Training in progress, epoch 1, checkpoint

last-checkpoint/global_step313/bf16_zero_pp_rank_0_mp_rank_00_optim_states.pt

@@ -1,3 +1,3 @@
 version https://git-lfs.github.com/spec/v1
-oid sha256:2ba276ff131c339916e52cb67158d2a04461c6fe832c629c212cb7bdb074409f
+oid sha256:c6d9f91b8139c9e79b5d87ea33dd7f4d0e9e2d153ab712609579fd81dc67a7f5
 size 14483468456

last-checkpoint/global_step313/bf16_zero_pp_rank_1_mp_rank_00_optim_states.pt

@@ -1,3 +1,3 @@
 version https://git-lfs.github.com/spec/v1
-oid sha256:b1004ac89f774ab79a9916ebb37e943c1b3aa2271816e21e357f1732059ab2f1
+oid sha256:cf23767d73dbf16bf26c0c158679348e2b396bf20bfa76d6ff4ea056d1393402
 size 14483468456

last-checkpoint/global_step313/bf16_zero_pp_rank_2_mp_rank_00_optim_states.pt

@@ -1,3 +1,3 @@
 version https://git-lfs.github.com/spec/v1
-oid sha256:dc4b71867757571a50b62c4b3441dc64372ea231c85843bc2c9f6dc04d8f9ebd
+oid sha256:b8c2943a40e459f88c5f82020c41539025dc72a97bdd791e7864f7522bf9471a
 size 14483468456

last-checkpoint/global_step313/bf16_zero_pp_rank_3_mp_rank_00_optim_states.pt

@@ -1,3 +1,3 @@
 version https://git-lfs.github.com/spec/v1
-oid sha256:6569fe52a3107a1d4eb1b4625ba7c2ef9ed0556ab45f04c6cf9da38bfaf77395
+oid sha256:b8708e717ac6c0e66e5b6025a9856d8a7443d35eb1f8ffc788352d47febe16a4
 size 14483468456

last-checkpoint/latest

@@ -1 +1 @@
-global_step939
+global_step313

last-checkpoint/model-00001-of-00003.safetensors

@@ -1,3 +1,3 @@
 version https://git-lfs.github.com/spec/v1
-oid sha256:eca3b3f802b4e2c4a57eedfb70509258c2809baf1b47d4ca9f71c08d50554441
+oid sha256:47a89a06839841c7eb87dcbed4e8913f1671af0d8ed467692e12b0a36f7b66d5
 size 4943162336

last-checkpoint/model-00002-of-00003.safetensors

@@ -1,3 +1,3 @@
 version https://git-lfs.github.com/spec/v1
-oid sha256:bd61800d71ab4509ac872cd60d757f9115b3ba96f61e6774bb9b416737542cb5
+oid sha256:fd92879869d7067d8e76adbd5cc290284d880eddaf06396d04c9b4533088766b
 size 4999819336

last-checkpoint/model-00003-of-00003.safetensors

@@ -1,3 +1,3 @@
 version https://git-lfs.github.com/spec/v1
-oid sha256:938aeb2b57192138bb70e667bc639fd64d11ea5e3d20ea91aef236c86dfd9e4e
+oid sha256:3a56de2e70c5e45089154d64b749c86385b3aca73991e54c21bce0a5140a6d3a
 size 4540516344

last-checkpoint/rng_state_0.pth

@@ -1,3 +1,3 @@
 version https://git-lfs.github.com/spec/v1
-oid sha256:ab7cf0d34d60dfda516cf9661904550e2e294e723edd07c25c738f05e8ba92d1
+oid sha256:f1ce770862b76ceecb3453a5d0b9ef65311eab0f6a60f52997726a89d718de08
 size 15024

last-checkpoint/rng_state_1.pth

@@ -1,3 +1,3 @@
 version https://git-lfs.github.com/spec/v1
-oid sha256:f7b183324e8227a51a9556d86b2ad893a8c4c52205ed4a737356c6611dac7353
+oid sha256:2bcae6573442578b752be3e988d4ccb38056b45c31b5c02eb579ec6cebcfa62e
 size 15024

last-checkpoint/rng_state_2.pth

@@ -1,3 +1,3 @@
 version https://git-lfs.github.com/spec/v1
-oid sha256:ac69e994090f4818cb1fa6f6cefa363178552c3c731c6507ff195bcb07fd5bef
+oid sha256:398ff45072fee2975e88b9e078f915103d75d94bc08753303cf855b915973623
 size 15024

last-checkpoint/rng_state_3.pth

@@ -1,3 +1,3 @@
 version https://git-lfs.github.com/spec/v1
-oid sha256:f68e61b63402f8afb1f69c960f7944965655dac11e3ccf29919c282f23931f86
+oid sha256:7f4b210bc832d4c648e3bee8c21dca26e5a8b365d6ec90c638062005a052e57b
 size 15024

last-checkpoint/scheduler.pt

@@ -1,3 +1,3 @@
 version https://git-lfs.github.com/spec/v1
-oid sha256:640747ca69b57ef58ae079eec14158d13652c8bf30fb7c733fab30d65edaeb50
+oid sha256:c019f2bd3222f739db75f458caa9c28765649010b6fa70739851be6f2b00c68b
 size 1064

last-checkpoint/trainer_state.json

@@ -1,21 +1,21 @@
 {
   "best_metric": null,
   "best_model_checkpoint": null,
-  "epoch": 3.0,
+  "epoch": 1.0,
   "eval_steps": 500,
-  "global_step": 939,
+  "global_step": 313,
   "is_hyper_param_search": false,
   "is_local_process_zero": true,
   "is_world_process_zero": true,
   "log_history": [
     {
       "epoch": 0.003194888178913738,
-      "grad_norm": 63.76829168210548,
+      "grad_norm": 106.83801953543872,
       "learning_rate": 5.3191489361702125e-09,
-      "logits/chosen": -3.703125,
-      "logits/rejected": -3.859375,
-      "logps/chosen": -139.0,
-      "logps/rejected": -80.0,
+      "logits/chosen": -3.578125,
+      "logits/rejected": -3.421875,
+      "logps/chosen": -282.0,
+      "logps/rejected": -290.0,
       "loss": 0.6914,
       "rewards/accuracies": 0.0,
       "rewards/chosen": 0.0,
@@ -25,1446 +25,484 @@
     },
     {
       "epoch": 0.03194888178913738,
-      "grad_norm": 92.94577098387109,
+      "grad_norm": 98.60082387787162,
       "learning_rate": 5.3191489361702123e-08,
-      "logits/chosen": -3.625,
-      "logits/rejected": -3.53125,
-      "logps/chosen": -99.5,
-      "logps/rejected": -96.0,
-      "loss": 0.6939,
-      "rewards/accuracies": 0.1527777761220932,
-      "rewards/chosen": -0.0010528564453125,
-      "rewards/margins": -0.001739501953125,
-      "rewards/rejected": 0.0006866455078125,
+      "logits/chosen": -3.484375,
+      "logits/rejected": -3.390625,
+      "logps/chosen": -274.0,
+      "logps/rejected": -278.0,
+      "loss": 0.6868,
+      "rewards/accuracies": 0.1388888955116272,
+      "rewards/chosen": -0.00970458984375,
+      "rewards/margins": 6.794929504394531e-06,
+      "rewards/rejected": -0.009765625,
       "step": 10
     },
     {
       "epoch": 0.06389776357827476,
-      "grad_norm": 70.79325392606067,
+      "grad_norm": 116.57506460735225,
       "learning_rate": 1.0638297872340425e-07,
-      "logits/chosen": -3.375,
-      "logits/rejected": -3.5625,
-      "logps/chosen": -96.0,
-      "logps/rejected": -105.0,
-      "loss": 0.6872,
-      "rewards/accuracies": 0.4375,
-      "rewards/chosen": -0.0242919921875,
-      "rewards/margins": 0.017333984375,
-      "rewards/rejected": -0.04150390625,
+      "logits/chosen": -3.421875,
+      "logits/rejected": -3.421875,
+      "logps/chosen": -280.0,
+      "logps/rejected": -266.0,
+      "loss": 0.6537,
+      "rewards/accuracies": 0.637499988079071,
+      "rewards/chosen": -0.03125,
+      "rewards/margins": 0.12158203125,
+      "rewards/rejected": -0.15234375,
       "step": 20
     },
     {
       "epoch": 0.09584664536741214,
-      "grad_norm": 63.190582339280105,
+      "grad_norm": 78.04564900745318,
       "learning_rate": 1.5957446808510638e-07,
-      "logits/chosen": -3.5625,
-      "logits/rejected": -3.453125,
-      "logps/chosen": -99.5,
-      "logps/rejected": -108.0,
-      "loss": 0.671,
-      "rewards/accuracies": 0.5625,
-      "rewards/chosen": -0.08935546875,
-      "rewards/margins": 0.0361328125,
-      "rewards/rejected": -0.125,
+      "logits/chosen": -3.34375,
+      "logits/rejected": -3.484375,
+      "logps/chosen": -276.0,
+      "logps/rejected": -268.0,
+      "loss": 0.6477,
+      "rewards/accuracies": 0.625,
+      "rewards/chosen": -0.1552734375,
+      "rewards/margins": 0.1826171875,
+      "rewards/rejected": -0.337890625,
       "step": 30
     },
     {
       "epoch": 0.12779552715654952,
-      "grad_norm": 70.46556001874696,
+      "grad_norm": 104.09831779598169,
       "learning_rate": 2.127659574468085e-07,
-      "logits/chosen": -3.578125,
-      "logits/rejected": -3.59375,
-      "logps/chosen": -110.0,
-      "logps/rejected": -95.5,
-      "loss": 0.6578,
-      "rewards/accuracies": 0.48750001192092896,
-      "rewards/chosen": -0.1669921875,
-      "rewards/margins": 0.0164794921875,
-      "rewards/rejected": -0.1826171875,
+      "logits/chosen": -3.46875,
+      "logits/rejected": -3.421875,
+      "logps/chosen": -272.0,
+      "logps/rejected": -276.0,
+      "loss": 0.5849,
+      "rewards/accuracies": 0.6625000238418579,
+      "rewards/chosen": -0.1923828125,
+      "rewards/margins": 0.341796875,
+      "rewards/rejected": -0.53515625,
       "step": 40
     },
     {
       "epoch": 0.1597444089456869,
-      "grad_norm": 53.92661468897121,
+      "grad_norm": 83.06087118508516,
       "learning_rate": 2.659574468085106e-07,
-      "logits/chosen": -3.6875,
-      "logits/rejected": -3.640625,
-      "logps/chosen": -95.0,
-      "logps/rejected": -106.5,
-      "loss": 0.6153,
-      "rewards/accuracies": 0.574999988079071,
-      "rewards/chosen": -0.251953125,
-      "rewards/margins": 0.1376953125,
-      "rewards/rejected": -0.388671875,
+      "logits/chosen": -3.375,
+      "logits/rejected": -3.40625,
+      "logps/chosen": -278.0,
+      "logps/rejected": -274.0,
+      "loss": 0.6062,
+      "rewards/accuracies": 0.6000000238418579,
+      "rewards/chosen": -0.34375,
+      "rewards/margins": 0.244140625,
+      "rewards/rejected": -0.5859375,
       "step": 50
     },
     {
       "epoch": 0.19169329073482427,
-      "grad_norm": 61.31819398041802,
+      "grad_norm": 85.2326419243345,
       "learning_rate": 3.1914893617021275e-07,
-      "logits/chosen": -3.453125,
-      "logits/rejected": -3.328125,
-      "logps/chosen": -101.0,
-      "logps/rejected": -107.0,
-      "loss": 0.7335,
-      "rewards/accuracies": 0.6875,
-      "rewards/chosen": -0.283203125,
-      "rewards/margins": 0.478515625,
-      "rewards/rejected": -0.76171875,
+      "logits/chosen": -3.28125,
+      "logits/rejected": -3.3125,
+      "logps/chosen": -280.0,
+      "logps/rejected": -276.0,
+      "loss": 0.5718,
+      "rewards/accuracies": 0.7124999761581421,
+      "rewards/chosen": -0.314453125,
+      "rewards/margins": 0.5546875,
+      "rewards/rejected": -0.87109375,
       "step": 60
     },
     {
       "epoch": 0.22364217252396165,
-      "grad_norm": 55.44332583758203,
+      "grad_norm": 85.35363076925235,
       "learning_rate": 3.7234042553191484e-07,
-      "logits/chosen": -3.421875,
-      "logits/rejected": -3.546875,
-      "logps/chosen": -103.0,
-      "logps/rejected": -112.0,
-      "loss": 0.6863,
-      "rewards/accuracies": 0.6875,
-      "rewards/chosen": -0.1220703125,
-      "rewards/margins": 0.6328125,
-      "rewards/rejected": -0.75390625,
+      "logits/chosen": -3.28125,
+      "logits/rejected": -3.296875,
+      "logps/chosen": -278.0,
+      "logps/rejected": -278.0,
+      "loss": 0.5627,
+      "rewards/accuracies": 0.6625000238418579,
+      "rewards/chosen": -0.244140625,
+      "rewards/margins": 0.451171875,
+      "rewards/rejected": -0.6953125,
       "step": 70
     },
     {
       "epoch": 0.25559105431309903,
-      "grad_norm": 51.149001898192076,
+      "grad_norm": 80.13630527816748,
       "learning_rate": 4.25531914893617e-07,
-      "logits/chosen": -3.546875,
-      "logits/rejected": -3.53125,
-      "logps/chosen": -110.5,
-      "logps/rejected": -110.5,
-      "loss": 0.5805,
-      "rewards/accuracies": 0.75,
-      "rewards/chosen": -0.0986328125,
-      "rewards/margins": 0.62890625,
-      "rewards/rejected": -0.7265625,
+      "logits/chosen": -3.328125,
+      "logits/rejected": -3.296875,
+      "logps/chosen": -280.0,
+      "logps/rejected": -274.0,
+      "loss": 0.5984,
+      "rewards/accuracies": 0.699999988079071,
+      "rewards/chosen": -0.3203125,
+      "rewards/margins": 0.546875,
+      "rewards/rejected": -0.8671875,
       "step": 80
     },
     {
       "epoch": 0.28753993610223644,
-      "grad_norm": 55.06999142181094,
+      "grad_norm": 64.74916812090929,
       "learning_rate": 4.787234042553192e-07,
-      "logits/chosen": -3.671875,
-      "logits/rejected": -3.59375,
-      "logps/chosen": -107.0,
-      "logps/rejected": -116.0,
-      "loss": 0.5765,
-      "rewards/accuracies": 0.7124999761581421,
-      "rewards/chosen": -0.1494140625,
-      "rewards/margins": 0.7734375,
+      "logits/chosen": -3.375,
+      "logits/rejected": -3.296875,
+      "logps/chosen": -274.0,
+      "logps/rejected": -278.0,
+      "loss": 0.5594,
+      "rewards/accuracies": 0.6875,
+      "rewards/chosen": -0.224609375,
+      "rewards/margins": 0.69921875,
       "rewards/rejected": -0.921875,
       "step": 90
     },
     {
       "epoch": 0.3194888178913738,
-      "grad_norm": 33.54407800330637,
+      "grad_norm": 62.874343395319244,
       "learning_rate": 4.964497041420119e-07,
-      "logits/chosen": -3.546875,
-      "logits/rejected": -3.53125,
-      "logps/chosen": -105.0,
-      "logps/rejected": -112.5,
-      "loss": 0.5228,
-      "rewards/accuracies": 0.824999988079071,
-      "rewards/chosen": -0.1162109375,
-      "rewards/margins": 0.61328125,
-      "rewards/rejected": -0.7265625,
+      "logits/chosen": -3.34375,
+      "logits/rejected": -3.34375,
+      "logps/chosen": -280.0,
+      "logps/rejected": -268.0,
+      "loss": 0.5807,
+      "rewards/accuracies": 0.675000011920929,
+      "rewards/chosen": -0.3203125,
+      "rewards/margins": 0.55078125,
+      "rewards/rejected": -0.87109375,
       "step": 100
     },
     {
       "epoch": 0.3514376996805112,
-      "grad_norm": 71.5169022650209,
+      "grad_norm": 65.70150692167816,
       "learning_rate": 4.905325443786982e-07,
-      "logits/chosen": -3.5625,
-      "logits/rejected": -3.515625,
-      "logps/chosen": -101.5,
-      "logps/rejected": -123.0,
-      "loss": 0.5408,
-      "rewards/accuracies": 0.6875,
-      "rewards/chosen": -0.1337890625,
-      "rewards/margins": 0.8515625,
-      "rewards/rejected": -0.984375,
+      "logits/chosen": -3.28125,
+      "logits/rejected": -3.328125,
+      "logps/chosen": -276.0,
+      "logps/rejected": -286.0,
+      "loss": 0.5878,
+      "rewards/accuracies": 0.7124999761581421,
+      "rewards/chosen": -0.1162109375,
+      "rewards/margins": 0.75,
+      "rewards/rejected": -0.8671875,
       "step": 110
     },
     {
       "epoch": 0.38338658146964855,
-      "grad_norm": 65.33074960619584,
+      "grad_norm": 66.45326771352913,
       "learning_rate": 4.846153846153846e-07,
-      "logits/chosen": -3.546875,
-      "logits/rejected": -3.5625,
-      "logps/chosen": -100.5,
-      "logps/rejected": -139.0,
-      "loss": 0.5381,
-      "rewards/accuracies": 0.800000011920929,
-      "rewards/chosen": 0.021484375,
-      "rewards/margins": 1.671875,
-      "rewards/rejected": -1.6484375,
+      "logits/chosen": -3.296875,
+      "logits/rejected": -3.375,
+      "logps/chosen": -280.0,
+      "logps/rejected": -276.0,
+      "loss": 0.5633,
+      "rewards/accuracies": 0.6875,
+      "rewards/chosen": -0.1328125,
+      "rewards/margins": 0.67578125,
+      "rewards/rejected": -0.80859375,
       "step": 120
     },
     {
       "epoch": 0.41533546325878595,
-      "grad_norm": 49.04091358646178,
+      "grad_norm": 83.00130651053203,
       "learning_rate": 4.78698224852071e-07,
-      "logits/chosen": -3.515625,
-      "logits/rejected": -3.5,
-      "logps/chosen": -95.0,
-      "logps/rejected": -116.5,
-      "loss": 0.5387,
-      "rewards/accuracies": 0.762499988079071,
-      "rewards/chosen": -0.10791015625,
-      "rewards/margins": 0.99609375,
-      "rewards/rejected": -1.1015625,
+      "logits/chosen": -3.3125,
+      "logits/rejected": -3.296875,
+      "logps/chosen": -282.0,
+      "logps/rejected": -276.0,
+      "loss": 0.5831,
+      "rewards/accuracies": 0.737500011920929,
+      "rewards/chosen": 0.05224609375,
+      "rewards/margins": 1.0625,
+      "rewards/rejected": -1.0078125,
       "step": 130
     },
     {
       "epoch": 0.4472843450479233,
-      "grad_norm": 49.48725125304273,
+      "grad_norm": 65.14922163628061,
       "learning_rate": 4.727810650887574e-07,
-      "logits/chosen": -3.671875,
-      "logits/rejected": -3.515625,
-      "logps/chosen": -101.0,
-      "logps/rejected": -121.0,
-      "loss": 0.4801,
-      "rewards/accuracies": 0.762499988079071,
-      "rewards/chosen": -0.0260009765625,
-      "rewards/margins": 1.609375,
-      "rewards/rejected": -1.6328125,
+      "logits/chosen": -3.328125,
+      "logits/rejected": -3.34375,
+      "logps/chosen": -276.0,
+      "logps/rejected": -284.0,
+      "loss": 0.5408,
+      "rewards/accuracies": 0.737500011920929,
+      "rewards/chosen": 0.02001953125,
+      "rewards/margins": 0.97265625,
+      "rewards/rejected": -0.953125,
       "step": 140
     },
     {
       "epoch": 0.4792332268370607,
-      "grad_norm": 88.63757443724184,
+      "grad_norm": 69.5805163425697,
       "learning_rate": 4.668639053254438e-07,
-      "logits/chosen": -3.671875,
-      "logits/rejected": -3.53125,
-      "logps/chosen": -104.5,
-      "logps/rejected": -107.5,
-      "loss": 0.5525,
-      "rewards/accuracies": 0.75,
-      "rewards/chosen": -0.0185546875,
-      "rewards/margins": 0.88671875,
-      "rewards/rejected": -0.90625,
+      "logits/chosen": -3.3125,
+      "logits/rejected": -3.328125,
+      "logps/chosen": -266.0,
+      "logps/rejected": -278.0,
+      "loss": 0.5394,
+      "rewards/accuracies": 0.7124999761581421,
+      "rewards/chosen": 0.047607421875,
+      "rewards/margins": 0.890625,
+      "rewards/rejected": -0.84375,
       "step": 150
     },
     {
       "epoch": 0.5111821086261981,
-      "grad_norm": 65.41898185778713,
+      "grad_norm": 67.64968360037635,
       "learning_rate": 4.6094674556213014e-07,
-      "logits/chosen": -3.53125,
-      "logits/rejected": -3.375,
-      "logps/chosen": -102.5,
-      "logps/rejected": -110.0,
-      "loss": 0.4824,
-      "rewards/accuracies": 0.737500011920929,
-      "rewards/chosen": -0.20703125,
-      "rewards/margins": 1.0390625,
-      "rewards/rejected": -1.2421875,
+      "logits/chosen": -3.328125,
+      "logits/rejected": -3.171875,
+      "logps/chosen": -270.0,
+      "logps/rejected": -288.0,
+      "loss": 0.6143,
+      "rewards/accuracies": 0.6625000238418579,
+      "rewards/chosen": -0.26171875,
+      "rewards/margins": 0.83984375,
+      "rewards/rejected": -1.1015625,
       "step": 160
     },
     {
       "epoch": 0.5431309904153354,
-      "grad_norm": 64.3627134555641,
+      "grad_norm": 75.9892459922825,
       "learning_rate": 4.5502958579881655e-07,
-      "logits/chosen": -3.671875,
-      "logits/rejected": -3.5625,
-      "logps/chosen": -95.5,
-      "logps/rejected": -114.0,
-      "loss": 0.4837,
-      "rewards/accuracies": 0.800000011920929,
-      "rewards/chosen": 0.04736328125,
-      "rewards/margins": 1.3984375,
-      "rewards/rejected": -1.3515625,
+      "logits/chosen": -3.265625,
+      "logits/rejected": -3.171875,
+      "logps/chosen": -278.0,
+      "logps/rejected": -290.0,
+      "loss": 0.5603,
+      "rewards/accuracies": 0.6625000238418579,
+      "rewards/chosen": -0.44140625,
+      "rewards/margins": 0.796875,
+      "rewards/rejected": -1.234375,
       "step": 170
     },
     {
       "epoch": 0.5750798722044729,
-      "grad_norm": 50.45380597678468,
+      "grad_norm": 90.65062968036268,
       "learning_rate": 4.491124260355029e-07,
-      "logits/chosen": -3.53125,
-      "logits/rejected": -3.453125,
-      "logps/chosen": -100.5,
-      "logps/rejected": -114.5,
-      "loss": 0.4732,
-      "rewards/accuracies": 0.8125,
-      "rewards/chosen": -0.031005859375,
-      "rewards/margins": 1.1328125,
-      "rewards/rejected": -1.1640625,
+      "logits/chosen": -3.203125,
+      "logits/rejected": -3.1875,
+      "logps/chosen": -278.0,
+      "logps/rejected": -282.0,
+      "loss": 0.6629,
+      "rewards/accuracies": 0.675000011920929,
+      "rewards/chosen": -0.306640625,
+      "rewards/margins": 0.75,
+      "rewards/rejected": -1.0546875,
       "step": 180
     },
     {
       "epoch": 0.6070287539936102,
-      "grad_norm": 47.054283179898476,
+      "grad_norm": 108.28762791772537,
       "learning_rate": 4.4319526627218936e-07,
-      "logits/chosen": -3.59375,
-      "logits/rejected": -3.53125,
-      "logps/chosen": -103.5,
-      "logps/rejected": -112.0,
-      "loss": 0.4777,
-      "rewards/accuracies": 0.7250000238418579,
-      "rewards/chosen": 0.08154296875,
-      "rewards/margins": 1.203125,
-      "rewards/rejected": -1.125,
+      "logits/chosen": -3.234375,
+      "logits/rejected": -3.28125,
+      "logps/chosen": -280.0,
+      "logps/rejected": -280.0,
+      "loss": 0.5251,
+      "rewards/accuracies": 0.699999988079071,
+      "rewards/chosen": -0.208984375,
+      "rewards/margins": 1.0390625,
+      "rewards/rejected": -1.25,
       "step": 190
     },
     {
       "epoch": 0.6389776357827476,
-      "grad_norm": 60.50447069634289,
+      "grad_norm": 78.98605565885474,
       "learning_rate": 4.372781065088757e-07,
-      "logits/chosen": -3.625,
-      "logits/rejected": -3.46875,
-      "logps/chosen": -100.0,
-      "logps/rejected": -112.0,
-      "loss": 0.4606,
-      "rewards/accuracies": 0.7749999761581421,
-      "rewards/chosen": 0.162109375,
-      "rewards/margins": 1.375,
-      "rewards/rejected": -1.2109375,
+      "logits/chosen": -3.3125,
+      "logits/rejected": -3.171875,
+      "logps/chosen": -278.0,
+      "logps/rejected": -288.0,
+      "loss": 0.5118,
+      "rewards/accuracies": 0.7124999761581421,
+      "rewards/chosen": -0.296875,
+      "rewards/margins": 1.015625,
+      "rewards/rejected": -1.3125,
       "step": 200
     },
     {
       "epoch": 0.670926517571885,
-      "grad_norm": 42.756358795311726,
+      "grad_norm": 77.91455442318774,
       "learning_rate": 4.313609467455621e-07,
-      "logits/chosen": -3.59375,
-      "logits/rejected": -3.46875,
-      "logps/chosen": -92.5,
-      "logps/rejected": -119.0,
-      "loss": 0.3956,
-      "rewards/accuracies": 0.7875000238418579,
-      "rewards/chosen": -0.058837890625,
-      "rewards/margins": 1.7578125,
-      "rewards/rejected": -1.8203125,
+      "logits/chosen": -3.234375,
+      "logits/rejected": -3.1875,
+      "logps/chosen": -272.0,
+      "logps/rejected": -288.0,
+      "loss": 0.5157,
+      "rewards/accuracies": 0.7124999761581421,
+      "rewards/chosen": -0.14453125,
+      "rewards/margins": 1.015625,
+      "rewards/rejected": -1.15625,
       "step": 210
     },
     {
       "epoch": 0.7028753993610224,
-      "grad_norm": 37.12579027519649,
+      "grad_norm": 63.1172852867484,
       "learning_rate": 4.2544378698224847e-07,
-      "logits/chosen": -3.453125,
-      "logits/rejected": -3.484375,
-      "logps/chosen": -101.5,
-      "logps/rejected": -112.0,
-      "loss": 0.4621,
-      "rewards/accuracies": 0.7875000238418579,
-      "rewards/chosen": 0.06884765625,
-      "rewards/margins": 1.3046875,
-      "rewards/rejected": -1.234375,
+      "logits/chosen": -3.1875,
+      "logits/rejected": -3.28125,
+      "logps/chosen": -268.0,
+      "logps/rejected": -280.0,
+      "loss": 0.5155,
+      "rewards/accuracies": 0.6875,
+      "rewards/chosen": -0.0301513671875,
+      "rewards/margins": 0.9375,
+      "rewards/rejected": -0.96875,
       "step": 220
     },
     {
       "epoch": 0.7348242811501597,
-      "grad_norm": 34.37003470756731,
+      "grad_norm": 75.1179190067346,
       "learning_rate": 4.195266272189349e-07,
-      "logits/chosen": -3.40625,
-      "logits/rejected": -3.421875,
-      "logps/chosen": -102.0,
-      "logps/rejected": -119.0,
-      "loss": 0.4478,
-      "rewards/accuracies": 0.762499988079071,
-      "rewards/chosen": -0.06640625,
-      "rewards/margins": 1.59375,
-      "rewards/rejected": -1.6640625,
+      "logits/chosen": -3.21875,
+      "logits/rejected": -3.3125,
+      "logps/chosen": -280.0,
+      "logps/rejected": -280.0,
+      "loss": 0.5115,
+      "rewards/accuracies": 0.737500011920929,
+      "rewards/chosen": -0.072265625,
+      "rewards/margins": 1.1953125,
+      "rewards/rejected": -1.265625,
       "step": 230
     },
     {
       "epoch": 0.7667731629392971,
-      "grad_norm": 39.975701583088714,
+      "grad_norm": 68.71250790991336,
       "learning_rate": 4.1360946745562133e-07,
-      "logits/chosen": -3.484375,
-      "logits/rejected": -3.4375,
-      "logps/chosen": -95.5,
-      "logps/rejected": -123.0,
-      "loss": 0.4043,
-      "rewards/accuracies": 0.762499988079071,
-      "rewards/chosen": 0.0322265625,
-      "rewards/margins": 1.5546875,
-      "rewards/rejected": -1.5234375,
+      "logits/chosen": -3.265625,
+      "logits/rejected": -3.203125,
+      "logps/chosen": -272.0,
+      "logps/rejected": -290.0,
+      "loss": 0.4746,
+      "rewards/accuracies": 0.8374999761581421,
+      "rewards/chosen": 0.029052734375,
+      "rewards/margins": 1.6796875,
+      "rewards/rejected": -1.6484375,
       "step": 240
     },
     {
       "epoch": 0.7987220447284346,
-      "grad_norm": 40.31896174589532,
+      "grad_norm": 55.46878046949881,
       "learning_rate": 4.076923076923077e-07,
-      "logits/chosen": -3.53125,
-      "logits/rejected": -3.484375,
-      "logps/chosen": -107.0,
-      "logps/rejected": -122.0,
-      "loss": 0.4742,
-      "rewards/accuracies": 0.887499988079071,
-      "rewards/chosen": 0.54296875,
-      "rewards/margins": 2.46875,
-      "rewards/rejected": -1.921875,
+      "logits/chosen": -3.265625,
+      "logits/rejected": -3.328125,
+      "logps/chosen": -280.0,
+      "logps/rejected": -278.0,
+      "loss": 0.4711,
+      "rewards/accuracies": 0.800000011920929,
+      "rewards/chosen": -0.150390625,
+      "rewards/margins": 1.140625,
+      "rewards/rejected": -1.2890625,
       "step": 250
     },
     {
       "epoch": 0.8306709265175719,
-      "grad_norm": 32.53695466017645,
+      "grad_norm": 60.69436763385765,
       "learning_rate": 4.017751479289941e-07,
-      "logits/chosen": -3.578125,
-      "logits/rejected": -3.5,
-      "logps/chosen": -81.0,
-      "logps/rejected": -135.0,
-      "loss": 0.3528,
-      "rewards/accuracies": 0.8500000238418579,
-      "rewards/chosen": 0.044921875,
-      "rewards/margins": 2.765625,
-      "rewards/rejected": -2.71875,
+      "logits/chosen": -3.25,
+      "logits/rejected": -3.28125,
+      "logps/chosen": -274.0,
+      "logps/rejected": -286.0,
+      "loss": 0.5292,
+      "rewards/accuracies": 0.762499988079071,
+      "rewards/chosen": -0.1650390625,
+      "rewards/margins": 1.25,
+      "rewards/rejected": -1.4140625,
       "step": 260
     },
     {
       "epoch": 0.8626198083067093,
-      "grad_norm": 35.58565159605434,
+      "grad_norm": 61.755235473426524,
       "learning_rate": 3.9585798816568044e-07,
-      "logits/chosen": -3.625,
-      "logits/rejected": -3.46875,
-      "logps/chosen": -106.5,
-      "logps/rejected": -107.5,
-      "loss": 0.3774,
-      "rewards/accuracies": 0.862500011920929,
-      "rewards/chosen": 0.318359375,
-      "rewards/margins": 1.8984375,
-      "rewards/rejected": -1.578125,
+      "logits/chosen": -3.28125,
+      "logits/rejected": -3.3125,
+      "logps/chosen": -274.0,
+      "logps/rejected": -282.0,
+      "loss": 0.4873,
+      "rewards/accuracies": 0.8125,
+      "rewards/chosen": -0.044921875,
+      "rewards/margins": 1.4375,
+      "rewards/rejected": -1.484375,
       "step": 270
     },
     {
       "epoch": 0.8945686900958466,
-      "grad_norm": 50.853618748566205,
+      "grad_norm": 64.8584152127513,
       "learning_rate": 3.8994082840236685e-07,
-      "logits/chosen": -3.515625,
-      "logits/rejected": -3.484375,
-      "logps/chosen": -99.5,
-      "logps/rejected": -119.0,
-      "loss": 0.4586,
-      "rewards/accuracies": 0.7124999761581421,
-      "rewards/chosen": 0.26953125,
-      "rewards/margins": 1.671875,
-      "rewards/rejected": -1.40625,
+      "logits/chosen": -3.328125,
+      "logits/rejected": -3.265625,
+      "logps/chosen": -284.0,
+      "logps/rejected": -280.0,
+      "loss": 0.5342,
+      "rewards/accuracies": 0.6625000238418579,
+      "rewards/chosen": -0.1640625,
+      "rewards/margins": 0.87109375,
+      "rewards/rejected": -1.03125,
       "step": 280
     },
     {
       "epoch": 0.9265175718849841,
-      "grad_norm": 49.95783582965489,
+      "grad_norm": 71.13313796916569,
       "learning_rate": 3.840236686390532e-07,
-      "logits/chosen": -3.640625,
-      "logits/rejected": -3.453125,
-      "logps/chosen": -100.5,
-      "logps/rejected": -113.0,
-      "loss": 0.4365,
-      "rewards/accuracies": 0.7875000238418579,
-      "rewards/chosen": 0.37109375,
-      "rewards/margins": 1.8671875,
-      "rewards/rejected": -1.5,
+      "logits/chosen": -3.296875,
+      "logits/rejected": -3.28125,
+      "logps/chosen": -268.0,
+      "logps/rejected": -284.0,
+      "loss": 0.455,
+      "rewards/accuracies": 0.824999988079071,
+      "rewards/chosen": -0.0546875,
+      "rewards/margins": 1.65625,
+      "rewards/rejected": -1.7109375,
       "step": 290
     },
     {
       "epoch": 0.9584664536741214,
-      "grad_norm": 26.77848071496543,
+      "grad_norm": 147.24265223128356,
       "learning_rate": 3.7810650887573966e-07,
-      "logits/chosen": -3.5625,
-      "logits/rejected": -3.59375,
-      "logps/chosen": -99.5,
-      "logps/rejected": -135.0,
-      "loss": 0.4772,
-      "rewards/accuracies": 0.8374999761581421,
-      "rewards/chosen": 0.25390625,
-      "rewards/margins": 2.046875,
-      "rewards/rejected": -1.7890625,
+      "logits/chosen": -3.296875,
+      "logits/rejected": -3.234375,
+      "logps/chosen": -284.0,
+      "logps/rejected": -284.0,
+      "loss": 0.5612,
+      "rewards/accuracies": 0.800000011920929,
+      "rewards/chosen": -0.27734375,
+      "rewards/margins": 1.140625,
+      "rewards/rejected": -1.421875,
       "step": 300
     },
     {
       "epoch": 0.9904153354632588,
-      "grad_norm": 44.271261713875845,
+      "grad_norm": 57.83608472539185,
       "learning_rate": 3.72189349112426e-07,
-      "logits/chosen": -3.5625,
-      "logits/rejected": -3.484375,
-      "logps/chosen": -108.5,
-      "logps/rejected": -122.5,
-      "loss": 0.3872,
-      "rewards/accuracies": 0.800000011920929,
-      "rewards/chosen": 0.0625,
-      "rewards/margins": 1.625,
-      "rewards/rejected": -1.5703125,
+      "logits/chosen": -3.25,
+      "logits/rejected": -3.25,
+      "logps/chosen": -274.0,
+      "logps/rejected": -282.0,
+      "loss": 0.4624,
+      "rewards/accuracies": 0.8374999761581421,
+      "rewards/chosen": 0.05712890625,
+      "rewards/margins": 1.40625,
+      "rewards/rejected": -1.3515625,
       "step": 310
     },
     {
       "epoch": 1.0,
-      "eval_logits/chosen": -3.5625,
-      "eval_logits/rejected": -3.5,
-      "eval_logps/chosen": -98.5,
-      "eval_logps/rejected": -125.0,
-      "eval_loss": 0.5349218845367432,
-      "eval_rewards/accuracies": 0.7857142686843872,
-      "eval_rewards/chosen": -0.12060546875,
-      "eval_rewards/margins": 1.203125,
-      "eval_rewards/rejected": -1.3203125,
-      "eval_runtime": 12.3829,
-      "eval_samples_per_second": 16.151,
-      "eval_steps_per_second": 0.565,
-      "step": 313
-    },
-    {
-      "epoch": 1.0223642172523961,
-      "grad_norm": 33.843167460384336,
-      "learning_rate": 3.662721893491124e-07,
-      "logits/chosen": -3.5,
-      "logits/rejected": -3.328125,
-      "logps/chosen": -98.0,
-      "logps/rejected": -138.0,
-      "loss": 0.19,
-      "rewards/accuracies": 0.9375,
-      "rewards/chosen": 0.236328125,
-      "rewards/margins": 2.96875,
-      "rewards/rejected": -2.734375,
-      "step": 320
-    },
-    {
-      "epoch": 1.0543130990415335,
-      "grad_norm": 14.421132534916795,
-      "learning_rate": 3.603550295857988e-07,
-      "logits/chosen": -3.546875,
-      "logits/rejected": -3.4375,
-      "logps/chosen": -101.0,
-      "logps/rejected": -125.0,
-      "loss": 0.148,
-      "rewards/accuracies": 0.949999988079071,
-      "rewards/chosen": 0.70703125,
-      "rewards/margins": 2.953125,
-      "rewards/rejected": -2.234375,
-      "step": 330
-    },
-    {
-      "epoch": 1.0862619808306708,
-      "grad_norm": 18.495628015771892,
-      "learning_rate": 3.544378698224852e-07,
-      "logits/chosen": -3.484375,
-      "logits/rejected": -3.4375,
-      "logps/chosen": -95.5,
-      "logps/rejected": -125.5,
-      "loss": 0.1841,
-      "rewards/accuracies": 0.949999988079071,
-      "rewards/chosen": 0.71484375,
-      "rewards/margins": 2.875,
-      "rewards/rejected": -2.15625,
-      "step": 340
-    },
-    {
-      "epoch": 1.1182108626198084,
-      "grad_norm": 24.09426865119445,
-      "learning_rate": 3.485207100591716e-07,
-      "logits/chosen": -3.5625,
-      "logits/rejected": -3.34375,
-      "logps/chosen": -101.0,
-      "logps/rejected": -124.5,
-      "loss": 0.1584,
-      "rewards/accuracies": 0.9750000238418579,
-      "rewards/chosen": 0.89453125,
-      "rewards/margins": 3.421875,
-      "rewards/rejected": -2.53125,
-      "step": 350
-    },
-    {
-      "epoch": 1.1501597444089458,
-      "grad_norm": 25.308356305497963,
-      "learning_rate": 3.42603550295858e-07,
-      "logits/chosen": -3.53125,
-      "logits/rejected": -3.234375,
-      "logps/chosen": -99.0,
-      "logps/rejected": -116.5,
-      "loss": 0.1927,
-      "rewards/accuracies": 0.9125000238418579,
-      "rewards/chosen": 0.80078125,
-      "rewards/margins": 3.078125,
-      "rewards/rejected": -2.28125,
-      "step": 360
-    },
-    {
-      "epoch": 1.182108626198083,
-      "grad_norm": 20.87093717123946,
-      "learning_rate": 3.366863905325444e-07,
-      "logits/chosen": -3.5,
-      "logits/rejected": -3.390625,
-      "logps/chosen": -91.5,
-      "logps/rejected": -129.0,
-      "loss": 0.1332,
-      "rewards/accuracies": 0.949999988079071,
-      "rewards/chosen": 0.734375,
-      "rewards/margins": 3.484375,
-      "rewards/rejected": -2.75,
-      "step": 370
-    },
-    {
-      "epoch": 1.2140575079872205,
-      "grad_norm": 33.11495125656759,
-      "learning_rate": 3.3076923076923075e-07,
-      "logits/chosen": -3.515625,
-      "logits/rejected": -3.375,
-      "logps/chosen": -99.0,
-      "logps/rejected": -148.0,
-      "loss": 0.1292,
-      "rewards/accuracies": 0.949999988079071,
-      "rewards/chosen": 0.578125,
-      "rewards/margins": 4.4375,
-      "rewards/rejected": -3.859375,
-      "step": 380
-    },
-    {
-      "epoch": 1.2460063897763578,
-      "grad_norm": 67.10626864365061,
-      "learning_rate": 3.2485207100591716e-07,
-      "logits/chosen": -3.515625,
-      "logits/rejected": -3.296875,
-      "logps/chosen": -95.0,
-      "logps/rejected": -138.0,
-      "loss": 0.2115,
-      "rewards/accuracies": 0.925000011920929,
-      "rewards/chosen": 0.765625,
-      "rewards/margins": 3.9375,
-      "rewards/rejected": -3.171875,
-      "step": 390
-    },
-    {
-      "epoch": 1.2779552715654952,
-      "grad_norm": 18.06304275803513,
-      "learning_rate": 3.189349112426035e-07,
-      "logits/chosen": -3.421875,
-      "logits/rejected": -3.28125,
-      "logps/chosen": -100.0,
-      "logps/rejected": -136.0,
-      "loss": 0.1318,
-      "rewards/accuracies": 0.949999988079071,
-      "rewards/chosen": 0.5234375,
-      "rewards/margins": 3.828125,
-      "rewards/rejected": -3.3125,
-      "step": 400
-    },
-    {
-      "epoch": 1.3099041533546325,
-      "grad_norm": 38.28309104781088,
-      "learning_rate": 3.130177514792899e-07,
-      "logits/chosen": -3.40625,
-      "logits/rejected": -3.46875,
-      "logps/chosen": -90.0,
-      "logps/rejected": -130.0,
-      "loss": 0.138,
-      "rewards/accuracies": 0.9125000238418579,
-      "rewards/chosen": 0.53125,
-      "rewards/margins": 3.578125,
-      "rewards/rejected": -3.046875,
-      "step": 410
-    },
-    {
-      "epoch": 1.34185303514377,
-      "grad_norm": 15.178754681689357,
-      "learning_rate": 3.071005917159763e-07,
-      "logits/chosen": -3.4375,
-      "logits/rejected": -3.40625,
-      "logps/chosen": -92.5,
-      "logps/rejected": -137.0,
-      "loss": 0.1377,
-      "rewards/accuracies": 0.925000011920929,
-      "rewards/chosen": 0.388671875,
-      "rewards/margins": 3.890625,
-      "rewards/rejected": -3.5,
-      "step": 420
-    },
-    {
-      "epoch": 1.3738019169329074,
-      "grad_norm": 38.06252515615029,
-      "learning_rate": 3.011834319526627e-07,
-      "logits/chosen": -3.5,
-      "logits/rejected": -3.484375,
-      "logps/chosen": -102.5,
-      "logps/rejected": -136.0,
-      "loss": 0.1334,
-      "rewards/accuracies": 0.949999988079071,
-      "rewards/chosen": 0.5078125,
-      "rewards/margins": 3.71875,
-      "rewards/rejected": -3.203125,
-      "step": 430
-    },
-    {
-      "epoch": 1.4057507987220448,
-      "grad_norm": 19.397383343646702,
-      "learning_rate": 2.952662721893491e-07,
-      "logits/chosen": -3.609375,
-      "logits/rejected": -3.53125,
-      "logps/chosen": -102.0,
-      "logps/rejected": -131.0,
-      "loss": 0.1618,
-      "rewards/accuracies": 0.987500011920929,
-      "rewards/chosen": 0.95703125,
-      "rewards/margins": 3.984375,
-      "rewards/rejected": -3.03125,
-      "step": 440
-    },
-    {
-      "epoch": 1.4376996805111821,
-      "grad_norm": 21.51659900182813,
-      "learning_rate": 2.893491124260355e-07,
-      "logits/chosen": -3.4375,
-      "logits/rejected": -3.359375,
-      "logps/chosen": -97.0,
-      "logps/rejected": -147.0,
-      "loss": 0.1678,
-      "rewards/accuracies": 0.9375,
-      "rewards/chosen": 0.291015625,
-      "rewards/margins": 4.28125,
-      "rewards/rejected": -3.96875,
-      "step": 450
-    },
-    {
-      "epoch": 1.4696485623003195,
-      "grad_norm": 4.91974930909796,
-      "learning_rate": 2.834319526627219e-07,
-      "logits/chosen": -3.453125,
-      "logits/rejected": -3.46875,
-      "logps/chosen": -87.0,
-      "logps/rejected": -159.0,
-      "loss": 0.1759,
-      "rewards/accuracies": 0.925000011920929,
-      "rewards/chosen": 0.140625,
-      "rewards/margins": 4.0625,
-      "rewards/rejected": -3.921875,
-      "step": 460
-    },
-    {
-      "epoch": 1.5015974440894568,
-      "grad_norm": 65.85227279676343,
-      "learning_rate": 2.7751479289940824e-07,
-      "logits/chosen": -3.515625,
-      "logits/rejected": -3.3125,
-      "logps/chosen": -94.0,
-      "logps/rejected": -149.0,
-      "loss": 0.142,
-      "rewards/accuracies": 0.949999988079071,
-      "rewards/chosen": 0.2373046875,
-      "rewards/margins": 4.03125,
-      "rewards/rejected": -3.8125,
-      "step": 470
-    },
-    {
-      "epoch": 1.5335463258785942,
-      "grad_norm": 18.354377975306473,
-      "learning_rate": 2.715976331360947e-07,
-      "logits/chosen": -3.46875,
-      "logits/rejected": -3.375,
-      "logps/chosen": -98.0,
-      "logps/rejected": -155.0,
-      "loss": 0.1047,
-      "rewards/accuracies": 0.9750000238418579,
-      "rewards/chosen": 0.287109375,
-      "rewards/margins": 4.625,
-      "rewards/rejected": -4.34375,
-      "step": 480
-    },
-    {
-      "epoch": 1.5654952076677318,
-      "grad_norm": 16.420773639490726,
-      "learning_rate": 2.6568047337278106e-07,
-      "logits/chosen": -3.5625,
-      "logits/rejected": -3.390625,
-      "logps/chosen": -100.0,
-      "logps/rejected": -130.0,
-      "loss": 0.1373,
-      "rewards/accuracies": 0.949999988079071,
-      "rewards/chosen": 0.76171875,
-      "rewards/margins": 4.125,
-      "rewards/rejected": -3.359375,
-      "step": 490
-    },
-    {
-      "epoch": 1.5974440894568689,
-      "grad_norm": 43.48032898852333,
-      "learning_rate": 2.5976331360946746e-07,
-      "logits/chosen": -3.375,
-      "logits/rejected": -3.359375,
-      "logps/chosen": -94.0,
-      "logps/rejected": -151.0,
-      "loss": 0.1149,
-      "rewards/accuracies": 0.9624999761581421,
-      "rewards/chosen": 0.1669921875,
-      "rewards/margins": 4.75,
-      "rewards/rejected": -4.59375,
-      "step": 500
-    },
-    {
-      "epoch": 1.6293929712460065,
-      "grad_norm": 39.88152223306465,
-      "learning_rate": 2.538461538461538e-07,
-      "logits/chosen": -3.390625,
-      "logits/rejected": -3.328125,
-      "logps/chosen": -84.5,
-      "logps/rejected": -157.0,
-      "loss": 0.1571,
-      "rewards/accuracies": 0.949999988079071,
-      "rewards/chosen": -0.201171875,
-      "rewards/margins": 4.1875,
-      "rewards/rejected": -4.375,
-      "step": 510
-    },
-    {
-      "epoch": 1.6613418530351438,
-      "grad_norm": 16.23543571951759,
-      "learning_rate": 2.479289940828402e-07,
-      "logits/chosen": -3.53125,
-      "logits/rejected": -3.390625,
-      "logps/chosen": -97.0,
-      "logps/rejected": -145.0,
-      "loss": 0.1344,
-      "rewards/accuracies": 0.8999999761581421,
-      "rewards/chosen": 0.40234375,
-      "rewards/margins": 4.15625,
-      "rewards/rejected": -3.765625,
-      "step": 520
-    },
-    {
-      "epoch": 1.6932907348242812,
-      "grad_norm": 30.6571237995008,
-      "learning_rate": 2.420118343195266e-07,
-      "logits/chosen": -3.390625,
-      "logits/rejected": -3.1875,
-      "logps/chosen": -87.5,
-      "logps/rejected": -152.0,
-      "loss": 0.1264,
-      "rewards/accuracies": 0.949999988079071,
-      "rewards/chosen": 0.0791015625,
-      "rewards/margins": 4.5,
-      "rewards/rejected": -4.40625,
-      "step": 530
-    },
-    {
-      "epoch": 1.7252396166134185,
-      "grad_norm": 13.821155728848668,
-      "learning_rate": 2.36094674556213e-07,
-      "logits/chosen": -3.453125,
-      "logits/rejected": -3.25,
-      "logps/chosen": -108.0,
-      "logps/rejected": -135.0,
-      "loss": 0.1009,
-      "rewards/accuracies": 0.9750000238418579,
-      "rewards/chosen": 0.71875,
-      "rewards/margins": 4.46875,
-      "rewards/rejected": -3.75,
-      "step": 540
-    },
-    {
-      "epoch": 1.7571884984025559,
-      "grad_norm": 25.951691798178203,
-      "learning_rate": 2.301775147928994e-07,
-      "logits/chosen": -3.453125,
-      "logits/rejected": -3.25,
-      "logps/chosen": -99.5,
-      "logps/rejected": -141.0,
-      "loss": 0.1441,
-      "rewards/accuracies": 0.925000011920929,
-      "rewards/chosen": 0.4921875,
-      "rewards/margins": 5.0,
-      "rewards/rejected": -4.5,
-      "step": 550
-    },
-    {
-      "epoch": 1.7891373801916934,
-      "grad_norm": 12.667867070291107,
-      "learning_rate": 2.242603550295858e-07,
-      "logits/chosen": -3.34375,
-      "logits/rejected": -3.25,
-      "logps/chosen": -103.0,
-      "logps/rejected": -158.0,
-      "loss": 0.1124,
-      "rewards/accuracies": 0.9375,
-      "rewards/chosen": 0.07958984375,
-      "rewards/margins": 4.6875,
-      "rewards/rejected": -4.59375,
-      "step": 560
-    },
-    {
-      "epoch": 1.8210862619808306,
-      "grad_norm": 66.98788513046767,
-      "learning_rate": 2.1834319526627217e-07,
-      "logits/chosen": -3.484375,
-      "logits/rejected": -3.359375,
-      "logps/chosen": -100.5,
-      "logps/rejected": -148.0,
-      "loss": 0.1575,
-      "rewards/accuracies": 0.887499988079071,
-      "rewards/chosen": 0.44140625,
-      "rewards/margins": 4.125,
-      "rewards/rejected": -3.6875,
-      "step": 570
-    },
-    {
-      "epoch": 1.8530351437699681,
-      "grad_norm": 22.80146314369706,
-      "learning_rate": 2.1242603550295858e-07,
-      "logits/chosen": -3.515625,
-      "logits/rejected": -3.28125,
-      "logps/chosen": -103.5,
-      "logps/rejected": -146.0,
-      "loss": 0.146,
-      "rewards/accuracies": 0.8999999761581421,
-      "rewards/chosen": 0.392578125,
-      "rewards/margins": 4.375,
-      "rewards/rejected": -3.984375,
-      "step": 580
-    },
-    {
-      "epoch": 1.8849840255591053,
-      "grad_norm": 22.73841977980904,
-      "learning_rate": 2.0650887573964496e-07,
-      "logits/chosen": -3.53125,
-      "logits/rejected": -3.34375,
-      "logps/chosen": -103.5,
-      "logps/rejected": -149.0,
-      "loss": 0.1338,
-      "rewards/accuracies": 0.949999988079071,
-      "rewards/chosen": 0.30078125,
-      "rewards/margins": 4.5625,
-      "rewards/rejected": -4.25,
-      "step": 590
-    },
-    {
-      "epoch": 1.9169329073482428,
-      "grad_norm": 38.211271754825084,
-      "learning_rate": 2.0059171597633133e-07,
-      "logits/chosen": -3.5625,
-      "logits/rejected": -3.296875,
-      "logps/chosen": -101.5,
-      "logps/rejected": -138.0,
-      "loss": 0.1455,
-      "rewards/accuracies": 0.9624999761581421,
-      "rewards/chosen": 0.578125,
-      "rewards/margins": 4.53125,
-      "rewards/rejected": -3.96875,
-      "step": 600
-    },
-    {
-      "epoch": 1.9488817891373802,
-      "grad_norm": 20.886358667126427,
-      "learning_rate": 1.9467455621301774e-07,
-      "logits/chosen": -3.21875,
-      "logits/rejected": -3.296875,
-      "logps/chosen": -86.5,
-      "logps/rejected": -165.0,
-      "loss": 0.1428,
-      "rewards/accuracies": 0.9750000238418579,
-      "rewards/chosen": -0.004638671875,
-      "rewards/margins": 4.40625,
-      "rewards/rejected": -4.40625,
-      "step": 610
-    },
-    {
-      "epoch": 1.9808306709265175,
-      "grad_norm": 26.719930163743193,
-      "learning_rate": 1.8875739644970412e-07,
-      "logits/chosen": -3.421875,
-      "logits/rejected": -3.28125,
-      "logps/chosen": -101.5,
-      "logps/rejected": -154.0,
-      "loss": 0.1191,
-      "rewards/accuracies": 0.9750000238418579,
-      "rewards/chosen": 0.29296875,
-      "rewards/margins": 5.15625,
-      "rewards/rejected": -4.875,
-      "step": 620
-    },
-    {
-      "epoch": 2.0,
-      "eval_logits/chosen": -3.4375,
+      "eval_logits/chosen": -3.3125,
       "eval_logits/rejected": -3.34375,
-      "eval_logps/chosen": -108.5,
-      "eval_logps/rejected": -137.0,
-      "eval_loss": 0.6070312261581421,
-      "eval_rewards/accuracies": 0.6964285969734192,
-      "eval_rewards/chosen": -1.125,
-      "eval_rewards/margins": 1.3828125,
-      "eval_rewards/rejected": -2.515625,
-      "eval_runtime": 12.3074,
-      "eval_samples_per_second": 16.25,
-      "eval_steps_per_second": 0.569,
-      "step": 626
-    },
-    {
-      "epoch": 2.012779552715655,
-      "grad_norm": 233.3927856017294,
-      "learning_rate": 1.8284023668639053e-07,
-      "logits/chosen": -3.453125,
-      "logits/rejected": -3.390625,
-      "logps/chosen": -90.0,
-      "logps/rejected": -156.0,
-      "loss": 0.1624,
-      "rewards/accuracies": 0.925000011920929,
-      "rewards/chosen": 0.060791015625,
-      "rewards/margins": 4.25,
-      "rewards/rejected": -4.21875,
-      "step": 630
-    },
-    {
-      "epoch": 2.0447284345047922,
-      "grad_norm": 6.8655748966408305,
-      "learning_rate": 1.7692307692307693e-07,
-      "logits/chosen": -3.40625,
-      "logits/rejected": -3.28125,
-      "logps/chosen": -102.0,
-      "logps/rejected": -150.0,
-      "loss": 0.0546,
-      "rewards/accuracies": 0.987500011920929,
-      "rewards/chosen": 0.51953125,
-      "rewards/margins": 5.3125,
-      "rewards/rejected": -4.8125,
-      "step": 640
-    },
-    {
-      "epoch": 2.07667731629393,
-      "grad_norm": 5.52148734018718,
-      "learning_rate": 1.710059171597633e-07,
-      "logits/chosen": -3.46875,
-      "logits/rejected": -3.125,
-      "logps/chosen": -103.0,
-      "logps/rejected": -152.0,
-      "loss": 0.0275,
-      "rewards/accuracies": 0.987500011920929,
-      "rewards/chosen": 0.64453125,
-      "rewards/margins": 5.8125,
-      "rewards/rejected": -5.1875,
-      "step": 650
-    },
-    {
-      "epoch": 2.108626198083067,
-      "grad_norm": 5.945489873996584,
-      "learning_rate": 1.650887573964497e-07,
-      "logits/chosen": -3.53125,
-      "logits/rejected": -3.140625,
-      "logps/chosen": -94.0,
-      "logps/rejected": -163.0,
-      "loss": 0.0346,
-      "rewards/accuracies": 0.987500011920929,
-      "rewards/chosen": 0.3125,
-      "rewards/margins": 6.28125,
-      "rewards/rejected": -5.96875,
-      "step": 660
-    },
-    {
-      "epoch": 2.1405750798722045,
-      "grad_norm": 15.641153990271073,
-      "learning_rate": 1.591715976331361e-07,
-      "logits/chosen": -3.46875,
-      "logits/rejected": -3.203125,
-      "logps/chosen": -95.0,
-      "logps/rejected": -172.0,
-      "loss": 0.0404,
-      "rewards/accuracies": 0.987500011920929,
-      "rewards/chosen": 0.28125,
-      "rewards/margins": 6.5625,
-      "rewards/rejected": -6.28125,
-      "step": 670
-    },
-    {
-      "epoch": 2.1725239616613417,
-      "grad_norm": 12.428538162453988,
-      "learning_rate": 1.5325443786982248e-07,
-      "logits/chosen": -3.265625,
-      "logits/rejected": -3.109375,
-      "logps/chosen": -107.5,
-      "logps/rejected": -170.0,
-      "loss": 0.037,
-      "rewards/accuracies": 0.9750000238418579,
-      "rewards/chosen": 0.265625,
-      "rewards/margins": 6.625,
-      "rewards/rejected": -6.375,
-      "step": 680
-    },
-    {
-      "epoch": 2.2044728434504792,
-      "grad_norm": 4.017973551269618,
-      "learning_rate": 1.4733727810650885e-07,
-      "logits/chosen": -3.25,
-      "logits/rejected": -3.109375,
-      "logps/chosen": -108.0,
-      "logps/rejected": -173.0,
-      "loss": 0.037,
-      "rewards/accuracies": 0.987500011920929,
-      "rewards/chosen": 0.302734375,
-      "rewards/margins": 7.0,
-      "rewards/rejected": -6.6875,
-      "step": 690
-    },
-    {
-      "epoch": 2.236421725239617,
-      "grad_norm": 86.48671040112693,
-      "learning_rate": 1.4142011834319526e-07,
-      "logits/chosen": -3.28125,
-      "logits/rejected": -3.0625,
-      "logps/chosen": -97.0,
-      "logps/rejected": -173.0,
-      "loss": 0.0335,
-      "rewards/accuracies": 0.9750000238418579,
-      "rewards/chosen": -0.2236328125,
-      "rewards/margins": 7.4375,
-      "rewards/rejected": -7.65625,
-      "step": 700
-    },
-    {
-      "epoch": 2.268370607028754,
-      "grad_norm": 3.430888312255222,
-      "learning_rate": 1.3550295857988164e-07,
-      "logits/chosen": -3.15625,
-      "logits/rejected": -3.125,
-      "logps/chosen": -108.0,
-      "logps/rejected": -192.0,
-      "loss": 0.0355,
-      "rewards/accuracies": 0.9750000238418579,
-      "rewards/chosen": -0.9765625,
-      "rewards/margins": 7.53125,
-      "rewards/rejected": -8.5,
-      "step": 710
-    },
-    {
-      "epoch": 2.3003194888178915,
-      "grad_norm": 7.379623913134848,
-      "learning_rate": 1.2958579881656802e-07,
-      "logits/chosen": -3.34375,
-      "logits/rejected": -3.25,
-      "logps/chosen": -101.5,
-      "logps/rejected": -163.0,
-      "loss": 0.0452,
-      "rewards/accuracies": 0.9750000238418579,
-      "rewards/chosen": 0.0230712890625,
-      "rewards/margins": 6.53125,
-      "rewards/rejected": -6.5,
-      "step": 720
-    },
-    {
-      "epoch": 2.3322683706070286,
-      "grad_norm": 16.067536241252174,
-      "learning_rate": 1.2366863905325443e-07,
-      "logits/chosen": -3.40625,
-      "logits/rejected": -3.25,
-      "logps/chosen": -104.0,
-      "logps/rejected": -183.0,
-      "loss": 0.0485,
-      "rewards/accuracies": 0.987500011920929,
-      "rewards/chosen": -0.265625,
-      "rewards/margins": 7.0625,
-      "rewards/rejected": -7.3125,
-      "step": 730
-    },
-    {
-      "epoch": 2.364217252396166,
-      "grad_norm": 8.775321084313918,
-      "learning_rate": 1.1775147928994082e-07,
-      "logits/chosen": -3.25,
-      "logits/rejected": -3.1875,
-      "logps/chosen": -101.5,
-      "logps/rejected": -198.0,
-      "loss": 0.0573,
-      "rewards/accuracies": 0.949999988079071,
-      "rewards/chosen": -1.109375,
-      "rewards/margins": 7.1875,
-      "rewards/rejected": -8.3125,
-      "step": 740
-    },
-    {
-      "epoch": 2.3961661341853033,
-      "grad_norm": 2.7323617705921777,
-      "learning_rate": 1.1183431952662721e-07,
-      "logits/chosen": -3.203125,
-      "logits/rejected": -3.125,
-      "logps/chosen": -101.0,
-      "logps/rejected": -190.0,
-      "loss": 0.0264,
-      "rewards/accuracies": 1.0,
-      "rewards/chosen": -0.60546875,
-      "rewards/margins": 8.0,
-      "rewards/rejected": -8.625,
-      "step": 750
-    },
-    {
-      "epoch": 2.428115015974441,
-      "grad_norm": 26.885338666699912,
-      "learning_rate": 1.059171597633136e-07,
-      "logits/chosen": -3.296875,
-      "logits/rejected": -3.15625,
-      "logps/chosen": -100.0,
-      "logps/rejected": -168.0,
-      "loss": 0.0437,
-      "rewards/accuracies": 0.949999988079071,
-      "rewards/chosen": -0.12255859375,
-      "rewards/margins": 6.75,
-      "rewards/rejected": -6.875,
-      "step": 760
-    },
-    {
-      "epoch": 2.460063897763578,
-      "grad_norm": 4.85033080386058,
-      "learning_rate": 1e-07,
-      "logits/chosen": -3.171875,
-      "logits/rejected": -3.265625,
-      "logps/chosen": -97.0,
-      "logps/rejected": -188.0,
-      "loss": 0.0788,
-      "rewards/accuracies": 1.0,
-      "rewards/chosen": -0.474609375,
-      "rewards/margins": 7.6875,
-      "rewards/rejected": -8.125,
-      "step": 770
-    },
-    {
-      "epoch": 2.4920127795527156,
-      "grad_norm": 205.54438620209856,
-      "learning_rate": 9.408284023668639e-08,
-      "logits/chosen": -3.25,
-      "logits/rejected": -3.234375,
-      "logps/chosen": -96.0,
-      "logps/rejected": -186.0,
-      "loss": 0.0678,
-      "rewards/accuracies": 0.9750000238418579,
-      "rewards/chosen": -0.177734375,
-      "rewards/margins": 7.3125,
-      "rewards/rejected": -7.5,
-      "step": 780
-    },
-    {
-      "epoch": 2.523961661341853,
-      "grad_norm": 28.433975567152338,
-      "learning_rate": 8.816568047337278e-08,
-      "logits/chosen": -3.15625,
-      "logits/rejected": -3.03125,
-      "logps/chosen": -98.5,
-      "logps/rejected": -188.0,
-      "loss": 0.031,
-      "rewards/accuracies": 1.0,
-      "rewards/chosen": -0.91796875,
-      "rewards/margins": 7.5,
-      "rewards/rejected": -8.4375,
-      "step": 790
-    },
-    {
-      "epoch": 2.5559105431309903,
-      "grad_norm": 4.2386789501934965,
-      "learning_rate": 8.224852071005916e-08,
-      "logits/chosen": -3.265625,
-      "logits/rejected": -3.03125,
-      "logps/chosen": -110.5,
-      "logps/rejected": -175.0,
-      "loss": 0.0499,
-      "rewards/accuracies": 1.0,
-      "rewards/chosen": -0.353515625,
-      "rewards/margins": 7.15625,
-      "rewards/rejected": -7.5,
-      "step": 800
-    },
-    {
-      "epoch": 2.587859424920128,
-      "grad_norm": 6.261571908946508,
-      "learning_rate": 7.633136094674555e-08,
-      "logits/chosen": -3.234375,
-      "logits/rejected": -3.09375,
-      "logps/chosen": -104.5,
-      "logps/rejected": -185.0,
-      "loss": 0.0425,
-      "rewards/accuracies": 0.9750000238418579,
-      "rewards/chosen": -0.7890625,
-      "rewards/margins": 7.3125,
-      "rewards/rejected": -8.125,
-      "step": 810
-    },
-    {
-      "epoch": 2.619808306709265,
-      "grad_norm": 43.7884576040152,
-      "learning_rate": 7.041420118343195e-08,
-      "logits/chosen": -3.15625,
-      "logits/rejected": -3.140625,
-      "logps/chosen": -107.0,
-      "logps/rejected": -193.0,
-      "loss": 0.0592,
-      "rewards/accuracies": 1.0,
-      "rewards/chosen": -0.404296875,
-      "rewards/margins": 7.59375,
-      "rewards/rejected": -8.0,
-      "step": 820
-    },
-    {
-      "epoch": 2.6517571884984026,
-      "grad_norm": 7.063030707335778,
-      "learning_rate": 6.449704142011835e-08,
-      "logits/chosen": -3.203125,
-      "logits/rejected": -3.25,
-      "logps/chosen": -107.5,
-      "logps/rejected": -219.0,
-      "loss": 0.0132,
-      "rewards/accuracies": 0.987500011920929,
-      "rewards/chosen": -1.0,
-      "rewards/margins": 8.6875,
-      "rewards/rejected": -9.6875,
-      "step": 830
-    },
-    {
-      "epoch": 2.68370607028754,
-      "grad_norm": 31.69985105836376,
-      "learning_rate": 5.857988165680473e-08,
-      "logits/chosen": -3.234375,
-      "logits/rejected": -3.125,
-      "logps/chosen": -110.0,
-      "logps/rejected": -194.0,
-      "loss": 0.0767,
-      "rewards/accuracies": 0.9624999761581421,
-      "rewards/chosen": -1.40625,
-      "rewards/margins": 7.375,
-      "rewards/rejected": -8.8125,
-      "step": 840
-    },
-    {
-      "epoch": 2.7156549520766773,
-      "grad_norm": 36.344404095239184,
-      "learning_rate": 5.266272189349112e-08,
-      "logits/chosen": -3.171875,
-      "logits/rejected": -3.15625,
-      "logps/chosen": -110.0,
-      "logps/rejected": -196.0,
-      "loss": 0.0383,
-      "rewards/accuracies": 0.9750000238418579,
-      "rewards/chosen": -1.234375,
-      "rewards/margins": 7.71875,
-      "rewards/rejected": -8.9375,
-      "step": 850
-    },
-    {
-      "epoch": 2.747603833865815,
-      "grad_norm": 26.561166415223443,
-      "learning_rate": 4.674556213017751e-08,
-      "logits/chosen": -3.296875,
-      "logits/rejected": -3.328125,
-      "logps/chosen": -109.5,
-      "logps/rejected": -183.0,
-      "loss": 0.0619,
-      "rewards/accuracies": 0.9624999761581421,
-      "rewards/chosen": -1.28125,
-      "rewards/margins": 6.5625,
-      "rewards/rejected": -7.84375,
-      "step": 860
-    },
-    {
-      "epoch": 2.779552715654952,
-      "grad_norm": 51.634391271169434,
-      "learning_rate": 4.082840236686391e-08,
-      "logits/chosen": -3.34375,
-      "logits/rejected": -3.078125,
-      "logps/chosen": -115.5,
-      "logps/rejected": -175.0,
-      "loss": 0.0302,
-      "rewards/accuracies": 0.9624999761581421,
-      "rewards/chosen": -0.75,
-      "rewards/margins": 7.78125,
-      "rewards/rejected": -8.5625,
-      "step": 870
-    },
-    {
-      "epoch": 2.8115015974440896,
-      "grad_norm": 4.862668702951866,
-      "learning_rate": 3.4911242603550294e-08,
-      "logits/chosen": -3.109375,
-      "logits/rejected": -3.1875,
-      "logps/chosen": -110.0,
-      "logps/rejected": -192.0,
-      "loss": 0.0318,
-      "rewards/accuracies": 0.987500011920929,
-      "rewards/chosen": -1.3125,
-      "rewards/margins": 8.125,
-      "rewards/rejected": -9.375,
-      "step": 880
-    },
-    {
-      "epoch": 2.8434504792332267,
-      "grad_norm": 5.808427284559429,
-      "learning_rate": 2.8994082840236687e-08,
-      "logits/chosen": -3.09375,
-      "logits/rejected": -3.21875,
-      "logps/chosen": -118.5,
-      "logps/rejected": -202.0,
-      "loss": 0.0413,
-      "rewards/accuracies": 0.987500011920929,
-      "rewards/chosen": -1.6171875,
-      "rewards/margins": 8.25,
-      "rewards/rejected": -9.8125,
-      "step": 890
-    },
-    {
-      "epoch": 2.8753993610223643,
-      "grad_norm": 1.47350840913393,
-      "learning_rate": 2.3076923076923076e-08,
-      "logits/chosen": -3.21875,
-      "logits/rejected": -3.171875,
-      "logps/chosen": -102.5,
-      "logps/rejected": -202.0,
-      "loss": 0.024,
-      "rewards/accuracies": 0.987500011920929,
-      "rewards/chosen": -1.3046875,
-      "rewards/margins": 8.4375,
-      "rewards/rejected": -9.75,
-      "step": 900
-    },
-    {
-      "epoch": 2.9073482428115014,
-      "grad_norm": 35.85817350730659,
-      "learning_rate": 1.7159763313609465e-08,
-      "logits/chosen": -3.265625,
-      "logits/rejected": -3.234375,
-      "logps/chosen": -116.5,
-      "logps/rejected": -181.0,
-      "loss": 0.0377,
-      "rewards/accuracies": 0.987500011920929,
-      "rewards/chosen": -0.474609375,
-      "rewards/margins": 7.96875,
-      "rewards/rejected": -8.4375,
-      "step": 910
-    },
-    {
-      "epoch": 2.939297124600639,
-      "grad_norm": 2.972228065765212,
-      "learning_rate": 1.1242603550295858e-08,
-      "logits/chosen": -3.1875,
-      "logits/rejected": -3.28125,
-      "logps/chosen": -107.5,
-      "logps/rejected": -200.0,
-      "loss": 0.0497,
-      "rewards/accuracies": 0.9750000238418579,
-      "rewards/chosen": -1.109375,
-      "rewards/margins": 8.0625,
-      "rewards/rejected": -9.1875,
-      "step": 920
-    },
-    {
-      "epoch": 2.9712460063897765,
-      "grad_norm": 41.76203114287835,
-      "learning_rate": 5.325443786982248e-09,
-      "logits/chosen": -3.125,
-      "logits/rejected": -3.03125,
-      "logps/chosen": -96.5,
-      "logps/rejected": -183.0,
-      "loss": 0.0546,
-      "rewards/accuracies": 0.9624999761581421,
-      "rewards/chosen": -1.0234375,
-      "rewards/margins": 7.71875,
-      "rewards/rejected": -8.75,
-      "step": 930
-    },
-    {
-      "epoch": 3.0,
-      "eval_logits/chosen": -3.328125,
-      "eval_logits/rejected": -3.1875,
-      "eval_logps/chosen": -133.0,
-      "eval_logps/rejected": -166.0,
-      "eval_loss": 0.8457812666893005,
-      "eval_rewards/accuracies": 0.6785714030265808,
-      "eval_rewards/chosen": -3.546875,
-      "eval_rewards/margins": 1.8359375,
-      "eval_rewards/rejected": -5.40625,
-      "eval_runtime": 15.4802,
-      "eval_samples_per_second": 12.92,
-      "eval_steps_per_second": 0.452,
-      "step": 939
+      "eval_logps/chosen": -272.0,
+      "eval_logps/rejected": -276.0,
+      "eval_loss": 0.5753124952316284,
+      "eval_rewards/accuracies": 0.75,
+      "eval_rewards/chosen": -0.039306640625,
+      "eval_rewards/margins": 0.703125,
+      "eval_rewards/rejected": -0.7421875,
+      "eval_runtime": 12.4952,
+      "eval_samples_per_second": 16.006,
+      "eval_steps_per_second": 0.56,
+      "step": 313
     }
   ],
   "logging_steps": 10,
@@ -1479,7 +517,7 @@
         "should_evaluate": false,
         "should_log": false,
         "should_save": true,
-        "should_training_stop": true
+        "should_training_stop": false
       },
       "attributes": {}
     }

last-checkpoint/training_args.bin

@@ -1,3 +1,3 @@
 version https://git-lfs.github.com/spec/v1
-oid sha256:1d48dffd8128dc85e053c0e19e9adf47b4da180a2d5453636b16aa5ad71de8d6
-size 7800
+oid sha256:a10177d0b6c8197647a260127a52a1a99a27e2288b639bfcd33f32ab84d343a0
+size 7864