Model save

README.md

@@ -32,7 +32,7 @@ More information needed
 ### Training hyperparameters
 
 The following hyperparameters were used during training:
-- learning_rate: 5e-08
+- learning_rate: 5e-07
 - train_batch_size: 8
 - eval_batch_size: 8
 - seed: 42
@@ -52,7 +52,7 @@ The following hyperparameters were used during training:
 
 ### Framework versions
 
-- Transformers 4.41.1
+- Transformers 4.38.2
 - Pytorch 2.1.2+cu118
 - Datasets 2.16.1
-- Tokenizers 0.19.1
+- Tokenizers 0.15.2

all_results.json

@@ -1,9 +1,8 @@
 {
-    "epoch": 0.9984,
-    "total_flos": 0.0,
-    "train_loss": 0.4014626894241724,
-    "train_runtime": 5052.0322,
-    "train_samples": 39942,
-    "train_samples_per_second": 7.906,
-    "train_steps_per_second": 0.031
+    "epoch": 1.0,
+    "train_loss": 0.4889662015822626,
+    "train_runtime": 12705.4599,
+    "train_samples": 111132,
+    "train_samples_per_second": 8.747,
+    "train_steps_per_second": 0.034
 }

config.json

@@ -1,5 +1,5 @@
 {
-  "_name_or_path": "/mnt/bn/xuruijie-llm/checkpoints/new_world/v1-distill-round2-0.5-lr58",
+  "_name_or_path": "/mnt/bn/xuruijie-llm/checkpoints/new_world/v1-ultral",
   "architectures": [
     "MistralForCausalLM"
   ],
@@ -20,7 +20,7 @@
   "sliding_window": 4096,
   "tie_word_embeddings": false,
   "torch_dtype": "bfloat16",
-  "transformers_version": "4.41.1",
+  "transformers_version": "4.38.2",
   "use_cache": false,
   "vocab_size": 32002
 }

generation_config.json

@@ -2,5 +2,5 @@
   "_from_model_config": true,
   "bos_token_id": 1,
   "eos_token_id": 32000,
-  "transformers_version": "4.41.1"
+  "transformers_version": "4.38.2"
 }

model-00001-of-00003.safetensors

@@ -1,3 +1,3 @@
 version https://git-lfs.github.com/spec/v1
-oid sha256:7f512a730a26a61a1c16526c02b35b83222d091cd9296f1da7e71affd7995410
+oid sha256:ba82277808bebcc49034306497093d16422a339ebc807caf3b62d8187af13671
 size 4943178720

model-00002-of-00003.safetensors

@@ -1,3 +1,3 @@
 version https://git-lfs.github.com/spec/v1
-oid sha256:ac0006e40eab93ba96e9c4ac961c6bd73d92585b3f54f427b278b0fb7b9767b2
+oid sha256:f2614e2a34b78dae61710e0a0cf7eeba46229562764f73882cf8bb8fa34e1b45
 size 4999819336

model-00003-of-00003.safetensors

@@ -1,3 +1,3 @@
 version https://git-lfs.github.com/spec/v1
-oid sha256:a220e930529efb9b6318941d96516bbbb5fa2c9d14cf80d2860389fe098f2e86
+oid sha256:e86adf394ab9a7dfc3ae66370db0589217e01eb8e16aaf03919807ccfa6307b6
 size 4540532728

runs/Jun05_21-05-40_n136-129-074/events.out.tfevents.1717593358.n136-129-074.764974.0

@@ -1,3 +1,3 @@
 version https://git-lfs.github.com/spec/v1
-oid sha256:ac407ca9090f32e38aaeeff755a3e1cd779d81a0e0ca3d3e6fa6863155fcf075
-size 32906
+oid sha256:a3e65fa8603a34da15af751381fb85ddf30d99aac6c873f8076d444675944de6
+size 35324

tokenizer.json

@@ -152,7 +152,6 @@
     "end_of_word_suffix": null,
     "fuse_unk": true,
     "byte_fallback": true,
-    "ignore_merges": false,
     "vocab": {
       "<unk>": 0,
       "<s>": 1,

train_results.json

@@ -1,9 +1,8 @@
 {
-    "epoch": 0.9984,
-    "total_flos": 0.0,
-    "train_loss": 0.4014626894241724,
-    "train_runtime": 5052.0322,
-    "train_samples": 39942,
-    "train_samples_per_second": 7.906,
-    "train_steps_per_second": 0.031
+    "epoch": 1.0,
+    "train_loss": 0.4889662015822626,
+    "train_runtime": 12705.4599,
+    "train_samples": 111132,
+    "train_samples_per_second": 8.747,
+    "train_steps_per_second": 0.034
 }

trainer_state.json

@@ -1,21 +1,21 @@
 {
   "best_metric": null,
   "best_model_checkpoint": null,
-  "epoch": 0.9984,
+  "epoch": 0.9994242947610823,
   "eval_steps": 500,
-  "global_step": 156,
+  "global_step": 434,
   "is_hyper_param_search": false,
   "is_local_process_zero": true,
   "is_world_process_zero": true,
   "log_history": [
     {
-      "epoch": 0.0064,
-      "grad_norm": 1341.948853106591,
-      "learning_rate": 3.125e-09,
-      "logits/chosen": -3.9499800205230713,
-      "logits/rejected": -4.237819194793701,
-      "logps/chosen": -300.693115234375,
-      "logps/rejected": -249.96307373046875,
+      "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,
       "rewards/accuracies": 0.0,
       "rewards/chosen": 0.0,
@@ -24,257 +24,665 @@
       "step": 1
     },
     {
-      "epoch": 0.064,
-      "grad_norm": 1338.549586928997,
-      "learning_rate": 3.125e-08,
-      "logits/chosen": -4.129465579986572,
-      "logits/rejected": -4.351754188537598,
-      "logps/chosen": -351.4585266113281,
-      "logps/rejected": -308.89202880859375,
-      "loss": 0.7278,
-      "rewards/accuracies": 0.375,
-      "rewards/chosen": -0.016052477061748505,
-      "rewards/margins": -0.04994054138660431,
-      "rewards/rejected": 0.033888060599565506,
+      "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,
       "step": 10
     },
     {
-      "epoch": 0.128,
-      "grad_norm": 1285.3502810697903,
-      "learning_rate": 4.9899357349880975e-08,
-      "logits/chosen": -4.195886611938477,
-      "logits/rejected": -4.383217811584473,
-      "logps/chosen": -334.9483337402344,
-      "logps/rejected": -293.82867431640625,
-      "loss": 0.6803,
-      "rewards/accuracies": 0.550000011920929,
-      "rewards/chosen": 0.20189008116722107,
-      "rewards/margins": 0.08841639757156372,
-      "rewards/rejected": 0.11347369104623795,
+      "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,
       "step": 20
     },
     {
-      "epoch": 0.192,
-      "grad_norm": 938.2936106934807,
-      "learning_rate": 4.877641290737884e-08,
-      "logits/chosen": -4.2334418296813965,
-      "logits/rejected": -4.3670477867126465,
-      "logps/chosen": -327.77716064453125,
-      "logps/rejected": -295.27203369140625,
-      "loss": 0.5536,
-      "rewards/accuracies": 0.699999988079071,
-      "rewards/chosen": 0.9395688772201538,
-      "rewards/margins": 0.4487342834472656,
-      "rewards/rejected": 0.49083465337753296,
+      "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,
       "step": 30
     },
     {
-      "epoch": 0.256,
-      "grad_norm": 899.5533271284629,
-      "learning_rate": 4.646121984004665e-08,
-      "logits/chosen": -4.14696741104126,
-      "logits/rejected": -4.348996162414551,
-      "logps/chosen": -330.3467712402344,
-      "logps/rejected": -288.37066650390625,
-      "loss": 0.4332,
-      "rewards/accuracies": 0.8218749761581421,
-      "rewards/chosen": 1.8148998022079468,
-      "rewards/margins": 1.0534685850143433,
-      "rewards/rejected": 0.7614310383796692,
+      "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,
       "step": 40
     },
     {
-      "epoch": 0.32,
-      "grad_norm": 737.601892593307,
-      "learning_rate": 4.3069871595684784e-08,
-      "logits/chosen": -4.238839149475098,
-      "logits/rejected": -4.417731285095215,
-      "logps/chosen": -329.99560546875,
-      "logps/rejected": -291.286865234375,
-      "loss": 0.3783,
-      "rewards/accuracies": 0.8531249761581421,
-      "rewards/chosen": 2.4285898208618164,
-      "rewards/margins": 1.3909423351287842,
-      "rewards/rejected": 1.0376476049423218,
+      "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,
       "step": 50
     },
     {
-      "epoch": 0.384,
-      "grad_norm": 725.6613657693058,
-      "learning_rate": 3.8772424536302564e-08,
-      "logits/chosen": -4.256905555725098,
-      "logits/rejected": -4.42824125289917,
-      "logps/chosen": -321.1641845703125,
-      "logps/rejected": -291.2255859375,
-      "loss": 0.3554,
-      "rewards/accuracies": 0.846875011920929,
-      "rewards/chosen": 2.799910306930542,
-      "rewards/margins": 1.64876389503479,
-      "rewards/rejected": 1.151146411895752,
+      "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,
       "step": 60
     },
     {
-      "epoch": 0.448,
-      "grad_norm": 650.256008502672,
-      "learning_rate": 3.378437060203357e-08,
-      "logits/chosen": -4.1877875328063965,
-      "logits/rejected": -4.377414703369141,
-      "logps/chosen": -320.77117919921875,
-      "logps/rejected": -288.512451171875,
-      "loss": 0.3377,
-      "rewards/accuracies": 0.7906249761581421,
-      "rewards/chosen": 3.0348763465881348,
-      "rewards/margins": 1.8613855838775635,
-      "rewards/rejected": 1.1734905242919922,
+      "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,
       "step": 70
     },
     {
-      "epoch": 0.512,
-      "grad_norm": 643.4652639779088,
-      "learning_rate": 2.8355831645441387e-08,
-      "logits/chosen": -4.054490089416504,
-      "logits/rejected": -4.3433709144592285,
-      "logps/chosen": -346.53253173828125,
-      "logps/rejected": -307.34930419921875,
-      "loss": 0.3369,
-      "rewards/accuracies": 0.8656250238418579,
-      "rewards/chosen": 3.375593662261963,
-      "rewards/margins": 2.1429190635681152,
-      "rewards/rejected": 1.2326747179031372,
+      "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,
       "step": 80
     },
     {
-      "epoch": 0.576,
-      "grad_norm": 658.4601656800384,
-      "learning_rate": 2.2759017277414164e-08,
-      "logits/chosen": -4.182233810424805,
-      "logits/rejected": -4.391860485076904,
-      "logps/chosen": -333.4163513183594,
-      "logps/rejected": -295.03515625,
-      "loss": 0.3301,
-      "rewards/accuracies": 0.859375,
-      "rewards/chosen": 2.9584574699401855,
-      "rewards/margins": 1.9474881887435913,
-      "rewards/rejected": 1.0109691619873047,
+      "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,
       "step": 90
     },
     {
-      "epoch": 0.64,
-      "grad_norm": 716.6222925151064,
-      "learning_rate": 1.7274575140626317e-08,
-      "logits/chosen": -4.168593406677246,
-      "logits/rejected": -4.387479305267334,
-      "logps/chosen": -330.75775146484375,
-      "logps/rejected": -285.5531311035156,
-      "loss": 0.3315,
-      "rewards/accuracies": 0.862500011920929,
-      "rewards/chosen": 3.267747163772583,
-      "rewards/margins": 2.245748281478882,
-      "rewards/rejected": 1.0219987630844116,
+      "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,
       "step": 100
     },
     {
-      "epoch": 0.704,
-      "grad_norm": 660.4496184068971,
-      "learning_rate": 1.217751806485235e-08,
-      "logits/chosen": -4.1479597091674805,
-      "logits/rejected": -4.388964653015137,
-      "logps/chosen": -312.3727722167969,
-      "logps/rejected": -276.13372802734375,
-      "loss": 0.3225,
-      "rewards/accuracies": 0.8656250238418579,
-      "rewards/chosen": 3.277716875076294,
-      "rewards/margins": 2.2097737789154053,
-      "rewards/rejected": 1.0679429769515991,
+      "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,
       "step": 110
     },
     {
-      "epoch": 0.768,
-      "grad_norm": 637.7261383555538,
-      "learning_rate": 7.723433775328384e-09,
-      "logits/chosen": -4.145724296569824,
-      "logits/rejected": -4.354761123657227,
-      "logps/chosen": -326.25091552734375,
-      "logps/rejected": -280.4151916503906,
-      "loss": 0.3134,
-      "rewards/accuracies": 0.8531249761581421,
-      "rewards/chosen": 3.3363654613494873,
-      "rewards/margins": 2.202853202819824,
-      "rewards/rejected": 1.1335121393203735,
+      "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,
       "step": 120
     },
     {
-      "epoch": 0.832,
-      "grad_norm": 768.5710075733069,
-      "learning_rate": 4.135668656967433e-09,
-      "logits/chosen": -4.2324018478393555,
-      "logits/rejected": -4.393607139587402,
-      "logps/chosen": -331.71051025390625,
-      "logps/rejected": -286.67938232421875,
-      "loss": 0.3203,
-      "rewards/accuracies": 0.875,
-      "rewards/chosen": 3.384364604949951,
-      "rewards/margins": 2.2786831855773926,
-      "rewards/rejected": 1.1056816577911377,
+      "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,
       "step": 130
     },
     {
-      "epoch": 0.896,
-      "grad_norm": 742.2155895768275,
-      "learning_rate": 1.5941282340065698e-09,
-      "logits/chosen": -4.18549919128418,
-      "logits/rejected": -4.4001359939575195,
-      "logps/chosen": -333.18035888671875,
-      "logps/rejected": -303.45220947265625,
-      "loss": 0.3282,
-      "rewards/accuracies": 0.84375,
-      "rewards/chosen": 3.2540221214294434,
-      "rewards/margins": 2.2057249546051025,
-      "rewards/rejected": 1.048297643661499,
+      "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,
       "step": 140
     },
     {
-      "epoch": 0.96,
-      "grad_norm": 668.7483012920786,
-      "learning_rate": 2.262559558016325e-10,
-      "logits/chosen": -4.1221604347229,
-      "logits/rejected": -4.351316452026367,
-      "logps/chosen": -339.6897888183594,
-      "logps/rejected": -295.03692626953125,
-      "loss": 0.3208,
-      "rewards/accuracies": 0.859375,
-      "rewards/chosen": 3.4081547260284424,
-      "rewards/margins": 2.2413132190704346,
-      "rewards/rejected": 1.1668416261672974,
+      "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,
       "step": 150
     },
     {
-      "epoch": 0.9984,
-      "step": 156,
+      "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,
       "total_flos": 0.0,
-      "train_loss": 0.4014626894241724,
-      "train_runtime": 5052.0322,
-      "train_samples_per_second": 7.906,
-      "train_steps_per_second": 0.031
+      "train_loss": 0.4889662015822626,
+      "train_runtime": 12705.4599,
+      "train_samples_per_second": 8.747,
+      "train_steps_per_second": 0.034
     }
   ],
   "logging_steps": 10,
-  "max_steps": 156,
+  "max_steps": 434,
   "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": 8,
   "trial_name": null,

training_args.bin

@@ -1,3 +1,3 @@
 version https://git-lfs.github.com/spec/v1
-oid sha256:2d92a249952e57df29e57ec30d0088226bf91a9212b8928552a333e4b9069864
-size 6520
+oid sha256:b92372e760a727f6582979303a96b389cee04d00a2f4a6c534c36c4b19bfdc3f
+size 6264