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 the None dataset.
+This model was trained from scratch on an unknown dataset.
 
 ## Model description
 
@@ -52,7 +52,7 @@ The following hyperparameters were used during training:
 
 ### Framework versions
 
-- Transformers 4.41.1
+- Transformers 4.39.3
 - Pytorch 2.1.2+cu118
-- Datasets 2.16.1
-- Tokenizers 0.19.1
+- Datasets 2.19.1
+- Tokenizers 0.15.2

all_results.json

@@ -1,9 +1,8 @@
 {
-    "epoch": 0.9980806142034548,
-    "total_flos": 0.0,
-    "train_loss": 0.3119094243416419,
-    "train_runtime": 6443.086,
+    "epoch": 1.0,
+    "train_loss": 0.22171047773116676,
+    "train_runtime": 5890.091,
     "train_samples": 50000,
-    "train_samples_per_second": 7.76,
-    "train_steps_per_second": 0.061
+    "train_samples_per_second": 8.489,
+    "train_steps_per_second": 0.066
 }

generation_config.json

@@ -2,5 +2,5 @@
   "_from_model_config": true,
   "bos_token_id": 1,
   "eos_token_id": 2,
-  "transformers_version": "4.41.1"
+  "transformers_version": "4.39.3"
 }

model-00001-of-00003.safetensors

@@ -1,3 +1,3 @@
 version https://git-lfs.github.com/spec/v1
-oid sha256:5bc88b0288a40faf21a1e24a96384ccf683d38e8246d3b0cab22d31c505d8b9f
+oid sha256:d0abc7bcb57ebd4c85bca100446495fa1f35516ea4513f630bb3534e01a3e139
 size 4943162336

model-00002-of-00003.safetensors

@@ -1,3 +1,3 @@
 version https://git-lfs.github.com/spec/v1
-oid sha256:c430b926ebd01f65609ab319292fa842772b87fc92486924b50fdd7065be6155
+oid sha256:9901d8efdf08230c45749f927c9bcca0fc53e1bb330c70a7989c85951effcb3e
 size 4999819336

model-00003-of-00003.safetensors

@@ -1,3 +1,3 @@
 version https://git-lfs.github.com/spec/v1
-oid sha256:4d5049e6472b9a1956a056ed3a8a66408f55efa3d2ea2bd03073317eac4eec19
+oid sha256:07758d9f276a4287d05416d1f89f26070bd0cd4f6fe426596e860df642b6c6e8
 size 4540516344

runs/Jun22_09-40-11_n136-112-146/events.out.tfevents.1719021114.n136-112-146.3387325.0

@@ -1,3 +1,3 @@
 version https://git-lfs.github.com/spec/v1
-oid sha256:ae2e602011d96a1d261b995a1e371de2d2d68ae27891840cc7c7ae421436694c
-size 26098
+oid sha256:0f47c1118994624212cf7c63f077ee6e2ea3cd246ed37d29b5d4571eca132c8b
+size 32644

train_results.json

@@ -1,9 +1,8 @@
 {
-    "epoch": 0.9980806142034548,
-    "total_flos": 0.0,
-    "train_loss": 0.3119094243416419,
-    "train_runtime": 6443.086,
+    "epoch": 1.0,
+    "train_loss": 0.22171047773116676,
+    "train_runtime": 5890.091,
     "train_samples": 50000,
-    "train_samples_per_second": 7.76,
-    "train_steps_per_second": 0.061
+    "train_samples_per_second": 8.489,
+    "train_steps_per_second": 0.066
 }

trainer_state.json

@@ -9,13 +9,13 @@
   "is_world_process_zero": true,
   "log_history": [
     {
-      "epoch": 0.0025591810620601407,
-      "grad_norm": 709.5459154500938,
+      "epoch": 0.0,
+      "grad_norm": 746.9239950470109,
       "learning_rate": 1.0256410256410256e-08,
-      "logits/chosen": -2.5583817958831787,
-      "logits/rejected": -2.4487552642822266,
+      "logits/chosen": -2.5617921352386475,
+      "logits/rejected": -2.415619373321533,
       "logps/chosen": -258.1644592285156,
-      "logps/rejected": -216.25729370117188,
+      "logps/rejected": -191.65736389160156,
       "loss": 0.6931,
       "rewards/accuracies": 0.0,
       "rewards/chosen": 0.0,
@@ -24,598 +24,598 @@
       "step": 1
     },
     {
-      "epoch": 0.025591810620601407,
-      "grad_norm": 684.196668594934,
+      "epoch": 0.03,
+      "grad_norm": 712.2130131356953,
       "learning_rate": 1.0256410256410255e-07,
-      "logits/chosen": -2.605860471725464,
-      "logits/rejected": -2.5528433322906494,
-      "logps/chosen": -267.56048583984375,
-      "logps/rejected": -217.59442138671875,
-      "loss": 0.7027,
-      "rewards/accuracies": 0.4166666567325592,
-      "rewards/chosen": 0.0020057493820786476,
-      "rewards/margins": -0.020002255216240883,
-      "rewards/rejected": 0.022008005529642105,
+      "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,
       "step": 10
     },
     {
-      "epoch": 0.05118362124120281,
-      "grad_norm": 515.3907221989738,
+      "epoch": 0.05,
+      "grad_norm": 343.2600415646243,
       "learning_rate": 2.051282051282051e-07,
-      "logits/chosen": -2.6285552978515625,
-      "logits/rejected": -2.566141128540039,
-      "logps/chosen": -260.791259765625,
-      "logps/rejected": -207.073486328125,
-      "loss": 0.5875,
-      "rewards/accuracies": 0.71875,
-      "rewards/chosen": 0.2564736604690552,
-      "rewards/margins": 0.2661496698856354,
-      "rewards/rejected": -0.00967598520219326,
+      "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,
       "step": 20
     },
     {
-      "epoch": 0.07677543186180422,
-      "grad_norm": 608.3322300561331,
+      "epoch": 0.08,
+      "grad_norm": 460.51807031992456,
       "learning_rate": 3.076923076923077e-07,
-      "logits/chosen": -2.6465039253234863,
-      "logits/rejected": -2.5724854469299316,
-      "logps/chosen": -251.90835571289062,
-      "logps/rejected": -198.9151611328125,
-      "loss": 0.3737,
-      "rewards/accuracies": 0.8500000238418579,
-      "rewards/chosen": 1.6194490194320679,
-      "rewards/margins": 1.6427723169326782,
-      "rewards/rejected": -0.02332335151731968,
+      "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,
       "step": 30
     },
     {
-      "epoch": 0.10236724248240563,
-      "grad_norm": 286.23056215212245,
+      "epoch": 0.1,
+      "grad_norm": 386.50821557830585,
       "learning_rate": 3.999919890759704e-07,
-      "logits/chosen": -2.650782823562622,
-      "logits/rejected": -2.571556329727173,
-      "logps/chosen": -244.2300567626953,
-      "logps/rejected": -193.5239715576172,
-      "loss": 0.3167,
-      "rewards/accuracies": 0.793749988079071,
-      "rewards/chosen": 3.0596303939819336,
-      "rewards/margins": 2.9784700870513916,
-      "rewards/rejected": 0.08116073161363602,
+      "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,
       "step": 40
     },
     {
-      "epoch": 0.12795905310300704,
-      "grad_norm": 355.8436814450708,
+      "epoch": 0.13,
+      "grad_norm": 288.39653383863197,
       "learning_rate": 3.9903145446619833e-07,
-      "logits/chosen": -2.6738078594207764,
-      "logits/rejected": -2.601217031478882,
-      "logps/chosen": -257.42144775390625,
-      "logps/rejected": -202.82069396972656,
-      "loss": 0.3214,
-      "rewards/accuracies": 0.875,
-      "rewards/chosen": 5.407862663269043,
-      "rewards/margins": 4.2437334060668945,
-      "rewards/rejected": 1.164129376411438,
+      "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,
       "step": 50
     },
     {
-      "epoch": 0.15355086372360843,
-      "grad_norm": 419.99201519842893,
+      "epoch": 0.15,
+      "grad_norm": 288.38654209122495,
       "learning_rate": 3.9647754748313294e-07,
-      "logits/chosen": -2.6383204460144043,
-      "logits/rejected": -2.5659520626068115,
-      "logps/chosen": -259.7900695800781,
-      "logps/rejected": -207.2904510498047,
-      "loss": 0.3176,
-      "rewards/accuracies": 0.8500000238418579,
-      "rewards/chosen": 4.518810272216797,
-      "rewards/margins": 4.13478422164917,
-      "rewards/rejected": 0.3840256929397583,
+      "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,
       "step": 60
     },
     {
-      "epoch": 0.17914267434420986,
-      "grad_norm": 305.786716914146,
+      "epoch": 0.18,
+      "grad_norm": 271.6769920625721,
       "learning_rate": 3.9235071376357917e-07,
-      "logits/chosen": -2.6047496795654297,
-      "logits/rejected": -2.5317835807800293,
-      "logps/chosen": -290.00616455078125,
-      "logps/rejected": -227.17697143554688,
-      "loss": 0.301,
-      "rewards/accuracies": 0.90625,
-      "rewards/chosen": 3.9522290229797363,
-      "rewards/margins": 4.945390701293945,
-      "rewards/rejected": -0.9931615591049194,
+      "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,
       "step": 70
     },
     {
-      "epoch": 0.20473448496481125,
-      "grad_norm": 512.2762241729247,
+      "epoch": 0.2,
+      "grad_norm": 381.16800166622966,
       "learning_rate": 3.8668399121531677e-07,
-      "logits/chosen": -2.552280902862549,
-      "logits/rejected": -2.476268768310547,
-      "logps/chosen": -260.35906982421875,
-      "logps/rejected": -206.0688018798828,
-      "loss": 0.2615,
-      "rewards/accuracies": 0.875,
-      "rewards/chosen": 3.5995476245880127,
-      "rewards/margins": 5.095798969268799,
-      "rewards/rejected": -1.4962517023086548,
+      "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,
       "step": 80
     },
     {
-      "epoch": 0.23032629558541268,
-      "grad_norm": 515.1909309248075,
+      "epoch": 0.23,
+      "grad_norm": 609.4338656498198,
       "learning_rate": 3.795227455278028e-07,
-      "logits/chosen": -2.5692882537841797,
-      "logits/rejected": -2.510436534881592,
-      "logps/chosen": -256.6405334472656,
-      "logps/rejected": -217.69192504882812,
-      "loss": 0.264,
-      "rewards/accuracies": 0.862500011920929,
-      "rewards/chosen": 2.5523829460144043,
-      "rewards/margins": 4.150609016418457,
-      "rewards/rejected": -1.5982261896133423,
+      "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,
       "step": 90
     },
     {
-      "epoch": 0.2559181062060141,
-      "grad_norm": 587.7494413414474,
+      "epoch": 0.26,
+      "grad_norm": 259.36924157384516,
       "learning_rate": 3.7092430699120403e-07,
-      "logits/chosen": -2.608997344970703,
-      "logits/rejected": -2.530529260635376,
-      "logps/chosen": -235.7121124267578,
-      "logps/rejected": -201.58700561523438,
-      "loss": 0.3994,
-      "rewards/accuracies": 0.84375,
-      "rewards/chosen": 3.2284069061279297,
-      "rewards/margins": 4.6865153312683105,
-      "rewards/rejected": -1.4581090211868286,
+      "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,
       "step": 100
     },
     {
-      "epoch": 0.28150991682661547,
-      "grad_norm": 383.9666636386822,
+      "epoch": 0.28,
+      "grad_norm": 344.20413939741326,
       "learning_rate": 3.6095751153125107e-07,
-      "logits/chosen": -2.605412483215332,
-      "logits/rejected": -2.5209498405456543,
-      "logps/chosen": -257.01318359375,
-      "logps/rejected": -210.5634002685547,
-      "loss": 0.3274,
-      "rewards/accuracies": 0.862500011920929,
-      "rewards/chosen": 4.330618381500244,
-      "rewards/margins": 5.40033483505249,
-      "rewards/rejected": -1.069716453552246,
+      "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,
       "step": 110
     },
     {
-      "epoch": 0.30710172744721687,
-      "grad_norm": 452.32780946070125,
+      "epoch": 0.31,
+      "grad_norm": 298.19415193137047,
       "learning_rate": 3.497021496342202e-07,
-      "logits/chosen": -2.5948054790496826,
-      "logits/rejected": -2.5218663215637207,
-      "logps/chosen": -261.63739013671875,
-      "logps/rejected": -209.2183074951172,
-      "loss": 0.3015,
-      "rewards/accuracies": 0.887499988079071,
-      "rewards/chosen": 3.912487030029297,
-      "rewards/margins": 5.338441371917725,
-      "rewards/rejected": -1.4259542226791382,
+      "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,
       "step": 120
     },
     {
-      "epoch": 0.3326935380678183,
-      "grad_norm": 325.92030560656247,
+      "epoch": 0.33,
+      "grad_norm": 143.10334660096893,
       "learning_rate": 3.372483275737467e-07,
-      "logits/chosen": -2.574514389038086,
-      "logits/rejected": -2.505225896835327,
-      "logps/chosen": -250.41543579101562,
-      "logps/rejected": -196.6962127685547,
-      "loss": 0.2826,
-      "rewards/accuracies": 0.8374999761581421,
-      "rewards/chosen": 3.443896532058716,
-      "rewards/margins": 4.925398826599121,
-      "rewards/rejected": -1.4815022945404053,
+      "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,
       "step": 130
     },
     {
-      "epoch": 0.3582853486884197,
-      "grad_norm": 286.3934639339848,
+      "epoch": 0.36,
+      "grad_norm": 396.95461669932075,
       "learning_rate": 3.2369574605325094e-07,
-      "logits/chosen": -2.598271369934082,
-      "logits/rejected": -2.5281975269317627,
-      "logps/chosen": -268.22943115234375,
-      "logps/rejected": -216.2914581298828,
-      "loss": 0.3028,
-      "rewards/accuracies": 0.856249988079071,
-      "rewards/chosen": 3.4734299182891846,
-      "rewards/margins": 4.84061336517334,
-      "rewards/rejected": -1.3671828508377075,
+      "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,
       "step": 140
     },
     {
-      "epoch": 0.3838771593090211,
-      "grad_norm": 326.69658537939597,
+      "epoch": 0.38,
+      "grad_norm": 248.0981758080623,
       "learning_rate": 3.091529020389009e-07,
-      "logits/chosen": -2.5794036388397217,
-      "logits/rejected": -2.5112969875335693,
-      "logps/chosen": -262.27850341796875,
-      "logps/rejected": -213.2576141357422,
-      "loss": 0.3047,
-      "rewards/accuracies": 0.887499988079071,
-      "rewards/chosen": 4.1566081047058105,
-      "rewards/margins": 5.587738990783691,
-      "rewards/rejected": -1.4311310052871704,
+      "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,
       "step": 150
     },
     {
-      "epoch": 0.4094689699296225,
-      "grad_norm": 334.0349111221283,
+      "epoch": 0.41,
+      "grad_norm": 531.5498022382194,
       "learning_rate": 2.937362201729407e-07,
-      "logits/chosen": -2.569304943084717,
-      "logits/rejected": -2.4969332218170166,
-      "logps/chosen": -241.8642120361328,
-      "logps/rejected": -196.44754028320312,
-      "loss": 0.2947,
-      "rewards/accuracies": 0.893750011920929,
-      "rewards/chosen": 3.53686785697937,
-      "rewards/margins": 5.24989128112793,
-      "rewards/rejected": -1.7130241394042969,
+      "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,
       "step": 160
     },
     {
-      "epoch": 0.4350607805502239,
-      "grad_norm": 385.1324260357516,
+      "epoch": 0.44,
+      "grad_norm": 204.85526693169226,
       "learning_rate": 2.77569120720971e-07,
-      "logits/chosen": -2.5677335262298584,
-      "logits/rejected": -2.499415874481201,
-      "logps/chosen": -263.1123046875,
-      "logps/rejected": -209.7564239501953,
-      "loss": 0.2546,
-      "rewards/accuracies": 0.875,
-      "rewards/chosen": 3.789729595184326,
-      "rewards/margins": 5.479173183441162,
-      "rewards/rejected": -1.6894439458847046,
+      "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,
       "step": 170
     },
     {
-      "epoch": 0.46065259117082535,
-      "grad_norm": 298.5381373642981,
+      "epoch": 0.46,
+      "grad_norm": 346.922389452496,
       "learning_rate": 2.6078103151484935e-07,
-      "logits/chosen": -2.632197856903076,
-      "logits/rejected": -2.5449578762054443,
-      "logps/chosen": -234.98526000976562,
-      "logps/rejected": -185.88897705078125,
-      "loss": 0.3076,
-      "rewards/accuracies": 0.8812500238418579,
-      "rewards/chosen": 3.5240702629089355,
-      "rewards/margins": 4.843334674835205,
-      "rewards/rejected": -1.3192641735076904,
+      "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,
       "step": 180
     },
     {
-      "epoch": 0.48624440179142675,
-      "grad_norm": 275.15385748139875,
+      "epoch": 0.49,
+      "grad_norm": 469.7545752914211,
       "learning_rate": 2.435063518012335e-07,
-      "logits/chosen": -2.592048168182373,
-      "logits/rejected": -2.5161328315734863,
-      "logps/chosen": -257.56207275390625,
-      "logps/rejected": -202.87477111816406,
-      "loss": 0.2628,
-      "rewards/accuracies": 0.8812500238418579,
-      "rewards/chosen": 3.6292481422424316,
-      "rewards/margins": 5.401675224304199,
-      "rewards/rejected": -1.7724273204803467,
+      "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,
       "step": 190
     },
     {
-      "epoch": 0.5118362124120281,
-      "grad_norm": 441.7914810102142,
+      "epoch": 0.51,
+      "grad_norm": 532.7452067278298,
       "learning_rate": 2.2588337629081105e-07,
-      "logits/chosen": -2.6288845539093018,
-      "logits/rejected": -2.539705753326416,
-      "logps/chosen": -272.2264709472656,
-      "logps/rejected": -217.74020385742188,
-      "loss": 0.2691,
-      "rewards/accuracies": 0.8812500238418579,
-      "rewards/chosen": 4.148322105407715,
-      "rewards/margins": 6.209266662597656,
-      "rewards/rejected": -2.0609450340270996,
+      "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,
       "step": 200
     },
     {
-      "epoch": 0.5374280230326296,
-      "grad_norm": 578.0076877712855,
+      "epoch": 0.54,
+      "grad_norm": 489.8003677533757,
       "learning_rate": 2.0805318802188303e-07,
-      "logits/chosen": -2.6119117736816406,
-      "logits/rejected": -2.5380859375,
-      "logps/chosen": -265.3163146972656,
-      "logps/rejected": -211.01132202148438,
-      "loss": 0.4145,
-      "rewards/accuracies": 0.824999988079071,
-      "rewards/chosen": 2.868762969970703,
-      "rewards/margins": 5.194035530090332,
-      "rewards/rejected": -2.32527232170105,
+      "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,
       "step": 210
     },
     {
-      "epoch": 0.5630198336532309,
-      "grad_norm": 435.59772463261663,
+      "epoch": 0.56,
+      "grad_norm": 398.2303862264868,
       "learning_rate": 1.9015852890162436e-07,
-      "logits/chosen": -2.628199577331543,
-      "logits/rejected": -2.5644900798797607,
-      "logps/chosen": -256.01708984375,
-      "logps/rejected": -213.6505126953125,
-      "loss": 0.302,
-      "rewards/accuracies": 0.8812500238418579,
-      "rewards/chosen": 2.8298840522766113,
-      "rewards/margins": 4.771709442138672,
-      "rewards/rejected": -1.9418258666992188,
+      "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,
       "step": 220
     },
     {
-      "epoch": 0.5886116442738324,
-      "grad_norm": 341.5067091186977,
+      "epoch": 0.59,
+      "grad_norm": 345.5121551134584,
       "learning_rate": 1.723426569670534e-07,
-      "logits/chosen": -2.660271644592285,
-      "logits/rejected": -2.5745387077331543,
-      "logps/chosen": -265.7137145996094,
-      "logps/rejected": -216.05184936523438,
-      "loss": 0.2709,
-      "rewards/accuracies": 0.862500011920929,
-      "rewards/chosen": 3.8446857929229736,
-      "rewards/margins": 5.413482666015625,
-      "rewards/rejected": -1.5687971115112305,
+      "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,
       "step": 230
     },
     {
-      "epoch": 0.6142034548944337,
-      "grad_norm": 518.1474749065682,
+      "epoch": 0.61,
+      "grad_norm": 366.69298678156184,
       "learning_rate": 1.547481995140556e-07,
-      "logits/chosen": -2.6460232734680176,
-      "logits/rejected": -2.5979976654052734,
-      "logps/chosen": -260.61566162109375,
-      "logps/rejected": -219.18740844726562,
-      "loss": 0.2732,
-      "rewards/accuracies": 0.8812500238418579,
-      "rewards/chosen": 3.5578017234802246,
-      "rewards/margins": 6.3658447265625,
-      "rewards/rejected": -2.808042287826538,
+      "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,
       "step": 240
     },
     {
-      "epoch": 0.6397952655150352,
-      "grad_norm": 488.31668996973804,
+      "epoch": 0.64,
+      "grad_norm": 355.57895419726833,
       "learning_rate": 1.375160112758885e-07,
-      "logits/chosen": -2.657061815261841,
-      "logits/rejected": -2.5865418910980225,
-      "logps/chosen": -265.78448486328125,
-      "logps/rejected": -221.42001342773438,
-      "loss": 0.2763,
-      "rewards/accuracies": 0.875,
-      "rewards/chosen": 2.9970998764038086,
-      "rewards/margins": 5.516369819641113,
-      "rewards/rejected": -2.519270420074463,
+      "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,
       "step": 250
     },
     {
-      "epoch": 0.6653870761356366,
-      "grad_norm": 266.72245551152963,
+      "epoch": 0.67,
+      "grad_norm": 284.33502896066835,
       "learning_rate": 1.2078404679216862e-07,
-      "logits/chosen": -2.648825168609619,
-      "logits/rejected": -2.589503526687622,
-      "logps/chosen": -260.22369384765625,
-      "logps/rejected": -208.0442352294922,
-      "loss": 0.2613,
-      "rewards/accuracies": 0.875,
-      "rewards/chosen": 2.818129301071167,
-      "rewards/margins": 5.237301826477051,
-      "rewards/rejected": -2.419172525405884,
+      "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,
       "step": 260
     },
     {
-      "epoch": 0.690978886756238,
-      "grad_norm": 410.4072035856202,
+      "epoch": 0.69,
+      "grad_norm": 281.9194802132945,
       "learning_rate": 1.0468625599573841e-07,
-      "logits/chosen": -2.6281862258911133,
-      "logits/rejected": -2.5414164066314697,
-      "logps/chosen": -269.1915283203125,
-      "logps/rejected": -228.3280792236328,
-      "loss": 0.3168,
-      "rewards/accuracies": 0.893750011920929,
-      "rewards/chosen": 3.6961987018585205,
-      "rewards/margins": 7.5373077392578125,
-      "rewards/rejected": -3.841109037399292,
+      "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,
       "step": 270
     },
     {
-      "epoch": 0.7165706973768394,
-      "grad_norm": 230.87089413037836,
+      "epoch": 0.72,
+      "grad_norm": 262.1937325706126,
       "learning_rate": 8.935151185893727e-08,
-      "logits/chosen": -2.653529644012451,
-      "logits/rejected": -2.5968880653381348,
-      "logps/chosen": -268.2784729003906,
-      "logps/rejected": -218.708984375,
-      "loss": 0.2585,
-      "rewards/accuracies": 0.893750011920929,
-      "rewards/chosen": 2.713934898376465,
-      "rewards/margins": 5.596070289611816,
-      "rewards/rejected": -2.8821353912353516,
+      "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,
       "step": 280
     },
     {
-      "epoch": 0.7421625079974408,
-      "grad_norm": 355.41024702928314,
+      "epoch": 0.74,
+      "grad_norm": 267.1170614305438,
       "learning_rate": 7.490257868414449e-08,
-      "logits/chosen": -2.613981246948242,
-      "logits/rejected": -2.5670533180236816,
-      "logps/chosen": -266.4203186035156,
-      "logps/rejected": -216.7047119140625,
-      "loss": 0.2561,
-      "rewards/accuracies": 0.875,
-      "rewards/chosen": 2.7628657817840576,
-      "rewards/margins": 5.556154727935791,
-      "rewards/rejected": -2.793288469314575,
+      "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,
       "step": 290
     },
     {
-      "epoch": 0.7677543186180422,
-      "grad_norm": 334.3998771127154,
+      "epoch": 0.77,
+      "grad_norm": 337.4409958446907,
       "learning_rate": 6.145512929808013e-08,
-      "logits/chosen": -2.6336982250213623,
-      "logits/rejected": -2.579585552215576,
-      "logps/chosen": -273.02044677734375,
-      "logps/rejected": -220.3231964111328,
-      "loss": 0.274,
-      "rewards/accuracies": 0.893750011920929,
-      "rewards/chosen": 2.7771525382995605,
-      "rewards/margins": 5.427972316741943,
-      "rewards/rejected": -2.650820016860962,
+      "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,
       "step": 300
     },
     {
-      "epoch": 0.7933461292386437,
-      "grad_norm": 248.8225361148432,
+      "epoch": 0.79,
+      "grad_norm": 161.29540411078162,
       "learning_rate": 4.911681901784197e-08,
-      "logits/chosen": -2.5988125801086426,
-      "logits/rejected": -2.5305018424987793,
-      "logps/chosen": -276.5069885253906,
-      "logps/rejected": -215.4237518310547,
-      "loss": 0.3069,
-      "rewards/accuracies": 0.893750011920929,
-      "rewards/chosen": 2.883484125137329,
-      "rewards/margins": 5.80063533782959,
-      "rewards/rejected": -2.917151927947998,
+      "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,
       "step": 310
     },
     {
-      "epoch": 0.818937939859245,
-      "grad_norm": 378.21338988259265,
+      "epoch": 0.82,
+      "grad_norm": 394.51415258074036,
       "learning_rate": 3.79864238021667e-08,
-      "logits/chosen": -2.621037006378174,
-      "logits/rejected": -2.5753254890441895,
-      "logps/chosen": -259.85833740234375,
-      "logps/rejected": -204.21798706054688,
-      "loss": 0.2691,
-      "rewards/accuracies": 0.8999999761581421,
-      "rewards/chosen": 2.5267233848571777,
-      "rewards/margins": 5.2202558517456055,
-      "rewards/rejected": -2.6935324668884277,
+      "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,
       "step": 320
     },
     {
-      "epoch": 0.8445297504798465,
-      "grad_norm": 354.9475645655247,
+      "epoch": 0.84,
+      "grad_norm": 235.27856577918993,
       "learning_rate": 2.8153049487556634e-08,
-      "logits/chosen": -2.6436188220977783,
-      "logits/rejected": -2.585820436477661,
-      "logps/chosen": -263.09808349609375,
-      "logps/rejected": -217.03341674804688,
-      "loss": 0.2695,
-      "rewards/accuracies": 0.8999999761581421,
-      "rewards/chosen": 3.3003220558166504,
-      "rewards/margins": 5.452363014221191,
-      "rewards/rejected": -2.1520419120788574,
+      "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,
       "step": 330
     },
     {
-      "epoch": 0.8701215611004478,
-      "grad_norm": 298.79503920742496,
+      "epoch": 0.87,
+      "grad_norm": 326.32410341900527,
       "learning_rate": 1.9695418439836796e-08,
-      "logits/chosen": -2.6284804344177246,
-      "logits/rejected": -2.5717878341674805,
-      "logps/chosen": -258.5690002441406,
-      "logps/rejected": -210.4982147216797,
-      "loss": 0.2354,
-      "rewards/accuracies": 0.9312499761581421,
-      "rewards/chosen": 3.1211013793945312,
-      "rewards/margins": 5.934812545776367,
-      "rewards/rejected": -2.813711643218994,
+      "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,
       "step": 340
     },
     {
-      "epoch": 0.8957133717210493,
-      "grad_norm": 256.56016530768545,
+      "epoch": 0.9,
+      "grad_norm": 356.14147259437647,
       "learning_rate": 1.2681239331945692e-08,
-      "logits/chosen": -2.6408398151397705,
-      "logits/rejected": -2.5925440788269043,
-      "logps/chosen": -268.15728759765625,
-      "logps/rejected": -226.1197967529297,
-      "loss": 0.2666,
-      "rewards/accuracies": 0.8812500238418579,
-      "rewards/chosen": 3.046782970428467,
-      "rewards/margins": 5.363730430603027,
-      "rewards/rejected": -2.3169472217559814,
+      "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,
       "step": 350
     },
     {
-      "epoch": 0.9213051823416507,
-      "grad_norm": 316.87066881593194,
+      "epoch": 0.92,
+      "grad_norm": 330.15606711737206,
       "learning_rate": 7.166665093287538e-09,
-      "logits/chosen": -2.6406874656677246,
-      "logits/rejected": -2.5731348991394043,
-      "logps/chosen": -273.30157470703125,
-      "logps/rejected": -218.8077392578125,
-      "loss": 0.3057,
-      "rewards/accuracies": 0.90625,
-      "rewards/chosen": 2.9576709270477295,
-      "rewards/margins": 5.49710750579834,
-      "rewards/rejected": -2.539436101913452,
+      "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,
       "step": 360
     },
     {
-      "epoch": 0.946896992962252,
-      "grad_norm": 569.0448969619218,
+      "epoch": 0.95,
+      "grad_norm": 244.55642297419294,
       "learning_rate": 3.1958433701019694e-09,
-      "logits/chosen": -2.6003308296203613,
-      "logits/rejected": -2.5643393993377686,
-      "logps/chosen": -272.14678955078125,
-      "logps/rejected": -228.38943481445312,
-      "loss": 0.3023,
-      "rewards/accuracies": 0.856249988079071,
-      "rewards/chosen": 3.5620276927948,
-      "rewards/margins": 5.470287322998047,
-      "rewards/rejected": -1.908259391784668,
+      "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,
       "step": 370
     },
     {
-      "epoch": 0.9724888035828535,
-      "grad_norm": 206.4373391747023,
+      "epoch": 0.97,
+      "grad_norm": 197.28407297465554,
       "learning_rate": 8.005630957010012e-10,
-      "logits/chosen": -2.6491971015930176,
-      "logits/rejected": -2.6083197593688965,
-      "logps/chosen": -261.6040344238281,
-      "logps/rejected": -229.88302612304688,
-      "loss": 0.273,
-      "rewards/accuracies": 0.8812500238418579,
-      "rewards/chosen": 3.3906753063201904,
-      "rewards/margins": 5.430436611175537,
-      "rewards/rejected": -2.039761543273926,
+      "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,
       "step": 380
     },
     {
-      "epoch": 0.9980806142034548,
-      "grad_norm": 529.0802276796763,
+      "epoch": 1.0,
+      "grad_norm": 504.7364263980203,
       "learning_rate": 0.0,
-      "logits/chosen": -2.6596872806549072,
-      "logits/rejected": -2.598179578781128,
-      "logps/chosen": -249.18356323242188,
-      "logps/rejected": -209.78622436523438,
-      "loss": 0.2803,
-      "rewards/accuracies": 0.893750011920929,
-      "rewards/chosen": 2.9054455757141113,
-      "rewards/margins": 5.3939642906188965,
-      "rewards/rejected": -2.488518476486206,
+      "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,
       "step": 390
     },
     {
-      "epoch": 0.9980806142034548,
+      "epoch": 1.0,
       "step": 390,
       "total_flos": 0.0,
-      "train_loss": 0.3119094243416419,
-      "train_runtime": 6443.086,
-      "train_samples_per_second": 7.76,
-      "train_steps_per_second": 0.061
+      "train_loss": 0.22171047773116676,
+      "train_runtime": 5890.091,
+      "train_samples_per_second": 8.489,
+      "train_steps_per_second": 0.066
     }
   ],
   "logging_steps": 10,
@@ -623,18 +623,6 @@
   "num_input_tokens_seen": 0,
   "num_train_epochs": 1,
   "save_steps": 100,
-  "stateful_callbacks": {
-    "TrainerControl": {
-      "args": {
-        "should_epoch_stop": false,
-        "should_evaluate": false,
-        "should_log": false,
-        "should_save": true,
-        "should_training_stop": false
-      },
-      "attributes": {}
-    }
-  },
   "total_flos": 0.0,
   "train_batch_size": 4,
   "trial_name": null,

training_args.bin

@@ -1,3 +1,3 @@
 version https://git-lfs.github.com/spec/v1
-oid sha256:0a0444b55e899bda2b049d57e26db7ba9934f12da1827bec93bef3d39b6e277a
-size 6264
+oid sha256:aa0e00d3aa1ab0a0ef1dd07b8b8bd52c64ef5747a57df47646eee4b22e879a38
+size 6328