RyanYr commited on
Commit
783a57d
1 Parent(s): b19db67

Training in progress, step 1500, checkpoint

Browse files
last-checkpoint/global_step1500/bf16_zero_pp_rank_0_mp_rank_00_optim_states.pt ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:380c5b3d4a50b2c96e9e9cbb9c39c7ba002cf4734c56dd71ec52a7009ca7d7ab
3
+ size 7843036668
last-checkpoint/global_step1500/bf16_zero_pp_rank_1_mp_rank_00_optim_states.pt ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:09bb2b8e1858310c41a7d6c72ec5c75bdf4dd7060cf6f257dbb00cf1a0a9b1fc
3
+ size 7843043580
last-checkpoint/global_step1500/bf16_zero_pp_rank_2_mp_rank_00_optim_states.pt ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:2bdee7d2a297dca2ab84aca341f8f877f2d914a24734386b5ff6b922a3a6f385
3
+ size 7843043004
last-checkpoint/global_step1500/bf16_zero_pp_rank_3_mp_rank_00_optim_states.pt ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:c91214c04de8b5db131ad5404dcba6e217f724f41d64fb874167ee3d31d9475d
3
+ size 7843043388
last-checkpoint/global_step1500/mp_rank_00_model_states.pt ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:dd4d869fa72945b74786f1315130651eeacb9591d23fe9c00187e45a556fc278
3
+ size 5228775200
last-checkpoint/latest CHANGED
@@ -1 +1 @@
1
- global_step1200
 
1
+ global_step1500
last-checkpoint/model-00001-of-00002.safetensors CHANGED
@@ -1,3 +1,3 @@
1
  version https://git-lfs.github.com/spec/v1
2
- oid sha256:7b5467af04ff29012437a7517c2dc67421fbc898979e0a9158a5699267e14db6
3
  size 4988030368
 
1
  version https://git-lfs.github.com/spec/v1
2
+ oid sha256:617221b5a3979bb5c195e80814b940ad0fd5e4ea46fcf53c004738f7521b9b05
3
  size 4988030368
last-checkpoint/model-00002-of-00002.safetensors CHANGED
@@ -1,3 +1,3 @@
1
  version https://git-lfs.github.com/spec/v1
2
- oid sha256:e12b943204d51eacdf53c41136bae9c021c0b7ac57fce30d459ef73fbfc2983c
3
  size 1420344488
 
1
  version https://git-lfs.github.com/spec/v1
2
+ oid sha256:53b46ffbe9750f068af7cff31ad24813da6ce5bbc66f559f4dcbf3d434d5e8f7
3
  size 1420344488
last-checkpoint/scheduler.pt CHANGED
@@ -1,3 +1,3 @@
1
  version https://git-lfs.github.com/spec/v1
2
- oid sha256:973efa8e69c59defd765e5150ce17bd7b1970481a388708a880f067c876a3880
3
  size 1064
 
1
  version https://git-lfs.github.com/spec/v1
2
+ oid sha256:cb1f8e086c96cde9498cc8372841552a3b3d37b7449d73d2153f92624f5efc96
3
  size 1064
last-checkpoint/trainer_state.json CHANGED
@@ -1,9 +1,9 @@
1
  {
2
  "best_metric": null,
3
  "best_model_checkpoint": null,
4
- "epoch": 1.4931176607823313,
5
  "eval_steps": 999999,
6
- "global_step": 1200,
7
  "is_hyper_param_search": false,
8
  "is_local_process_zero": true,
9
  "is_world_process_zero": true,
@@ -9007,6 +9007,2256 @@
9007
  "rewards/margins": 3.2304723262786865,
9008
  "rewards/rejected": -3.3748857975006104,
9009
  "step": 1200
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
9010
  }
9011
  ],
9012
  "logging_steps": 2,
 
1
  {
2
  "best_metric": null,
3
  "best_model_checkpoint": null,
4
+ "epoch": 1.8663970759779143,
5
  "eval_steps": 999999,
6
+ "global_step": 1500,
7
  "is_hyper_param_search": false,
8
  "is_local_process_zero": true,
9
  "is_world_process_zero": true,
 
9007
  "rewards/margins": 3.2304723262786865,
9008
  "rewards/rejected": -3.3748857975006104,
9009
  "step": 1200
9010
+ },
9011
+ {
9012
+ "epoch": 1.4956061902169686,
9013
+ "grad_norm": 30.88616371154785,
9014
+ "learning_rate": 3.345983350831798e-08,
9015
+ "logits/chosen": -10.752206802368164,
9016
+ "logits/rejected": -10.766036987304688,
9017
+ "logps/chosen": -19.481313705444336,
9018
+ "logps/rejected": -54.68231201171875,
9019
+ "loss": 0.3279,
9020
+ "rewards/accuracies": 0.65625,
9021
+ "rewards/chosen": 0.06634411960840225,
9022
+ "rewards/margins": 3.3881375789642334,
9023
+ "rewards/rejected": -3.321793556213379,
9024
+ "step": 1202
9025
+ },
9026
+ {
9027
+ "epoch": 1.498094719651606,
9028
+ "grad_norm": 4.875926971435547,
9029
+ "learning_rate": 3.3148972168516734e-08,
9030
+ "logits/chosen": -10.793999671936035,
9031
+ "logits/rejected": -10.790190696716309,
9032
+ "logps/chosen": -15.71826457977295,
9033
+ "logps/rejected": -53.232261657714844,
9034
+ "loss": 0.3231,
9035
+ "rewards/accuracies": 0.59375,
9036
+ "rewards/chosen": 0.3771316409111023,
9037
+ "rewards/margins": 3.606487512588501,
9038
+ "rewards/rejected": -3.229356050491333,
9039
+ "step": 1204
9040
+ },
9041
+ {
9042
+ "epoch": 1.500583249086243,
9043
+ "grad_norm": 1.0279144048690796,
9044
+ "learning_rate": 3.2839274464991854e-08,
9045
+ "logits/chosen": -10.80120849609375,
9046
+ "logits/rejected": -10.804399490356445,
9047
+ "logps/chosen": -19.204309463500977,
9048
+ "logps/rejected": -54.1872673034668,
9049
+ "loss": 0.387,
9050
+ "rewards/accuracies": 0.65625,
9051
+ "rewards/chosen": 0.022899247705936432,
9052
+ "rewards/margins": 3.3754444122314453,
9053
+ "rewards/rejected": -3.3525447845458984,
9054
+ "step": 1206
9055
+ },
9056
+ {
9057
+ "epoch": 1.5030717785208805,
9058
+ "grad_norm": 9.367308616638184,
9059
+ "learning_rate": 3.253074578846805e-08,
9060
+ "logits/chosen": -10.806297302246094,
9061
+ "logits/rejected": -10.80675983428955,
9062
+ "logps/chosen": -20.797101974487305,
9063
+ "logps/rejected": -50.487945556640625,
9064
+ "loss": 0.3453,
9065
+ "rewards/accuracies": 0.6875,
9066
+ "rewards/chosen": -0.15202410519123077,
9067
+ "rewards/margins": 2.7633230686187744,
9068
+ "rewards/rejected": -2.915347099304199,
9069
+ "step": 1208
9070
+ },
9071
+ {
9072
+ "epoch": 1.5055603079555175,
9073
+ "grad_norm": 42.94970703125,
9074
+ "learning_rate": 3.222339150932133e-08,
9075
+ "logits/chosen": -10.7854585647583,
9076
+ "logits/rejected": -10.784134864807129,
9077
+ "logps/chosen": -19.761821746826172,
9078
+ "logps/rejected": -61.553802490234375,
9079
+ "loss": 0.3576,
9080
+ "rewards/accuracies": 0.71875,
9081
+ "rewards/chosen": 0.026075001806020737,
9082
+ "rewards/margins": 4.108028888702393,
9083
+ "rewards/rejected": -4.081954002380371,
9084
+ "step": 1210
9085
+ },
9086
+ {
9087
+ "epoch": 1.5080488373901546,
9088
+ "grad_norm": 33.7689323425293,
9089
+ "learning_rate": 3.191721697748576e-08,
9090
+ "logits/chosen": -10.799263954162598,
9091
+ "logits/rejected": -10.799744606018066,
9092
+ "logps/chosen": -23.306804656982422,
9093
+ "logps/rejected": -64.52201843261719,
9094
+ "loss": 0.2794,
9095
+ "rewards/accuracies": 0.75,
9096
+ "rewards/chosen": -0.3046596050262451,
9097
+ "rewards/margins": 4.032251834869385,
9098
+ "rewards/rejected": -4.336911201477051,
9099
+ "step": 1212
9100
+ },
9101
+ {
9102
+ "epoch": 1.5105373668247921,
9103
+ "grad_norm": 6.178465366363525,
9104
+ "learning_rate": 3.161222752236024e-08,
9105
+ "logits/chosen": -10.777413368225098,
9106
+ "logits/rejected": -10.791770935058594,
9107
+ "logps/chosen": -24.492530822753906,
9108
+ "logps/rejected": -61.961761474609375,
9109
+ "loss": 0.3106,
9110
+ "rewards/accuracies": 0.71875,
9111
+ "rewards/chosen": -0.4530527591705322,
9112
+ "rewards/margins": 3.6912074089050293,
9113
+ "rewards/rejected": -4.144260406494141,
9114
+ "step": 1214
9115
+ },
9116
+ {
9117
+ "epoch": 1.5130258962594292,
9118
+ "grad_norm": 11.556970596313477,
9119
+ "learning_rate": 3.130842845271564e-08,
9120
+ "logits/chosen": -10.765844345092773,
9121
+ "logits/rejected": -10.766853332519531,
9122
+ "logps/chosen": -18.27448272705078,
9123
+ "logps/rejected": -67.04450225830078,
9124
+ "loss": 0.3224,
9125
+ "rewards/accuracies": 0.8125,
9126
+ "rewards/chosen": 0.13868612051010132,
9127
+ "rewards/margins": 4.729941368103027,
9128
+ "rewards/rejected": -4.591255187988281,
9129
+ "step": 1216
9130
+ },
9131
+ {
9132
+ "epoch": 1.5155144256940662,
9133
+ "grad_norm": 5.874917984008789,
9134
+ "learning_rate": 3.100582505660263e-08,
9135
+ "logits/chosen": -10.817387580871582,
9136
+ "logits/rejected": -10.818646430969238,
9137
+ "logps/chosen": -22.664018630981445,
9138
+ "logps/rejected": -61.78883361816406,
9139
+ "loss": 0.3315,
9140
+ "rewards/accuracies": 0.71875,
9141
+ "rewards/chosen": -0.26951882243156433,
9142
+ "rewards/margins": 3.7975308895111084,
9143
+ "rewards/rejected": -4.067049980163574,
9144
+ "step": 1218
9145
+ },
9146
+ {
9147
+ "epoch": 1.5180029551287038,
9148
+ "grad_norm": 7.716071128845215,
9149
+ "learning_rate": 3.0704422601259386e-08,
9150
+ "logits/chosen": -10.804115295410156,
9151
+ "logits/rejected": -10.803618431091309,
9152
+ "logps/chosen": -20.043956756591797,
9153
+ "logps/rejected": -59.39397430419922,
9154
+ "loss": 0.3588,
9155
+ "rewards/accuracies": 0.71875,
9156
+ "rewards/chosen": -0.025287901982665062,
9157
+ "rewards/margins": 3.833770990371704,
9158
+ "rewards/rejected": -3.8590588569641113,
9159
+ "step": 1220
9160
+ },
9161
+ {
9162
+ "epoch": 1.5204914845633408,
9163
+ "grad_norm": 5.23994779586792,
9164
+ "learning_rate": 3.0404226333020114e-08,
9165
+ "logits/chosen": -10.771346092224121,
9166
+ "logits/rejected": -10.723176002502441,
9167
+ "logps/chosen": -21.367895126342773,
9168
+ "logps/rejected": -51.024864196777344,
9169
+ "loss": 0.2626,
9170
+ "rewards/accuracies": 0.59375,
9171
+ "rewards/chosen": -0.20692677795886993,
9172
+ "rewards/margins": 2.8495285511016846,
9173
+ "rewards/rejected": -3.056455135345459,
9174
+ "step": 1222
9175
+ },
9176
+ {
9177
+ "epoch": 1.522980013997978,
9178
+ "grad_norm": 17.1359920501709,
9179
+ "learning_rate": 3.010524147722353e-08,
9180
+ "logits/chosen": -10.765385627746582,
9181
+ "logits/rejected": -10.76011848449707,
9182
+ "logps/chosen": -25.971342086791992,
9183
+ "logps/rejected": -48.49481201171875,
9184
+ "loss": 0.3097,
9185
+ "rewards/accuracies": 0.625,
9186
+ "rewards/chosen": -0.6388694047927856,
9187
+ "rewards/margins": 2.1466176509857178,
9188
+ "rewards/rejected": -2.785486936569214,
9189
+ "step": 1224
9190
+ },
9191
+ {
9192
+ "epoch": 1.5254685434326154,
9193
+ "grad_norm": 5.530016899108887,
9194
+ "learning_rate": 2.9807473238122096e-08,
9195
+ "logits/chosen": -10.834047317504883,
9196
+ "logits/rejected": -10.830108642578125,
9197
+ "logps/chosen": -23.04602813720703,
9198
+ "logps/rejected": -76.07076263427734,
9199
+ "loss": 0.2867,
9200
+ "rewards/accuracies": 0.84375,
9201
+ "rewards/chosen": -0.3454991281032562,
9202
+ "rewards/margins": 5.049150466918945,
9203
+ "rewards/rejected": -5.394649028778076,
9204
+ "step": 1226
9205
+ },
9206
+ {
9207
+ "epoch": 1.5279570728672525,
9208
+ "grad_norm": 7.995519161224365,
9209
+ "learning_rate": 2.951092679879136e-08,
9210
+ "logits/chosen": -10.817156791687012,
9211
+ "logits/rejected": -10.819764137268066,
9212
+ "logps/chosen": -26.958938598632812,
9213
+ "logps/rejected": -63.7525634765625,
9214
+ "loss": 0.3123,
9215
+ "rewards/accuracies": 0.78125,
9216
+ "rewards/chosen": -0.7338520288467407,
9217
+ "rewards/margins": 3.5241026878356934,
9218
+ "rewards/rejected": -4.2579545974731445,
9219
+ "step": 1228
9220
+ },
9221
+ {
9222
+ "epoch": 1.5304456023018898,
9223
+ "grad_norm": 2.934828519821167,
9224
+ "learning_rate": 2.9215607321039604e-08,
9225
+ "logits/chosen": -10.804816246032715,
9226
+ "logits/rejected": -10.802118301391602,
9227
+ "logps/chosen": -30.615299224853516,
9228
+ "logps/rejected": -78.92305755615234,
9229
+ "loss": 0.2861,
9230
+ "rewards/accuracies": 0.78125,
9231
+ "rewards/chosen": -1.078827142715454,
9232
+ "rewards/margins": 4.707486629486084,
9233
+ "rewards/rejected": -5.786314010620117,
9234
+ "step": 1230
9235
+ },
9236
+ {
9237
+ "epoch": 1.532934131736527,
9238
+ "grad_norm": 10.17944622039795,
9239
+ "learning_rate": 2.8921519945318274e-08,
9240
+ "logits/chosen": -10.815936088562012,
9241
+ "logits/rejected": -10.81978702545166,
9242
+ "logps/chosen": -22.286001205444336,
9243
+ "logps/rejected": -50.4647331237793,
9244
+ "loss": 0.3433,
9245
+ "rewards/accuracies": 0.65625,
9246
+ "rewards/chosen": -0.28749099373817444,
9247
+ "rewards/margins": 2.67037296295166,
9248
+ "rewards/rejected": -2.9578638076782227,
9249
+ "step": 1232
9250
+ },
9251
+ {
9252
+ "epoch": 1.535422661171164,
9253
+ "grad_norm": 22.162673950195312,
9254
+ "learning_rate": 2.8628669790632188e-08,
9255
+ "logits/chosen": -10.832916259765625,
9256
+ "logits/rejected": -10.841936111450195,
9257
+ "logps/chosen": -24.979412078857422,
9258
+ "logps/rejected": -69.00961303710938,
9259
+ "loss": 0.2422,
9260
+ "rewards/accuracies": 0.71875,
9261
+ "rewards/chosen": -0.5436473488807678,
9262
+ "rewards/margins": 4.181175231933594,
9263
+ "rewards/rejected": -4.724822521209717,
9264
+ "step": 1234
9265
+ },
9266
+ {
9267
+ "epoch": 1.5379111906058014,
9268
+ "grad_norm": 10.095711708068848,
9269
+ "learning_rate": 2.8337061954450748e-08,
9270
+ "logits/chosen": -10.793304443359375,
9271
+ "logits/rejected": -10.794236183166504,
9272
+ "logps/chosen": -24.665653228759766,
9273
+ "logps/rejected": -65.7728271484375,
9274
+ "loss": 0.3002,
9275
+ "rewards/accuracies": 0.625,
9276
+ "rewards/chosen": -0.5008203387260437,
9277
+ "rewards/margins": 3.981152057647705,
9278
+ "rewards/rejected": -4.481971740722656,
9279
+ "step": 1236
9280
+ },
9281
+ {
9282
+ "epoch": 1.5403997200404387,
9283
+ "grad_norm": 50.88451385498047,
9284
+ "learning_rate": 2.804670151261891e-08,
9285
+ "logits/chosen": -10.825129508972168,
9286
+ "logits/rejected": -10.825586318969727,
9287
+ "logps/chosen": -28.969776153564453,
9288
+ "logps/rejected": -54.09687042236328,
9289
+ "loss": 0.3774,
9290
+ "rewards/accuracies": 0.65625,
9291
+ "rewards/chosen": -0.9328697323799133,
9292
+ "rewards/margins": 2.505774974822998,
9293
+ "rewards/rejected": -3.4386448860168457,
9294
+ "step": 1238
9295
+ },
9296
+ {
9297
+ "epoch": 1.5428882494750757,
9298
+ "grad_norm": 10.008930206298828,
9299
+ "learning_rate": 2.7757593519269084e-08,
9300
+ "logits/chosen": -10.8192138671875,
9301
+ "logits/rejected": -10.820106506347656,
9302
+ "logps/chosen": -24.02053451538086,
9303
+ "logps/rejected": -57.27711868286133,
9304
+ "loss": 0.2994,
9305
+ "rewards/accuracies": 0.71875,
9306
+ "rewards/chosen": -0.46384507417678833,
9307
+ "rewards/margins": 3.1840813159942627,
9308
+ "rewards/rejected": -3.6479263305664062,
9309
+ "step": 1240
9310
+ },
9311
+ {
9312
+ "epoch": 1.545376778909713,
9313
+ "grad_norm": 8.246393203735352,
9314
+ "learning_rate": 2.746974300673296e-08,
9315
+ "logits/chosen": -10.833887100219727,
9316
+ "logits/rejected": -10.839536666870117,
9317
+ "logps/chosen": -34.37361145019531,
9318
+ "logps/rejected": -72.40042114257812,
9319
+ "loss": 0.3491,
9320
+ "rewards/accuracies": 0.59375,
9321
+ "rewards/chosen": -1.4933593273162842,
9322
+ "rewards/margins": 3.673614501953125,
9323
+ "rewards/rejected": -5.16697359085083,
9324
+ "step": 1242
9325
+ },
9326
+ {
9327
+ "epoch": 1.5478653083443503,
9328
+ "grad_norm": 9.959875106811523,
9329
+ "learning_rate": 2.718315498545407e-08,
9330
+ "logits/chosen": -10.775245666503906,
9331
+ "logits/rejected": -10.781604766845703,
9332
+ "logps/chosen": -27.01799201965332,
9333
+ "logps/rejected": -58.5137825012207,
9334
+ "loss": 0.3782,
9335
+ "rewards/accuracies": 0.71875,
9336
+ "rewards/chosen": -0.739520788192749,
9337
+ "rewards/margins": 3.0444231033325195,
9338
+ "rewards/rejected": -3.7839434146881104,
9339
+ "step": 1244
9340
+ },
9341
+ {
9342
+ "epoch": 1.5503538377789874,
9343
+ "grad_norm": 6.7338480949401855,
9344
+ "learning_rate": 2.6897834443900524e-08,
9345
+ "logits/chosen": -10.767149925231934,
9346
+ "logits/rejected": -10.770130157470703,
9347
+ "logps/chosen": -27.426719665527344,
9348
+ "logps/rejected": -58.368736267089844,
9349
+ "loss": 0.3511,
9350
+ "rewards/accuracies": 0.625,
9351
+ "rewards/chosen": -0.7709727883338928,
9352
+ "rewards/margins": 3.010869264602661,
9353
+ "rewards/rejected": -3.7818422317504883,
9354
+ "step": 1246
9355
+ },
9356
+ {
9357
+ "epoch": 1.5528423672136247,
9358
+ "grad_norm": 4.529770374298096,
9359
+ "learning_rate": 2.661378634847805e-08,
9360
+ "logits/chosen": -10.790437698364258,
9361
+ "logits/rejected": -10.792939186096191,
9362
+ "logps/chosen": -28.72300910949707,
9363
+ "logps/rejected": -63.083126068115234,
9364
+ "loss": 0.2706,
9365
+ "rewards/accuracies": 0.625,
9366
+ "rewards/chosen": -0.8835013508796692,
9367
+ "rewards/margins": 3.3935039043426514,
9368
+ "rewards/rejected": -4.277005195617676,
9369
+ "step": 1248
9370
+ },
9371
+ {
9372
+ "epoch": 1.555330896648262,
9373
+ "grad_norm": 4.372297763824463,
9374
+ "learning_rate": 2.633101564344381e-08,
9375
+ "logits/chosen": -10.801713943481445,
9376
+ "logits/rejected": -10.806915283203125,
9377
+ "logps/chosen": -35.876853942871094,
9378
+ "logps/rejected": -73.22596740722656,
9379
+ "loss": 0.3021,
9380
+ "rewards/accuracies": 0.75,
9381
+ "rewards/chosen": -1.6220556497573853,
9382
+ "rewards/margins": 3.583003044128418,
9383
+ "rewards/rejected": -5.205059051513672,
9384
+ "step": 1250
9385
+ },
9386
+ {
9387
+ "epoch": 1.557819426082899,
9388
+ "grad_norm": 78.30384063720703,
9389
+ "learning_rate": 2.6049527250820048e-08,
9390
+ "logits/chosen": -10.82096004486084,
9391
+ "logits/rejected": -10.825112342834473,
9392
+ "logps/chosen": -24.951976776123047,
9393
+ "logps/rejected": -56.246063232421875,
9394
+ "loss": 0.2652,
9395
+ "rewards/accuracies": 0.625,
9396
+ "rewards/chosen": -0.5131027698516846,
9397
+ "rewards/margins": 3.009666919708252,
9398
+ "rewards/rejected": -3.5227696895599365,
9399
+ "step": 1252
9400
+ },
9401
+ {
9402
+ "epoch": 1.5603079555175363,
9403
+ "grad_norm": 26.9722900390625,
9404
+ "learning_rate": 2.5769326070308673e-08,
9405
+ "logits/chosen": -10.822922706604004,
9406
+ "logits/rejected": -10.821035385131836,
9407
+ "logps/chosen": -29.673250198364258,
9408
+ "logps/rejected": -67.98855590820312,
9409
+ "loss": 0.3846,
9410
+ "rewards/accuracies": 0.71875,
9411
+ "rewards/chosen": -1.0159591436386108,
9412
+ "rewards/margins": 3.564460039138794,
9413
+ "rewards/rejected": -4.580419063568115,
9414
+ "step": 1254
9415
+ },
9416
+ {
9417
+ "epoch": 1.5627964849521736,
9418
+ "grad_norm": 7.283795356750488,
9419
+ "learning_rate": 2.5490416979205754e-08,
9420
+ "logits/chosen": -10.802266120910645,
9421
+ "logits/rejected": -10.805842399597168,
9422
+ "logps/chosen": -24.364179611206055,
9423
+ "logps/rejected": -71.14039611816406,
9424
+ "loss": 0.3523,
9425
+ "rewards/accuracies": 0.71875,
9426
+ "rewards/chosen": -0.4461979866027832,
9427
+ "rewards/margins": 4.550894737243652,
9428
+ "rewards/rejected": -4.9970927238464355,
9429
+ "step": 1256
9430
+ },
9431
+ {
9432
+ "epoch": 1.5652850143868107,
9433
+ "grad_norm": 34.32064437866211,
9434
+ "learning_rate": 2.521280483231678e-08,
9435
+ "logits/chosen": -10.794498443603516,
9436
+ "logits/rejected": -10.793449401855469,
9437
+ "logps/chosen": -28.45277214050293,
9438
+ "logps/rejected": -65.07192993164062,
9439
+ "loss": 0.2844,
9440
+ "rewards/accuracies": 0.71875,
9441
+ "rewards/chosen": -0.8511688113212585,
9442
+ "rewards/margins": 3.4714159965515137,
9443
+ "rewards/rejected": -4.322584629058838,
9444
+ "step": 1258
9445
+ },
9446
+ {
9447
+ "epoch": 1.5677735438214482,
9448
+ "grad_norm": 4.81486701965332,
9449
+ "learning_rate": 2.4936494461872125e-08,
9450
+ "logits/chosen": -10.802877426147461,
9451
+ "logits/rejected": -10.80418586730957,
9452
+ "logps/chosen": -26.037837982177734,
9453
+ "logps/rejected": -66.15322875976562,
9454
+ "loss": 0.2786,
9455
+ "rewards/accuracies": 0.71875,
9456
+ "rewards/chosen": -0.6298637390136719,
9457
+ "rewards/margins": 3.9220423698425293,
9458
+ "rewards/rejected": -4.551905632019043,
9459
+ "step": 1260
9460
+ },
9461
+ {
9462
+ "epoch": 1.5702620732560852,
9463
+ "grad_norm": 10.898475646972656,
9464
+ "learning_rate": 2.4661490677442832e-08,
9465
+ "logits/chosen": -10.785269737243652,
9466
+ "logits/rejected": -10.790578842163086,
9467
+ "logps/chosen": -26.191368103027344,
9468
+ "logps/rejected": -51.842777252197266,
9469
+ "loss": 0.3908,
9470
+ "rewards/accuracies": 0.78125,
9471
+ "rewards/chosen": -0.6927503943443298,
9472
+ "rewards/margins": 2.401498794555664,
9473
+ "rewards/rejected": -3.0942492485046387,
9474
+ "step": 1262
9475
+ },
9476
+ {
9477
+ "epoch": 1.5727506026907223,
9478
+ "grad_norm": 12.538436889648438,
9479
+ "learning_rate": 2.4387798265857075e-08,
9480
+ "logits/chosen": -10.842425346374512,
9481
+ "logits/rejected": -10.846158981323242,
9482
+ "logps/chosen": -28.756683349609375,
9483
+ "logps/rejected": -64.9261245727539,
9484
+ "loss": 0.3156,
9485
+ "rewards/accuracies": 0.625,
9486
+ "rewards/chosen": -0.8623592257499695,
9487
+ "rewards/margins": 3.4886474609375,
9488
+ "rewards/rejected": -4.351006984710693,
9489
+ "step": 1264
9490
+ },
9491
+ {
9492
+ "epoch": 1.5752391321253598,
9493
+ "grad_norm": 12.948237419128418,
9494
+ "learning_rate": 2.4115421991116603e-08,
9495
+ "logits/chosen": -10.809388160705566,
9496
+ "logits/rejected": -10.816155433654785,
9497
+ "logps/chosen": -26.006319046020508,
9498
+ "logps/rejected": -77.81361389160156,
9499
+ "loss": 0.2467,
9500
+ "rewards/accuracies": 0.84375,
9501
+ "rewards/chosen": -0.5785080194473267,
9502
+ "rewards/margins": 5.077298641204834,
9503
+ "rewards/rejected": -5.655807018280029,
9504
+ "step": 1266
9505
+ },
9506
+ {
9507
+ "epoch": 1.5777276615599969,
9508
+ "grad_norm": 43.29991149902344,
9509
+ "learning_rate": 2.3844366594314092e-08,
9510
+ "logits/chosen": -10.77228832244873,
9511
+ "logits/rejected": -10.772639274597168,
9512
+ "logps/chosen": -20.349361419677734,
9513
+ "logps/rejected": -52.78181076049805,
9514
+ "loss": 0.3604,
9515
+ "rewards/accuracies": 0.65625,
9516
+ "rewards/chosen": -0.04957125708460808,
9517
+ "rewards/margins": 3.16991925239563,
9518
+ "rewards/rejected": -3.2194907665252686,
9519
+ "step": 1268
9520
+ },
9521
+ {
9522
+ "epoch": 1.580216190994634,
9523
+ "grad_norm": 5.082618236541748,
9524
+ "learning_rate": 2.3574636793550375e-08,
9525
+ "logits/chosen": -10.785139083862305,
9526
+ "logits/rejected": -10.78280258178711,
9527
+ "logps/chosen": -25.451231002807617,
9528
+ "logps/rejected": -53.280418395996094,
9529
+ "loss": 0.352,
9530
+ "rewards/accuracies": 0.5625,
9531
+ "rewards/chosen": -0.5985268354415894,
9532
+ "rewards/margins": 2.6375532150268555,
9533
+ "rewards/rejected": -3.2360801696777344,
9534
+ "step": 1270
9535
+ },
9536
+ {
9537
+ "epoch": 1.5827047204292715,
9538
+ "grad_norm": 141.70098876953125,
9539
+ "learning_rate": 2.330623728385246e-08,
9540
+ "logits/chosen": -10.739248275756836,
9541
+ "logits/rejected": -10.738280296325684,
9542
+ "logps/chosen": -24.47553253173828,
9543
+ "logps/rejected": -61.69512939453125,
9544
+ "loss": 0.4036,
9545
+ "rewards/accuracies": 0.71875,
9546
+ "rewards/chosen": -0.4863220751285553,
9547
+ "rewards/margins": 3.448641300201416,
9548
+ "rewards/rejected": -3.9349637031555176,
9549
+ "step": 1272
9550
+ },
9551
+ {
9552
+ "epoch": 1.5851932498639085,
9553
+ "grad_norm": 6.310169219970703,
9554
+ "learning_rate": 2.3039172737091807e-08,
9555
+ "logits/chosen": -10.816719055175781,
9556
+ "logits/rejected": -10.82149887084961,
9557
+ "logps/chosen": -23.517425537109375,
9558
+ "logps/rejected": -44.70637512207031,
9559
+ "loss": 0.3769,
9560
+ "rewards/accuracies": 0.59375,
9561
+ "rewards/chosen": -0.39209693670272827,
9562
+ "rewards/margins": 2.0814270973205566,
9563
+ "rewards/rejected": -2.4735240936279297,
9564
+ "step": 1274
9565
+ },
9566
+ {
9567
+ "epoch": 1.5876817792985458,
9568
+ "grad_norm": 13.863208770751953,
9569
+ "learning_rate": 2.2773447801902855e-08,
9570
+ "logits/chosen": -10.832393646240234,
9571
+ "logits/rejected": -10.82877254486084,
9572
+ "logps/chosen": -18.8774471282959,
9573
+ "logps/rejected": -49.65419006347656,
9574
+ "loss": 0.354,
9575
+ "rewards/accuracies": 0.6875,
9576
+ "rewards/chosen": 0.06631821393966675,
9577
+ "rewards/margins": 2.8835067749023438,
9578
+ "rewards/rejected": -2.8171889781951904,
9579
+ "step": 1276
9580
+ },
9581
+ {
9582
+ "epoch": 1.590170308733183,
9583
+ "grad_norm": 15.649863243103027,
9584
+ "learning_rate": 2.250906710360235e-08,
9585
+ "logits/chosen": -10.807013511657715,
9586
+ "logits/rejected": -10.8053560256958,
9587
+ "logps/chosen": -22.7423038482666,
9588
+ "logps/rejected": -49.1067008972168,
9589
+ "loss": 0.3119,
9590
+ "rewards/accuracies": 0.59375,
9591
+ "rewards/chosen": -0.3249521255493164,
9592
+ "rewards/margins": 2.519008159637451,
9593
+ "rewards/rejected": -2.8439598083496094,
9594
+ "step": 1278
9595
+ },
9596
+ {
9597
+ "epoch": 1.5926588381678202,
9598
+ "grad_norm": 15.802570343017578,
9599
+ "learning_rate": 2.2246035244108586e-08,
9600
+ "logits/chosen": -10.775257110595703,
9601
+ "logits/rejected": -10.774992942810059,
9602
+ "logps/chosen": -20.724456787109375,
9603
+ "logps/rejected": -60.901023864746094,
9604
+ "loss": 0.3112,
9605
+ "rewards/accuracies": 0.75,
9606
+ "rewards/chosen": -0.12378637492656708,
9607
+ "rewards/margins": 3.8235392570495605,
9608
+ "rewards/rejected": -3.9473254680633545,
9609
+ "step": 1280
9610
+ },
9611
+ {
9612
+ "epoch": 1.5951473676024575,
9613
+ "grad_norm": 18.153547286987305,
9614
+ "learning_rate": 2.1984356801861502e-08,
9615
+ "logits/chosen": -10.780445098876953,
9616
+ "logits/rejected": -10.779111862182617,
9617
+ "logps/chosen": -22.182926177978516,
9618
+ "logps/rejected": -62.560401916503906,
9619
+ "loss": 0.2616,
9620
+ "rewards/accuracies": 0.65625,
9621
+ "rewards/chosen": -0.21629738807678223,
9622
+ "rewards/margins": 3.8685829639434814,
9623
+ "rewards/rejected": -4.084880352020264,
9624
+ "step": 1282
9625
+ },
9626
+ {
9627
+ "epoch": 1.5976358970370947,
9628
+ "grad_norm": 5.7486443519592285,
9629
+ "learning_rate": 2.1724036331742834e-08,
9630
+ "logits/chosen": -10.770216941833496,
9631
+ "logits/rejected": -10.772773742675781,
9632
+ "logps/chosen": -21.333688735961914,
9633
+ "logps/rejected": -49.32101058959961,
9634
+ "loss": 0.3264,
9635
+ "rewards/accuracies": 0.6875,
9636
+ "rewards/chosen": -0.12258227914571762,
9637
+ "rewards/margins": 2.6417171955108643,
9638
+ "rewards/rejected": -2.7642996311187744,
9639
+ "step": 1284
9640
+ },
9641
+ {
9642
+ "epoch": 1.6001244264717318,
9643
+ "grad_norm": 72.83959197998047,
9644
+ "learning_rate": 2.1465078364996968e-08,
9645
+ "logits/chosen": -10.802704811096191,
9646
+ "logits/rejected": -10.801321029663086,
9647
+ "logps/chosen": -23.669212341308594,
9648
+ "logps/rejected": -56.59158706665039,
9649
+ "loss": 0.3146,
9650
+ "rewards/accuracies": 0.71875,
9651
+ "rewards/chosen": -0.41491052508354187,
9652
+ "rewards/margins": 3.080634117126465,
9653
+ "rewards/rejected": -3.495544672012329,
9654
+ "step": 1286
9655
+ },
9656
+ {
9657
+ "epoch": 1.602612955906369,
9658
+ "grad_norm": 34.2864875793457,
9659
+ "learning_rate": 2.120748740915198e-08,
9660
+ "logits/chosen": -10.832742691040039,
9661
+ "logits/rejected": -10.830432891845703,
9662
+ "logps/chosen": -21.441856384277344,
9663
+ "logps/rejected": -63.40147399902344,
9664
+ "loss": 0.2765,
9665
+ "rewards/accuracies": 0.78125,
9666
+ "rewards/chosen": -0.2095116674900055,
9667
+ "rewards/margins": 3.9784107208251953,
9668
+ "rewards/rejected": -4.187922477722168,
9669
+ "step": 1288
9670
+ },
9671
+ {
9672
+ "epoch": 1.6051014853410064,
9673
+ "grad_norm": 9.747991561889648,
9674
+ "learning_rate": 2.0951267947941143e-08,
9675
+ "logits/chosen": -10.8104248046875,
9676
+ "logits/rejected": -10.815136909484863,
9677
+ "logps/chosen": -23.734474182128906,
9678
+ "logps/rejected": -60.64533233642578,
9679
+ "loss": 0.3503,
9680
+ "rewards/accuracies": 0.65625,
9681
+ "rewards/chosen": -0.4252595007419586,
9682
+ "rewards/margins": 3.603969097137451,
9683
+ "rewards/rejected": -4.029229164123535,
9684
+ "step": 1290
9685
+ },
9686
+ {
9687
+ "epoch": 1.6075900147756434,
9688
+ "grad_norm": 4.882062911987305,
9689
+ "learning_rate": 2.0696424441225036e-08,
9690
+ "logits/chosen": -10.807500839233398,
9691
+ "logits/rejected": -10.805845260620117,
9692
+ "logps/chosen": -21.90993309020996,
9693
+ "logps/rejected": -65.13987731933594,
9694
+ "loss": 0.24,
9695
+ "rewards/accuracies": 0.75,
9696
+ "rewards/chosen": -0.2415923923254013,
9697
+ "rewards/margins": 4.038329124450684,
9698
+ "rewards/rejected": -4.279921531677246,
9699
+ "step": 1292
9700
+ },
9701
+ {
9702
+ "epoch": 1.6100785442102807,
9703
+ "grad_norm": 3.259903907775879,
9704
+ "learning_rate": 2.0442961324913686e-08,
9705
+ "logits/chosen": -10.808277130126953,
9706
+ "logits/rejected": -10.814513206481934,
9707
+ "logps/chosen": -21.407821655273438,
9708
+ "logps/rejected": -55.829368591308594,
9709
+ "loss": 0.3427,
9710
+ "rewards/accuracies": 0.75,
9711
+ "rewards/chosen": -0.20267558097839355,
9712
+ "rewards/margins": 3.1913931369781494,
9713
+ "rewards/rejected": -3.394068717956543,
9714
+ "step": 1294
9715
+ },
9716
+ {
9717
+ "epoch": 1.612567073644918,
9718
+ "grad_norm": 9.57787036895752,
9719
+ "learning_rate": 2.0190883010889615e-08,
9720
+ "logits/chosen": -10.820773124694824,
9721
+ "logits/rejected": -10.818120956420898,
9722
+ "logps/chosen": -15.357433319091797,
9723
+ "logps/rejected": -41.842857360839844,
9724
+ "loss": 0.3415,
9725
+ "rewards/accuracies": 0.71875,
9726
+ "rewards/chosen": 0.40116557478904724,
9727
+ "rewards/margins": 2.474684715270996,
9728
+ "rewards/rejected": -2.073519229888916,
9729
+ "step": 1296
9730
+ },
9731
+ {
9732
+ "epoch": 1.615055603079555,
9733
+ "grad_norm": 8.003713607788086,
9734
+ "learning_rate": 1.9940193886930777e-08,
9735
+ "logits/chosen": -10.783833503723145,
9736
+ "logits/rejected": -10.784221649169922,
9737
+ "logps/chosen": -24.231639862060547,
9738
+ "logps/rejected": -71.94761657714844,
9739
+ "loss": 0.2483,
9740
+ "rewards/accuracies": 0.8125,
9741
+ "rewards/chosen": -0.4585307240486145,
9742
+ "rewards/margins": 4.653621673583984,
9743
+ "rewards/rejected": -5.112152576446533,
9744
+ "step": 1298
9745
+ },
9746
+ {
9747
+ "epoch": 1.6175441325141924,
9748
+ "grad_norm": 8.922944068908691,
9749
+ "learning_rate": 1.969089831663443e-08,
9750
+ "logits/chosen": -10.830660820007324,
9751
+ "logits/rejected": -10.817536354064941,
9752
+ "logps/chosen": -23.25543785095215,
9753
+ "logps/rejected": -53.022544860839844,
9754
+ "loss": 0.3277,
9755
+ "rewards/accuracies": 0.65625,
9756
+ "rewards/chosen": -0.34474945068359375,
9757
+ "rewards/margins": 2.839651346206665,
9758
+ "rewards/rejected": -3.1844005584716797,
9759
+ "step": 1300
9760
+ },
9761
+ {
9762
+ "epoch": 1.6200326619488297,
9763
+ "grad_norm": 9.474493980407715,
9764
+ "learning_rate": 1.9443000639341045e-08,
9765
+ "logits/chosen": -10.812176704406738,
9766
+ "logits/rejected": -10.77513599395752,
9767
+ "logps/chosen": -20.876327514648438,
9768
+ "logps/rejected": -52.44502258300781,
9769
+ "loss": 0.2945,
9770
+ "rewards/accuracies": 0.5625,
9771
+ "rewards/chosen": -0.11885665357112885,
9772
+ "rewards/margins": 3.071859121322632,
9773
+ "rewards/rejected": -3.190715789794922,
9774
+ "step": 1302
9775
+ },
9776
+ {
9777
+ "epoch": 1.6225211913834667,
9778
+ "grad_norm": 24.406034469604492,
9779
+ "learning_rate": 1.919650517005872e-08,
9780
+ "logits/chosen": -10.835334777832031,
9781
+ "logits/rejected": -10.845525741577148,
9782
+ "logps/chosen": -28.61736297607422,
9783
+ "logps/rejected": -68.78744506835938,
9784
+ "loss": 0.3318,
9785
+ "rewards/accuracies": 0.625,
9786
+ "rewards/chosen": -0.8773050308227539,
9787
+ "rewards/margins": 3.880265712738037,
9788
+ "rewards/rejected": -4.757571220397949,
9789
+ "step": 1304
9790
+ },
9791
+ {
9792
+ "epoch": 1.625009720818104,
9793
+ "grad_norm": 29.102094650268555,
9794
+ "learning_rate": 1.895141619938825e-08,
9795
+ "logits/chosen": -10.814085006713867,
9796
+ "logits/rejected": -10.81423568725586,
9797
+ "logps/chosen": -21.529321670532227,
9798
+ "logps/rejected": -57.59548568725586,
9799
+ "loss": 0.3405,
9800
+ "rewards/accuracies": 0.65625,
9801
+ "rewards/chosen": -0.19233551621437073,
9802
+ "rewards/margins": 3.4261107444763184,
9803
+ "rewards/rejected": -3.618446111679077,
9804
+ "step": 1306
9805
+ },
9806
+ {
9807
+ "epoch": 1.6274982502527413,
9808
+ "grad_norm": 18.181053161621094,
9809
+ "learning_rate": 1.8707737993448247e-08,
9810
+ "logits/chosen": -10.777032852172852,
9811
+ "logits/rejected": -10.783935546875,
9812
+ "logps/chosen": -21.770933151245117,
9813
+ "logps/rejected": -70.67555236816406,
9814
+ "loss": 0.2861,
9815
+ "rewards/accuracies": 0.75,
9816
+ "rewards/chosen": -0.20613786578178406,
9817
+ "rewards/margins": 4.698180675506592,
9818
+ "rewards/rejected": -4.904318332672119,
9819
+ "step": 1308
9820
+ },
9821
+ {
9822
+ "epoch": 1.6299867796873784,
9823
+ "grad_norm": 18.842073440551758,
9824
+ "learning_rate": 1.8465474793801085e-08,
9825
+ "logits/chosen": -10.790057182312012,
9826
+ "logits/rejected": -10.789998054504395,
9827
+ "logps/chosen": -25.611600875854492,
9828
+ "logps/rejected": -61.528900146484375,
9829
+ "loss": 0.3441,
9830
+ "rewards/accuracies": 0.5625,
9831
+ "rewards/chosen": -0.5731645226478577,
9832
+ "rewards/margins": 3.542222499847412,
9833
+ "rewards/rejected": -4.115387439727783,
9834
+ "step": 1310
9835
+ },
9836
+ {
9837
+ "epoch": 1.6324753091220157,
9838
+ "grad_norm": 2.9410083293914795,
9839
+ "learning_rate": 1.8224630817378827e-08,
9840
+ "logits/chosen": -10.817235946655273,
9841
+ "logits/rejected": -10.816126823425293,
9842
+ "logps/chosen": -23.82830047607422,
9843
+ "logps/rejected": -59.01426696777344,
9844
+ "loss": 0.2962,
9845
+ "rewards/accuracies": 0.65625,
9846
+ "rewards/chosen": -0.39294612407684326,
9847
+ "rewards/margins": 3.3668603897094727,
9848
+ "rewards/rejected": -3.7598063945770264,
9849
+ "step": 1312
9850
+ },
9851
+ {
9852
+ "epoch": 1.634963838556653,
9853
+ "grad_norm": 34.97274398803711,
9854
+ "learning_rate": 1.7985210256410088e-08,
9855
+ "logits/chosen": -10.75239372253418,
9856
+ "logits/rejected": -10.75178337097168,
9857
+ "logps/chosen": -26.019506454467773,
9858
+ "logps/rejected": -63.77971267700195,
9859
+ "loss": 0.4301,
9860
+ "rewards/accuracies": 0.65625,
9861
+ "rewards/chosen": -0.6273402571678162,
9862
+ "rewards/margins": 3.5424182415008545,
9863
+ "rewards/rejected": -4.1697587966918945,
9864
+ "step": 1314
9865
+ },
9866
+ {
9867
+ "epoch": 1.63745236799129,
9868
+ "grad_norm": 2.5021934509277344,
9869
+ "learning_rate": 1.7747217278346838e-08,
9870
+ "logits/chosen": -10.83252239227295,
9871
+ "logits/rejected": -10.830848693847656,
9872
+ "logps/chosen": -23.322322845458984,
9873
+ "logps/rejected": -55.16864013671875,
9874
+ "loss": 0.3096,
9875
+ "rewards/accuracies": 0.75,
9876
+ "rewards/chosen": -0.384496808052063,
9877
+ "rewards/margins": 2.9950833320617676,
9878
+ "rewards/rejected": -3.37958025932312,
9879
+ "step": 1316
9880
+ },
9881
+ {
9882
+ "epoch": 1.6399408974259275,
9883
+ "grad_norm": 33.00211715698242,
9884
+ "learning_rate": 1.7510656025792004e-08,
9885
+ "logits/chosen": -10.799005508422852,
9886
+ "logits/rejected": -10.798194885253906,
9887
+ "logps/chosen": -26.886280059814453,
9888
+ "logps/rejected": -58.32006072998047,
9889
+ "loss": 0.3002,
9890
+ "rewards/accuracies": 0.71875,
9891
+ "rewards/chosen": -0.7233449816703796,
9892
+ "rewards/margins": 3.111297130584717,
9893
+ "rewards/rejected": -3.8346424102783203,
9894
+ "step": 1318
9895
+ },
9896
+ {
9897
+ "epoch": 1.6424294268605646,
9898
+ "grad_norm": 7.195638179779053,
9899
+ "learning_rate": 1.7275530616427335e-08,
9900
+ "logits/chosen": -10.778608322143555,
9901
+ "logits/rejected": -10.783539772033691,
9902
+ "logps/chosen": -27.22859001159668,
9903
+ "logps/rejected": -60.26847839355469,
9904
+ "loss": 0.3174,
9905
+ "rewards/accuracies": 0.59375,
9906
+ "rewards/chosen": -0.7433955669403076,
9907
+ "rewards/margins": 3.1764450073242188,
9908
+ "rewards/rejected": -3.9198405742645264,
9909
+ "step": 1320
9910
+ },
9911
+ {
9912
+ "epoch": 1.6449179562952017,
9913
+ "grad_norm": 14.944016456604004,
9914
+ "learning_rate": 1.7041845142941612e-08,
9915
+ "logits/chosen": -10.831104278564453,
9916
+ "logits/rejected": -10.8374605178833,
9917
+ "logps/chosen": -22.72991371154785,
9918
+ "logps/rejected": -69.8420181274414,
9919
+ "loss": 0.317,
9920
+ "rewards/accuracies": 0.84375,
9921
+ "rewards/chosen": -0.288443922996521,
9922
+ "rewards/margins": 4.563740253448486,
9923
+ "rewards/rejected": -4.852183818817139,
9924
+ "step": 1322
9925
+ },
9926
+ {
9927
+ "epoch": 1.6474064857298392,
9928
+ "grad_norm": 17.540937423706055,
9929
+ "learning_rate": 1.6809603672959615e-08,
9930
+ "logits/chosen": -10.862117767333984,
9931
+ "logits/rejected": -10.859867095947266,
9932
+ "logps/chosen": -25.939712524414062,
9933
+ "logps/rejected": -51.00326156616211,
9934
+ "loss": 0.3688,
9935
+ "rewards/accuracies": 0.5,
9936
+ "rewards/chosen": -0.6566795110702515,
9937
+ "rewards/margins": 2.3946855068206787,
9938
+ "rewards/rejected": -3.051365375518799,
9939
+ "step": 1324
9940
+ },
9941
+ {
9942
+ "epoch": 1.6498950151644762,
9943
+ "grad_norm": 12.726731300354004,
9944
+ "learning_rate": 1.6578810248971142e-08,
9945
+ "logits/chosen": -10.80691909790039,
9946
+ "logits/rejected": -10.802922248840332,
9947
+ "logps/chosen": -20.270612716674805,
9948
+ "logps/rejected": -53.08592987060547,
9949
+ "loss": 0.329,
9950
+ "rewards/accuracies": 0.71875,
9951
+ "rewards/chosen": -0.04774314910173416,
9952
+ "rewards/margins": 3.1508731842041016,
9953
+ "rewards/rejected": -3.1986162662506104,
9954
+ "step": 1326
9955
+ },
9956
+ {
9957
+ "epoch": 1.6523835445991133,
9958
+ "grad_norm": 63.046714782714844,
9959
+ "learning_rate": 1.6349468888260764e-08,
9960
+ "logits/chosen": -10.805728912353516,
9961
+ "logits/rejected": -10.807948112487793,
9962
+ "logps/chosen": -22.131887435913086,
9963
+ "logps/rejected": -50.19975662231445,
9964
+ "loss": 0.3256,
9965
+ "rewards/accuracies": 0.625,
9966
+ "rewards/chosen": -0.2647934556007385,
9967
+ "rewards/margins": 2.623716354370117,
9968
+ "rewards/rejected": -2.88850998878479,
9969
+ "step": 1328
9970
+ },
9971
+ {
9972
+ "epoch": 1.6548720740337508,
9973
+ "grad_norm": 18.786680221557617,
9974
+ "learning_rate": 1.6121583582837772e-08,
9975
+ "logits/chosen": -10.854766845703125,
9976
+ "logits/rejected": -10.862516403198242,
9977
+ "logps/chosen": -25.881559371948242,
9978
+ "logps/rejected": -60.02587127685547,
9979
+ "loss": 0.3334,
9980
+ "rewards/accuracies": 0.59375,
9981
+ "rewards/chosen": -0.6450967788696289,
9982
+ "rewards/margins": 3.2605414390563965,
9983
+ "rewards/rejected": -3.9056379795074463,
9984
+ "step": 1330
9985
+ },
9986
+ {
9987
+ "epoch": 1.6573606034683879,
9988
+ "grad_norm": 5.466382026672363,
9989
+ "learning_rate": 1.589515829936684e-08,
9990
+ "logits/chosen": -10.840960502624512,
9991
+ "logits/rejected": -10.838695526123047,
9992
+ "logps/chosen": -21.14974594116211,
9993
+ "logps/rejected": -54.80947494506836,
9994
+ "loss": 0.3163,
9995
+ "rewards/accuracies": 0.625,
9996
+ "rewards/chosen": -0.16069187223911285,
9997
+ "rewards/margins": 3.2574622631073,
9998
+ "rewards/rejected": -3.418154239654541,
9999
+ "step": 1332
10000
+ },
10001
+ {
10002
+ "epoch": 1.6598491329030252,
10003
+ "grad_norm": 3.6651666164398193,
10004
+ "learning_rate": 1.5670196979098837e-08,
10005
+ "logits/chosen": -10.87842845916748,
10006
+ "logits/rejected": -10.880535125732422,
10007
+ "logps/chosen": -25.185527801513672,
10008
+ "logps/rejected": -65.58544921875,
10009
+ "loss": 0.3046,
10010
+ "rewards/accuracies": 0.53125,
10011
+ "rewards/chosen": -0.5343016386032104,
10012
+ "rewards/margins": 3.9803366661071777,
10013
+ "rewards/rejected": -4.514638423919678,
10014
+ "step": 1334
10015
+ },
10016
+ {
10017
+ "epoch": 1.6623376623376624,
10018
+ "grad_norm": 22.037355422973633,
10019
+ "learning_rate": 1.5446703537802342e-08,
10020
+ "logits/chosen": -10.828059196472168,
10021
+ "logits/rejected": -10.813085556030273,
10022
+ "logps/chosen": -24.3900089263916,
10023
+ "logps/rejected": -57.214691162109375,
10024
+ "loss": 0.3129,
10025
+ "rewards/accuracies": 0.75,
10026
+ "rewards/chosen": -0.4744144380092621,
10027
+ "rewards/margins": 3.0794365406036377,
10028
+ "rewards/rejected": -3.5538506507873535,
10029
+ "step": 1336
10030
+ },
10031
+ {
10032
+ "epoch": 1.6648261917722995,
10033
+ "grad_norm": 4.078120231628418,
10034
+ "learning_rate": 1.5224681865695422e-08,
10035
+ "logits/chosen": -10.837578773498535,
10036
+ "logits/rejected": -10.836234092712402,
10037
+ "logps/chosen": -22.567678451538086,
10038
+ "logps/rejected": -62.71944808959961,
10039
+ "loss": 0.3147,
10040
+ "rewards/accuracies": 0.71875,
10041
+ "rewards/chosen": -0.32626694440841675,
10042
+ "rewards/margins": 3.8645825386047363,
10043
+ "rewards/rejected": -4.190849304199219,
10044
+ "step": 1338
10045
+ },
10046
+ {
10047
+ "epoch": 1.6673147212069368,
10048
+ "grad_norm": 12.054242134094238,
10049
+ "learning_rate": 1.5004135827377905e-08,
10050
+ "logits/chosen": -10.865391731262207,
10051
+ "logits/rejected": -10.868090629577637,
10052
+ "logps/chosen": -24.837133407592773,
10053
+ "logps/rejected": -87.59353637695312,
10054
+ "loss": 0.2993,
10055
+ "rewards/accuracies": 0.78125,
10056
+ "rewards/chosen": -0.4641028940677643,
10057
+ "rewards/margins": 6.047280311584473,
10058
+ "rewards/rejected": -6.511383533477783,
10059
+ "step": 1340
10060
+ },
10061
+ {
10062
+ "epoch": 1.669803250641574,
10063
+ "grad_norm": 33.8203125,
10064
+ "learning_rate": 1.4785069261764182e-08,
10065
+ "logits/chosen": -10.816351890563965,
10066
+ "logits/rejected": -10.816112518310547,
10067
+ "logps/chosen": -27.667259216308594,
10068
+ "logps/rejected": -46.164127349853516,
10069
+ "loss": 0.331,
10070
+ "rewards/accuracies": 0.46875,
10071
+ "rewards/chosen": -0.7790037393569946,
10072
+ "rewards/margins": 1.784177303314209,
10073
+ "rewards/rejected": -2.563180923461914,
10074
+ "step": 1342
10075
+ },
10076
+ {
10077
+ "epoch": 1.6722917800762112,
10078
+ "grad_norm": 5.524724960327148,
10079
+ "learning_rate": 1.4567485982016258e-08,
10080
+ "logits/chosen": -10.838247299194336,
10081
+ "logits/rejected": -10.837743759155273,
10082
+ "logps/chosen": -24.21609878540039,
10083
+ "logps/rejected": -47.889774322509766,
10084
+ "loss": 0.3116,
10085
+ "rewards/accuracies": 0.5,
10086
+ "rewards/chosen": -0.45775747299194336,
10087
+ "rewards/margins": 2.266995429992676,
10088
+ "rewards/rejected": -2.7247531414031982,
10089
+ "step": 1344
10090
+ },
10091
+ {
10092
+ "epoch": 1.6747803095108484,
10093
+ "grad_norm": 16.984832763671875,
10094
+ "learning_rate": 1.4351389775477573e-08,
10095
+ "logits/chosen": -10.84825611114502,
10096
+ "logits/rejected": -10.851517677307129,
10097
+ "logps/chosen": -28.410385131835938,
10098
+ "logps/rejected": -76.4227294921875,
10099
+ "loss": 0.2681,
10100
+ "rewards/accuracies": 0.84375,
10101
+ "rewards/chosen": -0.8531373143196106,
10102
+ "rewards/margins": 4.609513282775879,
10103
+ "rewards/rejected": -5.462650299072266,
10104
+ "step": 1346
10105
+ },
10106
+ {
10107
+ "epoch": 1.6772688389454857,
10108
+ "grad_norm": 7.401856422424316,
10109
+ "learning_rate": 1.4136784403606839e-08,
10110
+ "logits/chosen": -10.8283109664917,
10111
+ "logits/rejected": -10.829707145690918,
10112
+ "logps/chosen": -26.35080337524414,
10113
+ "logps/rejected": -65.35711669921875,
10114
+ "loss": 0.2665,
10115
+ "rewards/accuracies": 0.71875,
10116
+ "rewards/chosen": -0.6642946004867554,
10117
+ "rewards/margins": 3.7643990516662598,
10118
+ "rewards/rejected": -4.428694248199463,
10119
+ "step": 1348
10120
+ },
10121
+ {
10122
+ "epoch": 1.6797573683801228,
10123
+ "grad_norm": 6.8130621910095215,
10124
+ "learning_rate": 1.3923673601912777e-08,
10125
+ "logits/chosen": -10.753373146057129,
10126
+ "logits/rejected": -10.752891540527344,
10127
+ "logps/chosen": -25.6737060546875,
10128
+ "logps/rejected": -65.83049011230469,
10129
+ "loss": 0.3014,
10130
+ "rewards/accuracies": 0.78125,
10131
+ "rewards/chosen": -0.5438342094421387,
10132
+ "rewards/margins": 3.9245076179504395,
10133
+ "rewards/rejected": -4.468341827392578,
10134
+ "step": 1350
10135
+ },
10136
+ {
10137
+ "epoch": 1.68224589781476,
10138
+ "grad_norm": 20.65499496459961,
10139
+ "learning_rate": 1.3712061079889014e-08,
10140
+ "logits/chosen": -10.788976669311523,
10141
+ "logits/rejected": -10.791670799255371,
10142
+ "logps/chosen": -22.58700942993164,
10143
+ "logps/rejected": -53.30634689331055,
10144
+ "loss": 0.3424,
10145
+ "rewards/accuracies": 0.65625,
10146
+ "rewards/chosen": -0.31589654088020325,
10147
+ "rewards/margins": 2.8593664169311523,
10148
+ "rewards/rejected": -3.175262928009033,
10149
+ "step": 1352
10150
+ },
10151
+ {
10152
+ "epoch": 1.6847344272493974,
10153
+ "grad_norm": 11.894336700439453,
10154
+ "learning_rate": 1.3501950520949434e-08,
10155
+ "logits/chosen": -10.869431495666504,
10156
+ "logits/rejected": -10.864688873291016,
10157
+ "logps/chosen": -24.943134307861328,
10158
+ "logps/rejected": -54.439842224121094,
10159
+ "loss": 0.3573,
10160
+ "rewards/accuracies": 0.71875,
10161
+ "rewards/chosen": -0.5166460275650024,
10162
+ "rewards/margins": 2.882528781890869,
10163
+ "rewards/rejected": -3.3991751670837402,
10164
+ "step": 1354
10165
+ },
10166
+ {
10167
+ "epoch": 1.6872229566840344,
10168
+ "grad_norm": 8.13518238067627,
10169
+ "learning_rate": 1.3293345582364224e-08,
10170
+ "logits/chosen": -10.843225479125977,
10171
+ "logits/rejected": -10.838338851928711,
10172
+ "logps/chosen": -23.392269134521484,
10173
+ "logps/rejected": -65.21499633789062,
10174
+ "loss": 0.3396,
10175
+ "rewards/accuracies": 0.71875,
10176
+ "rewards/chosen": -0.34925830364227295,
10177
+ "rewards/margins": 4.064693927764893,
10178
+ "rewards/rejected": -4.413951873779297,
10179
+ "step": 1356
10180
+ },
10181
+ {
10182
+ "epoch": 1.6897114861186717,
10183
+ "grad_norm": 5.220905303955078,
10184
+ "learning_rate": 1.3086249895196043e-08,
10185
+ "logits/chosen": -10.816040992736816,
10186
+ "logits/rejected": -10.818885803222656,
10187
+ "logps/chosen": -21.512014389038086,
10188
+ "logps/rejected": -65.09014129638672,
10189
+ "loss": 0.3114,
10190
+ "rewards/accuracies": 0.6875,
10191
+ "rewards/chosen": -0.15071852505207062,
10192
+ "rewards/margins": 4.254648208618164,
10193
+ "rewards/rejected": -4.40536642074585,
10194
+ "step": 1358
10195
+ },
10196
+ {
10197
+ "epoch": 1.692200015553309,
10198
+ "grad_norm": 85.90353393554688,
10199
+ "learning_rate": 1.2880667064237004e-08,
10200
+ "logits/chosen": -10.824080467224121,
10201
+ "logits/rejected": -10.832862854003906,
10202
+ "logps/chosen": -23.073331832885742,
10203
+ "logps/rejected": -58.96718978881836,
10204
+ "loss": 0.3724,
10205
+ "rewards/accuracies": 0.5,
10206
+ "rewards/chosen": -0.34528830647468567,
10207
+ "rewards/margins": 3.405324935913086,
10208
+ "rewards/rejected": -3.750613212585449,
10209
+ "step": 1360
10210
+ },
10211
+ {
10212
+ "epoch": 1.694688544987946,
10213
+ "grad_norm": 6.757018566131592,
10214
+ "learning_rate": 1.2676600667945714e-08,
10215
+ "logits/chosen": -10.774547576904297,
10216
+ "logits/rejected": -10.776646614074707,
10217
+ "logps/chosen": -27.236501693725586,
10218
+ "logps/rejected": -47.35329055786133,
10219
+ "loss": 0.3518,
10220
+ "rewards/accuracies": 0.65625,
10221
+ "rewards/chosen": -0.6959034204483032,
10222
+ "rewards/margins": 2.0354065895080566,
10223
+ "rewards/rejected": -2.7313101291656494,
10224
+ "step": 1362
10225
+ },
10226
+ {
10227
+ "epoch": 1.6971770744225834,
10228
+ "grad_norm": 6.0155487060546875,
10229
+ "learning_rate": 1.2474054258385225e-08,
10230
+ "logits/chosen": -10.811131477355957,
10231
+ "logits/rejected": -10.817914962768555,
10232
+ "logps/chosen": -29.25282859802246,
10233
+ "logps/rejected": -74.38288879394531,
10234
+ "loss": 0.276,
10235
+ "rewards/accuracies": 0.9375,
10236
+ "rewards/chosen": -0.927348256111145,
10237
+ "rewards/margins": 4.406147003173828,
10238
+ "rewards/rejected": -5.333494663238525,
10239
+ "step": 1364
10240
+ },
10241
+ {
10242
+ "epoch": 1.6996656038572207,
10243
+ "grad_norm": 60.83378601074219,
10244
+ "learning_rate": 1.2273031361160957e-08,
10245
+ "logits/chosen": -10.810754776000977,
10246
+ "logits/rejected": -10.807637214660645,
10247
+ "logps/chosen": -25.723983764648438,
10248
+ "logps/rejected": -63.92774200439453,
10249
+ "loss": 0.3274,
10250
+ "rewards/accuracies": 0.78125,
10251
+ "rewards/chosen": -0.5800298452377319,
10252
+ "rewards/margins": 3.7809581756591797,
10253
+ "rewards/rejected": -4.360988140106201,
10254
+ "step": 1366
10255
+ },
10256
+ {
10257
+ "epoch": 1.7021541332918577,
10258
+ "grad_norm": 5.306698799133301,
10259
+ "learning_rate": 1.207353547535953e-08,
10260
+ "logits/chosen": -10.826414108276367,
10261
+ "logits/rejected": -10.828743934631348,
10262
+ "logps/chosen": -22.830368041992188,
10263
+ "logps/rejected": -55.398826599121094,
10264
+ "loss": 0.2879,
10265
+ "rewards/accuracies": 0.6875,
10266
+ "rewards/chosen": -0.31796175241470337,
10267
+ "rewards/margins": 3.1197590827941895,
10268
+ "rewards/rejected": -3.437720775604248,
10269
+ "step": 1368
10270
+ },
10271
+ {
10272
+ "epoch": 1.7046426627264952,
10273
+ "grad_norm": 51.56437301635742,
10274
+ "learning_rate": 1.1875570073487785e-08,
10275
+ "logits/chosen": -10.848183631896973,
10276
+ "logits/rejected": -10.847176551818848,
10277
+ "logps/chosen": -25.83056640625,
10278
+ "logps/rejected": -67.20542907714844,
10279
+ "loss": 0.2889,
10280
+ "rewards/accuracies": 0.8125,
10281
+ "rewards/chosen": -0.63761967420578,
10282
+ "rewards/margins": 3.966122627258301,
10283
+ "rewards/rejected": -4.603742599487305,
10284
+ "step": 1370
10285
+ },
10286
+ {
10287
+ "epoch": 1.7071311921611323,
10288
+ "grad_norm": 9.324451446533203,
10289
+ "learning_rate": 1.1679138601412253e-08,
10290
+ "logits/chosen": -10.814537048339844,
10291
+ "logits/rejected": -10.819119453430176,
10292
+ "logps/chosen": -21.487300872802734,
10293
+ "logps/rejected": -64.1480712890625,
10294
+ "loss": 0.3161,
10295
+ "rewards/accuracies": 0.625,
10296
+ "rewards/chosen": -0.13521310687065125,
10297
+ "rewards/margins": 4.159755706787109,
10298
+ "rewards/rejected": -4.294968605041504,
10299
+ "step": 1372
10300
+ },
10301
+ {
10302
+ "epoch": 1.7096197215957694,
10303
+ "grad_norm": 18.857152938842773,
10304
+ "learning_rate": 1.1484244478299366e-08,
10305
+ "logits/chosen": -10.806600570678711,
10306
+ "logits/rejected": -10.810232162475586,
10307
+ "logps/chosen": -22.422306060791016,
10308
+ "logps/rejected": -55.91395950317383,
10309
+ "loss": 0.342,
10310
+ "rewards/accuracies": 0.75,
10311
+ "rewards/chosen": -0.2938164174556732,
10312
+ "rewards/margins": 3.184936761856079,
10313
+ "rewards/rejected": -3.4787533283233643,
10314
+ "step": 1374
10315
+ },
10316
+ {
10317
+ "epoch": 1.7121082510304069,
10318
+ "grad_norm": 5.337169647216797,
10319
+ "learning_rate": 1.1290891096555744e-08,
10320
+ "logits/chosen": -10.812167167663574,
10321
+ "logits/rejected": -10.815803527832031,
10322
+ "logps/chosen": -22.376619338989258,
10323
+ "logps/rejected": -64.20872497558594,
10324
+ "loss": 0.3014,
10325
+ "rewards/accuracies": 0.75,
10326
+ "rewards/chosen": -0.24579763412475586,
10327
+ "rewards/margins": 4.0266289710998535,
10328
+ "rewards/rejected": -4.272426605224609,
10329
+ "step": 1376
10330
+ },
10331
+ {
10332
+ "epoch": 1.714596780465044,
10333
+ "grad_norm": 30.906925201416016,
10334
+ "learning_rate": 1.1099081821769296e-08,
10335
+ "logits/chosen": -10.828575134277344,
10336
+ "logits/rejected": -10.829512596130371,
10337
+ "logps/chosen": -24.83142852783203,
10338
+ "logps/rejected": -53.37904357910156,
10339
+ "loss": 0.2964,
10340
+ "rewards/accuracies": 0.65625,
10341
+ "rewards/chosen": -0.4947197735309601,
10342
+ "rewards/margins": 2.6709203720092773,
10343
+ "rewards/rejected": -3.165640354156494,
10344
+ "step": 1378
10345
+ },
10346
+ {
10347
+ "epoch": 1.717085309899681,
10348
+ "grad_norm": 11.849616050720215,
10349
+ "learning_rate": 1.090881999265051e-08,
10350
+ "logits/chosen": -10.887191772460938,
10351
+ "logits/rejected": -10.888714790344238,
10352
+ "logps/chosen": -29.812423706054688,
10353
+ "logps/rejected": -71.98737335205078,
10354
+ "loss": 0.2891,
10355
+ "rewards/accuracies": 0.78125,
10356
+ "rewards/chosen": -0.9166967868804932,
10357
+ "rewards/margins": 4.209402561187744,
10358
+ "rewards/rejected": -5.126099109649658,
10359
+ "step": 1380
10360
+ },
10361
+ {
10362
+ "epoch": 1.7195738393343185,
10363
+ "grad_norm": 12.149803161621094,
10364
+ "learning_rate": 1.0720108920974469e-08,
10365
+ "logits/chosen": -10.778029441833496,
10366
+ "logits/rejected": -10.781806945800781,
10367
+ "logps/chosen": -23.731386184692383,
10368
+ "logps/rejected": -68.31604766845703,
10369
+ "loss": 0.3045,
10370
+ "rewards/accuracies": 0.65625,
10371
+ "rewards/chosen": -0.35708463191986084,
10372
+ "rewards/margins": 4.310543060302734,
10373
+ "rewards/rejected": -4.667627811431885,
10374
+ "step": 1382
10375
+ },
10376
+ {
10377
+ "epoch": 1.7220623687689556,
10378
+ "grad_norm": 12.090971946716309,
10379
+ "learning_rate": 1.0532951891523123e-08,
10380
+ "logits/chosen": -10.829472541809082,
10381
+ "logits/rejected": -10.826078414916992,
10382
+ "logps/chosen": -26.86117172241211,
10383
+ "logps/rejected": -62.79159927368164,
10384
+ "loss": 0.2627,
10385
+ "rewards/accuracies": 0.71875,
10386
+ "rewards/chosen": -0.7374318838119507,
10387
+ "rewards/margins": 3.4001479148864746,
10388
+ "rewards/rejected": -4.137579917907715,
10389
+ "step": 1384
10390
+ },
10391
+ {
10392
+ "epoch": 1.7245508982035929,
10393
+ "grad_norm": 129.79551696777344,
10394
+ "learning_rate": 1.0347352162028088e-08,
10395
+ "logits/chosen": -10.817343711853027,
10396
+ "logits/rejected": -10.824586868286133,
10397
+ "logps/chosen": -23.552154541015625,
10398
+ "logps/rejected": -61.207462310791016,
10399
+ "loss": 0.3297,
10400
+ "rewards/accuracies": 0.65625,
10401
+ "rewards/chosen": -0.3887484073638916,
10402
+ "rewards/margins": 3.627763271331787,
10403
+ "rewards/rejected": -4.0165114402771,
10404
+ "step": 1386
10405
+ },
10406
+ {
10407
+ "epoch": 1.7270394276382302,
10408
+ "grad_norm": 3.878584146499634,
10409
+ "learning_rate": 1.0163312963114035e-08,
10410
+ "logits/chosen": -10.860746383666992,
10411
+ "logits/rejected": -10.86077880859375,
10412
+ "logps/chosen": -28.208192825317383,
10413
+ "logps/rejected": -57.151390075683594,
10414
+ "loss": 0.3755,
10415
+ "rewards/accuracies": 0.6875,
10416
+ "rewards/chosen": -0.8361051082611084,
10417
+ "rewards/margins": 2.7554969787597656,
10418
+ "rewards/rejected": -3.591602087020874,
10419
+ "step": 1388
10420
+ },
10421
+ {
10422
+ "epoch": 1.7295279570728672,
10423
+ "grad_norm": 2.7282052040100098,
10424
+ "learning_rate": 9.980837498242357e-09,
10425
+ "logits/chosen": -10.849210739135742,
10426
+ "logits/rejected": -10.844944953918457,
10427
+ "logps/chosen": -26.760292053222656,
10428
+ "logps/rejected": -63.53279495239258,
10429
+ "loss": 0.2893,
10430
+ "rewards/accuracies": 0.78125,
10431
+ "rewards/chosen": -0.6709414720535278,
10432
+ "rewards/margins": 3.6953835487365723,
10433
+ "rewards/rejected": -4.366325378417969,
10434
+ "step": 1390
10435
+ },
10436
+ {
10437
+ "epoch": 1.7320164865075045,
10438
+ "grad_norm": 13.552352905273438,
10439
+ "learning_rate": 9.799928943655488e-09,
10440
+ "logits/chosen": -10.841174125671387,
10441
+ "logits/rejected": -10.840930938720703,
10442
+ "logps/chosen": -22.428604125976562,
10443
+ "logps/rejected": -58.52720642089844,
10444
+ "loss": 0.297,
10445
+ "rewards/accuracies": 0.75,
10446
+ "rewards/chosen": -0.2888602912425995,
10447
+ "rewards/margins": 3.4409823417663574,
10448
+ "rewards/rejected": -3.7298426628112793,
10449
+ "step": 1392
10450
+ },
10451
+ {
10452
+ "epoch": 1.7345050159421418,
10453
+ "grad_norm": 5.00348424911499,
10454
+ "learning_rate": 9.620590448321553e-09,
10455
+ "logits/chosen": -10.81922435760498,
10456
+ "logits/rejected": -10.820487976074219,
10457
+ "logps/chosen": -29.074127197265625,
10458
+ "logps/rejected": -61.92654800415039,
10459
+ "loss": 0.3292,
10460
+ "rewards/accuracies": 0.6875,
10461
+ "rewards/chosen": -0.9312417507171631,
10462
+ "rewards/margins": 3.181020498275757,
10463
+ "rewards/rejected": -4.11226224899292,
10464
+ "step": 1394
10465
+ },
10466
+ {
10467
+ "epoch": 1.7369935453767789,
10468
+ "grad_norm": 8.92225456237793,
10469
+ "learning_rate": 9.442825133879607e-09,
10470
+ "logits/chosen": -10.817237854003906,
10471
+ "logits/rejected": -10.821587562561035,
10472
+ "logps/chosen": -28.017192840576172,
10473
+ "logps/rejected": -59.26190185546875,
10474
+ "loss": 0.3484,
10475
+ "rewards/accuracies": 0.59375,
10476
+ "rewards/chosen": -0.82759690284729,
10477
+ "rewards/margins": 3.0587477684020996,
10478
+ "rewards/rejected": -3.8863444328308105,
10479
+ "step": 1396
10480
+ },
10481
+ {
10482
+ "epoch": 1.7394820748114161,
10483
+ "grad_norm": 14.537239074707031,
10484
+ "learning_rate": 9.2666360945853e-09,
10485
+ "logits/chosen": -10.792232513427734,
10486
+ "logits/rejected": -10.797674179077148,
10487
+ "logps/chosen": -28.772098541259766,
10488
+ "logps/rejected": -68.55633544921875,
10489
+ "loss": 0.3414,
10490
+ "rewards/accuracies": 0.71875,
10491
+ "rewards/chosen": -0.8768026828765869,
10492
+ "rewards/margins": 3.842247486114502,
10493
+ "rewards/rejected": -4.719050407409668,
10494
+ "step": 1398
10495
+ },
10496
+ {
10497
+ "epoch": 1.7419706042460534,
10498
+ "grad_norm": 7.152594566345215,
10499
+ "learning_rate": 9.092026397256913e-09,
10500
+ "logits/chosen": -10.825244903564453,
10501
+ "logits/rejected": -10.82767105102539,
10502
+ "logps/chosen": -19.60363006591797,
10503
+ "logps/rejected": -46.500396728515625,
10504
+ "loss": 0.3422,
10505
+ "rewards/accuracies": 0.71875,
10506
+ "rewards/chosen": -0.033207476139068604,
10507
+ "rewards/margins": 2.5526838302612305,
10508
+ "rewards/rejected": -2.5858914852142334,
10509
+ "step": 1400
10510
+ },
10511
+ {
10512
+ "epoch": 1.7444591336806905,
10513
+ "grad_norm": 10.673386573791504,
10514
+ "learning_rate": 8.918999081222156e-09,
10515
+ "logits/chosen": -10.80329418182373,
10516
+ "logits/rejected": -10.805293083190918,
10517
+ "logps/chosen": -24.16779899597168,
10518
+ "logps/rejected": -61.346763610839844,
10519
+ "loss": 0.2918,
10520
+ "rewards/accuracies": 0.78125,
10521
+ "rewards/chosen": -0.43302100896835327,
10522
+ "rewards/margins": 3.563382625579834,
10523
+ "rewards/rejected": -3.996403694152832,
10524
+ "step": 1402
10525
+ },
10526
+ {
10527
+ "epoch": 1.7469476631153278,
10528
+ "grad_norm": 1.928539514541626,
10529
+ "learning_rate": 8.747557158265073e-09,
10530
+ "logits/chosen": -10.844294548034668,
10531
+ "logits/rejected": -10.835131645202637,
10532
+ "logps/chosen": -25.480384826660156,
10533
+ "logps/rejected": -57.03585433959961,
10534
+ "loss": 0.3263,
10535
+ "rewards/accuracies": 0.75,
10536
+ "rewards/chosen": -0.5597760677337646,
10537
+ "rewards/margins": 2.9487338066101074,
10538
+ "rewards/rejected": -3.508509874343872,
10539
+ "step": 1404
10540
+ },
10541
+ {
10542
+ "epoch": 1.749436192549965,
10543
+ "grad_norm": 12.497711181640625,
10544
+ "learning_rate": 8.577703612573783e-09,
10545
+ "logits/chosen": -10.82893180847168,
10546
+ "logits/rejected": -10.83089828491211,
10547
+ "logps/chosen": -24.069276809692383,
10548
+ "logps/rejected": -66.67253112792969,
10549
+ "loss": 0.3401,
10550
+ "rewards/accuracies": 0.6875,
10551
+ "rewards/chosen": -0.4053136110305786,
10552
+ "rewards/margins": 4.1401472091674805,
10553
+ "rewards/rejected": -4.545460224151611,
10554
+ "step": 1406
10555
+ },
10556
+ {
10557
+ "epoch": 1.7519247219846021,
10558
+ "grad_norm": 22.551191329956055,
10559
+ "learning_rate": 8.409441400688399e-09,
10560
+ "logits/chosen": -10.837417602539062,
10561
+ "logits/rejected": -10.837981224060059,
10562
+ "logps/chosen": -21.41533660888672,
10563
+ "logps/rejected": -61.951133728027344,
10564
+ "loss": 0.3606,
10565
+ "rewards/accuracies": 0.75,
10566
+ "rewards/chosen": -0.19989776611328125,
10567
+ "rewards/margins": 3.897480010986328,
10568
+ "rewards/rejected": -4.097377777099609,
10569
+ "step": 1408
10570
+ },
10571
+ {
10572
+ "epoch": 1.7544132514192394,
10573
+ "grad_norm": 28.929887771606445,
10574
+ "learning_rate": 8.24277345144967e-09,
10575
+ "logits/chosen": -10.846136093139648,
10576
+ "logits/rejected": -10.846277236938477,
10577
+ "logps/chosen": -26.540502548217773,
10578
+ "logps/rejected": -40.02751159667969,
10579
+ "loss": 0.3718,
10580
+ "rewards/accuracies": 0.5,
10581
+ "rewards/chosen": -0.6883711814880371,
10582
+ "rewards/margins": 1.336569905281067,
10583
+ "rewards/rejected": -2.0249409675598145,
10584
+ "step": 1410
10585
+ },
10586
+ {
10587
+ "epoch": 1.7569017808538767,
10588
+ "grad_norm": 8.628479957580566,
10589
+ "learning_rate": 8.077702665947973e-09,
10590
+ "logits/chosen": -10.839735984802246,
10591
+ "logits/rejected": -10.83474063873291,
10592
+ "logps/chosen": -22.964393615722656,
10593
+ "logps/rejected": -69.26115417480469,
10594
+ "loss": 0.3341,
10595
+ "rewards/accuracies": 0.75,
10596
+ "rewards/chosen": -0.2930600941181183,
10597
+ "rewards/margins": 4.422696590423584,
10598
+ "rewards/rejected": -4.715756416320801,
10599
+ "step": 1412
10600
+ },
10601
+ {
10602
+ "epoch": 1.7593903102885138,
10603
+ "grad_norm": 28.3585205078125,
10604
+ "learning_rate": 7.914231917472746e-09,
10605
+ "logits/chosen": -10.830848693847656,
10606
+ "logits/rejected": -10.833673477172852,
10607
+ "logps/chosen": -23.10784149169922,
10608
+ "logps/rejected": -69.79341125488281,
10609
+ "loss": 0.3478,
10610
+ "rewards/accuracies": 0.75,
10611
+ "rewards/chosen": -0.35468006134033203,
10612
+ "rewards/margins": 4.51253604888916,
10613
+ "rewards/rejected": -4.867216110229492,
10614
+ "step": 1414
10615
+ },
10616
+ {
10617
+ "epoch": 1.761878839723151,
10618
+ "grad_norm": 8.422154426574707,
10619
+ "learning_rate": 7.75236405146258e-09,
10620
+ "logits/chosen": -10.832451820373535,
10621
+ "logits/rejected": -10.834232330322266,
10622
+ "logps/chosen": -28.385543823242188,
10623
+ "logps/rejected": -70.90185546875,
10624
+ "loss": 0.3218,
10625
+ "rewards/accuracies": 0.65625,
10626
+ "rewards/chosen": -0.7882261276245117,
10627
+ "rewards/margins": 4.108382701873779,
10628
+ "rewards/rejected": -4.896608829498291,
10629
+ "step": 1416
10630
+ },
10631
+ {
10632
+ "epoch": 1.7643673691577884,
10633
+ "grad_norm": 29.272140502929688,
10634
+ "learning_rate": 7.592101885455593e-09,
10635
+ "logits/chosen": -10.832554817199707,
10636
+ "logits/rejected": -10.830697059631348,
10637
+ "logps/chosen": -25.028911590576172,
10638
+ "logps/rejected": -58.800750732421875,
10639
+ "loss": 0.3287,
10640
+ "rewards/accuracies": 0.65625,
10641
+ "rewards/chosen": -0.5799860954284668,
10642
+ "rewards/margins": 3.158944606781006,
10643
+ "rewards/rejected": -3.7389309406280518,
10644
+ "step": 1418
10645
+ },
10646
+ {
10647
+ "epoch": 1.7668558985924254,
10648
+ "grad_norm": 63.721107482910156,
10649
+ "learning_rate": 7.4334482090404935e-09,
10650
+ "logits/chosen": -10.820099830627441,
10651
+ "logits/rejected": -10.814040184020996,
10652
+ "logps/chosen": -25.835189819335938,
10653
+ "logps/rejected": -54.49407958984375,
10654
+ "loss": 0.3232,
10655
+ "rewards/accuracies": 0.65625,
10656
+ "rewards/chosen": -0.6330866813659668,
10657
+ "rewards/margins": 2.6440882682800293,
10658
+ "rewards/rejected": -3.2771754264831543,
10659
+ "step": 1420
10660
+ },
10661
+ {
10662
+ "epoch": 1.7693444280270627,
10663
+ "grad_norm": 7.655338764190674,
10664
+ "learning_rate": 7.276405783807893e-09,
10665
+ "logits/chosen": -10.817934036254883,
10666
+ "logits/rejected": -10.813355445861816,
10667
+ "logps/chosen": -23.41266632080078,
10668
+ "logps/rejected": -64.01387786865234,
10669
+ "loss": 0.3305,
10670
+ "rewards/accuracies": 0.6875,
10671
+ "rewards/chosen": -0.35830530524253845,
10672
+ "rewards/margins": 3.9317524433135986,
10673
+ "rewards/rejected": -4.29005765914917,
10674
+ "step": 1422
10675
+ },
10676
+ {
10677
+ "epoch": 1.7718329574617,
10678
+ "grad_norm": 16.495460510253906,
10679
+ "learning_rate": 7.120977343302359e-09,
10680
+ "logits/chosen": -10.812400817871094,
10681
+ "logits/rejected": -10.806700706481934,
10682
+ "logps/chosen": -26.264169692993164,
10683
+ "logps/rejected": -59.75008010864258,
10684
+ "loss": 0.316,
10685
+ "rewards/accuracies": 0.6875,
10686
+ "rewards/chosen": -0.6269797682762146,
10687
+ "rewards/margins": 3.2161946296691895,
10688
+ "rewards/rejected": -3.8431742191314697,
10689
+ "step": 1424
10690
+ },
10691
+ {
10692
+ "epoch": 1.774321486896337,
10693
+ "grad_norm": 52.430747985839844,
10694
+ "learning_rate": 6.9671655929747884e-09,
10695
+ "logits/chosen": -10.831871032714844,
10696
+ "logits/rejected": -10.833426475524902,
10697
+ "logps/chosen": -24.449472427368164,
10698
+ "logps/rejected": -45.51153564453125,
10699
+ "loss": 0.3439,
10700
+ "rewards/accuracies": 0.5,
10701
+ "rewards/chosen": -0.46670615673065186,
10702
+ "rewards/margins": 2.025547981262207,
10703
+ "rewards/rejected": -2.4922542572021484,
10704
+ "step": 1426
10705
+ },
10706
+ {
10707
+ "epoch": 1.7768100163309746,
10708
+ "grad_norm": 1.0906507968902588,
10709
+ "learning_rate": 6.814973210135255e-09,
10710
+ "logits/chosen": -10.85045051574707,
10711
+ "logits/rejected": -10.853729248046875,
10712
+ "logps/chosen": -23.105697631835938,
10713
+ "logps/rejected": -72.16358184814453,
10714
+ "loss": 0.285,
10715
+ "rewards/accuracies": 0.8125,
10716
+ "rewards/chosen": -0.36262983083724976,
10717
+ "rewards/margins": 4.786954879760742,
10718
+ "rewards/rejected": -5.149584770202637,
10719
+ "step": 1428
10720
+ },
10721
+ {
10722
+ "epoch": 1.7792985457656116,
10723
+ "grad_norm": 4.459844589233398,
10724
+ "learning_rate": 6.664402843906514e-09,
10725
+ "logits/chosen": -10.799229621887207,
10726
+ "logits/rejected": -10.7999267578125,
10727
+ "logps/chosen": -31.46733283996582,
10728
+ "logps/rejected": -65.57007598876953,
10729
+ "loss": 0.299,
10730
+ "rewards/accuracies": 0.78125,
10731
+ "rewards/chosen": -1.1061089038848877,
10732
+ "rewards/margins": 3.301386833190918,
10733
+ "rewards/rejected": -4.407495975494385,
10734
+ "step": 1430
10735
+ },
10736
+ {
10737
+ "epoch": 1.7817870752002487,
10738
+ "grad_norm": 15.899592399597168,
10739
+ "learning_rate": 6.515457115177802e-09,
10740
+ "logits/chosen": -10.796611785888672,
10741
+ "logits/rejected": -10.7985258102417,
10742
+ "logps/chosen": -28.31963539123535,
10743
+ "logps/rejected": -66.415771484375,
10744
+ "loss": 0.2964,
10745
+ "rewards/accuracies": 0.625,
10746
+ "rewards/chosen": -0.8590655326843262,
10747
+ "rewards/margins": 3.653881549835205,
10748
+ "rewards/rejected": -4.512947082519531,
10749
+ "step": 1432
10750
+ },
10751
+ {
10752
+ "epoch": 1.7842756046348862,
10753
+ "grad_norm": 19.590593338012695,
10754
+ "learning_rate": 6.368138616559282e-09,
10755
+ "logits/chosen": -10.814291000366211,
10756
+ "logits/rejected": -10.816354751586914,
10757
+ "logps/chosen": -27.344654083251953,
10758
+ "logps/rejected": -57.43811798095703,
10759
+ "loss": 0.3212,
10760
+ "rewards/accuracies": 0.75,
10761
+ "rewards/chosen": -0.7679904103279114,
10762
+ "rewards/margins": 2.95462703704834,
10763
+ "rewards/rejected": -3.7226176261901855,
10764
+ "step": 1434
10765
+ },
10766
+ {
10767
+ "epoch": 1.7867641340695233,
10768
+ "grad_norm": 114.24454498291016,
10769
+ "learning_rate": 6.2224499123368576e-09,
10770
+ "logits/chosen": -10.82932186126709,
10771
+ "logits/rejected": -10.82844066619873,
10772
+ "logps/chosen": -28.379507064819336,
10773
+ "logps/rejected": -47.598514556884766,
10774
+ "loss": 0.4161,
10775
+ "rewards/accuracies": 0.6875,
10776
+ "rewards/chosen": -0.8367791175842285,
10777
+ "rewards/margins": 1.8975491523742676,
10778
+ "rewards/rejected": -2.734328269958496,
10779
+ "step": 1436
10780
+ },
10781
+ {
10782
+ "epoch": 1.7892526635041603,
10783
+ "grad_norm": 6.041500091552734,
10784
+ "learning_rate": 6.078393538427573e-09,
10785
+ "logits/chosen": -10.825316429138184,
10786
+ "logits/rejected": -10.831624031066895,
10787
+ "logps/chosen": -28.40496063232422,
10788
+ "logps/rejected": -74.67303466796875,
10789
+ "loss": 0.3551,
10790
+ "rewards/accuracies": 0.65625,
10791
+ "rewards/chosen": -0.8339643478393555,
10792
+ "rewards/margins": 4.416720390319824,
10793
+ "rewards/rejected": -5.25068473815918,
10794
+ "step": 1438
10795
+ },
10796
+ {
10797
+ "epoch": 1.7917411929387979,
10798
+ "grad_norm": 5.8538737297058105,
10799
+ "learning_rate": 5.93597200233551e-09,
10800
+ "logits/chosen": -10.814746856689453,
10801
+ "logits/rejected": -10.814842224121094,
10802
+ "logps/chosen": -25.401987075805664,
10803
+ "logps/rejected": -66.26980590820312,
10804
+ "loss": 0.2539,
10805
+ "rewards/accuracies": 0.84375,
10806
+ "rewards/chosen": -0.5636205673217773,
10807
+ "rewards/margins": 3.9558775424957275,
10808
+ "rewards/rejected": -4.519497871398926,
10809
+ "step": 1440
10810
+ },
10811
+ {
10812
+ "epoch": 1.794229722373435,
10813
+ "grad_norm": 32.84010314941406,
10814
+ "learning_rate": 5.795187783108002e-09,
10815
+ "logits/chosen": -10.800264358520508,
10816
+ "logits/rejected": -10.800673484802246,
10817
+ "logps/chosen": -29.123065948486328,
10818
+ "logps/rejected": -67.73751831054688,
10819
+ "loss": 0.3369,
10820
+ "rewards/accuracies": 0.6875,
10821
+ "rewards/chosen": -0.9221575260162354,
10822
+ "rewards/margins": 3.7697463035583496,
10823
+ "rewards/rejected": -4.691904067993164,
10824
+ "step": 1442
10825
+ },
10826
+ {
10827
+ "epoch": 1.7967182518080722,
10828
+ "grad_norm": 7.772801399230957,
10829
+ "learning_rate": 5.656043331292681e-09,
10830
+ "logits/chosen": -10.818206787109375,
10831
+ "logits/rejected": -10.823054313659668,
10832
+ "logps/chosen": -23.672134399414062,
10833
+ "logps/rejected": -51.89259719848633,
10834
+ "loss": 0.3593,
10835
+ "rewards/accuracies": 0.71875,
10836
+ "rewards/chosen": -0.4020726680755615,
10837
+ "rewards/margins": 2.701967716217041,
10838
+ "rewards/rejected": -3.1040403842926025,
10839
+ "step": 1444
10840
+ },
10841
+ {
10842
+ "epoch": 1.7992067812427095,
10843
+ "grad_norm": 6.961812973022461,
10844
+ "learning_rate": 5.518541068894622e-09,
10845
+ "logits/chosen": -10.823866844177246,
10846
+ "logits/rejected": -10.824728012084961,
10847
+ "logps/chosen": -33.00081253051758,
10848
+ "logps/rejected": -69.7044906616211,
10849
+ "loss": 0.3134,
10850
+ "rewards/accuracies": 0.78125,
10851
+ "rewards/chosen": -1.3416486978530884,
10852
+ "rewards/margins": 3.5815513134002686,
10853
+ "rewards/rejected": -4.9232001304626465,
10854
+ "step": 1446
10855
+ },
10856
+ {
10857
+ "epoch": 1.8016953106773466,
10858
+ "grad_norm": 31.4688720703125,
10859
+ "learning_rate": 5.382683389334375e-09,
10860
+ "logits/chosen": -10.822563171386719,
10861
+ "logits/rejected": -10.822004318237305,
10862
+ "logps/chosen": -21.888166427612305,
10863
+ "logps/rejected": -52.723751068115234,
10864
+ "loss": 0.339,
10865
+ "rewards/accuracies": 0.59375,
10866
+ "rewards/chosen": -0.21833878755569458,
10867
+ "rewards/margins": 2.944748640060425,
10868
+ "rewards/rejected": -3.1630876064300537,
10869
+ "step": 1448
10870
+ },
10871
+ {
10872
+ "epoch": 1.8041838401119838,
10873
+ "grad_norm": 14.762541770935059,
10874
+ "learning_rate": 5.248472657406122e-09,
10875
+ "logits/chosen": -10.821406364440918,
10876
+ "logits/rejected": -10.816853523254395,
10877
+ "logps/chosen": -28.940649032592773,
10878
+ "logps/rejected": -57.142417907714844,
10879
+ "loss": 0.3214,
10880
+ "rewards/accuracies": 0.6875,
10881
+ "rewards/chosen": -0.9556154608726501,
10882
+ "rewards/margins": 2.6828153133392334,
10883
+ "rewards/rejected": -3.638430595397949,
10884
+ "step": 1450
10885
+ },
10886
+ {
10887
+ "epoch": 1.8066723695466211,
10888
+ "grad_norm": 11.76990032196045,
10889
+ "learning_rate": 5.1159112092366676e-09,
10890
+ "logits/chosen": -10.837207794189453,
10891
+ "logits/rejected": -10.829712867736816,
10892
+ "logps/chosen": -29.688844680786133,
10893
+ "logps/rejected": -66.00604248046875,
10894
+ "loss": 0.3339,
10895
+ "rewards/accuracies": 0.65625,
10896
+ "rewards/chosen": -0.9291807413101196,
10897
+ "rewards/margins": 3.522550106048584,
10898
+ "rewards/rejected": -4.451730728149414,
10899
+ "step": 1452
10900
+ },
10901
+ {
10902
+ "epoch": 1.8091608989812582,
10903
+ "grad_norm": 2.2030696868896484,
10904
+ "learning_rate": 4.985001352244666e-09,
10905
+ "logits/chosen": -10.811424255371094,
10906
+ "logits/rejected": -10.815189361572266,
10907
+ "logps/chosen": -24.620361328125,
10908
+ "logps/rejected": -64.42039489746094,
10909
+ "loss": 0.3568,
10910
+ "rewards/accuracies": 0.65625,
10911
+ "rewards/chosen": -0.49651896953582764,
10912
+ "rewards/margins": 3.8992347717285156,
10913
+ "rewards/rejected": -4.395754337310791,
10914
+ "step": 1454
10915
+ },
10916
+ {
10917
+ "epoch": 1.8116494284158955,
10918
+ "grad_norm": 32.40312194824219,
10919
+ "learning_rate": 4.855745365100538e-09,
10920
+ "logits/chosen": -10.825475692749023,
10921
+ "logits/rejected": -10.827003479003906,
10922
+ "logps/chosen": -33.1482048034668,
10923
+ "logps/rejected": -60.821807861328125,
10924
+ "loss": 0.3544,
10925
+ "rewards/accuracies": 0.71875,
10926
+ "rewards/chosen": -1.3415831327438354,
10927
+ "rewards/margins": 2.7061212062835693,
10928
+ "rewards/rejected": -4.047704696655273,
10929
+ "step": 1456
10930
+ },
10931
+ {
10932
+ "epoch": 1.8141379578505328,
10933
+ "grad_norm": 5.195944786071777,
10934
+ "learning_rate": 4.728145497686753e-09,
10935
+ "logits/chosen": -10.808531761169434,
10936
+ "logits/rejected": -10.80744743347168,
10937
+ "logps/chosen": -27.157238006591797,
10938
+ "logps/rejected": -56.74330139160156,
10939
+ "loss": 0.3454,
10940
+ "rewards/accuracies": 0.6875,
10941
+ "rewards/chosen": -0.7516356706619263,
10942
+ "rewards/margins": 2.846492290496826,
10943
+ "rewards/rejected": -3.598127841949463,
10944
+ "step": 1458
10945
+ },
10946
+ {
10947
+ "epoch": 1.8166264872851698,
10948
+ "grad_norm": 3.704167366027832,
10949
+ "learning_rate": 4.60220397105866e-09,
10950
+ "logits/chosen": -10.799447059631348,
10951
+ "logits/rejected": -10.80561351776123,
10952
+ "logps/chosen": -26.86786651611328,
10953
+ "logps/rejected": -77.04949188232422,
10954
+ "loss": 0.2995,
10955
+ "rewards/accuracies": 0.75,
10956
+ "rewards/chosen": -0.7065700888633728,
10957
+ "rewards/margins": 4.82396125793457,
10958
+ "rewards/rejected": -5.530531406402588,
10959
+ "step": 1460
10960
+ },
10961
+ {
10962
+ "epoch": 1.8191150167198071,
10963
+ "grad_norm": 4.062532424926758,
10964
+ "learning_rate": 4.477922977405912e-09,
10965
+ "logits/chosen": -10.818702697753906,
10966
+ "logits/rejected": -10.819602966308594,
10967
+ "logps/chosen": -21.75606346130371,
10968
+ "logps/rejected": -57.006935119628906,
10969
+ "loss": 0.297,
10970
+ "rewards/accuracies": 0.6875,
10971
+ "rewards/chosen": -0.16213519871234894,
10972
+ "rewards/margins": 3.5094478130340576,
10973
+ "rewards/rejected": -3.6715829372406006,
10974
+ "step": 1462
10975
+ },
10976
+ {
10977
+ "epoch": 1.8216035461544444,
10978
+ "grad_norm": 24.149431228637695,
10979
+ "learning_rate": 4.355304680014171e-09,
10980
+ "logits/chosen": -10.831608772277832,
10981
+ "logits/rejected": -10.84537410736084,
10982
+ "logps/chosen": -24.60963249206543,
10983
+ "logps/rejected": -69.60423278808594,
10984
+ "loss": 0.3599,
10985
+ "rewards/accuracies": 0.59375,
10986
+ "rewards/chosen": -0.5038677453994751,
10987
+ "rewards/margins": 4.358825206756592,
10988
+ "rewards/rejected": -4.862692832946777,
10989
+ "step": 1464
10990
+ },
10991
+ {
10992
+ "epoch": 1.8240920755890815,
10993
+ "grad_norm": 5.768329620361328,
10994
+ "learning_rate": 4.234351213227605e-09,
10995
+ "logits/chosen": -10.844436645507812,
10996
+ "logits/rejected": -10.842813491821289,
10997
+ "logps/chosen": -22.979904174804688,
10998
+ "logps/rejected": -51.99818801879883,
10999
+ "loss": 0.331,
11000
+ "rewards/accuracies": 0.59375,
11001
+ "rewards/chosen": -0.32598400115966797,
11002
+ "rewards/margins": 2.7477188110351562,
11003
+ "rewards/rejected": -3.073702812194824,
11004
+ "step": 1466
11005
+ },
11006
+ {
11007
+ "epoch": 1.8265806050237188,
11008
+ "grad_norm": 7.735873222351074,
11009
+ "learning_rate": 4.1150646824116064e-09,
11010
+ "logits/chosen": -10.821906089782715,
11011
+ "logits/rejected": -10.81991195678711,
11012
+ "logps/chosen": -23.383838653564453,
11013
+ "logps/rejected": -50.20802688598633,
11014
+ "loss": 0.3584,
11015
+ "rewards/accuracies": 0.59375,
11016
+ "rewards/chosen": -0.32233789563179016,
11017
+ "rewards/margins": 2.645047426223755,
11018
+ "rewards/rejected": -2.9673855304718018,
11019
+ "step": 1468
11020
+ },
11021
+ {
11022
+ "epoch": 1.829069134458356,
11023
+ "grad_norm": 9.88779067993164,
11024
+ "learning_rate": 3.997447163916223e-09,
11025
+ "logits/chosen": -10.780189514160156,
11026
+ "logits/rejected": -10.784330368041992,
11027
+ "logps/chosen": -26.26874542236328,
11028
+ "logps/rejected": -64.21620178222656,
11029
+ "loss": 0.315,
11030
+ "rewards/accuracies": 0.625,
11031
+ "rewards/chosen": -0.6374044418334961,
11032
+ "rewards/margins": 3.6502692699432373,
11033
+ "rewards/rejected": -4.287673473358154,
11034
+ "step": 1470
11035
+ },
11036
+ {
11037
+ "epoch": 1.8315576638929931,
11038
+ "grad_norm": 3.3723583221435547,
11039
+ "learning_rate": 3.8815007050399975e-09,
11040
+ "logits/chosen": -10.845205307006836,
11041
+ "logits/rejected": -10.850592613220215,
11042
+ "logps/chosen": -24.638830184936523,
11043
+ "logps/rejected": -62.163787841796875,
11044
+ "loss": 0.3005,
11045
+ "rewards/accuracies": 0.6875,
11046
+ "rewards/chosen": -0.4939250349998474,
11047
+ "rewards/margins": 3.6708312034606934,
11048
+ "rewards/rejected": -4.1647562980651855,
11049
+ "step": 1472
11050
+ },
11051
+ {
11052
+ "epoch": 1.8340461933276304,
11053
+ "grad_norm": 2.0998055934906006,
11054
+ "learning_rate": 3.767227323994293e-09,
11055
+ "logits/chosen": -10.836874008178711,
11056
+ "logits/rejected": -10.8333740234375,
11057
+ "logps/chosen": -20.189462661743164,
11058
+ "logps/rejected": -55.19818878173828,
11059
+ "loss": 0.3569,
11060
+ "rewards/accuracies": 0.46875,
11061
+ "rewards/chosen": -0.0895194411277771,
11062
+ "rewards/margins": 3.2878525257110596,
11063
+ "rewards/rejected": -3.3773720264434814,
11064
+ "step": 1474
11065
+ },
11066
+ {
11067
+ "epoch": 1.8365347227622677,
11068
+ "grad_norm": 18.310754776000977,
11069
+ "learning_rate": 3.6546290098682485e-09,
11070
+ "logits/chosen": -10.863044738769531,
11071
+ "logits/rejected": -10.86330509185791,
11072
+ "logps/chosen": -24.017717361450195,
11073
+ "logps/rejected": -58.69614028930664,
11074
+ "loss": 0.3456,
11075
+ "rewards/accuracies": 0.6875,
11076
+ "rewards/chosen": -0.43135157227516174,
11077
+ "rewards/margins": 3.3879075050354004,
11078
+ "rewards/rejected": -3.819258689880371,
11079
+ "step": 1476
11080
+ },
11081
+ {
11082
+ "epoch": 1.8390232521969048,
11083
+ "grad_norm": 3.3576934337615967,
11084
+ "learning_rate": 3.543707722594069e-09,
11085
+ "logits/chosen": -10.862442016601562,
11086
+ "logits/rejected": -10.859800338745117,
11087
+ "logps/chosen": -27.191797256469727,
11088
+ "logps/rejected": -71.88859558105469,
11089
+ "loss": 0.3095,
11090
+ "rewards/accuracies": 0.6875,
11091
+ "rewards/chosen": -0.7576779723167419,
11092
+ "rewards/margins": 4.214293479919434,
11093
+ "rewards/rejected": -4.971971035003662,
11094
+ "step": 1478
11095
+ },
11096
+ {
11097
+ "epoch": 1.841511781631542,
11098
+ "grad_norm": 6.704062461853027,
11099
+ "learning_rate": 3.4344653929129554e-09,
11100
+ "logits/chosen": -10.799659729003906,
11101
+ "logits/rejected": -10.798733711242676,
11102
+ "logps/chosen": -19.196088790893555,
11103
+ "logps/rejected": -47.406768798828125,
11104
+ "loss": 0.3276,
11105
+ "rewards/accuracies": 0.6875,
11106
+ "rewards/chosen": 0.06628356873989105,
11107
+ "rewards/margins": 2.775508403778076,
11108
+ "rewards/rejected": -2.7092249393463135,
11109
+ "step": 1480
11110
+ },
11111
+ {
11112
+ "epoch": 1.8440003110661793,
11113
+ "grad_norm": 7.278040409088135,
11114
+ "learning_rate": 3.326903922341473e-09,
11115
+ "logits/chosen": -10.815013885498047,
11116
+ "logits/rejected": -10.810517311096191,
11117
+ "logps/chosen": -23.322364807128906,
11118
+ "logps/rejected": -60.02410888671875,
11119
+ "loss": 0.3409,
11120
+ "rewards/accuracies": 0.71875,
11121
+ "rewards/chosen": -0.3628728985786438,
11122
+ "rewards/margins": 3.500354290008545,
11123
+ "rewards/rejected": -3.863227367401123,
11124
+ "step": 1482
11125
+ },
11126
+ {
11127
+ "epoch": 1.8464888405008164,
11128
+ "grad_norm": 18.574892044067383,
11129
+ "learning_rate": 3.221025183138493e-09,
11130
+ "logits/chosen": -10.82407283782959,
11131
+ "logits/rejected": -10.823734283447266,
11132
+ "logps/chosen": -25.93548011779785,
11133
+ "logps/rejected": -61.550941467285156,
11134
+ "loss": 0.2756,
11135
+ "rewards/accuracies": 0.6875,
11136
+ "rewards/chosen": -0.6312334537506104,
11137
+ "rewards/margins": 3.374722957611084,
11138
+ "rewards/rejected": -4.005956649780273,
11139
+ "step": 1484
11140
+ },
11141
+ {
11142
+ "epoch": 1.848977369935454,
11143
+ "grad_norm": 28.313570022583008,
11144
+ "learning_rate": 3.116831018272581e-09,
11145
+ "logits/chosen": -10.880552291870117,
11146
+ "logits/rejected": -10.877769470214844,
11147
+ "logps/chosen": -22.949565887451172,
11148
+ "logps/rejected": -65.82709503173828,
11149
+ "loss": 0.2973,
11150
+ "rewards/accuracies": 0.78125,
11151
+ "rewards/chosen": -0.3210931420326233,
11152
+ "rewards/margins": 4.221843719482422,
11153
+ "rewards/rejected": -4.5429368019104,
11154
+ "step": 1486
11155
+ },
11156
+ {
11157
+ "epoch": 1.851465899370091,
11158
+ "grad_norm": 50.520294189453125,
11159
+ "learning_rate": 3.0143232413898602e-09,
11160
+ "logits/chosen": -10.797541618347168,
11161
+ "logits/rejected": -10.794352531433105,
11162
+ "logps/chosen": -28.862197875976562,
11163
+ "logps/rejected": -75.26273345947266,
11164
+ "loss": 0.2318,
11165
+ "rewards/accuracies": 0.8125,
11166
+ "rewards/chosen": -0.8763686418533325,
11167
+ "rewards/margins": 4.540627956390381,
11168
+ "rewards/rejected": -5.416996479034424,
11169
+ "step": 1488
11170
+ },
11171
+ {
11172
+ "epoch": 1.853954428804728,
11173
+ "grad_norm": 8.633639335632324,
11174
+ "learning_rate": 2.913503636782577e-09,
11175
+ "logits/chosen": -10.839570999145508,
11176
+ "logits/rejected": -10.838019371032715,
11177
+ "logps/chosen": -22.53410530090332,
11178
+ "logps/rejected": -46.19023513793945,
11179
+ "loss": 0.3611,
11180
+ "rewards/accuracies": 0.5,
11181
+ "rewards/chosen": -0.3035596013069153,
11182
+ "rewards/margins": 2.2712674140930176,
11183
+ "rewards/rejected": -2.574826955795288,
11184
+ "step": 1490
11185
+ },
11186
+ {
11187
+ "epoch": 1.8564429582393656,
11188
+ "grad_norm": 5.074777126312256,
11189
+ "learning_rate": 2.8143739593578853e-09,
11190
+ "logits/chosen": -10.803709030151367,
11191
+ "logits/rejected": -10.803434371948242,
11192
+ "logps/chosen": -23.481767654418945,
11193
+ "logps/rejected": -63.84934997558594,
11194
+ "loss": 0.3185,
11195
+ "rewards/accuracies": 0.65625,
11196
+ "rewards/chosen": -0.3606772720813751,
11197
+ "rewards/margins": 3.896145820617676,
11198
+ "rewards/rejected": -4.2568230628967285,
11199
+ "step": 1492
11200
+ },
11201
+ {
11202
+ "epoch": 1.8589314876740026,
11203
+ "grad_norm": 8.714546203613281,
11204
+ "learning_rate": 2.716935934607434e-09,
11205
+ "logits/chosen": -10.823909759521484,
11206
+ "logits/rejected": -10.820356369018555,
11207
+ "logps/chosen": -23.897640228271484,
11208
+ "logps/rejected": -52.00101089477539,
11209
+ "loss": 0.3194,
11210
+ "rewards/accuracies": 0.6875,
11211
+ "rewards/chosen": -0.44381412863731384,
11212
+ "rewards/margins": 2.6014151573181152,
11213
+ "rewards/rejected": -3.045229196548462,
11214
+ "step": 1494
11215
+ },
11216
+ {
11217
+ "epoch": 1.86142001710864,
11218
+ "grad_norm": 4.605140209197998,
11219
+ "learning_rate": 2.6211912585772377e-09,
11220
+ "logits/chosen": -10.820908546447754,
11221
+ "logits/rejected": -10.817206382751465,
11222
+ "logps/chosen": -22.388513565063477,
11223
+ "logps/rejected": -52.58741760253906,
11224
+ "loss": 0.3037,
11225
+ "rewards/accuracies": 0.59375,
11226
+ "rewards/chosen": -0.27936896681785583,
11227
+ "rewards/margins": 2.8748831748962402,
11228
+ "rewards/rejected": -3.154252052307129,
11229
+ "step": 1496
11230
+ },
11231
+ {
11232
+ "epoch": 1.8639085465432772,
11233
+ "grad_norm": 6.918066024780273,
11234
+ "learning_rate": 2.5271415978382116e-09,
11235
+ "logits/chosen": -10.764432907104492,
11236
+ "logits/rejected": -10.769204139709473,
11237
+ "logps/chosen": -21.751548767089844,
11238
+ "logps/rejected": -47.14215087890625,
11239
+ "loss": 0.3786,
11240
+ "rewards/accuracies": 0.65625,
11241
+ "rewards/chosen": -0.1537971943616867,
11242
+ "rewards/margins": 2.537064790725708,
11243
+ "rewards/rejected": -2.690861701965332,
11244
+ "step": 1498
11245
+ },
11246
+ {
11247
+ "epoch": 1.8663970759779143,
11248
+ "grad_norm": 12.986876487731934,
11249
+ "learning_rate": 2.4347885894571484e-09,
11250
+ "logits/chosen": -10.837455749511719,
11251
+ "logits/rejected": -10.839674949645996,
11252
+ "logps/chosen": -28.00076675415039,
11253
+ "logps/rejected": -66.6291275024414,
11254
+ "loss": 0.3554,
11255
+ "rewards/accuracies": 0.6875,
11256
+ "rewards/chosen": -0.8600501418113708,
11257
+ "rewards/margins": 3.6425986289978027,
11258
+ "rewards/rejected": -4.502648830413818,
11259
+ "step": 1500
11260
  }
11261
  ],
11262
  "logging_steps": 2,