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
 
@@ -32,7 +32,7 @@ More information needed
 ### Training hyperparameters
 
 The following hyperparameters were used during training:
-- learning_rate: 1e-07
+- learning_rate: 1e-08
 - train_batch_size: 8
 - eval_batch_size: 8
 - seed: 42
@@ -44,7 +44,7 @@ The following hyperparameters were used during training:
 - optimizer: Adam with betas=(0.9,0.999) and epsilon=1e-08
 - lr_scheduler_type: cosine
 - lr_scheduler_warmup_ratio: 0.1
-- num_epochs: 1
+- num_epochs: 2
 
 ### Training results
 
@@ -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": 0.9972602739726028,
-    "total_flos": 0.0,
-    "train_loss": 0.41821049595927146,
-    "train_runtime": 5820.0845,
+    "epoch": 1.99,
+    "train_loss": 0.5661277400923299,
+    "train_runtime": 10889.322,
     "train_samples": 46672,
-    "train_samples_per_second": 8.019,
-    "train_steps_per_second": 0.031
+    "train_samples_per_second": 8.572,
+    "train_steps_per_second": 0.033
 }

config.json

@@ -20,7 +20,7 @@
   "sliding_window": 4096,
   "tie_word_embeddings": false,
   "torch_dtype": "bfloat16",
-  "transformers_version": "4.41.1",
+  "transformers_version": "4.39.3",
   "use_cache": false,
   "vocab_size": 32000
 }

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 sha256:7290f1e2c26695e1d8c538d9fe41285a5e52f43708406dcae0878541a4f6d105
+oid sha256:272835aec6a939eddb8f3c6761a09722f8293bf2e2879f737a53fca1bdafb26a
 size 4943162336

model-00002-of-00003.safetensors

@@ -1,3 +1,3 @@
 version https://git-lfs.github.com/spec/v1
-oid sha256:55ca4a716c63ab1b6533bd8ba1f04ef1b977d004fe97544fdad8b7fcb3020a04
+oid sha256:42495f2a7c8291991dc1126b3da9fe9851e33187e0aaaeef6cdd654ba5d4a54c
 size 4999819336

model-00003-of-00003.safetensors

@@ -1,3 +1,3 @@
 version https://git-lfs.github.com/spec/v1
-oid sha256:c07c8434b14ba8d74fb0863d0c4cafada3c494cf0444a8a71edfb894b7e87540
+oid sha256:28778bd034ca6343f4ed1bdeedff7d601a33547dcaa0c4f056c9501d69a1ad0c
 size 4540516344

runs/Jun18_02-21-39_n136-112-146/events.out.tfevents.1718649051.n136-112-146.1324127.0

@@ -1,3 +1,3 @@
 version https://git-lfs.github.com/spec/v1
-oid sha256:e33fb55159c48dd57933203bc5c1b84424e7200d6e7c29ab81f607cd58c2d34a
-size 26083
+oid sha256:9aaf65a375a0fcc4f5e04faacf884f629c130c493ea7c1b858876c4aff51ff1b
+size 30565

tokenizer.json

@@ -134,7 +134,6 @@
     "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": 0.9972602739726028,
-    "total_flos": 0.0,
-    "train_loss": 0.41821049595927146,
-    "train_runtime": 5820.0845,
+    "epoch": 1.99,
+    "train_loss": 0.5661277400923299,
+    "train_runtime": 10889.322,
     "train_samples": 46672,
-    "train_samples_per_second": 8.019,
-    "train_steps_per_second": 0.031
+    "train_samples_per_second": 8.572,
+    "train_steps_per_second": 0.033
 }

trainer_state.json

@@ -1,17 +1,17 @@
 {
   "best_metric": null,
   "best_model_checkpoint": null,
-  "epoch": 0.9972602739726028,
+  "epoch": 1.9945205479452055,
   "eval_steps": 500,
-  "global_step": 182,
+  "global_step": 364,
   "is_hyper_param_search": false,
   "is_local_process_zero": true,
   "is_world_process_zero": true,
   "log_history": [
     {
-      "epoch": 0.005479452054794521,
-      "grad_norm": 749.8344019634822,
-      "learning_rate": 5.263157894736841e-09,
+      "epoch": 0.01,
+      "grad_norm": 749.8044037682902,
+      "learning_rate": 2.702702702702703e-10,
       "logits/chosen": -0.7030794620513916,
       "logits/rejected": -0.3951629400253296,
       "logps/chosen": -341.73382568359375,
@@ -24,302 +24,560 @@
       "step": 1
     },
     {
-      "epoch": 0.0547945205479452,
-      "grad_norm": 712.5612854291265,
-      "learning_rate": 5.2631578947368416e-08,
-      "logits/chosen": -0.8543994426727295,
-      "logits/rejected": -0.5004442930221558,
-      "logps/chosen": -343.0475158691406,
-      "logps/rejected": -297.69866943359375,
-      "loss": 0.7047,
-      "rewards/accuracies": 0.4652777910232544,
-      "rewards/chosen": 0.028890050947666168,
-      "rewards/margins": 0.015952428802847862,
-      "rewards/rejected": 0.012937622144818306,
+      "epoch": 0.05,
+      "grad_norm": 728.4025747943551,
+      "learning_rate": 2.702702702702703e-09,
+      "logits/chosen": -0.8550878167152405,
+      "logits/rejected": -0.5013476610183716,
+      "logps/chosen": -343.0583801269531,
+      "logps/rejected": -297.6263427734375,
+      "loss": 0.7129,
+      "rewards/accuracies": 0.40625,
+      "rewards/chosen": 0.02346261963248253,
+      "rewards/margins": -0.025641746819019318,
+      "rewards/rejected": 0.049104366451501846,
       "step": 10
     },
     {
-      "epoch": 0.1095890410958904,
-      "grad_norm": 633.132107951183,
-      "learning_rate": 9.999071352056673e-08,
-      "logits/chosen": -0.8784568905830383,
-      "logits/rejected": -0.5061537027359009,
-      "logps/chosen": -333.73223876953125,
-      "logps/rejected": -288.03704833984375,
-      "loss": 0.6532,
-      "rewards/accuracies": 0.59375,
-      "rewards/chosen": -0.058163970708847046,
-      "rewards/margins": 0.1349145770072937,
-      "rewards/rejected": -0.19307854771614075,
+      "epoch": 0.11,
+      "grad_norm": 769.1100121309379,
+      "learning_rate": 5.405405405405406e-09,
+      "logits/chosen": -0.89165860414505,
+      "logits/rejected": -0.5214177966117859,
+      "logps/chosen": -333.6175231933594,
+      "logps/rejected": -287.69482421875,
+      "loss": 0.7088,
+      "rewards/accuracies": 0.512499988079071,
+      "rewards/chosen": -0.0008101940038613975,
+      "rewards/margins": 0.021183893084526062,
+      "rewards/rejected": -0.02199408784508705,
       "step": 20
     },
     {
-      "epoch": 0.1643835616438356,
-      "grad_norm": 498.73085461545014,
-      "learning_rate": 9.888050389939171e-08,
-      "logits/chosen": -0.9048949480056763,
-      "logits/rejected": -0.5001607537269592,
-      "logps/chosen": -331.8179626464844,
-      "logps/rejected": -290.59527587890625,
-      "loss": 0.5232,
-      "rewards/accuracies": 0.7593749761581421,
-      "rewards/chosen": 0.2456372082233429,
-      "rewards/margins": 0.7688120603561401,
-      "rewards/rejected": -0.5231748819351196,
+      "epoch": 0.16,
+      "grad_norm": 782.2455729371497,
+      "learning_rate": 8.108108108108109e-09,
+      "logits/chosen": -0.9365941882133484,
+      "logits/rejected": -0.5470438599586487,
+      "logps/chosen": -332.33966064453125,
+      "logps/rejected": -289.5589904785156,
+      "loss": 0.7083,
+      "rewards/accuracies": 0.5,
+      "rewards/chosen": -0.015217331238090992,
+      "rewards/margins": -0.010174614377319813,
+      "rewards/rejected": -0.005042716860771179,
       "step": 30
     },
     {
-      "epoch": 0.2191780821917808,
-      "grad_norm": 468.38328488591065,
-      "learning_rate": 9.596014080776422e-08,
-      "logits/chosen": -0.8753167986869812,
-      "logits/rejected": -0.5111785531044006,
-      "logps/chosen": -335.57745361328125,
-      "logps/rejected": -306.9455261230469,
-      "loss": 0.4522,
-      "rewards/accuracies": 0.8218749761581421,
-      "rewards/chosen": 1.0787924528121948,
-      "rewards/margins": 1.3148835897445679,
-      "rewards/rejected": -0.23609113693237305,
+      "epoch": 0.22,
+      "grad_norm": 736.3539975893915,
+      "learning_rate": 9.997923381619256e-09,
+      "logits/chosen": -0.9017000198364258,
+      "logits/rejected": -0.555226743221283,
+      "logps/chosen": -337.62115478515625,
+      "logps/rejected": -306.5302429199219,
+      "loss": 0.7072,
+      "rewards/accuracies": 0.5687500238418579,
+      "rewards/chosen": 0.05695560574531555,
+      "rewards/margins": 0.08540164679288864,
+      "rewards/rejected": -0.02844604291021824,
       "step": 40
     },
     {
-      "epoch": 0.273972602739726,
-      "grad_norm": 405.49075569087563,
-      "learning_rate": 9.133777176798013e-08,
-      "logits/chosen": -0.9362503886222839,
-      "logits/rejected": -0.493760883808136,
-      "logps/chosen": -329.0995178222656,
-      "logps/rejected": -294.7619323730469,
-      "loss": 0.4065,
-      "rewards/accuracies": 0.8374999761581421,
-      "rewards/chosen": 2.0607025623321533,
-      "rewards/margins": 1.7504669427871704,
-      "rewards/rejected": 0.3102358281612396,
+      "epoch": 0.27,
+      "grad_norm": 730.7089871756754,
+      "learning_rate": 9.961053687802851e-09,
+      "logits/chosen": -0.9294391870498657,
+      "logits/rejected": -0.5007705688476562,
+      "logps/chosen": -333.2597961425781,
+      "logps/rejected": -295.4553527832031,
+      "loss": 0.6941,
+      "rewards/accuracies": 0.53125,
+      "rewards/chosen": -0.01942090317606926,
+      "rewards/margins": 0.017053481191396713,
+      "rewards/rejected": -0.03647438436746597,
       "step": 50
     },
     {
-      "epoch": 0.3287671232876712,
-      "grad_norm": 531.0849294440985,
-      "learning_rate": 8.518457335743925e-08,
-      "logits/chosen": -0.8332949876785278,
-      "logits/rejected": -0.44049328565597534,
-      "logps/chosen": -328.05059814453125,
-      "logps/rejected": -295.0623474121094,
-      "loss": 0.397,
-      "rewards/accuracies": 0.824999988079071,
-      "rewards/chosen": 1.8471581935882568,
-      "rewards/margins": 1.8996388912200928,
-      "rewards/rejected": -0.05248071998357773,
+      "epoch": 0.33,
+      "grad_norm": 730.7067078437899,
+      "learning_rate": 9.878428410862483e-09,
+      "logits/chosen": -0.8303865194320679,
+      "logits/rejected": -0.45489081740379333,
+      "logps/chosen": -331.8045349121094,
+      "logps/rejected": -295.13055419921875,
+      "loss": 0.6832,
+      "rewards/accuracies": 0.6031249761581421,
+      "rewards/chosen": -0.029834549874067307,
+      "rewards/margins": 0.05674406886100769,
+      "rewards/rejected": -0.0865786224603653,
       "step": 60
     },
     {
-      "epoch": 0.3835616438356164,
-      "grad_norm": 410.01119928479034,
-      "learning_rate": 7.772841216033532e-08,
-      "logits/chosen": -1.053856611251831,
-      "logits/rejected": -0.6151344776153564,
-      "logps/chosen": -347.8240966796875,
-      "logps/rejected": -304.7304382324219,
-      "loss": 0.3992,
-      "rewards/accuracies": 0.793749988079071,
-      "rewards/chosen": 1.6472972631454468,
-      "rewards/margins": 1.7940982580184937,
-      "rewards/rejected": -0.14680106937885284,
+      "epoch": 0.38,
+      "grad_norm": 674.6624722402527,
+      "learning_rate": 9.750809600145954e-09,
+      "logits/chosen": -1.0535982847213745,
+      "logits/rejected": -0.6284598112106323,
+      "logps/chosen": -351.1028747558594,
+      "logps/rejected": -304.6434631347656,
+      "loss": 0.6804,
+      "rewards/accuracies": 0.625,
+      "rewards/chosen": 0.0078976359218359,
+      "rewards/margins": 0.11121608316898346,
+      "rewards/rejected": -0.10331843793392181,
       "step": 70
     },
     {
-      "epoch": 0.4383561643835616,
-      "grad_norm": 422.6505304779972,
-      "learning_rate": 6.924540636266271e-08,
-      "logits/chosen": -0.9396562576293945,
-      "logits/rejected": -0.3743361532688141,
-      "logps/chosen": -341.07342529296875,
-      "logps/rejected": -294.79632568359375,
-      "loss": 0.3788,
-      "rewards/accuracies": 0.8374999761581421,
-      "rewards/chosen": 1.4693505764007568,
-      "rewards/margins": 1.931475043296814,
-      "rewards/rejected": -0.46212440729141235,
+      "epoch": 0.44,
+      "grad_norm": 640.6205441818687,
+      "learning_rate": 9.579374278412817e-09,
+      "logits/chosen": -0.9343639612197876,
+      "logits/rejected": -0.3806554079055786,
+      "logps/chosen": -344.04443359375,
+      "logps/rejected": -294.193359375,
+      "loss": 0.6487,
+      "rewards/accuracies": 0.659375011920929,
+      "rewards/chosen": -0.01617306098341942,
+      "rewards/margins": 0.14446857571601868,
+      "rewards/rejected": -0.1606416255235672,
       "step": 80
     },
     {
-      "epoch": 0.4931506849315068,
-      "grad_norm": 435.4943837818754,
-      "learning_rate": 6.004970048339225e-08,
-      "logits/chosen": -1.0359487533569336,
-      "logits/rejected": -0.6516568064689636,
-      "logps/chosen": -337.1309509277344,
-      "logps/rejected": -307.5055847167969,
-      "loss": 0.3792,
-      "rewards/accuracies": 0.8218749761581421,
-      "rewards/chosen": 1.8926198482513428,
-      "rewards/margins": 1.9282394647598267,
-      "rewards/rejected": -0.03561948612332344,
+      "epoch": 0.49,
+      "grad_norm": 629.6802089882239,
+      "learning_rate": 9.365703586204494e-09,
+      "logits/chosen": -1.0297319889068604,
+      "logits/rejected": -0.654572606086731,
+      "logps/chosen": -340.8974609375,
+      "logps/rejected": -307.7338562011719,
+      "loss": 0.639,
+      "rewards/accuracies": 0.621874988079071,
+      "rewards/chosen": 0.009362801909446716,
+      "rewards/margins": 0.15911515057086945,
+      "rewards/rejected": -0.14975234866142273,
       "step": 90
     },
     {
-      "epoch": 0.547945205479452,
-      "grad_norm": 424.3608116994891,
-      "learning_rate": 5.0481831906199036e-08,
-      "logits/chosen": -0.8561469316482544,
-      "logits/rejected": -0.5456556081771851,
-      "logps/chosen": -315.7440185546875,
-      "logps/rejected": -284.4901428222656,
-      "loss": 0.3607,
-      "rewards/accuracies": 0.796875,
-      "rewards/chosen": 1.5119119882583618,
-      "rewards/margins": 1.8389956951141357,
-      "rewards/rejected": -0.3270837962627411,
+      "epoch": 0.55,
+      "grad_norm": 654.9042346572915,
+      "learning_rate": 9.111768199053588e-09,
+      "logits/chosen": -0.8516262173652649,
+      "logits/rejected": -0.5497715473175049,
+      "logps/chosen": -318.9173889160156,
+      "logps/rejected": -284.2920837402344,
+      "loss": 0.6303,
+      "rewards/accuracies": 0.609375,
+      "rewards/chosen": -0.07476776093244553,
+      "rewards/margins": 0.15329131484031677,
+      "rewards/rejected": -0.2280590981245041,
       "step": 100
     },
     {
-      "epoch": 0.6027397260273972,
-      "grad_norm": 376.9480565248481,
-      "learning_rate": 4.0896120024874283e-08,
-      "logits/chosen": -0.7713192105293274,
-      "logits/rejected": -0.4743874967098236,
-      "logps/chosen": -332.10638427734375,
-      "logps/rejected": -302.2787780761719,
-      "loss": 0.3675,
-      "rewards/accuracies": 0.831250011920929,
-      "rewards/chosen": 1.8271310329437256,
-      "rewards/margins": 1.9777942895889282,
-      "rewards/rejected": -0.15066327154636383,
+      "epoch": 0.6,
+      "grad_norm": 568.2894000484466,
+      "learning_rate": 8.819910152028871e-09,
+      "logits/chosen": -0.7658065557479858,
+      "logits/rejected": -0.4776793122291565,
+      "logps/chosen": -335.7439270019531,
+      "logps/rejected": -302.370849609375,
+      "loss": 0.6175,
+      "rewards/accuracies": 0.640625,
+      "rewards/chosen": 0.008360857143998146,
+      "rewards/margins": 0.20505543053150177,
+      "rewards/rejected": -0.19669455289840698,
       "step": 110
     },
     {
-      "epoch": 0.6575342465753424,
-      "grad_norm": 452.4360162863762,
-      "learning_rate": 3.1647545010335394e-08,
-      "logits/chosen": -0.8479810953140259,
-      "logits/rejected": -0.47955965995788574,
-      "logps/chosen": -345.54071044921875,
-      "logps/rejected": -308.65179443359375,
-      "loss": 0.3527,
-      "rewards/accuracies": 0.840624988079071,
-      "rewards/chosen": 1.8864860534667969,
-      "rewards/margins": 2.1840381622314453,
-      "rewards/rejected": -0.2975522577762604,
+      "epoch": 0.66,
+      "grad_norm": 569.4638082407581,
+      "learning_rate": 8.492821239247364e-09,
+      "logits/chosen": -0.851113498210907,
+      "logits/rejected": -0.49246683716773987,
+      "logps/chosen": -349.08477783203125,
+      "logps/rejected": -308.5341491699219,
+      "loss": 0.5935,
+      "rewards/accuracies": 0.7437499761581421,
+      "rewards/chosen": 0.11443378776311874,
+      "rewards/margins": 0.3531562387943268,
+      "rewards/rejected": -0.23872247338294983,
       "step": 120
     },
     {
-      "epoch": 0.7123287671232876,
-      "grad_norm": 315.5021095438557,
-      "learning_rate": 2.3078602107593897e-08,
-      "logits/chosen": -0.8250066041946411,
-      "logits/rejected": -0.49061331152915955,
-      "logps/chosen": -325.8185119628906,
-      "logps/rejected": -298.6694030761719,
-      "loss": 0.3638,
-      "rewards/accuracies": 0.7906249761581421,
-      "rewards/chosen": 1.3761605024337769,
-      "rewards/margins": 1.7671406269073486,
-      "rewards/rejected": -0.3909802734851837,
+      "epoch": 0.71,
+      "grad_norm": 581.8336234971125,
+      "learning_rate": 8.133518187573863e-09,
+      "logits/chosen": -0.8286861181259155,
+      "logits/rejected": -0.5028008818626404,
+      "logps/chosen": -328.44915771484375,
+      "logps/rejected": -298.2557373046875,
+      "loss": 0.5959,
+      "rewards/accuracies": 0.643750011920929,
+      "rewards/chosen": 0.06084731966257095,
+      "rewards/margins": 0.24498350918293,
+      "rewards/rejected": -0.18413618206977844,
       "step": 130
     },
     {
-      "epoch": 0.7671232876712328,
-      "grad_norm": 420.91275678465917,
-      "learning_rate": 1.5506618277219406e-08,
-      "logits/chosen": -1.0089247226715088,
-      "logits/rejected": -0.5239464044570923,
-      "logps/chosen": -334.70111083984375,
-      "logps/rejected": -293.3878479003906,
-      "loss": 0.3607,
-      "rewards/accuracies": 0.8343750238418579,
-      "rewards/chosen": 1.8048830032348633,
-      "rewards/margins": 2.049487829208374,
-      "rewards/rejected": -0.2446049153804779,
+      "epoch": 0.77,
+      "grad_norm": 566.985090918969,
+      "learning_rate": 7.745314833479833e-09,
+      "logits/chosen": -1.0121524333953857,
+      "logits/rejected": -0.5360809564590454,
+      "logps/chosen": -338.01531982421875,
+      "logps/rejected": -293.3200378417969,
+      "loss": 0.5864,
+      "rewards/accuracies": 0.7281249761581421,
+      "rewards/chosen": 0.14776429533958435,
+      "rewards/margins": 0.35846027731895447,
+      "rewards/rejected": -0.21069595217704773,
       "step": 140
     },
     {
-      "epoch": 0.821917808219178,
-      "grad_norm": 450.70186106616944,
-      "learning_rate": 9.212000874196951e-09,
-      "logits/chosen": -0.9020761251449585,
-      "logits/rejected": -0.5833565592765808,
-      "logps/chosen": -328.1164245605469,
-      "logps/rejected": -298.4231872558594,
-      "loss": 0.349,
-      "rewards/accuracies": 0.84375,
-      "rewards/chosen": 1.6165488958358765,
-      "rewards/margins": 1.8227107524871826,
-      "rewards/rejected": -0.20616202056407928,
+      "epoch": 0.82,
+      "grad_norm": 585.5307136387664,
+      "learning_rate": 7.33179155967327e-09,
+      "logits/chosen": -0.9024865031242371,
+      "logits/rejected": -0.5908719301223755,
+      "logps/chosen": -331.0682373046875,
+      "logps/rejected": -298.38677978515625,
+      "loss": 0.5718,
+      "rewards/accuracies": 0.675000011920929,
+      "rewards/chosen": 0.1406635046005249,
+      "rewards/margins": 0.32861852645874023,
+      "rewards/rejected": -0.18795502185821533,
       "step": 150
     },
     {
-      "epoch": 0.8767123287671232,
-      "grad_norm": 399.4981855057766,
-      "learning_rate": 4.427853541662091e-09,
-      "logits/chosen": -0.9409669041633606,
-      "logits/rejected": -0.5335644483566284,
-      "logps/chosen": -337.31488037109375,
-      "logps/rejected": -302.06365966796875,
-      "loss": 0.3548,
-      "rewards/accuracies": 0.846875011920929,
-      "rewards/chosen": 1.994319200515747,
-      "rewards/margins": 2.289970636367798,
-      "rewards/rejected": -0.29565146565437317,
+      "epoch": 0.88,
+      "grad_norm": 608.4424256652742,
+      "learning_rate": 6.896762273384178e-09,
+      "logits/chosen": -0.9404014348983765,
+      "logits/rejected": -0.5414465069770813,
+      "logps/chosen": -340.7105407714844,
+      "logps/rejected": -301.7948303222656,
+      "loss": 0.5625,
+      "rewards/accuracies": 0.7406250238418579,
+      "rewards/chosen": 0.29647737741470337,
+      "rewards/margins": 0.45771312713623047,
+      "rewards/rejected": -0.1612357348203659,
       "step": 160
     },
     {
-      "epoch": 0.9315068493150684,
-      "grad_norm": 381.8223301695622,
-      "learning_rate": 1.3313438659999398e-09,
-      "logits/chosen": -0.6689538955688477,
-      "logits/rejected": -0.37201908230781555,
-      "logps/chosen": -323.4323425292969,
-      "logps/rejected": -292.47857666015625,
-      "loss": 0.3629,
-      "rewards/accuracies": 0.840624988079071,
-      "rewards/chosen": 2.0121166706085205,
-      "rewards/margins": 2.282526731491089,
-      "rewards/rejected": -0.2704099416732788,
+      "epoch": 0.93,
+      "grad_norm": 571.5105228832468,
+      "learning_rate": 6.444239230863505e-09,
+      "logits/chosen": -0.6664124131202698,
+      "logits/rejected": -0.3767244815826416,
+      "logps/chosen": -326.7174377441406,
+      "logps/rejected": -292.18304443359375,
+      "loss": 0.5516,
+      "rewards/accuracies": 0.75,
+      "rewards/chosen": 0.3695653975009918,
+      "rewards/margins": 0.49222081899642944,
+      "rewards/rejected": -0.12265541404485703,
       "step": 170
     },
     {
-      "epoch": 0.9863013698630136,
-      "grad_norm": 380.7644210750603,
-      "learning_rate": 3.7142468185014095e-11,
-      "logits/chosen": -0.8619197010993958,
-      "logits/rejected": -0.5339924097061157,
-      "logps/chosen": -319.94158935546875,
-      "logps/rejected": -290.5792236328125,
-      "loss": 0.367,
-      "rewards/accuracies": 0.8500000238418579,
-      "rewards/chosen": 1.9236304759979248,
-      "rewards/margins": 2.2062630653381348,
-      "rewards/rejected": -0.2826324701309204,
+      "epoch": 0.99,
+      "grad_norm": 486.04829364539904,
+      "learning_rate": 5.97839603251764e-09,
+      "logits/chosen": -0.8600166440010071,
+      "logits/rejected": -0.5389689803123474,
+      "logps/chosen": -323.1247863769531,
+      "logps/rejected": -290.40802001953125,
+      "loss": 0.5408,
+      "rewards/accuracies": 0.784375011920929,
+      "rewards/chosen": 0.3320290446281433,
+      "rewards/margins": 0.5290471315383911,
+      "rewards/rejected": -0.19701813161373138,
       "step": 180
     },
     {
-      "epoch": 0.9972602739726028,
-      "step": 182,
+      "epoch": 1.04,
+      "grad_norm": 524.5665762946036,
+      "learning_rate": 5.5035291299727915e-09,
+      "logits/chosen": -1.0205037593841553,
+      "logits/rejected": -0.6305373907089233,
+      "logps/chosen": -351.1410827636719,
+      "logps/rejected": -313.13812255859375,
+      "loss": 0.5258,
+      "rewards/accuracies": 0.731249988079071,
+      "rewards/chosen": 0.40980952978134155,
+      "rewards/margins": 0.5211332440376282,
+      "rewards/rejected": -0.11132367700338364,
+      "step": 190
+    },
+    {
+      "epoch": 1.1,
+      "grad_norm": 474.67341086960977,
+      "learning_rate": 5.024018200087854e-09,
+      "logits/chosen": -0.836185097694397,
+      "logits/rejected": -0.33586248755455017,
+      "logps/chosen": -321.29168701171875,
+      "logps/rejected": -280.28753662109375,
+      "loss": 0.512,
+      "rewards/accuracies": 0.7562500238418579,
+      "rewards/chosen": 0.3657322824001312,
+      "rewards/margins": 0.5857880711555481,
+      "rewards/rejected": -0.22005578875541687,
+      "step": 200
+    },
+    {
+      "epoch": 1.15,
+      "grad_norm": 480.69197567309544,
+      "learning_rate": 4.544285751384584e-09,
+      "logits/chosen": -0.9048161506652832,
+      "logits/rejected": -0.6336894631385803,
+      "logps/chosen": -329.5954895019531,
+      "logps/rejected": -297.79901123046875,
+      "loss": 0.5067,
+      "rewards/accuracies": 0.721875011920929,
+      "rewards/chosen": 0.41357460618019104,
+      "rewards/margins": 0.5688098669052124,
+      "rewards/rejected": -0.1552352011203766,
+      "step": 210
+    },
+    {
+      "epoch": 1.21,
+      "grad_norm": 498.26717868032813,
+      "learning_rate": 4.068756335443199e-09,
+      "logits/chosen": -0.8500760793685913,
+      "logits/rejected": -0.501125693321228,
+      "logps/chosen": -339.07012939453125,
+      "logps/rejected": -299.48095703125,
+      "loss": 0.514,
+      "rewards/accuracies": 0.7906249761581421,
+      "rewards/chosen": 0.5280696153640747,
+      "rewards/margins": 0.6353160738945007,
+      "rewards/rejected": -0.10724644362926483,
+      "step": 220
+    },
+    {
+      "epoch": 1.26,
+      "grad_norm": 522.8370181213115,
+      "learning_rate": 3.6018157394549284e-09,
+      "logits/chosen": -0.8433824777603149,
+      "logits/rejected": -0.4627400040626526,
+      "logps/chosen": -332.1639709472656,
+      "logps/rejected": -297.5298156738281,
+      "loss": 0.5207,
+      "rewards/accuracies": 0.737500011920929,
+      "rewards/chosen": 0.4683798849582672,
+      "rewards/margins": 0.5844911336898804,
+      "rewards/rejected": -0.11611130088567734,
+      "step": 230
+    },
+    {
+      "epoch": 1.32,
+      "grad_norm": 521.1145144981034,
+      "learning_rate": 3.14777053629687e-09,
+      "logits/chosen": -1.0663576126098633,
+      "logits/rejected": -0.6788166761398315,
+      "logps/chosen": -336.2250671386719,
+      "logps/rejected": -296.2794494628906,
+      "loss": 0.5124,
+      "rewards/accuracies": 0.7593749761581421,
+      "rewards/chosen": 0.5691989660263062,
+      "rewards/margins": 0.702237069606781,
+      "rewards/rejected": -0.13303814828395844,
+      "step": 240
+    },
+    {
+      "epoch": 1.37,
+      "grad_norm": 487.7677359334702,
+      "learning_rate": 2.7108083651970002e-09,
+      "logits/chosen": -0.7713569402694702,
+      "logits/rejected": -0.4026997983455658,
+      "logps/chosen": -327.7779235839844,
+      "logps/rejected": -290.0624084472656,
+      "loss": 0.5045,
+      "rewards/accuracies": 0.7406250238418579,
+      "rewards/chosen": 0.5247091054916382,
+      "rewards/margins": 0.6842475533485413,
+      "rewards/rejected": -0.1595384031534195,
+      "step": 250
+    },
+    {
+      "epoch": 1.42,
+      "grad_norm": 524.4673235373901,
+      "learning_rate": 2.294959309319086e-09,
+      "logits/chosen": -0.8702203035354614,
+      "logits/rejected": -0.6658123135566711,
+      "logps/chosen": -328.04638671875,
+      "logps/rejected": -302.77777099609375,
+      "loss": 0.5103,
+      "rewards/accuracies": 0.737500011920929,
+      "rewards/chosen": 0.6019162535667419,
+      "rewards/margins": 0.5868763327598572,
+      "rewards/rejected": 0.015039998106658459,
+      "step": 260
+    },
+    {
+      "epoch": 1.48,
+      "grad_norm": 553.8708647904535,
+      "learning_rate": 1.9040587264803672e-09,
+      "logits/chosen": -0.9641194343566895,
+      "logits/rejected": -0.4711450934410095,
+      "logps/chosen": -340.99169921875,
+      "logps/rejected": -293.9861145019531,
+      "loss": 0.4939,
+      "rewards/accuracies": 0.831250011920929,
+      "rewards/chosen": 0.56943279504776,
+      "rewards/margins": 0.7665254473686218,
+      "rewards/rejected": -0.1970926821231842,
+      "step": 270
+    },
+    {
+      "epoch": 1.53,
+      "grad_norm": 453.07875520502444,
+      "learning_rate": 1.5417118758126409e-09,
+      "logits/chosen": -0.8815242648124695,
+      "logits/rejected": -0.4348044991493225,
+      "logps/chosen": -342.6517028808594,
+      "logps/rejected": -296.92169189453125,
+      "loss": 0.4934,
+      "rewards/accuracies": 0.7906249761581421,
+      "rewards/chosen": 0.5705780982971191,
+      "rewards/margins": 0.8008774518966675,
+      "rewards/rejected": -0.23029938340187073,
+      "step": 280
+    },
+    {
+      "epoch": 1.59,
+      "grad_norm": 499.34891989016415,
+      "learning_rate": 1.2112606666135601e-09,
+      "logits/chosen": -0.9316185712814331,
+      "logits/rejected": -0.5389292240142822,
+      "logps/chosen": -333.5206604003906,
+      "logps/rejected": -297.6076354980469,
+      "loss": 0.4978,
+      "rewards/accuracies": 0.765625,
+      "rewards/chosen": 0.5433750748634338,
+      "rewards/margins": 0.6790112257003784,
+      "rewards/rejected": -0.135636106133461,
+      "step": 290
+    },
+    {
+      "epoch": 1.64,
+      "grad_norm": 475.94785473220173,
+      "learning_rate": 9.157528360620415e-10,
+      "logits/chosen": -0.8223875164985657,
+      "logits/rejected": -0.5513488054275513,
+      "logps/chosen": -312.7162780761719,
+      "logps/rejected": -280.6097106933594,
+      "loss": 0.4933,
+      "rewards/accuracies": 0.7593749761581421,
+      "rewards/chosen": 0.5738901495933533,
+      "rewards/margins": 0.6515259146690369,
+      "rewards/rejected": -0.07763569802045822,
+      "step": 300
+    },
+    {
+      "epoch": 1.7,
+      "grad_norm": 530.0260445258433,
+      "learning_rate": 6.579138400703716e-10,
+      "logits/chosen": -0.8587888479232788,
+      "logits/rejected": -0.4476223587989807,
+      "logps/chosen": -334.20648193359375,
+      "logps/rejected": -295.112548828125,
+      "loss": 0.4913,
+      "rewards/accuracies": 0.7906249761581421,
+      "rewards/chosen": 0.6548627018928528,
+      "rewards/margins": 0.7542473077774048,
+      "rewards/rejected": -0.09938466548919678,
+      "step": 310
+    },
+    {
+      "epoch": 1.75,
+      "grad_norm": 487.39892256915766,
+      "learning_rate": 4.401217165224564e-10,
+      "logits/chosen": -0.7274507284164429,
+      "logits/rejected": -0.45758286118507385,
+      "logps/chosen": -322.195068359375,
+      "logps/rejected": -289.47064208984375,
+      "loss": 0.4876,
+      "rewards/accuracies": 0.78125,
+      "rewards/chosen": 0.6047518849372864,
+      "rewards/margins": 0.6844498515129089,
+      "rewards/rejected": -0.07969798892736435,
+      "step": 320
+    },
+    {
+      "epoch": 1.81,
+      "grad_norm": 474.27503066220174,
+      "learning_rate": 2.643851527335006e-10,
+      "logits/chosen": -0.9080432653427124,
+      "logits/rejected": -0.5377383232116699,
+      "logps/chosen": -331.7919921875,
+      "logps/rejected": -301.7752685546875,
+      "loss": 0.4856,
+      "rewards/accuracies": 0.7718750238418579,
+      "rewards/chosen": 0.6742402911186218,
+      "rewards/margins": 0.7511230111122131,
+      "rewards/rejected": -0.07688269764184952,
+      "step": 330
+    },
+    {
+      "epoch": 1.86,
+      "grad_norm": 515.521744630777,
+      "learning_rate": 1.3232495941396638e-10,
+      "logits/chosen": -0.9773859977722168,
+      "logits/rejected": -0.6425100564956665,
+      "logps/chosen": -329.1192321777344,
+      "logps/rejected": -293.9158020019531,
+      "loss": 0.4902,
+      "rewards/accuracies": 0.78125,
+      "rewards/chosen": 0.5788453817367554,
+      "rewards/margins": 0.6871310472488403,
+      "rewards/rejected": -0.10828564316034317,
+      "step": 340
+    },
+    {
+      "epoch": 1.92,
+      "grad_norm": 474.472293237774,
+      "learning_rate": 4.515912200264427e-11,
+      "logits/chosen": -0.9029630422592163,
+      "logits/rejected": -0.4482271075248718,
+      "logps/chosen": -331.9292907714844,
+      "logps/rejected": -287.7762451171875,
+      "loss": 0.4845,
+      "rewards/accuracies": 0.7906249761581421,
+      "rewards/chosen": 0.5662914514541626,
+      "rewards/margins": 0.7054949998855591,
+      "rewards/rejected": -0.1392035186290741,
+      "step": 350
+    },
+    {
+      "epoch": 1.97,
+      "grad_norm": 466.2251927040291,
+      "learning_rate": 3.691567239743621e-12,
+      "logits/chosen": -0.866126537322998,
+      "logits/rejected": -0.38964423537254333,
+      "logps/chosen": -346.421875,
+      "logps/rejected": -304.1325988769531,
+      "loss": 0.4772,
+      "rewards/accuracies": 0.8031250238418579,
+      "rewards/chosen": 0.6508310437202454,
+      "rewards/margins": 0.8615090250968933,
+      "rewards/rejected": -0.21067801117897034,
+      "step": 360
+    },
+    {
+      "epoch": 1.99,
+      "step": 364,
       "total_flos": 0.0,
-      "train_loss": 0.41821049595927146,
-      "train_runtime": 5820.0845,
-      "train_samples_per_second": 8.019,
-      "train_steps_per_second": 0.031
+      "train_loss": 0.5661277400923299,
+      "train_runtime": 10889.322,
+      "train_samples_per_second": 8.572,
+      "train_steps_per_second": 0.033
     }
   ],
   "logging_steps": 10,
-  "max_steps": 182,
+  "max_steps": 364,
   "num_input_tokens_seen": 0,
-  "num_train_epochs": 1,
+  "num_train_epochs": 2,
   "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 sha256:91a455adcc1dd46693a6fa3188bc76c135951e0b0806108b876f1cac5c69def0
-size 6520
+oid sha256:fe7ad1d1edbbc30b461a2666bb026f02d74c0c127226e41e03c561f18f18d157
+size 6328