0.3923611044883728,
-      "rewards/chosen": -0.044815655797719955,
-      "rewards/margins": -0.03612741455435753,
-      "rewards/rejected": -0.008688241243362427,
+      "logits/chosen": -4.217168807983398,
+      "logits/rejected": -4.321505069732666,
+      "logps/chosen": -334.6739501953125,
+      "logps/rejected": -313.41986083984375,
+      "loss": 0.7234,
+      "rewards/accuracies": 0.3784722089767456,
+      "rewards/chosen": -0.04607396200299263,
+      "rewards/margins": -0.04357295483350754,
+      "rewards/rejected": -0.0025010076351463795,
       "step": 10
     },
     {
       "epoch": 0.13,
-      "grad_norm": 1253.9538230101048,
-      "learning_rate": 4.983095894354858e-09,
-      "logits/chosen": -4.266427040100098,
-      "logits/rejected": -4.4187798500061035,
-      "logps/chosen": -313.9143371582031,
-      "logps/rejected": -288.782470703125,
-      "loss": 0.732,
-      "rewards/accuracies": 0.4781250059604645,
-      "rewards/chosen": -0.000932177877984941,
-      "rewards/margins": -0.012739461846649647,
-      "rewards/rejected": 0.011807283386588097,
+      "grad_norm": 1270.1091311718026,
+      "learning_rate": 6.666666666666667e-09,
+      "logits/chosen": -4.26615571975708,
+      "logits/rejected": -4.41886043548584,
+      "logps/chosen": -313.93829345703125,
+      "logps/rejected": -288.78863525390625,
+      "loss": 0.7299,
+      "rewards/accuracies": 0.5062500238418579,
+      "rewards/chosen": -0.012899210676550865,
+      "rewards/margins": -0.021626513451337814,
+      "rewards/rejected": 0.00872730277478695,
       "step": 20
     },
     {
       "epoch": 0.2,
-      "grad_norm": 1147.308654200848,
-      "learning_rate": 4.849231551964771e-09,
-      "logits/chosen": -4.301184177398682,
-      "logits/rejected": -4.36545991897583,
-      "logps/chosen": -308.3140563964844,
-      "logps/rejected": -285.6829833984375,
-      "loss": 0.7352,
-      "rewards/accuracies": 0.46562498807907104,
-      "rewards/chosen": -0.018803134560585022,
-      "rewards/margins": 0.008298242464661598,
-      "rewards/rejected": -0.02710137702524662,
+      "grad_norm": 1195.4358872902142,
+      "learning_rate": 1e-08,
+      "logits/chosen": -4.3016462326049805,
+      "logits/rejected": -4.365716457366943,
+      "logps/chosen": -308.2979431152344,
+      "logps/rejected": -285.63018798828125,
+      "loss": 0.7311,
+      "rewards/accuracies": 0.49687498807907104,
+      "rewards/chosen": -0.010719490237534046,
+      "rewards/margins": -0.010031750425696373,
+      "rewards/rejected": -0.000687739229761064,
       "step": 30
     },
     {
       "epoch": 0.27,
-      "grad_norm": 1152.01480268075,
-      "learning_rate": 4.588719528532341e-09,
-      "logits/chosen": -4.186487674713135,
-      "logits/rejected": -4.270204544067383,
-      "logps/chosen": -332.467041015625,
-      "logps/rejected": -307.530517578125,
-      "loss": 0.722,
-      "rewards/accuracies": 0.5531250238418579,
-      "rewards/chosen": 0.012754167430102825,
-      "rewards/margins": 0.04999501258134842,
-      "rewards/rejected": -0.03724084421992302,
+      "grad_norm": 1144.0076561795222,
+      "learning_rate": 9.966191788709716e-09,
+      "logits/chosen": -4.187338829040527,
+      "logits/rejected": -4.271176338195801,
+      "logps/chosen": -332.39453125,
+      "logps/rejected": -307.4621276855469,
+      "loss": 0.7199,
+      "rewards/accuracies": 0.5375000238418579,
+      "rewards/chosen": 0.04900529980659485,
+      "rewards/margins": 0.05206792429089546,
+      "rewards/rejected": -0.0030626237858086824,
       "step": 40
     },
     {
       "epoch": 0.33,
-      "grad_norm": 1161.340877574017,
-      "learning_rate": 4.215604094671834e-09,
-      "logits/chosen": -4.197475910186768,
-      "logits/rejected": -4.366654396057129,
-      "logps/chosen": -333.69940185546875,
-      "logps/rejected": -309.05511474609375,
-      "loss": 0.7263,
-      "rewards/accuracies": 0.512499988079071,
-      "rewards/chosen": 0.022939234972000122,
-      "rewards/margins": -0.01653190515935421,
-      "rewards/rejected": 0.03947114199399948,
+      "grad_norm": 1135.1845096891816,
+      "learning_rate": 9.86522435289912e-09,
+      "logits/chosen": -4.197329044342041,
+      "logits/rejected": -4.366620063781738,
+      "logps/chosen": -333.64678955078125,
+      "logps/rejected": -309.0525817871094,
+      "loss": 0.7189,
+      "rewards/accuracies": 0.515625,
+      "rewards/chosen": 0.04922889173030853,
+      "rewards/margins": 0.008506924845278263,
+      "rewards/rejected": 0.040721967816352844,
       "step": 50
     },
     {
       "epoch": 0.4,
-      "grad_norm": 1135.5326964714952,
-      "learning_rate": 3.7500000000000005e-09,
-      "logits/chosen": -4.186155796051025,
-      "logits/rejected": -4.269167900085449,
-      "logps/chosen": -323.9843444824219,
-      "logps/rejected": -308.364990234375,
-      "loss": 0.7133,
-      "rewards/accuracies": 0.484375,
-      "rewards/chosen": 0.009113344363868237,
-      "rewards/margins": -0.005971288774162531,
-      "rewards/rejected": 0.015084633603692055,
+      "grad_norm": 1124.4015217482226,
+      "learning_rate": 9.698463103929542e-09,
+      "logits/chosen": -4.186924457550049,
+      "logits/rejected": -4.269418239593506,
+      "logps/chosen": -323.7723693847656,
+      "logps/rejected": -308.26971435546875,
+      "loss": 0.7082,
+      "rewards/accuracies": 0.5249999761581421,
+      "rewards/chosen": 0.11511299759149551,
+      "rewards/margins": 0.052407026290893555,
+      "rewards/rejected": 0.06270597130060196,
       "step": 60
     },
     {
       "epoch": 0.47,
-      "grad_norm": 1170.598014061488,
-      "learning_rate": 3.2170080817777257e-09,
-      "logits/chosen": -4.14601469039917,
-      "logits/rejected": -4.301178932189941,
-      "logps/chosen": -344.8753967285156,
-      "logps/rejected": -314.6096496582031,
-      "loss": 0.7182,
-      "rewards/accuracies": 0.581250011920929,
-      "rewards/chosen": 0.08767497539520264,
-      "rewards/margins": 0.08200599253177643,
-      "rewards/rejected": 0.005668987520039082,
+      "grad_norm": 1133.1247984062034,
+      "learning_rate": 9.468163201617062e-09,
+      "logits/chosen": -4.147147178649902,
+      "logits/rejected": -4.3019304275512695,
+      "logps/chosen": -344.58563232421875,
+      "logps/rejected": -314.4212951660156,
+      "loss": 0.7005,
+      "rewards/accuracies": 0.590624988079071,
+      "rewards/chosen": 0.23254117369651794,
+      "rewards/margins": 0.13268980383872986,
+      "rewards/rejected": 0.09985135495662689,
       "step": 70
     },
     {
       "epoch": 0.53,
-      "grad_norm": 1088.822875661876,
-      "learning_rate": 2.6453620722761897e-09,
-      "logits/chosen": -4.220850944519043,
-      "logits/rejected": -4.3867106437683105,
-      "logps/chosen": -334.7141418457031,
-      "logps/rejected": -305.6958923339844,
-      "loss": 0.7175,
-      "rewards/accuracies": 0.515625,
-      "rewards/chosen": 0.06255482137203217,
-      "rewards/margins": 0.04721928387880325,
-      "rewards/rejected": 0.015335534699261189,
+      "grad_norm": 1057.1503651106625,
+      "learning_rate": 9.177439057064682e-09,
+      "logits/chosen": -4.223555564880371,
+      "logits/rejected": -4.389444828033447,
+      "logps/chosen": -334.3599853515625,
+      "logps/rejected": -305.4422607421875,
+      "loss": 0.6844,
+      "rewards/accuracies": 0.59375,
+      "rewards/chosen": 0.23962649703025818,
+      "rewards/margins": 0.09747296571731567,
+      "rewards/rejected": 0.1421535313129425,
       "step": 80
     },
     {
       "epoch": 0.6,
-      "grad_norm": 1120.1472438016,
-      "learning_rate": 2.0658795558326744e-09,
-      "logits/chosen": -4.226521968841553,
-      "logits/rejected": -4.347161293029785,
-      "logps/chosen": -327.5687561035156,
-      "logps/rejected": -308.19744873046875,
-      "loss": 0.7128,
-      "rewards/accuracies": 0.518750011920929,
-      "rewards/chosen": 0.03935312479734421,
-      "rewards/margins": 0.012340927496552467,
-      "rewards/rejected": 0.02701219543814659,
+      "grad_norm": 1055.57995971272,
+      "learning_rate": 8.830222215594889e-09,
+      "logits/chosen": -4.2292633056640625,
+      "logits/rejected": -4.349828243255615,
+      "logps/chosen": -327.07330322265625,
+      "logps/rejected": -307.81707763671875,
+      "loss": 0.6772,
+      "rewards/accuracies": 0.5562499761581421,
+      "rewards/chosen": 0.2870681881904602,
+      "rewards/margins": 0.06986425817012787,
+      "rewards/rejected": 0.21720390021800995,
       "step": 90
     },
     {
       "epoch": 0.67,
-      "grad_norm": 1147.49194388717,
-      "learning_rate": 1.5098005849021078e-09,
-      "logits/chosen": -4.208071708679199,
-      "logits/rejected": -4.417771339416504,
-      "logps/chosen": -320.9470520019531,
-      "logps/rejected": -290.018310546875,
-      "loss": 0.712,
-      "rewards/accuracies": 0.5218750238418579,
-      "rewards/chosen": 0.060337893664836884,
-      "rewards/margins": 0.03545590117573738,
-      "rewards/rejected": 0.0248819962143898,
+      "grad_norm": 1124.231150209405,
+      "learning_rate": 8.431208189343668e-09,
+      "logits/chosen": -4.214221000671387,
+      "logits/rejected": -4.424824237823486,
+      "logps/chosen": -320.45928955078125,
+      "logps/rejected": -289.71466064453125,
+      "loss": 0.6738,
+      "rewards/accuracies": 0.581250011920929,
+      "rewards/chosen": 0.3042110502719879,
+      "rewards/margins": 0.12749677896499634,
+      "rewards/rejected": 0.1767142415046692,
       "step": 100
     },
     {
       "epoch": 0.73,
-      "grad_norm": 1131.801124941311,
-      "learning_rate": 1.0071035207430352e-09,
-      "logits/chosen": -4.222798824310303,
-      "logits/rejected": -4.275721549987793,
-      "logps/chosen": -315.00030517578125,
-      "logps/rejected": -303.7385559082031,
-      "loss": 0.7022,
-      "rewards/accuracies": 0.543749988079071,
-      "rewards/chosen": 0.08098876476287842,
-      "rewards/margins": 0.037886131554841995,
-      "rewards/rejected": 0.04310264065861702,
+      "grad_norm": 1045.9935050945592,
+      "learning_rate": 7.98579295851393e-09,
+      "logits/chosen": -4.233684062957764,
+      "logits/rejected": -4.286379814147949,
+      "logps/chosen": -314.39251708984375,
+      "logps/rejected": -303.2556457519531,
+      "loss": 0.6623,
+      "rewards/accuracies": 0.5687500238418579,
+      "rewards/chosen": 0.3848657011985779,
+      "rewards/margins": 0.10032075643539429,
+      "rewards/rejected": 0.2845449447631836,
       "step": 110
     },
     {
       "epoch": 0.8,
-      "grad_norm": 1201.8232728734756,
-      "learning_rate": 5.848888922025553e-10,
-      "logits/chosen": -4.147335052490234,
-      "logits/rejected": -4.3030242919921875,
-      "logps/chosen": -339.2781677246094,
-      "logps/rejected": -313.5749206542969,
-      "loss": 0.7129,
-      "rewards/accuracies": 0.543749988079071,
-      "rewards/chosen": 0.08869560062885284,
-      "rewards/margins": 0.048744406551122665,
-      "rewards/rejected": 0.03995119035243988,
+      "grad_norm": 1107.1293654591734,
+      "learning_rate": 7.500000000000001e-09,
+      "logits/chosen": -4.151052951812744,
+      "logits/rejected": -4.307383060455322,
+      "logps/chosen": -338.46759033203125,
+      "logps/rejected": -313.0832214355469,
+      "loss": 0.6634,
+      "rewards/accuracies": 0.625,
+      "rewards/chosen": 0.49397316575050354,
+      "rewards/margins": 0.20816774666309357,
+      "rewards/rejected": 0.2858053743839264,
       "step": 120
     },
     {
       "epoch": 0.87,
-      "grad_norm": 1182.4206440855232,
-      "learning_rate": 2.659183991914696e-10,
-      "logits/chosen": -4.173482418060303,
-      "logits/rejected": -4.367284297943115,
-      "logps/chosen": -324.4493713378906,
-      "logps/rejected": -305.2012634277344,
-      "loss": 0.7,
-      "rewards/accuracies": 0.5062500238418579,
-      "rewards/chosen": 0.08546491712331772,
-      "rewards/margins": 0.01908993348479271,
-      "rewards/rejected": 0.06637498736381531,
+      "grad_norm": 1113.1665283582252,
+      "learning_rate": 6.980398830195784e-09,
+      "logits/chosen": -4.174288272857666,
+      "logits/rejected": -4.36756706237793,
+      "logps/chosen": -323.51800537109375,
+      "logps/rejected": -304.7021484375,
+      "loss": 0.6392,
+      "rewards/accuracies": 0.637499988079071,
+      "rewards/chosen": 0.5511430501937866,
+      "rewards/margins": 0.23519937694072723,
+      "rewards/rejected": 0.31594371795654297,
       "step": 130
     },
     {
       "epoch": 0.93,
-      "grad_norm": 1158.9214495930962,
-      "learning_rate": 6.738782355044049e-11,
-      "logits/chosen": -4.290203094482422,
-      "logits/rejected": -4.35637092590332,
-      "logps/chosen": -308.0101318359375,
-      "logps/rejected": -297.37701416015625,
-      "loss": 0.7082,
-      "rewards/accuracies": 0.5406249761581421,
-      "rewards/chosen": 0.12010886520147324,
-      "rewards/margins": 0.05074785277247429,
-      "rewards/rejected": 0.06936100870370865,
+      "grad_norm": 1022.5201031232197,
+      "learning_rate": 6.434016163555451e-09,
+      "logits/chosen": -4.292388916015625,
+      "logits/rejected": -4.358359336853027,
+      "logps/chosen": -307.1917419433594,
+      "logps/rejected": -296.78192138671875,
+      "loss": 0.6402,
+      "rewards/accuracies": 0.6000000238418579,
+      "rewards/chosen": 0.5292906761169434,
+      "rewards/margins": 0.16237936913967133,
+      "rewards/rejected": 0.36691129207611084,
       "step": 140
     },
     {
       "epoch": 1.0,
-      "grad_norm": 1208.2615961890754,
-      "learning_rate": 0.0,
-      "logits/chosen": -4.257304668426514,
-      "logits/rejected": -4.3310017585754395,
-      "logps/chosen": -312.7613830566406,
-      "logps/rejected": -300.3546447753906,
-      "loss": 0.7025,
-      "rewards/accuracies": 0.5562499761581421,
-      "rewards/chosen": 0.10821112245321274,
-      "rewards/margins": 0.047078292816877365,
-      "rewards/rejected": 0.06113281846046448,
+      "grad_norm": 1051.3521300860295,
+      "learning_rate": 5.868240888334653e-09,
+      "logits/chosen": -4.260380268096924,
+      "logits/rejected": -4.333888053894043,
+      "logps/chosen": -311.844482421875,
+      "logps/rejected": -299.74261474609375,
+      "loss": 0.6387,
+      "rewards/accuracies": 0.6187499761581421,
+      "rewards/chosen": 0.5666605830192566,
+      "rewards/margins": 0.1995116025209427,
+      "rewards/rejected": 0.3671489655971527,
       "step": 150
     },
     {
-      "epoch": 1.0,
-      "step": 150,
+      "epoch": 1.06,
+      "grad_norm": 1029.1548288848146,
+      "learning_rate": 5.290724144552379e-09,
+      "logits/chosen": -4.191515922546387,
+      "logits/rejected": -4.348960876464844,
+      "logps/chosen": -326.90582275390625,
+      "logps/rejected": -305.37689208984375,
+      "loss": 0.6032,
+      "rewards/accuracies": 0.6937500238418579,
+      "rewards/chosen": 0.6744669079780579,
+      "rewards/margins": 0.37094420194625854,
+      "rewards/rejected": 0.3035227358341217,
+      "step": 160
+    },
+    {
+      "epoch": 1.13,
+      "grad_norm": 991.2604083787021,
+      "learning_rate": 4.709275855447621e-09,
+      "logits/chosen": -4.241927623748779,
+      "logits/rejected": -4.339847564697266,
+      "logps/chosen": -314.4874572753906,
+      "logps/rejected": -295.49151611328125,
+      "loss": 0.6151,
+      "rewards/accuracies": 0.6937500238418579,
+      "rewards/chosen": 0.6504887938499451,
+      "rewards/margins": 0.3018999397754669,
+      "rewards/rejected": 0.3485889136791229,
+      "step": 170
+    },
+    {
+      "epoch": 1.2,
+      "grad_norm": 976.0494957890444,
+      "learning_rate": 4.131759111665349e-09,
+      "logits/chosen": -4.174123287200928,
+      "logits/rejected": -4.313396453857422,
+      "logps/chosen": -325.78045654296875,
+      "logps/rejected": -304.82989501953125,
+      "loss": 0.6147,
+      "rewards/accuracies": 0.6656249761581421,
+      "rewards/chosen": 0.6252425909042358,
+      "rewards/margins": 0.27976280450820923,
+      "rewards/rejected": 0.3454797863960266,
+      "step": 180
+    },
+    {
+      "epoch": 1.26,
+      "grad_norm": 949.77837430784,
+      "learning_rate": 3.56598383644455e-09,
+      "logits/chosen": -4.231461524963379,
+      "logits/rejected": -4.367356777191162,
+      "logps/chosen": -326.8230895996094,
+      "logps/rejected": -299.7027587890625,
+      "loss": 0.5969,
+      "rewards/accuracies": 0.6968749761581421,
+      "rewards/chosen": 0.7501948475837708,
+      "rewards/margins": 0.35921385884284973,
+      "rewards/rejected": 0.39098095893859863,
+      "step": 190
+    },
+    {
+      "epoch": 1.33,
+      "grad_norm": 963.9694294192556,
+      "learning_rate": 3.0196011698042157e-09,
+      "logits/chosen": -4.249307155609131,
+      "logits/rejected": -4.317980766296387,
+      "logps/chosen": -313.9615783691406,
+      "logps/rejected": -293.1615905761719,
+      "loss": 0.5922,
+      "rewards/accuracies": 0.6781250238418579,
+      "rewards/chosen": 0.7425140142440796,
+      "rewards/margins": 0.3114904761314392,
+      "rewards/rejected": 0.4310235381126404,
+      "step": 200
+    },
+    {
+      "epoch": 1.4,
+      "grad_norm": 1038.4899227109647,
+      "learning_rate": 2.5000000000000013e-09,
+      "logits/chosen": -4.174517631530762,
+      "logits/rejected": -4.288783073425293,
+      "logps/chosen": -327.461181640625,
+      "logps/rejected": -304.9690246582031,
+      "loss": 0.5952,
+      "rewards/accuracies": 0.699999988079071,
+      "rewards/chosen": 0.7787834405899048,
+      "rewards/margins": 0.3926675319671631,
+      "rewards/rejected": 0.3861159086227417,
+      "step": 210
+    },
+    {
+      "epoch": 1.46,
+      "grad_norm": 992.6867793357102,
+      "learning_rate": 2.0142070414860704e-09,
+      "logits/chosen": -4.165514945983887,
+      "logits/rejected": -4.316833019256592,
+      "logps/chosen": -358.1416931152344,
+      "logps/rejected": -323.22998046875,
+      "loss": 0.5994,
+      "rewards/accuracies": 0.6781250238418579,
+      "rewards/chosen": 0.8851108551025391,
+      "rewards/margins": 0.35754603147506714,
+      "rewards/rejected": 0.5275647640228271,
+      "step": 220
+    },
+    {
+      "epoch": 1.53,
+      "grad_norm": 918.0076197376386,
+      "learning_rate": 1.5687918106563326e-09,
+      "logits/chosen": -4.234222412109375,
+      "logits/rejected": -4.369565486907959,
+      "logps/chosen": -328.3675537109375,
+      "logps/rejected": -303.726806640625,
+      "loss": 0.5919,
+      "rewards/accuracies": 0.65625,
+      "rewards/chosen": 0.8034540414810181,
+      "rewards/margins": 0.2974655032157898,
+      "rewards/rejected": 0.5059884786605835,
+      "step": 230
+    },
+    {
+      "epoch": 1.6,
+      "grad_norm": 944.3107996293463,
+      "learning_rate": 1.1697777844051105e-09,
+      "logits/chosen": -4.202380180358887,
+      "logits/rejected": -4.353602409362793,
+      "logps/chosen": -326.75604248046875,
+      "logps/rejected": -297.03619384765625,
+      "loss": 0.588,
+      "rewards/accuracies": 0.659375011920929,
+      "rewards/chosen": 0.8500032424926758,
+      "rewards/margins": 0.3546527922153473,
+      "rewards/rejected": 0.4953504502773285,
+      "step": 240
+    },
+    {
+      "epoch": 1.66,
+      "grad_norm": 944.2690060176395,
+      "learning_rate": 8.225609429353187e-10,
+      "logits/chosen": -4.173297882080078,
+      "logits/rejected": -4.368635654449463,
+      "logps/chosen": -321.2724914550781,
+      "logps/rejected": -293.50311279296875,
+      "loss": 0.5856,
+      "rewards/accuracies": 0.731249988079071,
+      "rewards/chosen": 0.8207529783248901,
+      "rewards/margins": 0.40889453887939453,
+      "rewards/rejected": 0.4118584990501404,
+      "step": 250
+    },
+    {
+      "epoch": 1.73,
+      "grad_norm": 957.8767493880096,
+      "learning_rate": 5.318367983829391e-10,
+      "logits/chosen": -4.18589448928833,
+      "logits/rejected": -4.284361839294434,
+      "logps/chosen": -348.6470031738281,
+      "logps/rejected": -326.4360046386719,
+      "loss": 0.5895,
+      "rewards/accuracies": 0.690625011920929,
+      "rewards/chosen": 0.8914071917533875,
+      "rewards/margins": 0.37524908781051636,
+      "rewards/rejected": 0.5161582231521606,
+      "step": 260
+    },
+    {
+      "epoch": 1.8,
+      "grad_norm": 949.249423709826,
+      "learning_rate": 3.015368960704584e-10,
+      "logits/chosen": -4.271857738494873,
+      "logits/rejected": -4.4089555740356445,
+      "logps/chosen": -325.19049072265625,
+      "logps/rejected": -302.7436828613281,
+      "loss": 0.5938,
+      "rewards/accuracies": 0.6781250238418579,
+      "rewards/chosen": 0.9044697880744934,
+      "rewards/margins": 0.38229116797447205,
+      "rewards/rejected": 0.522178590297699,
+      "step": 270
+    },
+    {
+      "epoch": 1.86,
+      "grad_norm": 932.1311904270524,
+      "learning_rate": 1.3477564710088098e-10,
+      "logits/chosen": -4.163296222686768,
+      "logits/rejected": -4.306557655334473,
+      "logps/chosen": -327.57257080078125,
+      "logps/rejected": -306.2264099121094,
+      "loss": 0.5865,
+      "rewards/accuracies": 0.715624988079071,
+      "rewards/chosen": 0.8713304400444031,
+      "rewards/margins": 0.4181024134159088,
+      "rewards/rejected": 0.45322805643081665,
+      "step": 280
+    },
+    {
+      "epoch": 1.93,
+      "grad_norm": 884.8124763448137,
+      "learning_rate": 3.380821129028488e-11,
+      "logits/chosen": -4.1040120124816895,
+      "logits/rejected": -4.260913372039795,
+      "logps/chosen": -334.97686767578125,
+      "logps/rejected": -312.01617431640625,
+      "loss": 0.5924,
+      "rewards/accuracies": 0.6812499761581421,
+      "rewards/chosen": 0.9016163945198059,
+      "rewards/margins": 0.4117993414402008,
+      "rewards/rejected": 0.48981720209121704,
+      "step": 290
+    },
+    {
+      "epoch": 2.0,
+      "grad_norm": 959.7419745329076,
+      "learning_rate": 0.0,
+      "logits/chosen": -4.31270170211792,
+      "logits/rejected": -4.468858242034912,
+      "logps/chosen": -314.4400939941406,
+      "logps/rejected": -284.8276672363281,
+      "loss": 0.5825,
+      "rewards/accuracies": 0.6875,
+      "rewards/chosen": 0.82757169008255,
+      "rewards/margins": 0.36657077074050903,
+      "rewards/rejected": 0.4610009789466858,
+      "step": 300
+    },
+    {
+      "epoch": 2.0,
+      "step": 300,
       "total_flos": 0.0,
-      "train_loss": 0.7162073644002279,
-      "train_runtime": 4512.4874,
-      "train_samples_per_second": 8.52,
-      "train_steps_per_second": 0.033
+      "train_loss": 0.6411590957641602,
+      "train_runtime": 8942.072,
+      "train_samples_per_second": 8.599,
+      "train_steps_per_second": 0.034
     }
   ],
   "logging_steps": 10,
-  "max_steps": 150,
+  "max_steps": 300,
   "num_input_tokens_seen": 0,
-  "num_train_epochs": 1,
+  "num_train_epochs": 2,
   "save_steps": 100,
   "total_flos": 0.0,
   "train_batch_size": 8,