Model save

README.md

@@ -32,7 +32,7 @@ More information needed
 ### Training hyperparameters
 
 The following hyperparameters were used during training:
-- learning_rate: 5e-07
+- learning_rate: 5e-08
 - train_batch_size: 8
 - eval_batch_size: 8
 - seed: 42

all_results.json

@@ -1,8 +1,8 @@
 {
     "epoch": 1.0,
-    "train_loss": 0.4889662015822626,
-    "train_runtime": 12705.4599,
-    "train_samples": 111132,
-    "train_samples_per_second": 8.747,
+    "train_loss": 0.5437000210468586,
+    "train_runtime": 4622.4032,
+    "train_samples": 39942,
+    "train_samples_per_second": 8.641,
     "train_steps_per_second": 0.034
 }

config.json

@@ -1,5 +1,5 @@
 {
-  "_name_or_path": "/mnt/bn/xuruijie-llm/checkpoints/new_world/v1-ultral",
+  "_name_or_path": "/mnt/bn/xuruijie-llm/checkpoints/new_world/v1-distill-round2-0.5-lr58",
   "architectures": [
     "MistralForCausalLM"
   ],

model-00001-of-00003.safetensors

@@ -1,3 +1,3 @@
 version https://git-lfs.github.com/spec/v1
-oid sha256:ba82277808bebcc49034306497093d16422a339ebc807caf3b62d8187af13671
+oid sha256:0f1145c8df98da5d886c710139d12133f030eeb6eac516ce616aa261c1a4af95
 size 4943178720

model-00002-of-00003.safetensors

@@ -1,3 +1,3 @@
 version https://git-lfs.github.com/spec/v1
-oid sha256:f2614e2a34b78dae61710e0a0cf7eeba46229562764f73882cf8bb8fa34e1b45
+oid sha256:dadcd0d623ac649e14077272450402fe741a935010b44760a1cf23f14bf4046a
 size 4999819336

model-00003-of-00003.safetensors

@@ -1,3 +1,3 @@
 version https://git-lfs.github.com/spec/v1
-oid sha256:e86adf394ab9a7dfc3ae66370db0589217e01eb8e16aaf03919807ccfa6307b6
+oid sha256:731c8da9d0797c68151c2f5e91ca88680335e53ef3600a69e45056d95a511ebc
 size 4540532728

runs/Jun05_23-40-57_n136-100-194/events.out.tfevents.1717602078.n136-100-194.602678.0

@@ -1,3 +1,3 @@
 version https://git-lfs.github.com/spec/v1
-oid sha256:8400baaa1bd4fba7cbdae4af38c389aa333390607c26bfe311862df46e072b6d
-size 12383
+oid sha256:07bd9ec8944a4ef7ae6f3f169a78c20386107a5bfd8095b158f73f36ec135e76
+size 16153

train_results.json

@@ -1,8 +1,8 @@
 {
     "epoch": 1.0,
-    "train_loss": 0.4889662015822626,
-    "train_runtime": 12705.4599,
-    "train_samples": 111132,
-    "train_samples_per_second": 8.747,
+    "train_loss": 0.5437000210468586,
+    "train_runtime": 4622.4032,
+    "train_samples": 39942,
+    "train_samples_per_second": 8.641,
     "train_steps_per_second": 0.034
 }

trainer_state.json

@@ -1,22 +1,22 @@
 {
   "best_metric": null,
   "best_model_checkpoint": null,
-  "epoch": 0.9994242947610823,
+  "epoch": 0.9984,
   "eval_steps": 500,
-  "global_step": 434,
+  "global_step": 156,
   "is_hyper_param_search": false,
   "is_local_process_zero": true,
   "is_world_process_zero": true,
   "log_history": [
     {
-      "epoch": 0.0,
-      "grad_norm": 38.91584534474776,
-      "learning_rate": 1.1363636363636363e-08,
-      "logits/chosen": -4.611147880554199,
-      "logits/rejected": -4.98025369644165,
-      "logps/chosen": -316.8686218261719,
-      "logps/rejected": -266.3260498046875,
-      "loss": 0.6931,
+      "epoch": 0.01,
+      "grad_norm": 1061.4292758130512,
+      "learning_rate": 3.125e-09,
+      "logits/chosen": -3.9499800205230713,
+      "logits/rejected": -4.237819194793701,
+      "logps/chosen": -300.693115234375,
+      "logps/rejected": -249.96307373046875,
+      "loss": 0.6941,
       "rewards/accuracies": 0.0,
       "rewards/chosen": 0.0,
       "rewards/margins": 0.0,
@@ -24,662 +24,242 @@
       "step": 1
     },
     {
-      "epoch": 0.02,
-      "grad_norm": 44.27169906285241,
-      "learning_rate": 1.1363636363636363e-07,
-      "logits/chosen": -4.698853015899658,
-      "logits/rejected": -4.950515270233154,
-      "logps/chosen": -315.6709899902344,
-      "logps/rejected": -330.23773193359375,
-      "loss": 0.6922,
-      "rewards/accuracies": 0.4756944477558136,
-      "rewards/chosen": -0.0019368077628314495,
-      "rewards/margins": 0.0017294860444962978,
-      "rewards/rejected": -0.003666294040158391,
+      "epoch": 0.06,
+      "grad_norm": 1100.0352396547798,
+      "learning_rate": 3.125e-08,
+      "logits/chosen": -4.129236221313477,
+      "logits/rejected": -4.351477146148682,
+      "logps/chosen": -351.48150634765625,
+      "logps/rejected": -308.8130187988281,
+      "loss": 0.7282,
+      "rewards/accuracies": 0.3611111044883728,
+      "rewards/chosen": -0.027542730793356895,
+      "rewards/margins": -0.10096076130867004,
+      "rewards/rejected": 0.073418028652668,
       "step": 10
     },
     {
-      "epoch": 0.05,
-      "grad_norm": 38.18848123177205,
-      "learning_rate": 2.2727272727272726e-07,
-      "logits/chosen": -4.717511177062988,
-      "logits/rejected": -4.855301380157471,
-      "logps/chosen": -322.32855224609375,
-      "logps/rejected": -361.94671630859375,
-      "loss": 0.6699,
-      "rewards/accuracies": 0.659375011920929,
-      "rewards/chosen": -0.09177067130804062,
-      "rewards/margins": 0.046934839338064194,
-      "rewards/rejected": -0.1387055218219757,
+      "epoch": 0.13,
+      "grad_norm": 1023.2979189429582,
+      "learning_rate": 4.9899357349880975e-08,
+      "logits/chosen": -4.195385932922363,
+      "logits/rejected": -4.382458686828613,
+      "logps/chosen": -334.9761962890625,
+      "logps/rejected": -293.8690185546875,
+      "loss": 0.6994,
+      "rewards/accuracies": 0.5625,
+      "rewards/chosen": 0.18794658780097961,
+      "rewards/margins": 0.09464438259601593,
+      "rewards/rejected": 0.09330219030380249,
       "step": 20
     },
     {
-      "epoch": 0.07,
-      "grad_norm": 43.45740176786646,
-      "learning_rate": 3.4090909090909085e-07,
-      "logits/chosen": -4.894489765167236,
-      "logits/rejected": -5.1233744621276855,
-      "logps/chosen": -348.95111083984375,
-      "logps/rejected": -423.50567626953125,
-      "loss": 0.6224,
-      "rewards/accuracies": 0.6312500238418579,
-      "rewards/chosen": -0.5039418935775757,
-      "rewards/margins": 0.27083495259284973,
-      "rewards/rejected": -0.7747769355773926,
+      "epoch": 0.19,
+      "grad_norm": 716.4790257172086,
+      "learning_rate": 4.877641290737884e-08,
+      "logits/chosen": -4.231568813323975,
+      "logits/rejected": -4.364924430847168,
+      "logps/chosen": -327.92413330078125,
+      "logps/rejected": -295.37176513671875,
+      "loss": 0.6055,
+      "rewards/accuracies": 0.684374988079071,
+      "rewards/chosen": 0.8660896420478821,
+      "rewards/margins": 0.425116628408432,
+      "rewards/rejected": 0.44097304344177246,
       "step": 30
     },
     {
-      "epoch": 0.09,
-      "grad_norm": 48.10105616024464,
-      "learning_rate": 4.545454545454545e-07,
-      "logits/chosen": -5.166990280151367,
-      "logits/rejected": -5.407620429992676,
-      "logps/chosen": -362.8187255859375,
-      "logps/rejected": -471.7628479003906,
-      "loss": 0.5774,
-      "rewards/accuracies": 0.6875,
-      "rewards/chosen": -0.6842997074127197,
-      "rewards/margins": 0.4909951090812683,
-      "rewards/rejected": -1.1752947568893433,
+      "epoch": 0.26,
+      "grad_norm": 820.682467230606,
+      "learning_rate": 4.646121984004665e-08,
+      "logits/chosen": -4.1439104080200195,
+      "logits/rejected": -4.345718860626221,
+      "logps/chosen": -330.7722473144531,
+      "logps/rejected": -288.60272216796875,
+      "loss": 0.5304,
+      "rewards/accuracies": 0.831250011920929,
+      "rewards/chosen": 1.6021575927734375,
+      "rewards/margins": 0.956769585609436,
+      "rewards/rejected": 0.6453880071640015,
       "step": 40
     },
     {
-      "epoch": 0.12,
-      "grad_norm": 47.781476451630894,
-      "learning_rate": 4.997080567080816e-07,
-      "logits/chosen": -5.179937839508057,
-      "logits/rejected": -5.452680587768555,
-      "logps/chosen": -368.87158203125,
-      "logps/rejected": -494.6646423339844,
-      "loss": 0.5513,
-      "rewards/accuracies": 0.715624988079071,
-      "rewards/chosen": -0.6543357372283936,
-      "rewards/margins": 0.5737784504890442,
-      "rewards/rejected": -1.228114366531372,
+      "epoch": 0.32,
+      "grad_norm": 587.6344375884266,
+      "learning_rate": 4.3069871595684784e-08,
+      "logits/chosen": -4.235655784606934,
+      "logits/rejected": -4.4143476486206055,
+      "logps/chosen": -330.8318786621094,
+      "logps/rejected": -291.75067138671875,
+      "loss": 0.525,
+      "rewards/accuracies": 0.8125,
+      "rewards/chosen": 2.010468006134033,
+      "rewards/margins": 1.2047243118286133,
+      "rewards/rejected": 0.8057435750961304,
       "step": 50
     },
     {
-      "epoch": 0.14,
-      "grad_norm": 49.51998331093021,
-      "learning_rate": 4.979264274553905e-07,
-      "logits/chosen": -5.6107635498046875,
-      "logits/rejected": -5.908527374267578,
-      "logps/chosen": -350.31158447265625,
-      "logps/rejected": -498.5696716308594,
-      "loss": 0.5316,
-      "rewards/accuracies": 0.75,
-      "rewards/chosen": -0.699661135673523,
-      "rewards/margins": 0.7308087348937988,
-      "rewards/rejected": -1.4304701089859009,
+      "epoch": 0.38,
+      "grad_norm": 586.1895231882175,
+      "learning_rate": 3.8772424536302564e-08,
+      "logits/chosen": -4.247714519500732,
+      "logits/rejected": -4.417937278747559,
+      "logps/chosen": -322.380126953125,
+      "logps/rejected": -291.76654052734375,
+      "loss": 0.517,
+      "rewards/accuracies": 0.8218749761581421,
+      "rewards/chosen": 2.19193696975708,
+      "rewards/margins": 1.311265230178833,
+      "rewards/rejected": 0.8806716203689575,
       "step": 60
     },
     {
-      "epoch": 0.16,
-      "grad_norm": 54.22311110544169,
-      "learning_rate": 4.945369001834514e-07,
-      "logits/chosen": -5.679194450378418,
-      "logits/rejected": -6.17197322845459,
-      "logps/chosen": -348.9252624511719,
-      "logps/rejected": -500.28570556640625,
-      "loss": 0.5218,
-      "rewards/accuracies": 0.746874988079071,
-      "rewards/chosen": -0.612450361251831,
-      "rewards/margins": 0.8183927536010742,
-      "rewards/rejected": -1.4308433532714844,
+      "epoch": 0.45,
+      "grad_norm": 528.8947001003231,
+      "learning_rate": 3.378437060203357e-08,
+      "logits/chosen": -4.170197486877441,
+      "logits/rejected": -4.359009742736816,
+      "logps/chosen": -322.4336242675781,
+      "logps/rejected": -289.197998046875,
+      "loss": 0.5096,
+      "rewards/accuracies": 0.815625011920929,
+      "rewards/chosen": 2.203657865524292,
+      "rewards/margins": 1.3729288578033447,
+      "rewards/rejected": 0.8307291269302368,
       "step": 70
     },
     {
-      "epoch": 0.18,
-      "grad_norm": 51.46924554201296,
-      "learning_rate": 4.895614572772916e-07,
-      "logits/chosen": -5.785242557525635,
-      "logits/rejected": -6.346865653991699,
-      "logps/chosen": -375.2938537597656,
-      "logps/rejected": -535.5648193359375,
-      "loss": 0.5071,
-      "rewards/accuracies": 0.746874988079071,
-      "rewards/chosen": -0.7518796324729919,
-      "rewards/margins": 1.0060392618179321,
-      "rewards/rejected": -1.7579187154769897,
+      "epoch": 0.51,
+      "grad_norm": 544.8408089721645,
+      "learning_rate": 2.8355831645441387e-08,
+      "logits/chosen": -4.038945198059082,
+      "logits/rejected": -4.326292991638184,
+      "logps/chosen": -348.65472412109375,
+      "logps/rejected": -308.1936950683594,
+      "loss": 0.5115,
+      "rewards/accuracies": 0.875,
+      "rewards/chosen": 2.3145015239715576,
+      "rewards/margins": 1.5040271282196045,
+      "rewards/rejected": 0.8104745149612427,
       "step": 80
     },
     {
-      "epoch": 0.21,
-      "grad_norm": 64.30139334546263,
-      "learning_rate": 4.830323663933919e-07,
-      "logits/chosen": -5.814670562744141,
-      "logits/rejected": -6.330450534820557,
-      "logps/chosen": -414.3651428222656,
-      "logps/rejected": -571.2876586914062,
-      "loss": 0.5095,
-      "rewards/accuracies": 0.746874988079071,
-      "rewards/chosen": -0.8821724057197571,
-      "rewards/margins": 0.963842511177063,
-      "rewards/rejected": -1.8460149765014648,
+      "epoch": 0.58,
+      "grad_norm": 526.2097671297138,
+      "learning_rate": 2.2759017277414164e-08,
+      "logits/chosen": -4.170694351196289,
+      "logits/rejected": -4.379502296447754,
+      "logps/chosen": -335.43402099609375,
+      "logps/rejected": -295.8087463378906,
+      "loss": 0.5066,
+      "rewards/accuracies": 0.859375,
+      "rewards/chosen": 1.9496158361434937,
+      "rewards/margins": 1.3254307508468628,
+      "rewards/rejected": 0.6241849660873413,
       "step": 90
     },
     {
-      "epoch": 0.23,
-      "grad_norm": 52.98329513030299,
-      "learning_rate": 4.74991971191553e-07,
-      "logits/chosen": -6.23067569732666,
-      "logits/rejected": -6.7989630699157715,
-      "logps/chosen": -387.9895324707031,
-      "logps/rejected": -506.8861389160156,
-      "loss": 0.5056,
-      "rewards/accuracies": 0.753125011920929,
-      "rewards/chosen": -0.8129897117614746,
-      "rewards/margins": 0.7770716547966003,
-      "rewards/rejected": -1.5900614261627197,
+      "epoch": 0.64,
+      "grad_norm": 595.4206000749634,
+      "learning_rate": 1.7274575140626317e-08,
+      "logits/chosen": -4.159605979919434,
+      "logits/rejected": -4.377494812011719,
+      "logps/chosen": -333.1622314453125,
+      "logps/rejected": -286.3451232910156,
+      "loss": 0.5142,
+      "rewards/accuracies": 0.871874988079071,
+      "rewards/chosen": 2.0654988288879395,
+      "rewards/margins": 1.4395055770874023,
+      "rewards/rejected": 0.6259931921958923,
       "step": 100
     },
     {
-      "epoch": 0.25,
-      "grad_norm": 59.87890171282983,
-      "learning_rate": 4.6549241672001225e-07,
-      "logits/chosen": -6.03170108795166,
-      "logits/rejected": -6.566573143005371,
-      "logps/chosen": -372.1159362792969,
-      "logps/rejected": -527.2628173828125,
-      "loss": 0.4885,
-      "rewards/accuracies": 0.7437499761581421,
-      "rewards/chosen": -0.8102461099624634,
-      "rewards/margins": 0.9041855931282043,
-      "rewards/rejected": -1.7144317626953125,
+      "epoch": 0.7,
+      "grad_norm": 528.6243658840883,
+      "learning_rate": 1.217751806485235e-08,
+      "logits/chosen": -4.138543128967285,
+      "logits/rejected": -4.379323959350586,
+      "logps/chosen": -314.89556884765625,
+      "logps/rejected": -277.0162658691406,
+      "loss": 0.5071,
+      "rewards/accuracies": 0.840624988079071,
+      "rewards/chosen": 2.0163025856018066,
+      "rewards/margins": 1.3896347284317017,
+      "rewards/rejected": 0.6266676783561707,
       "step": 110
     },
     {
-      "epoch": 0.28,
-      "grad_norm": 63.51832709179028,
-      "learning_rate": 4.5459531123479673e-07,
-      "logits/chosen": -6.038158893585205,
-      "logits/rejected": -6.699621677398682,
-      "logps/chosen": -401.9744873046875,
-      "logps/rejected": -594.4263916015625,
-      "loss": 0.4752,
-      "rewards/accuracies": 0.800000011920929,
-      "rewards/chosen": -0.8800112009048462,
-      "rewards/margins": 1.2125087976455688,
-      "rewards/rejected": -2.092519760131836,
+      "epoch": 0.77,
+      "grad_norm": 557.9046270518194,
+      "learning_rate": 7.723433775328384e-09,
+      "logits/chosen": -4.136676788330078,
+      "logits/rejected": -4.345128536224365,
+      "logps/chosen": -328.94482421875,
+      "logps/rejected": -281.4302673339844,
+      "loss": 0.5052,
+      "rewards/accuracies": 0.831250011920929,
+      "rewards/chosen": 1.9894015789031982,
+      "rewards/margins": 1.363441824913025,
+      "rewards/rejected": 0.625959575176239,
       "step": 120
     },
     {
-      "epoch": 0.3,
-      "grad_norm": 52.63677136901513,
-      "learning_rate": 4.4237132664654147e-07,
-      "logits/chosen": -6.488445281982422,
-      "logits/rejected": -7.092920780181885,
-      "logps/chosen": -405.026611328125,
-      "logps/rejected": -571.0286254882812,
-      "loss": 0.4707,
-      "rewards/accuracies": 0.778124988079071,
-      "rewards/chosen": -0.9200836420059204,
-      "rewards/margins": 0.9780105352401733,
-      "rewards/rejected": -1.8980941772460938,
+      "epoch": 0.83,
+      "grad_norm": 523.8022952711281,
+      "learning_rate": 4.135668656967433e-09,
+      "logits/chosen": -4.222664833068848,
+      "logits/rejected": -4.383507251739502,
+      "logps/chosen": -334.4266357421875,
+      "logps/rejected": -287.63067626953125,
+      "loss": 0.5059,
+      "rewards/accuracies": 0.8687499761581421,
+      "rewards/chosen": 2.026325225830078,
+      "rewards/margins": 1.3963022232055664,
+      "rewards/rejected": 0.6300228834152222,
       "step": 130
     },
     {
-      "epoch": 0.32,
-      "grad_norm": 57.84913429480699,
-      "learning_rate": 4.2889974018603024e-07,
-      "logits/chosen": -6.785845756530762,
-      "logits/rejected": -7.511507987976074,
-      "logps/chosen": -399.703125,
-      "logps/rejected": -555.4042358398438,
-      "loss": 0.4843,
-      "rewards/accuracies": 0.784375011920929,
-      "rewards/chosen": -0.9189607501029968,
-      "rewards/margins": 1.0339016914367676,
-      "rewards/rejected": -1.9528625011444092,
+      "epoch": 0.9,
+      "grad_norm": 601.0204054485596,
+      "learning_rate": 1.5941282340065698e-09,
+      "logits/chosen": -4.176892280578613,
+      "logits/rejected": -4.3910746574401855,
+      "logps/chosen": -335.8815612792969,
+      "logps/rejected": -304.35723876953125,
+      "loss": 0.5101,
+      "rewards/accuracies": 0.831250011920929,
+      "rewards/chosen": 1.9034252166748047,
+      "rewards/margins": 1.3076425790786743,
+      "rewards/rejected": 0.5957827568054199,
       "step": 140
     },
     {
-      "epoch": 0.35,
-      "grad_norm": 52.039202437687166,
-      "learning_rate": 4.142679202609327e-07,
-      "logits/chosen": -6.392851829528809,
-      "logits/rejected": -7.089343070983887,
-      "logps/chosen": -397.5931701660156,
-      "logps/rejected": -551.0249633789062,
-      "loss": 0.4862,
-      "rewards/accuracies": 0.7562500238418579,
-      "rewards/chosen": -0.8002569079399109,
-      "rewards/margins": 1.0864611864089966,
-      "rewards/rejected": -1.8867183923721313,
+      "epoch": 0.96,
+      "grad_norm": 496.1092972793449,
+      "learning_rate": 2.262559558016325e-10,
+      "logits/chosen": -4.114329814910889,
+      "logits/rejected": -4.342306137084961,
+      "logps/chosen": -342.4503479003906,
+      "logps/rejected": -296.07086181640625,
+      "loss": 0.506,
+      "rewards/accuracies": 0.840624988079071,
+      "rewards/chosen": 2.0278801918029785,
+      "rewards/margins": 1.378021001815796,
+      "rewards/rejected": 0.6498591303825378,
       "step": 150
     },
-    {
-      "epoch": 0.37,
-      "grad_norm": 60.69164650622946,
-      "learning_rate": 3.9857075983815435e-07,
-      "logits/chosen": -6.71727991104126,
-      "logits/rejected": -7.2690839767456055,
-      "logps/chosen": -344.13055419921875,
-      "logps/rejected": -493.87982177734375,
-      "loss": 0.4849,
-      "rewards/accuracies": 0.721875011920929,
-      "rewards/chosen": -0.6938045620918274,
-      "rewards/margins": 0.9113534092903137,
-      "rewards/rejected": -1.6051580905914307,
-      "step": 160
-    },
-    {
-      "epoch": 0.39,
-      "grad_norm": 65.68378818115322,
-      "learning_rate": 3.8191006102653317e-07,
-      "logits/chosen": -6.6441650390625,
-      "logits/rejected": -7.4138336181640625,
-      "logps/chosen": -401.4483947753906,
-      "logps/rejected": -552.6317138671875,
-      "loss": 0.4748,
-      "rewards/accuracies": 0.7406250238418579,
-      "rewards/chosen": -0.8133081197738647,
-      "rewards/margins": 1.0876938104629517,
-      "rewards/rejected": -1.9010016918182373,
-      "step": 170
-    },
-    {
-      "epoch": 0.41,
-      "grad_norm": 59.77164957437951,
-      "learning_rate": 3.6439387485109883e-07,
-      "logits/chosen": -7.071902275085449,
-      "logits/rejected": -7.657292366027832,
-      "logps/chosen": -400.84539794921875,
-      "logps/rejected": -545.3646850585938,
-      "loss": 0.492,
-      "rewards/accuracies": 0.737500011920929,
-      "rewards/chosen": -0.8680183291435242,
-      "rewards/margins": 1.0952622890472412,
-      "rewards/rejected": -1.963280439376831,
-      "step": 180
-    },
-    {
-      "epoch": 0.44,
-      "grad_norm": 61.16868629176781,
-      "learning_rate": 3.4613580050071274e-07,
-      "logits/chosen": -6.881494045257568,
-      "logits/rejected": -7.402888298034668,
-      "logps/chosen": -385.8741455078125,
-      "logps/rejected": -550.5389404296875,
-      "loss": 0.4856,
-      "rewards/accuracies": 0.753125011920929,
-      "rewards/chosen": -0.734123945236206,
-      "rewards/margins": 1.0420116186141968,
-      "rewards/rejected": -1.7761356830596924,
-      "step": 190
-    },
-    {
-      "epoch": 0.46,
-      "grad_norm": 55.469164225282256,
-      "learning_rate": 3.272542485937368e-07,
-      "logits/chosen": -7.093558311462402,
-      "logits/rejected": -7.623803615570068,
-      "logps/chosen": -439.5218811035156,
-      "logps/rejected": -535.7510986328125,
-      "loss": 0.4958,
-      "rewards/accuracies": 0.7406250238418579,
-      "rewards/chosen": -1.084832787513733,
-      "rewards/margins": 0.8213528394699097,
-      "rewards/rejected": -1.906185507774353,
-      "step": 200
-    },
-    {
-      "epoch": 0.48,
-      "grad_norm": 55.08397644622264,
-      "learning_rate": 3.078716732397358e-07,
-      "logits/chosen": -7.115843296051025,
-      "logits/rejected": -7.49353551864624,
-      "logps/chosen": -369.25775146484375,
-      "logps/rejected": -519.5389404296875,
-      "loss": 0.4802,
-      "rewards/accuracies": 0.734375,
-      "rewards/chosen": -0.7710889577865601,
-      "rewards/margins": 0.8956006765365601,
-      "rewards/rejected": -1.6666896343231201,
-      "step": 210
-    },
-    {
-      "epoch": 0.51,
-      "grad_norm": 57.42294921171166,
-      "learning_rate": 2.881137778775863e-07,
-      "logits/chosen": -7.241219997406006,
-      "logits/rejected": -7.856888771057129,
-      "logps/chosen": -418.5899353027344,
-      "logps/rejected": -585.8680419921875,
-      "loss": 0.4566,
-      "rewards/accuracies": 0.8187500238418579,
-      "rewards/chosen": -1.0504003763198853,
-      "rewards/margins": 1.1809632778167725,
-      "rewards/rejected": -2.2313637733459473,
-      "step": 220
-    },
-    {
-      "epoch": 0.53,
-      "grad_norm": 63.40362339583639,
-      "learning_rate": 2.681087000404406e-07,
-      "logits/chosen": -6.954517364501953,
-      "logits/rejected": -7.46679162979126,
-      "logps/chosen": -403.7725524902344,
-      "logps/rejected": -582.7224731445312,
-      "loss": 0.4504,
-      "rewards/accuracies": 0.8218749761581421,
-      "rewards/chosen": -0.9388762712478638,
-      "rewards/margins": 1.1833168268203735,
-      "rewards/rejected": -2.122192859649658,
-      "step": 230
-    },
-    {
-      "epoch": 0.55,
-      "grad_norm": 53.85424019884234,
-      "learning_rate": 2.4798618033465255e-07,
-      "logits/chosen": -7.1180739402771,
-      "logits/rejected": -7.677659034729004,
-      "logps/chosen": -380.5184326171875,
-      "logps/rejected": -603.7531127929688,
-      "loss": 0.4711,
-      "rewards/accuracies": 0.800000011920929,
-      "rewards/chosen": -0.8171305656433105,
-      "rewards/margins": 1.2705714702606201,
-      "rewards/rejected": -2.0877020359039307,
-      "step": 240
-    },
-    {
-      "epoch": 0.58,
-      "grad_norm": 57.06338659240929,
-      "learning_rate": 2.278767210221604e-07,
-      "logits/chosen": -7.427262783050537,
-      "logits/rejected": -7.926419258117676,
-      "logps/chosen": -406.00433349609375,
-      "logps/rejected": -564.0296630859375,
-      "loss": 0.4854,
-      "rewards/accuracies": 0.765625,
-      "rewards/chosen": -1.0326459407806396,
-      "rewards/margins": 1.145564079284668,
-      "rewards/rejected": -2.1782097816467285,
-      "step": 250
-    },
-    {
-      "epoch": 0.6,
-      "grad_norm": 50.302485784219535,
-      "learning_rate": 2.0791073966324034e-07,
-      "logits/chosen": -7.219435214996338,
-      "logits/rejected": -7.894481658935547,
-      "logps/chosen": -390.1949157714844,
-      "logps/rejected": -556.72900390625,
-      "loss": 0.4677,
-      "rewards/accuracies": 0.7562500238418579,
-      "rewards/chosen": -0.8992031216621399,
-      "rewards/margins": 1.1654497385025024,
-      "rewards/rejected": -2.064652919769287,
-      "step": 260
-    },
-    {
-      "epoch": 0.62,
-      "grad_norm": 56.08299211255023,
-      "learning_rate": 1.8821772330858257e-07,
-      "logits/chosen": -7.049578666687012,
-      "logits/rejected": -7.631491184234619,
-      "logps/chosen": -424.8095703125,
-      "logps/rejected": -596.650146484375,
-      "loss": 0.482,
-      "rewards/accuracies": 0.762499988079071,
-      "rewards/chosen": -1.0093523263931274,
-      "rewards/margins": 1.1269723176956177,
-      "rewards/rejected": -2.136324644088745,
-      "step": 270
-    },
-    {
-      "epoch": 0.64,
-      "grad_norm": 54.86550915852827,
-      "learning_rate": 1.6892538872607933e-07,
-      "logits/chosen": -7.0790252685546875,
-      "logits/rejected": -7.644691467285156,
-      "logps/chosen": -436.185791015625,
-      "logps/rejected": -609.986083984375,
-      "loss": 0.4494,
-      "rewards/accuracies": 0.800000011920929,
-      "rewards/chosen": -1.080344319343567,
-      "rewards/margins": 1.2829288244247437,
-      "rewards/rejected": -2.3632731437683105,
-      "step": 280
-    },
-    {
-      "epoch": 0.67,
-      "grad_norm": 61.27463111501325,
-      "learning_rate": 1.5015885410857614e-07,
-      "logits/chosen": -7.3400774002075195,
-      "logits/rejected": -7.88690710067749,
-      "logps/chosen": -387.97491455078125,
-      "logps/rejected": -569.4407958984375,
-      "loss": 0.445,
-      "rewards/accuracies": 0.765625,
-      "rewards/chosen": -0.9607879519462585,
-      "rewards/margins": 1.2261817455291748,
-      "rewards/rejected": -2.186969757080078,
-      "step": 290
-    },
-    {
-      "epoch": 0.69,
-      "grad_norm": 72.35584757362263,
-      "learning_rate": 1.320398276343795e-07,
-      "logits/chosen": -7.49501895904541,
-      "logits/rejected": -8.059203147888184,
-      "logps/chosen": -396.3390197753906,
-      "logps/rejected": -610.2857055664062,
-      "loss": 0.4408,
-      "rewards/accuracies": 0.815625011920929,
-      "rewards/chosen": -1.0337975025177002,
-      "rewards/margins": 1.399001121520996,
-      "rewards/rejected": -2.432798385620117,
-      "step": 300
-    },
-    {
-      "epoch": 0.71,
-      "grad_norm": 60.072786340008435,
-      "learning_rate": 1.1468581814301717e-07,
-      "logits/chosen": -7.159657955169678,
-      "logits/rejected": -7.803900718688965,
-      "logps/chosen": -392.6741638183594,
-      "logps/rejected": -566.23779296875,
-      "loss": 0.4572,
-      "rewards/accuracies": 0.796875,
-      "rewards/chosen": -0.9514274597167969,
-      "rewards/margins": 1.3155269622802734,
-      "rewards/rejected": -2.266954183578491,
-      "step": 310
-    },
-    {
-      "epoch": 0.74,
-      "grad_norm": 49.76656663928056,
-      "learning_rate": 9.82093730453222e-08,
-      "logits/chosen": -7.2263312339782715,
-      "logits/rejected": -7.8056640625,
-      "logps/chosen": -407.4519348144531,
-      "logps/rejected": -598.1903686523438,
-      "loss": 0.4484,
-      "rewards/accuracies": 0.7749999761581421,
-      "rewards/chosen": -1.0463659763336182,
-      "rewards/margins": 1.2292554378509521,
-      "rewards/rejected": -2.2756214141845703,
-      "step": 320
-    },
-    {
-      "epoch": 0.76,
-      "grad_norm": 53.042260153017224,
-      "learning_rate": 8.271734841028552e-08,
-      "logits/chosen": -7.178483009338379,
-      "logits/rejected": -7.908429145812988,
-      "logps/chosen": -397.34674072265625,
-      "logps/rejected": -566.149658203125,
-      "loss": 0.454,
-      "rewards/accuracies": 0.765625,
-      "rewards/chosen": -0.972808837890625,
-      "rewards/margins": 1.2322447299957275,
-      "rewards/rejected": -2.2050535678863525,
-      "step": 330
-    },
-    {
-      "epoch": 0.78,
-      "grad_norm": 52.07091389097818,
-      "learning_rate": 6.831021596244424e-08,
-      "logits/chosen": -7.447384834289551,
-      "logits/rejected": -7.893862724304199,
-      "logps/chosen": -401.7857360839844,
-      "logps/rejected": -565.5882568359375,
-      "loss": 0.4493,
-      "rewards/accuracies": 0.7749999761581421,
-      "rewards/chosen": -1.0928151607513428,
-      "rewards/margins": 1.0317901372909546,
-      "rewards/rejected": -2.124605655670166,
-      "step": 340
-    },
-    {
-      "epoch": 0.81,
-      "grad_norm": 61.51781141510449,
-      "learning_rate": 5.508141148419443e-08,
-      "logits/chosen": -7.330380439758301,
-      "logits/rejected": -7.86685848236084,
-      "logps/chosen": -420.15765380859375,
-      "logps/rejected": -605.4978637695312,
-      "loss": 0.4551,
-      "rewards/accuracies": 0.768750011920929,
-      "rewards/chosen": -1.1430304050445557,
-      "rewards/margins": 1.1955806016921997,
-      "rewards/rejected": -2.338611125946045,
-      "step": 350
-    },
-    {
-      "epoch": 0.83,
-      "grad_norm": 51.73537813299502,
-      "learning_rate": 4.311672884888756e-08,
-      "logits/chosen": -7.131661415100098,
-      "logits/rejected": -8.078085899353027,
-      "logps/chosen": -420.2433166503906,
-      "logps/rejected": -584.4310302734375,
-      "loss": 0.4624,
-      "rewards/accuracies": 0.7906249761581421,
-      "rewards/chosen": -1.0593332052230835,
-      "rewards/margins": 1.38120698928833,
-      "rewards/rejected": -2.440540075302124,
-      "step": 360
-    },
-    {
-      "epoch": 0.85,
-      "grad_norm": 56.350435408969204,
-      "learning_rate": 3.249376361464021e-08,
-      "logits/chosen": -7.158230781555176,
-      "logits/rejected": -7.789064884185791,
-      "logps/chosen": -407.6632995605469,
-      "logps/rejected": -590.59814453125,
-      "loss": 0.456,
-      "rewards/accuracies": 0.7906249761581421,
-      "rewards/chosen": -1.0117406845092773,
-      "rewards/margins": 1.3086559772491455,
-      "rewards/rejected": -2.3203964233398438,
-      "step": 370
-    },
-    {
-      "epoch": 0.88,
-      "grad_norm": 60.96351621879777,
-      "learning_rate": 2.3281409787363648e-08,
-      "logits/chosen": -7.183234214782715,
-      "logits/rejected": -7.844109535217285,
-      "logps/chosen": -393.98419189453125,
-      "logps/rejected": -574.4835815429688,
-      "loss": 0.4507,
-      "rewards/accuracies": 0.768750011920929,
-      "rewards/chosen": -1.0079011917114258,
-      "rewards/margins": 1.1862713098526,
-      "rewards/rejected": -2.1941723823547363,
-      "step": 380
-    },
-    {
-      "epoch": 0.9,
-      "grad_norm": 51.843711912736374,
-      "learning_rate": 1.553941301669892e-08,
-      "logits/chosen": -7.335826873779297,
-      "logits/rejected": -7.751254081726074,
-      "logps/chosen": -416.5980529785156,
-      "logps/rejected": -624.6790161132812,
-      "loss": 0.4582,
-      "rewards/accuracies": 0.753125011920929,
-      "rewards/chosen": -1.0918513536453247,
-      "rewards/margins": 1.3415181636810303,
-      "rewards/rejected": -2.4333693981170654,
-      "step": 390
-    },
-    {
-      "epoch": 0.92,
-      "grad_norm": 53.35049398455392,
-      "learning_rate": 9.31798312255233e-09,
-      "logits/chosen": -7.139242649078369,
-      "logits/rejected": -7.836366176605225,
-      "logps/chosen": -401.4609375,
-      "logps/rejected": -584.0276489257812,
-      "loss": 0.4488,
-      "rewards/accuracies": 0.762499988079071,
-      "rewards/chosen": -1.064345359802246,
-      "rewards/margins": 1.1964819431304932,
-      "rewards/rejected": -2.2608273029327393,
-      "step": 400
-    },
-    {
-      "epoch": 0.94,
-      "grad_norm": 54.66202048705383,
-      "learning_rate": 4.657468465146641e-09,
-      "logits/chosen": -7.338484287261963,
-      "logits/rejected": -7.767783164978027,
-      "logps/chosen": -407.49188232421875,
-      "logps/rejected": -591.4136352539062,
-      "loss": 0.4597,
-      "rewards/accuracies": 0.768750011920929,
-      "rewards/chosen": -1.0377376079559326,
-      "rewards/margins": 1.1812639236450195,
-      "rewards/rejected": -2.219001531600952,
-      "step": 410
-    },
-    {
-      "epoch": 0.97,
-      "grad_norm": 58.302245113936664,
-      "learning_rate": 1.5880942704217526e-09,
-      "logits/chosen": -7.232787132263184,
-      "logits/rejected": -7.815786838531494,
-      "logps/chosen": -418.7047424316406,
-      "logps/rejected": -592.791015625,
-      "loss": 0.4552,
-      "rewards/accuracies": 0.793749988079071,
-      "rewards/chosen": -1.077683448791504,
-      "rewards/margins": 1.2897697687149048,
-      "rewards/rejected": -2.3674533367156982,
-      "step": 420
-    },
-    {
-      "epoch": 0.99,
-      "grad_norm": 58.20386375203141,
-      "learning_rate": 1.297666078462767e-10,
-      "logits/chosen": -7.145885467529297,
-      "logits/rejected": -7.906874656677246,
-      "logps/chosen": -446.4364318847656,
-      "logps/rejected": -584.668701171875,
-      "loss": 0.457,
-      "rewards/accuracies": 0.793749988079071,
-      "rewards/chosen": -1.0853328704833984,
-      "rewards/margins": 1.1659003496170044,
-      "rewards/rejected": -2.2512331008911133,
-      "step": 430
-    },
     {
       "epoch": 1.0,
-      "step": 434,
+      "step": 156,
       "total_flos": 0.0,
-      "train_loss": 0.4889662015822626,
-      "train_runtime": 12705.4599,
-      "train_samples_per_second": 8.747,
+      "train_loss": 0.5437000210468586,
+      "train_runtime": 4622.4032,
+      "train_samples_per_second": 8.641,
       "train_steps_per_second": 0.034
     }
   ],
   "logging_steps": 10,
-  "max_steps": 434,
+  "max_steps": 156,
   "num_input_tokens_seen": 0,
   "num_train_epochs": 1,
   "save_steps": 100,

training_args.bin

@@ -1,3 +1,3 @@
 version https://git-lfs.github.com/spec/v1
-oid sha256:b92372e760a727f6582979303a96b389cee04d00a2f4a6c534c36c4b19bfdc3f
-size 6264
+oid sha256:e45318e4cbde50c1214320bcb3af5bfaf60bb98221a8335b87aca82b50f6806e
+size 6328