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 sha256:179965849f9d3009631e9cdbaaadf844b33bb60b697870238605605198f18e24
+oid sha256:89aa416e135b41f05720b557e0f9707d4deb2b420b406d1e953bfd371794893d
 size 14496051368

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:f71940459e78a610a093c9fba10b69c89e0f27e5dfa3b927cee675c2c5eeef54
+oid sha256:5c360ec7ef2ecd3bb6d9d09d931e8781f510db5f8b872ca5435597c491e3e09f
 size 14496051368

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:ab50f23c1aaf37539cceb8a6e037aeaddaf01b87e4fda8052fcb18b212675b63
+oid sha256:12136910eacd4c6e79b2166af73d8b9d967873b33a0b3166601f82755dab3512
 size 14496051368

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:d3f9740857f12a9e17a0bb2167b952ad636f373b53131ab9f12eea74b4f8e334
+oid sha256:d32b5002fb69f6281c17eaa037badfe008528c7d8e81c81b7cff115d555df612
 size 14496051368

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 sha256:6af24144f14ae2b3674e963fe51b610f2eb5c9418c64834e71ee2ea716d2e6fb
+oid sha256:61d5356d4a15061ea562977c5a21355be8902685c315c4a848e71d0b3c2850c1
 size 4949453792

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

@@ -1,3 +1,3 @@
 version https://git-lfs.github.com/spec/v1
-oid sha256:8b47166a94a13efeb16f350434d099c0f9bff78226bb64d1e9402fe9a586bbba
+oid sha256:437bf184ce59319aaa45a6aab6284a2006b10c6d045d56d64da7f44c15f91fbf
 size 4999819336

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

@@ -1,3 +1,3 @@
 version https://git-lfs.github.com/spec/v1
-oid sha256:558c83ccb161da0aca5d5429d81592a3d7a458e70dceb3f5d596667aa2024d9d
+oid sha256:86f7cb1d7aecd12d7aec857282a49e1bc06484bc16cd51a24ac3f525df9499b3
 size 4546807800

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": 39.256392372009124,
+      "grad_norm": 99.19212388021452,
       "learning_rate": 5.3191489361702125e-09,
-      "logits/chosen": -3.34375,
-      "logits/rejected": -3.609375,
-      "logps/chosen": -40.5,
-      "logps/rejected": -49.25,
+      "logits/chosen": -3.578125,
+      "logits/rejected": -3.46875,
+      "logps/chosen": -247.0,
+      "logps/rejected": -245.0,
       "loss": 0.6914,
       "rewards/accuracies": 0.0,
       "rewards/chosen": 0.0,
@@ -25,1446 +25,484 @@
     },
     {
       "epoch": 0.03194888178913738,
-      "grad_norm": 32.81412666334126,
+      "grad_norm": 100.8587700053837,
       "learning_rate": 5.3191489361702123e-08,
-      "logits/chosen": -3.09375,
-      "logits/rejected": -2.859375,
-      "logps/chosen": -41.25,
-      "logps/rejected": -38.25,
-      "loss": 0.6925,
-      "rewards/accuracies": 0.0972222238779068,
-      "rewards/chosen": -0.001739501953125,
-      "rewards/margins": -0.00017261505126953125,
-      "rewards/rejected": -0.00156402587890625,
+      "logits/chosen": -3.609375,
+      "logits/rejected": -3.4375,
+      "logps/chosen": -274.0,
+      "logps/rejected": -264.0,
+      "loss": 0.6877,
+      "rewards/accuracies": 0.125,
+      "rewards/chosen": 0.002777099609375,
+      "rewards/margins": 0.0111083984375,
+      "rewards/rejected": -0.00836181640625,
       "step": 10
     },
     {
       "epoch": 0.06389776357827476,
-      "grad_norm": 31.71771727943446,
+      "grad_norm": 83.15951987287777,
       "learning_rate": 1.0638297872340425e-07,
-      "logits/chosen": -3.03125,
-      "logits/rejected": -2.8125,
-      "logps/chosen": -42.0,
-      "logps/rejected": -36.75,
-      "loss": 0.6887,
-      "rewards/accuracies": 0.15000000596046448,
-      "rewards/chosen": -0.0001583099365234375,
-      "rewards/margins": 0.0050048828125,
-      "rewards/rejected": -0.005157470703125,
+      "logits/chosen": -3.609375,
+      "logits/rejected": -3.5625,
+      "logps/chosen": -274.0,
+      "logps/rejected": -264.0,
+      "loss": 0.6473,
+      "rewards/accuracies": 0.6000000238418579,
+      "rewards/chosen": 0.00872802734375,
+      "rewards/margins": 0.08544921875,
+      "rewards/rejected": -0.076171875,
       "step": 20
     },
     {
       "epoch": 0.09584664536741214,
-      "grad_norm": 33.89035773805732,
+      "grad_norm": 93.7637276950002,
       "learning_rate": 1.5957446808510638e-07,
-      "logits/chosen": -2.859375,
-      "logits/rejected": -3.0625,
-      "logps/chosen": -44.5,
-      "logps/rejected": -42.25,
-      "loss": 0.6852,
-      "rewards/accuracies": 0.32499998807907104,
-      "rewards/chosen": 0.0032958984375,
-      "rewards/margins": 0.0284423828125,
-      "rewards/rejected": -0.025146484375,
+      "logits/chosen": -3.609375,
+      "logits/rejected": -3.515625,
+      "logps/chosen": -284.0,
+      "logps/rejected": -252.0,
+      "loss": 0.5509,
+      "rewards/accuracies": 0.7749999761581421,
+      "rewards/chosen": 0.07275390625,
+      "rewards/margins": 0.390625,
+      "rewards/rejected": -0.318359375,
       "step": 30
     },
     {
       "epoch": 0.12779552715654952,
-      "grad_norm": 33.79231622581764,
+      "grad_norm": 74.20143960903192,
       "learning_rate": 2.127659574468085e-07,
-      "logits/chosen": -3.109375,
-      "logits/rejected": -2.921875,
-      "logps/chosen": -47.5,
-      "logps/rejected": -37.75,
-      "loss": 0.6768,
-      "rewards/accuracies": 0.26249998807907104,
-      "rewards/chosen": 0.03759765625,
-      "rewards/margins": 0.029296875,
-      "rewards/rejected": 0.00830078125,
+      "logits/chosen": -3.46875,
+      "logits/rejected": -3.375,
+      "logps/chosen": -282.0,
+      "logps/rejected": -268.0,
+      "loss": 0.5294,
+      "rewards/accuracies": 0.8500000238418579,
+      "rewards/chosen": 0.251953125,
+      "rewards/margins": 0.7578125,
+      "rewards/rejected": -0.50390625,
       "step": 40
     },
     {
       "epoch": 0.1597444089456869,
-      "grad_norm": 33.962249009797624,
+      "grad_norm": 71.1996189930263,
       "learning_rate": 2.659574468085106e-07,
-      "logits/chosen": -3.046875,
-      "logits/rejected": -2.828125,
-      "logps/chosen": -39.0,
-      "logps/rejected": -35.25,
-      "loss": 0.6589,
-      "rewards/accuracies": 0.32499998807907104,
-      "rewards/chosen": 0.04638671875,
-      "rewards/margins": 0.08740234375,
-      "rewards/rejected": -0.040771484375,
+      "logits/chosen": -3.578125,
+      "logits/rejected": -3.515625,
+      "logps/chosen": -274.0,
+      "logps/rejected": -260.0,
+      "loss": 0.5293,
+      "rewards/accuracies": 0.7749999761581421,
+      "rewards/chosen": 0.287109375,
+      "rewards/margins": 0.90234375,
+      "rewards/rejected": -0.6171875,
       "step": 50
     },
     {
       "epoch": 0.19169329073482427,
-      "grad_norm": 31.06082261346926,
+      "grad_norm": 59.86423059197552,
       "learning_rate": 3.1914893617021275e-07,
-      "logits/chosen": -2.9375,
-      "logits/rejected": -2.890625,
-      "logps/chosen": -36.0,
-      "logps/rejected": -41.5,
-      "loss": 0.6679,
-      "rewards/accuracies": 0.36250001192092896,
-      "rewards/chosen": 0.0947265625,
-      "rewards/margins": 0.087890625,
-      "rewards/rejected": 0.00640869140625,
+      "logits/chosen": -3.5625,
+      "logits/rejected": -3.453125,
+      "logps/chosen": -280.0,
+      "logps/rejected": -270.0,
+      "loss": 0.5133,
+      "rewards/accuracies": 0.8125,
+      "rewards/chosen": 0.263671875,
+      "rewards/margins": 0.94140625,
+      "rewards/rejected": -0.67578125,
       "step": 60
     },
     {
       "epoch": 0.22364217252396165,
-      "grad_norm": 33.61384586606748,
+      "grad_norm": 61.01100752295743,
       "learning_rate": 3.7234042553191484e-07,
-      "logits/chosen": -3.25,
-      "logits/rejected": -2.796875,
-      "logps/chosen": -54.0,
-      "logps/rejected": -39.5,
-      "loss": 0.6207,
-      "rewards/accuracies": 0.4375,
-      "rewards/chosen": 0.054931640625,
-      "rewards/margins": 0.296875,
-      "rewards/rejected": -0.2421875,
+      "logits/chosen": -3.5,
+      "logits/rejected": -3.421875,
+      "logps/chosen": -272.0,
+      "logps/rejected": -272.0,
+      "loss": 0.5364,
+      "rewards/accuracies": 0.7875000238418579,
+      "rewards/chosen": 0.2412109375,
+      "rewards/margins": 0.94140625,
+      "rewards/rejected": -0.69921875,
       "step": 70
     },
     {
       "epoch": 0.25559105431309903,
-      "grad_norm": 29.13019721736027,
+      "grad_norm": 69.23399604165094,
       "learning_rate": 4.25531914893617e-07,
-      "logits/chosen": -3.28125,
-      "logits/rejected": -3.0,
-      "logps/chosen": -49.0,
-      "logps/rejected": -47.0,
-      "loss": 0.6504,
-      "rewards/accuracies": 0.42500001192092896,
-      "rewards/chosen": -0.140625,
-      "rewards/margins": 0.2734375,
-      "rewards/rejected": -0.4140625,
+      "logits/chosen": -3.421875,
+      "logits/rejected": -3.40625,
+      "logps/chosen": -274.0,
+      "logps/rejected": -274.0,
+      "loss": 0.507,
+      "rewards/accuracies": 0.6625000238418579,
+      "rewards/chosen": -0.0260009765625,
+      "rewards/margins": 0.81640625,
+      "rewards/rejected": -0.84375,
       "step": 80
     },
     {
       "epoch": 0.28753993610223644,
-      "grad_norm": 40.06738628161476,
+      "grad_norm": 91.4754984927698,
       "learning_rate": 4.787234042553192e-07,
-      "logits/chosen": -3.390625,
-      "logits/rejected": -2.90625,
-      "logps/chosen": -47.75,
-      "logps/rejected": -38.25,
-      "loss": 0.5986,
-      "rewards/accuracies": 0.3125,
-      "rewards/chosen": -0.00396728515625,
-      "rewards/margins": 0.265625,
-      "rewards/rejected": -0.26953125,
+      "logits/chosen": -3.453125,
+      "logits/rejected": -3.40625,
+      "logps/chosen": -282.0,
+      "logps/rejected": -266.0,
+      "loss": 0.4801,
+      "rewards/accuracies": 0.800000011920929,
+      "rewards/chosen": 0.07421875,
+      "rewards/margins": 0.9453125,
+      "rewards/rejected": -0.87109375,
       "step": 90
     },
     {
       "epoch": 0.3194888178913738,
-      "grad_norm": 26.980268633866878,
+      "grad_norm": 72.3540416301487,
       "learning_rate": 4.964497041420119e-07,
-      "logits/chosen": -3.1875,
-      "logits/rejected": -3.0,
-      "logps/chosen": -43.25,
-      "logps/rejected": -43.25,
-      "loss": 0.6019,
-      "rewards/accuracies": 0.4124999940395355,
-      "rewards/chosen": -0.146484375,
-      "rewards/margins": 0.30859375,
-      "rewards/rejected": -0.455078125,
+      "logits/chosen": -3.421875,
+      "logits/rejected": -3.328125,
+      "logps/chosen": -286.0,
+      "logps/rejected": -264.0,
+      "loss": 0.4994,
+      "rewards/accuracies": 0.75,
+      "rewards/chosen": 0.0986328125,
+      "rewards/margins": 1.09375,
+      "rewards/rejected": -1.0,
       "step": 100
     },
     {
       "epoch": 0.3514376996805112,
-      "grad_norm": 42.49654986796296,
+      "grad_norm": 72.88717398749043,
       "learning_rate": 4.905325443786982e-07,
-      "logits/chosen": -2.984375,
-      "logits/rejected": -3.140625,
-      "logps/chosen": -45.75,
-      "logps/rejected": -53.5,
-      "loss": 0.5866,
-      "rewards/accuracies": 0.36250001192092896,
-      "rewards/chosen": -0.1298828125,
-      "rewards/margins": 0.546875,
-      "rewards/rejected": -0.67578125,
+      "logits/chosen": -3.453125,
+      "logits/rejected": -3.46875,
+      "logps/chosen": -282.0,
+      "logps/rejected": -276.0,
+      "loss": 0.5594,
+      "rewards/accuracies": 0.7124999761581421,
+      "rewards/chosen": -0.0693359375,
+      "rewards/margins": 0.84375,
+      "rewards/rejected": -0.9140625,
       "step": 110
     },
     {
       "epoch": 0.38338658146964855,
-      "grad_norm": 33.45575091341326,
+      "grad_norm": 48.378225872153784,
       "learning_rate": 4.846153846153846e-07,
-      "logits/chosen": -3.375,
-      "logits/rejected": -3.109375,
-      "logps/chosen": -55.75,
-      "logps/rejected": -53.25,
-      "loss": 0.5792,
-      "rewards/accuracies": 0.4375,
-      "rewards/chosen": -0.2890625,
-      "rewards/margins": 0.59375,
-      "rewards/rejected": -0.8828125,
+      "logits/chosen": -3.53125,
+      "logits/rejected": -3.375,
+      "logps/chosen": -280.0,
+      "logps/rejected": -270.0,
+      "loss": 0.4589,
+      "rewards/accuracies": 0.8374999761581421,
+      "rewards/chosen": 0.36328125,
+      "rewards/margins": 1.1875,
+      "rewards/rejected": -0.8203125,
       "step": 120
     },
     {
       "epoch": 0.41533546325878595,
-      "grad_norm": 35.64973041956874,
+      "grad_norm": 97.95130569742476,
       "learning_rate": 4.78698224852071e-07,
-      "logits/chosen": -3.09375,
-      "logits/rejected": -2.96875,
-      "logps/chosen": -45.25,
-      "logps/rejected": -47.0,
-      "loss": 0.5991,
-      "rewards/accuracies": 0.32499998807907104,
-      "rewards/chosen": -0.255859375,
-      "rewards/margins": 0.5,
-      "rewards/rejected": -0.75390625,
+      "logits/chosen": -3.40625,
+      "logits/rejected": -3.375,
+      "logps/chosen": -278.0,
+      "logps/rejected": -274.0,
+      "loss": 0.4837,
+      "rewards/accuracies": 0.762499988079071,
+      "rewards/chosen": 0.0101318359375,
+      "rewards/margins": 1.2421875,
+      "rewards/rejected": -1.234375,
       "step": 130
     },
     {
       "epoch": 0.4472843450479233,
-      "grad_norm": 33.67312459569186,
+      "grad_norm": 81.43086945039742,
       "learning_rate": 4.727810650887574e-07,
-      "logits/chosen": -3.328125,
-      "logits/rejected": -3.046875,
-      "logps/chosen": -48.25,
-      "logps/rejected": -52.75,
-      "loss": 0.544,
-      "rewards/accuracies": 0.44999998807907104,
-      "rewards/chosen": -0.2412109375,
-      "rewards/margins": 0.8359375,
-      "rewards/rejected": -1.078125,
+      "logits/chosen": -3.375,
+      "logits/rejected": -3.390625,
+      "logps/chosen": -276.0,
+      "logps/rejected": -270.0,
+      "loss": 0.5687,
+      "rewards/accuracies": 0.6875,
+      "rewards/chosen": -0.10107421875,
+      "rewards/margins": 1.0078125,
+      "rewards/rejected": -1.109375,
       "step": 140
     },
     {
       "epoch": 0.4792332268370607,
-      "grad_norm": 16.223843544431134,
+      "grad_norm": 73.82888993395747,
       "learning_rate": 4.668639053254438e-07,
-      "logits/chosen": -3.078125,
-      "logits/rejected": -3.03125,
-      "logps/chosen": -48.25,
-      "logps/rejected": -54.25,
-      "loss": 0.5827,
-      "rewards/accuracies": 0.44999998807907104,
-      "rewards/chosen": -0.4609375,
-      "rewards/margins": 0.8046875,
-      "rewards/rejected": -1.265625,
+      "logits/chosen": -3.390625,
+      "logits/rejected": -3.375,
+      "logps/chosen": -284.0,
+      "logps/rejected": -274.0,
+      "loss": 0.5149,
+      "rewards/accuracies": 0.762499988079071,
+      "rewards/chosen": -0.072265625,
+      "rewards/margins": 1.25,
+      "rewards/rejected": -1.3203125,
       "step": 150
     },
     {
       "epoch": 0.5111821086261981,
-      "grad_norm": 30.865326556253784,
+      "grad_norm": 76.07986743694904,
       "learning_rate": 4.6094674556213014e-07,
-      "logits/chosen": -2.984375,
-      "logits/rejected": -2.75,
-      "logps/chosen": -46.75,
-      "logps/rejected": -51.0,
-      "loss": 0.5442,
-      "rewards/accuracies": 0.4749999940395355,
-      "rewards/chosen": -0.4296875,
-      "rewards/margins": 0.875,
-      "rewards/rejected": -1.3046875,
+      "logits/chosen": -3.453125,
+      "logits/rejected": -3.390625,
+      "logps/chosen": -286.0,
+      "logps/rejected": -270.0,
+      "loss": 0.5211,
+      "rewards/accuracies": 0.7875000238418579,
+      "rewards/chosen": 0.023681640625,
+      "rewards/margins": 1.5,
+      "rewards/rejected": -1.4765625,
       "step": 160
     },
     {
       "epoch": 0.5431309904153354,
-      "grad_norm": 22.81312034142999,
+      "grad_norm": 54.39165794817972,
       "learning_rate": 4.5502958579881655e-07,
-      "logits/chosen": -3.09375,
-      "logits/rejected": -2.8125,
-      "logps/chosen": -44.0,
-      "logps/rejected": -49.5,
-      "loss": 0.5765,
-      "rewards/accuracies": 0.42500001192092896,
-      "rewards/chosen": -0.4609375,
-      "rewards/margins": 0.8203125,
-      "rewards/rejected": -1.28125,
+      "logits/chosen": -3.53125,
+      "logits/rejected": -3.421875,
+      "logps/chosen": -278.0,
+      "logps/rejected": -276.0,
+      "loss": 0.4454,
+      "rewards/accuracies": 0.762499988079071,
+      "rewards/chosen": 0.10009765625,
+      "rewards/margins": 1.6171875,
+      "rewards/rejected": -1.5234375,
       "step": 170
     },
     {
       "epoch": 0.5750798722044729,
-      "grad_norm": 50.05578499574947,
+      "grad_norm": 71.47811041177121,
       "learning_rate": 4.491124260355029e-07,
-      "logits/chosen": -3.140625,
-      "logits/rejected": -2.921875,
-      "logps/chosen": -49.25,
-      "logps/rejected": -56.25,
-      "loss": 0.5266,
-      "rewards/accuracies": 0.4375,
-      "rewards/chosen": -0.42578125,
-      "rewards/margins": 1.1796875,
-      "rewards/rejected": -1.609375,
+      "logits/chosen": -3.484375,
+      "logits/rejected": -3.25,
+      "logps/chosen": -278.0,
+      "logps/rejected": -278.0,
+      "loss": 0.5029,
+      "rewards/accuracies": 0.7250000238418579,
+      "rewards/chosen": 0.02392578125,
+      "rewards/margins": 1.34375,
+      "rewards/rejected": -1.3203125,
       "step": 180
     },
     {
       "epoch": 0.6070287539936102,
-      "grad_norm": 20.57671100687128,
+      "grad_norm": 63.19140078208231,
       "learning_rate": 4.4319526627218936e-07,
-      "logits/chosen": -3.3125,
-      "logits/rejected": -2.734375,
-      "logps/chosen": -56.25,
-      "logps/rejected": -45.5,
-      "loss": 0.5544,
-      "rewards/accuracies": 0.4000000059604645,
-      "rewards/chosen": -0.5234375,
-      "rewards/margins": 0.70703125,
-      "rewards/rejected": -1.2265625,
+      "logits/chosen": -3.421875,
+      "logits/rejected": -3.296875,
+      "logps/chosen": -284.0,
+      "logps/rejected": -278.0,
+      "loss": 0.4525,
+      "rewards/accuracies": 0.762499988079071,
+      "rewards/chosen": 0.0172119140625,
+      "rewards/margins": 1.703125,
+      "rewards/rejected": -1.6875,
       "step": 190
     },
     {
       "epoch": 0.6389776357827476,
-      "grad_norm": 35.51097905099775,
+      "grad_norm": 60.16768107522096,
       "learning_rate": 4.372781065088757e-07,
-      "logits/chosen": -3.09375,
-      "logits/rejected": -2.890625,
-      "logps/chosen": -49.75,
-      "logps/rejected": -53.0,
-      "loss": 0.6029,
-      "rewards/accuracies": 0.38749998807907104,
-      "rewards/chosen": -0.5546875,
-      "rewards/margins": 0.83984375,
-      "rewards/rejected": -1.390625,
+      "logits/chosen": -3.34375,
+      "logits/rejected": -3.34375,
+      "logps/chosen": -286.0,
+      "logps/rejected": -272.0,
+      "loss": 0.4489,
+      "rewards/accuracies": 0.7875000238418579,
+      "rewards/chosen": -0.03125,
+      "rewards/margins": 1.6015625,
+      "rewards/rejected": -1.625,
       "step": 200
     },
     {
       "epoch": 0.670926517571885,
-      "grad_norm": 17.899965646039988,
+      "grad_norm": 53.1463114314995,
       "learning_rate": 4.313609467455621e-07,
-      "logits/chosen": -3.140625,
-      "logits/rejected": -3.015625,
-      "logps/chosen": -44.5,
-      "logps/rejected": -50.0,
-      "loss": 0.5345,
-      "rewards/accuracies": 0.38749998807907104,
-      "rewards/chosen": -0.54296875,
-      "rewards/margins": 0.796875,
-      "rewards/rejected": -1.34375,
+      "logits/chosen": -3.359375,
+      "logits/rejected": -3.34375,
+      "logps/chosen": -284.0,
+      "logps/rejected": -276.0,
+      "loss": 0.5062,
+      "rewards/accuracies": 0.800000011920929,
+      "rewards/chosen": -0.158203125,
+      "rewards/margins": 1.40625,
+      "rewards/rejected": -1.5625,
       "step": 210
     },
     {
       "epoch": 0.7028753993610224,
-      "grad_norm": 21.641681357355413,
+      "grad_norm": 74.19788896651235,
       "learning_rate": 4.2544378698224847e-07,
-      "logits/chosen": -2.828125,
-      "logits/rejected": -2.890625,
-      "logps/chosen": -48.5,
-      "logps/rejected": -53.75,
-      "loss": 0.5107,
-      "rewards/accuracies": 0.4000000059604645,
-      "rewards/chosen": -0.546875,
-      "rewards/margins": 0.92578125,
-      "rewards/rejected": -1.4765625,
+      "logits/chosen": -3.359375,
+      "logits/rejected": -3.3125,
+      "logps/chosen": -288.0,
+      "logps/rejected": -272.0,
+      "loss": 0.5495,
+      "rewards/accuracies": 0.7124999761581421,
+      "rewards/chosen": -0.1962890625,
+      "rewards/margins": 1.3671875,
+      "rewards/rejected": -1.5625,
       "step": 220
     },
     {
       "epoch": 0.7348242811501597,
-      "grad_norm": 26.32082815714621,
+      "grad_norm": 37.264928663032826,
       "learning_rate": 4.195266272189349e-07,
-      "logits/chosen": -3.15625,
-      "logits/rejected": -2.828125,
-      "logps/chosen": -57.0,
-      "logps/rejected": -54.25,
-      "loss": 0.574,
-      "rewards/accuracies": 0.42500001192092896,
-      "rewards/chosen": -0.83984375,
-      "rewards/margins": 0.97265625,
-      "rewards/rejected": -1.8125,
+      "logits/chosen": -3.453125,
+      "logits/rejected": -3.390625,
+      "logps/chosen": -282.0,
+      "logps/rejected": -272.0,
+      "loss": 0.5136,
+      "rewards/accuracies": 0.762499988079071,
+      "rewards/chosen": -0.12109375,
+      "rewards/margins": 1.359375,
+      "rewards/rejected": -1.484375,
       "step": 230
     },
     {
       "epoch": 0.7667731629392971,
-      "grad_norm": 25.67855359237542,
+      "grad_norm": 51.9838491014926,
       "learning_rate": 4.1360946745562133e-07,
-      "logits/chosen": -3.0625,
-      "logits/rejected": -2.609375,
-      "logps/chosen": -47.75,
-      "logps/rejected": -47.5,
-      "loss": 0.5904,
-      "rewards/accuracies": 0.32499998807907104,
-      "rewards/chosen": -0.7109375,
-      "rewards/margins": 0.625,
-      "rewards/rejected": -1.3359375,
+      "logits/chosen": -3.46875,
+      "logits/rejected": -3.4375,
+      "logps/chosen": -280.0,
+      "logps/rejected": -276.0,
+      "loss": 0.4229,
+      "rewards/accuracies": 0.800000011920929,
+      "rewards/chosen": -0.0771484375,
+      "rewards/margins": 1.359375,
+      "rewards/rejected": -1.4375,
       "step": 240
     },
     {
       "epoch": 0.7987220447284346,
-      "grad_norm": 9.719975462174387,
+      "grad_norm": 66.37834170527424,
       "learning_rate": 4.076923076923077e-07,
-      "logits/chosen": -3.25,
-      "logits/rejected": -3.03125,
-      "logps/chosen": -52.0,
-      "logps/rejected": -62.5,
-      "loss": 0.4791,
-      "rewards/accuracies": 0.4749999940395355,
-      "rewards/chosen": -0.5078125,
-      "rewards/margins": 1.5234375,
-      "rewards/rejected": -2.03125,
+      "logits/chosen": -3.484375,
+      "logits/rejected": -3.46875,
+      "logps/chosen": -282.0,
+      "logps/rejected": -278.0,
+      "loss": 0.491,
+      "rewards/accuracies": 0.862500011920929,
+      "rewards/chosen": -0.058349609375,
+      "rewards/margins": 1.703125,
+      "rewards/rejected": -1.765625,
       "step": 250
     },
     {
       "epoch": 0.8306709265175719,
-      "grad_norm": 29.21855201998469,
+      "grad_norm": 56.1953898588245,
       "learning_rate": 4.017751479289941e-07,
-      "logits/chosen": -3.125,
-      "logits/rejected": -2.625,
-      "logps/chosen": -48.5,
-      "logps/rejected": -52.75,
-      "loss": 0.4803,
-      "rewards/accuracies": 0.4124999940395355,
-      "rewards/chosen": -0.953125,
-      "rewards/margins": 0.94140625,
-      "rewards/rejected": -1.890625,
+      "logits/chosen": -3.4375,
+      "logits/rejected": -3.421875,
+      "logps/chosen": -284.0,
+      "logps/rejected": -280.0,
+      "loss": 0.4617,
+      "rewards/accuracies": 0.824999988079071,
+      "rewards/chosen": 0.0272216796875,
+      "rewards/margins": 1.7734375,
+      "rewards/rejected": -1.7421875,
       "step": 260
     },
     {
       "epoch": 0.8626198083067093,
-      "grad_norm": 25.205926846344735,
+      "grad_norm": 55.24726339529526,
       "learning_rate": 3.9585798816568044e-07,
-      "logits/chosen": -3.28125,
-      "logits/rejected": -2.859375,
-      "logps/chosen": -50.75,
-      "logps/rejected": -60.25,
-      "loss": 0.5525,
-      "rewards/accuracies": 0.4749999940395355,
-      "rewards/chosen": -0.87890625,
-      "rewards/margins": 1.296875,
-      "rewards/rejected": -2.171875,
+      "logits/chosen": -3.484375,
+      "logits/rejected": -3.375,
+      "logps/chosen": -278.0,
+      "logps/rejected": -286.0,
+      "loss": 0.4634,
+      "rewards/accuracies": 0.7124999761581421,
+      "rewards/chosen": -0.2060546875,
+      "rewards/margins": 1.3671875,
+      "rewards/rejected": -1.578125,
       "step": 270
     },
     {
       "epoch": 0.8945686900958466,
-      "grad_norm": 36.1167561440604,
+      "grad_norm": 52.41202133207622,
       "learning_rate": 3.8994082840236685e-07,
-      "logits/chosen": -3.125,
-      "logits/rejected": -2.8125,
-      "logps/chosen": -53.5,
-      "logps/rejected": -54.5,
-      "loss": 0.5372,
-      "rewards/accuracies": 0.4625000059604645,
-      "rewards/chosen": -0.546875,
-      "rewards/margins": 1.265625,
-      "rewards/rejected": -1.8125,
+      "logits/chosen": -3.46875,
+      "logits/rejected": -3.4375,
+      "logps/chosen": -280.0,
+      "logps/rejected": -274.0,
+      "loss": 0.4426,
+      "rewards/accuracies": 0.8125,
+      "rewards/chosen": 0.1787109375,
+      "rewards/margins": 1.8359375,
+      "rewards/rejected": -1.65625,
       "step": 280
     },
     {
       "epoch": 0.9265175718849841,
-      "grad_norm": 30.60002887625237,
+      "grad_norm": 71.23457030042947,
       "learning_rate": 3.840236686390532e-07,
-      "logits/chosen": -3.234375,
-      "logits/rejected": -2.875,
-      "logps/chosen": -57.75,
-      "logps/rejected": -64.5,
-      "loss": 0.5178,
-      "rewards/accuracies": 0.44999998807907104,
-      "rewards/chosen": -0.8046875,
-      "rewards/margins": 1.6328125,
-      "rewards/rejected": -2.4375,
+      "logits/chosen": -3.53125,
+      "logits/rejected": -3.40625,
+      "logps/chosen": -276.0,
+      "logps/rejected": -280.0,
+      "loss": 0.4684,
+      "rewards/accuracies": 0.7749999761581421,
+      "rewards/chosen": -0.0234375,
+      "rewards/margins": 1.5234375,
+      "rewards/rejected": -1.546875,
       "step": 290
     },
     {
       "epoch": 0.9584664536741214,
-      "grad_norm": 8.950670028012786,
+      "grad_norm": 65.27022274437043,
       "learning_rate": 3.7810650887573966e-07,
-      "logits/chosen": -2.96875,
-      "logits/rejected": -3.03125,
-      "logps/chosen": -47.75,
-      "logps/rejected": -69.5,
-      "loss": 0.5006,
-      "rewards/accuracies": 0.4625000059604645,
-      "rewards/chosen": -0.96875,
-      "rewards/margins": 1.5078125,
-      "rewards/rejected": -2.484375,
+      "logits/chosen": -3.390625,
+      "logits/rejected": -3.375,
+      "logps/chosen": -282.0,
+      "logps/rejected": -280.0,
+      "loss": 0.5252,
+      "rewards/accuracies": 0.800000011920929,
+      "rewards/chosen": -0.11572265625,
+      "rewards/margins": 1.578125,
+      "rewards/rejected": -1.6953125,
       "step": 300
     },
     {
       "epoch": 0.9904153354632588,
-      "grad_norm": 30.917476600395073,
+      "grad_norm": 55.83163671794183,
       "learning_rate": 3.72189349112426e-07,
-      "logits/chosen": -3.09375,
-      "logits/rejected": -2.96875,
-      "logps/chosen": -55.75,
-      "logps/rejected": -67.0,
-      "loss": 0.5163,
-      "rewards/accuracies": 0.42500001192092896,
-      "rewards/chosen": -0.74609375,
-      "rewards/margins": 1.5546875,
-      "rewards/rejected": -2.296875,
+      "logits/chosen": -3.5625,
+      "logits/rejected": -3.46875,
+      "logps/chosen": -276.0,
+      "logps/rejected": -278.0,
+      "loss": 0.5266,
+      "rewards/accuracies": 0.7749999761581421,
+      "rewards/chosen": -0.041259765625,
+      "rewards/margins": 1.5859375,
+      "rewards/rejected": -1.625,
       "step": 310
     },
     {
       "epoch": 1.0,
-      "eval_logits/chosen": -3.296875,
-      "eval_logits/rejected": -2.671875,
-      "eval_logps/chosen": -62.25,
-      "eval_logps/rejected": -57.5,
-      "eval_loss": 0.5627734661102295,
-      "eval_rewards/accuracies": 0.4642857015132904,
-      "eval_rewards/chosen": -0.9453125,
-      "eval_rewards/margins": 1.0078125,
-      "eval_rewards/rejected": -1.953125,
-      "eval_runtime": 12.3094,
-      "eval_samples_per_second": 16.248,
-      "eval_steps_per_second": 0.569,
+      "eval_logits/chosen": -3.53125,
+      "eval_logits/rejected": -3.40625,
+      "eval_logps/chosen": -278.0,
+      "eval_logps/rejected": -278.0,
+      "eval_loss": 0.46781250834465027,
+      "eval_rewards/accuracies": 0.6785714030265808,
+      "eval_rewards/chosen": -0.1728515625,
+      "eval_rewards/margins": 1.09375,
+      "eval_rewards/rejected": -1.265625,
+      "eval_runtime": 12.4491,
+      "eval_samples_per_second": 16.065,
+      "eval_steps_per_second": 0.562,
       "step": 313
-    },
-    {
-      "epoch": 1.0223642172523961,
-      "grad_norm": 19.979402907011902,
-      "learning_rate": 3.662721893491124e-07,
-      "logits/chosen": -2.90625,
-      "logits/rejected": -2.921875,
-      "logps/chosen": -57.75,
-      "logps/rejected": -71.5,
-      "loss": 0.3885,
-      "rewards/accuracies": 0.550000011920929,
-      "rewards/chosen": -0.6875,
-      "rewards/margins": 2.171875,
-      "rewards/rejected": -2.859375,
-      "step": 320
-    },
-    {
-      "epoch": 1.0543130990415335,
-      "grad_norm": 19.525096175475955,
-      "learning_rate": 3.603550295857988e-07,
-      "logits/chosen": -3.21875,
-      "logits/rejected": -2.953125,
-      "logps/chosen": -54.25,
-      "logps/rejected": -83.0,
-      "loss": 0.3903,
-      "rewards/accuracies": 0.5874999761581421,
-      "rewards/chosen": -0.80859375,
-      "rewards/margins": 2.8125,
-      "rewards/rejected": -3.625,
-      "step": 330
-    },
-    {
-      "epoch": 1.0862619808306708,
-      "grad_norm": 9.288653666013474,
-      "learning_rate": 3.544378698224852e-07,
-      "logits/chosen": -2.875,
-      "logits/rejected": -3.0,
-      "logps/chosen": -54.0,
-      "logps/rejected": -79.0,
-      "loss": 0.4058,
-      "rewards/accuracies": 0.4625000059604645,
-      "rewards/chosen": -1.1953125,
-      "rewards/margins": 2.21875,
-      "rewards/rejected": -3.40625,
-      "step": 340
-    },
-    {
-      "epoch": 1.1182108626198084,
-      "grad_norm": 11.383357395078532,
-      "learning_rate": 3.485207100591716e-07,
-      "logits/chosen": -3.09375,
-      "logits/rejected": -2.90625,
-      "logps/chosen": -51.25,
-      "logps/rejected": -65.5,
-      "loss": 0.4021,
-      "rewards/accuracies": 0.4749999940395355,
-      "rewards/chosen": -0.3828125,
-      "rewards/margins": 2.171875,
-      "rewards/rejected": -2.5625,
-      "step": 350
-    },
-    {
-      "epoch": 1.1501597444089458,
-      "grad_norm": 9.539795711347603,
-      "learning_rate": 3.42603550295858e-07,
-      "logits/chosen": -3.125,
-      "logits/rejected": -2.71875,
-      "logps/chosen": -52.5,
-      "logps/rejected": -74.0,
-      "loss": 0.4098,
-      "rewards/accuracies": 0.4625000059604645,
-      "rewards/chosen": -0.828125,
-      "rewards/margins": 2.546875,
-      "rewards/rejected": -3.375,
-      "step": 360
-    },
-    {
-      "epoch": 1.182108626198083,
-      "grad_norm": 5.410467971043836,
-      "learning_rate": 3.366863905325444e-07,
-      "logits/chosen": -2.96875,
-      "logits/rejected": -2.625,
-      "logps/chosen": -55.5,
-      "logps/rejected": -77.5,
-      "loss": 0.412,
-      "rewards/accuracies": 0.4375,
-      "rewards/chosen": -1.3671875,
-      "rewards/margins": 2.546875,
-      "rewards/rejected": -3.921875,
-      "step": 370
-    },
-    {
-      "epoch": 1.2140575079872205,
-      "grad_norm": 42.894653920828944,
-      "learning_rate": 3.3076923076923075e-07,
-      "logits/chosen": -3.0,
-      "logits/rejected": -2.921875,
-      "logps/chosen": -56.0,
-      "logps/rejected": -90.5,
-      "loss": 0.3701,
-      "rewards/accuracies": 0.5625,
-      "rewards/chosen": -1.4296875,
-      "rewards/margins": 2.984375,
-      "rewards/rejected": -4.4375,
-      "step": 380
-    },
-    {
-      "epoch": 1.2460063897763578,
-      "grad_norm": 4.660091869361321,
-      "learning_rate": 3.2485207100591716e-07,
-      "logits/chosen": -3.25,
-      "logits/rejected": -2.828125,
-      "logps/chosen": -58.5,
-      "logps/rejected": -83.5,
-      "loss": 0.3882,
-      "rewards/accuracies": 0.5375000238418579,
-      "rewards/chosen": -1.078125,
-      "rewards/margins": 2.890625,
-      "rewards/rejected": -3.96875,
-      "step": 390
-    },
-    {
-      "epoch": 1.2779552715654952,
-      "grad_norm": 10.38551782674793,
-      "learning_rate": 3.189349112426035e-07,
-      "logits/chosen": -2.71875,
-      "logits/rejected": -2.9375,
-      "logps/chosen": -51.0,
-      "logps/rejected": -105.0,
-      "loss": 0.3654,
-      "rewards/accuracies": 0.5375000238418579,
-      "rewards/chosen": -1.8828125,
-      "rewards/margins": 3.15625,
-      "rewards/rejected": -5.03125,
-      "step": 400
-    },
-    {
-      "epoch": 1.3099041533546325,
-      "grad_norm": 39.726027217808905,
-      "learning_rate": 3.130177514792899e-07,
-      "logits/chosen": -2.65625,
-      "logits/rejected": -2.75,
-      "logps/chosen": -55.75,
-      "logps/rejected": -84.0,
-      "loss": 0.4834,
-      "rewards/accuracies": 0.4000000059604645,
-      "rewards/chosen": -2.234375,
-      "rewards/margins": 2.09375,
-      "rewards/rejected": -4.3125,
-      "step": 410
-    },
-    {
-      "epoch": 1.34185303514377,
-      "grad_norm": 24.27561181663484,
-      "learning_rate": 3.071005917159763e-07,
-      "logits/chosen": -2.953125,
-      "logits/rejected": -2.953125,
-      "logps/chosen": -64.5,
-      "logps/rejected": -93.5,
-      "loss": 0.3913,
-      "rewards/accuracies": 0.5,
-      "rewards/chosen": -2.28125,
-      "rewards/margins": 2.671875,
-      "rewards/rejected": -4.9375,
-      "step": 420
-    },
-    {
-      "epoch": 1.3738019169329074,
-      "grad_norm": 19.45683862793217,
-      "learning_rate": 3.011834319526627e-07,
-      "logits/chosen": -3.140625,
-      "logits/rejected": -2.84375,
-      "logps/chosen": -72.0,
-      "logps/rejected": -95.0,
-      "loss": 0.4083,
-      "rewards/accuracies": 0.512499988079071,
-      "rewards/chosen": -2.46875,
-      "rewards/margins": 2.828125,
-      "rewards/rejected": -5.3125,
-      "step": 430
-    },
-    {
-      "epoch": 1.4057507987220448,
-      "grad_norm": 1.2563354685102903,
-      "learning_rate": 2.952662721893491e-07,
-      "logits/chosen": -3.234375,
-      "logits/rejected": -2.859375,
-      "logps/chosen": -68.0,
-      "logps/rejected": -87.5,
-      "loss": 0.3984,
-      "rewards/accuracies": 0.44999998807907104,
-      "rewards/chosen": -2.3125,
-      "rewards/margins": 2.546875,
-      "rewards/rejected": -4.875,
-      "step": 440
-    },
-    {
-      "epoch": 1.4376996805111821,
-      "grad_norm": 4.240474199552514,
-      "learning_rate": 2.893491124260355e-07,
-      "logits/chosen": -3.109375,
-      "logits/rejected": -2.5625,
-      "logps/chosen": -73.0,
-      "logps/rejected": -79.5,
-      "loss": 0.3611,
-      "rewards/accuracies": 0.44999998807907104,
-      "rewards/chosen": -2.203125,
-      "rewards/margins": 2.3125,
-      "rewards/rejected": -4.53125,
-      "step": 450
-    },
-    {
-      "epoch": 1.4696485623003195,
-      "grad_norm": 3.2015246666235604,
-      "learning_rate": 2.834319526627219e-07,
-      "logits/chosen": -3.09375,
-      "logits/rejected": -2.75,
-      "logps/chosen": -69.0,
-      "logps/rejected": -91.5,
-      "loss": 0.3509,
-      "rewards/accuracies": 0.512499988079071,
-      "rewards/chosen": -2.265625,
-      "rewards/margins": 3.03125,
-      "rewards/rejected": -5.3125,
-      "step": 460
-    },
-    {
-      "epoch": 1.5015974440894568,
-      "grad_norm": 3.7256888631342133,
-      "learning_rate": 2.7751479289940824e-07,
-      "logits/chosen": -3.09375,
-      "logits/rejected": -2.828125,
-      "logps/chosen": -70.0,
-      "logps/rejected": -109.0,
-      "loss": 0.3974,
-      "rewards/accuracies": 0.5,
-      "rewards/chosen": -2.703125,
-      "rewards/margins": 3.734375,
-      "rewards/rejected": -6.4375,
-      "step": 470
-    },
-    {
-      "epoch": 1.5335463258785942,
-      "grad_norm": 30.783598585888882,
-      "learning_rate": 2.715976331360947e-07,
-      "logits/chosen": -2.984375,
-      "logits/rejected": -2.609375,
-      "logps/chosen": -73.5,
-      "logps/rejected": -97.0,
-      "loss": 0.3989,
-      "rewards/accuracies": 0.550000011920929,
-      "rewards/chosen": -2.28125,
-      "rewards/margins": 3.484375,
-      "rewards/rejected": -5.75,
-      "step": 480
-    },
-    {
-      "epoch": 1.5654952076677318,
-      "grad_norm": 21.790945391552757,
-      "learning_rate": 2.6568047337278106e-07,
-      "logits/chosen": -2.828125,
-      "logits/rejected": -2.40625,
-      "logps/chosen": -67.0,
-      "logps/rejected": -85.0,
-      "loss": 0.3987,
-      "rewards/accuracies": 0.4124999940395355,
-      "rewards/chosen": -2.4375,
-      "rewards/margins": 2.53125,
-      "rewards/rejected": -4.96875,
-      "step": 490
-    },
-    {
-      "epoch": 1.5974440894568689,
-      "grad_norm": 18.515662271373987,
-      "learning_rate": 2.5976331360946746e-07,
-      "logits/chosen": -3.140625,
-      "logits/rejected": -2.578125,
-      "logps/chosen": -64.5,
-      "logps/rejected": -88.5,
-      "loss": 0.3768,
-      "rewards/accuracies": 0.4625000059604645,
-      "rewards/chosen": -1.8671875,
-      "rewards/margins": 3.203125,
-      "rewards/rejected": -5.09375,
-      "step": 500
-    },
-    {
-      "epoch": 1.6293929712460065,
-      "grad_norm": 31.795221660155256,
-      "learning_rate": 2.538461538461538e-07,
-      "logits/chosen": -2.796875,
-      "logits/rejected": -2.65625,
-      "logps/chosen": -61.25,
-      "logps/rejected": -92.0,
-      "loss": 0.4063,
-      "rewards/accuracies": 0.4749999940395355,
-      "rewards/chosen": -2.28125,
-      "rewards/margins": 2.875,
-      "rewards/rejected": -5.15625,
-      "step": 510
-    },
-    {
-      "epoch": 1.6613418530351438,
-      "grad_norm": 29.964446819474464,
-      "learning_rate": 2.479289940828402e-07,
-      "logits/chosen": -3.328125,
-      "logits/rejected": -2.765625,
-      "logps/chosen": -70.0,
-      "logps/rejected": -86.5,
-      "loss": 0.3787,
-      "rewards/accuracies": 0.512499988079071,
-      "rewards/chosen": -1.625,
-      "rewards/margins": 3.171875,
-      "rewards/rejected": -4.8125,
-      "step": 520
-    },
-    {
-      "epoch": 1.6932907348242812,
-      "grad_norm": 2.216495724515744,
-      "learning_rate": 2.420118343195266e-07,
-      "logits/chosen": -3.25,
-      "logits/rejected": -2.828125,
-      "logps/chosen": -64.0,
-      "logps/rejected": -79.0,
-      "loss": 0.3663,
-      "rewards/accuracies": 0.4749999940395355,
-      "rewards/chosen": -1.7578125,
-      "rewards/margins": 2.625,
-      "rewards/rejected": -4.375,
-      "step": 530
-    },
-    {
-      "epoch": 1.7252396166134185,
-      "grad_norm": 76.57538198001816,
-      "learning_rate": 2.36094674556213e-07,
-      "logits/chosen": -3.265625,
-      "logits/rejected": -2.671875,
-      "logps/chosen": -74.5,
-      "logps/rejected": -90.5,
-      "loss": 0.4094,
-      "rewards/accuracies": 0.5,
-      "rewards/chosen": -1.9375,
-      "rewards/margins": 3.203125,
-      "rewards/rejected": -5.125,
-      "step": 540
-    },
-    {
-      "epoch": 1.7571884984025559,
-      "grad_norm": 20.192906544010256,
-      "learning_rate": 2.301775147928994e-07,
-      "logits/chosen": -2.90625,
-      "logits/rejected": -2.625,
-      "logps/chosen": -69.5,
-      "logps/rejected": -108.0,
-      "loss": 0.3854,
-      "rewards/accuracies": 0.5,
-      "rewards/chosen": -2.71875,
-      "rewards/margins": 3.828125,
-      "rewards/rejected": -6.5625,
-      "step": 550
-    },
-    {
-      "epoch": 1.7891373801916934,
-      "grad_norm": 12.963480738098086,
-      "learning_rate": 2.242603550295858e-07,
-      "logits/chosen": -3.0,
-      "logits/rejected": -2.890625,
-      "logps/chosen": -68.0,
-      "logps/rejected": -110.0,
-      "loss": 0.3581,
-      "rewards/accuracies": 0.5625,
-      "rewards/chosen": -2.46875,
-      "rewards/margins": 3.671875,
-      "rewards/rejected": -6.125,
-      "step": 560
-    },
-    {
-      "epoch": 1.8210862619808306,
-      "grad_norm": 1.229526550945938,
-      "learning_rate": 2.1834319526627217e-07,
-      "logits/chosen": -2.921875,
-      "logits/rejected": -2.78125,
-      "logps/chosen": -67.5,
-      "logps/rejected": -96.5,
-      "loss": 0.3482,
-      "rewards/accuracies": 0.48750001192092896,
-      "rewards/chosen": -1.9921875,
-      "rewards/margins": 3.375,
-      "rewards/rejected": -5.375,
-      "step": 570
-    },
-    {
-      "epoch": 1.8530351437699681,
-      "grad_norm": 28.7303260287631,
-      "learning_rate": 2.1242603550295858e-07,
-      "logits/chosen": -2.96875,
-      "logits/rejected": -2.75,
-      "logps/chosen": -68.5,
-      "logps/rejected": -97.0,
-      "loss": 0.4004,
-      "rewards/accuracies": 0.5625,
-      "rewards/chosen": -1.9921875,
-      "rewards/margins": 3.65625,
-      "rewards/rejected": -5.625,
-      "step": 580
-    },
-    {
-      "epoch": 1.8849840255591053,
-      "grad_norm": 19.014790872128135,
-      "learning_rate": 2.0650887573964496e-07,
-      "logits/chosen": -2.984375,
-      "logits/rejected": -2.65625,
-      "logps/chosen": -70.5,
-      "logps/rejected": -98.5,
-      "loss": 0.3753,
-      "rewards/accuracies": 0.5249999761581421,
-      "rewards/chosen": -2.375,
-      "rewards/margins": 3.390625,
-      "rewards/rejected": -5.78125,
-      "step": 590
-    },
-    {
-      "epoch": 1.9169329073482428,
-      "grad_norm": 35.984289038664066,
-      "learning_rate": 2.0059171597633133e-07,
-      "logits/chosen": -2.796875,
-      "logits/rejected": -2.4375,
-      "logps/chosen": -71.0,
-      "logps/rejected": -94.0,
-      "loss": 0.3799,
-      "rewards/accuracies": 0.42500001192092896,
-      "rewards/chosen": -3.0625,
-      "rewards/margins": 2.59375,
-      "rewards/rejected": -5.65625,
-      "step": 600
-    },
-    {
-      "epoch": 1.9488817891373802,
-      "grad_norm": 13.792657937454177,
-      "learning_rate": 1.9467455621301774e-07,
-      "logits/chosen": -2.546875,
-      "logits/rejected": -2.90625,
-      "logps/chosen": -66.5,
-      "logps/rejected": -107.0,
-      "loss": 0.3635,
-      "rewards/accuracies": 0.5375000238418579,
-      "rewards/chosen": -2.75,
-      "rewards/margins": 3.328125,
-      "rewards/rejected": -6.0625,
-      "step": 610
-    },
-    {
-      "epoch": 1.9808306709265175,
-      "grad_norm": 5.937775728533351,
-      "learning_rate": 1.8875739644970412e-07,
-      "logits/chosen": -2.59375,
-      "logits/rejected": -2.8125,
-      "logps/chosen": -72.5,
-      "logps/rejected": -110.0,
-      "loss": 0.421,
-      "rewards/accuracies": 0.5249999761581421,
-      "rewards/chosen": -2.796875,
-      "rewards/margins": 3.828125,
-      "rewards/rejected": -6.625,
-      "step": 620
-    },
-    {
-      "epoch": 2.0,
-      "eval_logits/chosen": -3.15625,
-      "eval_logits/rejected": -2.4375,
-      "eval_logps/chosen": -84.5,
-      "eval_logps/rejected": -89.5,
-      "eval_loss": 0.6398046612739563,
-      "eval_rewards/accuracies": 0.5,
-      "eval_rewards/chosen": -3.1875,
-      "eval_rewards/margins": 1.9765625,
-      "eval_rewards/rejected": -5.15625,
-      "eval_runtime": 12.4392,
-      "eval_samples_per_second": 16.078,
-      "eval_steps_per_second": 0.563,
-      "step": 626
-    },
-    {
-      "epoch": 2.012779552715655,
-      "grad_norm": 0.24081084446226275,
-      "learning_rate": 1.8284023668639053e-07,
-      "logits/chosen": -2.734375,
-      "logits/rejected": -2.84375,
-      "logps/chosen": -68.0,
-      "logps/rejected": -100.0,
-      "loss": 0.3339,
-      "rewards/accuracies": 0.4625000059604645,
-      "rewards/chosen": -2.578125,
-      "rewards/margins": 3.109375,
-      "rewards/rejected": -5.6875,
-      "step": 630
-    },
-    {
-      "epoch": 2.0447284345047922,
-      "grad_norm": 1.4786780156913717,
-      "learning_rate": 1.7692307692307693e-07,
-      "logits/chosen": -2.6875,
-      "logits/rejected": -2.46875,
-      "logps/chosen": -75.5,
-      "logps/rejected": -102.5,
-      "loss": 0.3557,
-      "rewards/accuracies": 0.5,
-      "rewards/chosen": -2.6875,
-      "rewards/margins": 3.6875,
-      "rewards/rejected": -6.375,
-      "step": 640
-    },
-    {
-      "epoch": 2.07667731629393,
-      "grad_norm": 0.43777076786049535,
-      "learning_rate": 1.710059171597633e-07,
-      "logits/chosen": -2.71875,
-      "logits/rejected": -2.328125,
-      "logps/chosen": -74.0,
-      "logps/rejected": -105.5,
-      "loss": 0.3302,
-      "rewards/accuracies": 0.4625000059604645,
-      "rewards/chosen": -2.890625,
-      "rewards/margins": 3.8125,
-      "rewards/rejected": -6.71875,
-      "step": 650
-    },
-    {
-      "epoch": 2.108626198083067,
-      "grad_norm": 0.7933794542571138,
-      "learning_rate": 1.650887573964497e-07,
-      "logits/chosen": -2.5625,
-      "logits/rejected": -2.078125,
-      "logps/chosen": -74.0,
-      "logps/rejected": -102.5,
-      "loss": 0.3764,
-      "rewards/accuracies": 0.4625000059604645,
-      "rewards/chosen": -3.109375,
-      "rewards/margins": 3.59375,
-      "rewards/rejected": -6.71875,
-      "step": 660
-    },
-    {
-      "epoch": 2.1405750798722045,
-      "grad_norm": 1.7952137818955538,
-      "learning_rate": 1.591715976331361e-07,
-      "logits/chosen": -2.8125,
-      "logits/rejected": -2.421875,
-      "logps/chosen": -75.5,
-      "logps/rejected": -115.5,
-      "loss": 0.3522,
-      "rewards/accuracies": 0.5375000238418579,
-      "rewards/chosen": -2.890625,
-      "rewards/margins": 4.5625,
-      "rewards/rejected": -7.46875,
-      "step": 670
-    },
-    {
-      "epoch": 2.1725239616613417,
-      "grad_norm": 23.23248789540149,
-      "learning_rate": 1.5325443786982248e-07,
-      "logits/chosen": -2.796875,
-      "logits/rejected": -2.4375,
-      "logps/chosen": -81.0,
-      "logps/rejected": -131.0,
-      "loss": 0.3342,
-      "rewards/accuracies": 0.550000011920929,
-      "rewards/chosen": -3.1875,
-      "rewards/margins": 5.4375,
-      "rewards/rejected": -8.625,
-      "step": 680
-    },
-    {
-      "epoch": 2.2044728434504792,
-      "grad_norm": 0.2830134057080892,
-      "learning_rate": 1.4733727810650885e-07,
-      "logits/chosen": -2.46875,
-      "logits/rejected": -2.734375,
-      "logps/chosen": -78.0,
-      "logps/rejected": -128.0,
-      "loss": 0.3383,
-      "rewards/accuracies": 0.5625,
-      "rewards/chosen": -3.71875,
-      "rewards/margins": 4.53125,
-      "rewards/rejected": -8.25,
-      "step": 690
-    },
-    {
-      "epoch": 2.236421725239617,
-      "grad_norm": 1.6720622855289147,
-      "learning_rate": 1.4142011834319526e-07,
-      "logits/chosen": -2.796875,
-      "logits/rejected": -2.109375,
-      "logps/chosen": -74.0,
-      "logps/rejected": -102.5,
-      "loss": 0.3513,
-      "rewards/accuracies": 0.48750001192092896,
-      "rewards/chosen": -2.703125,
-      "rewards/margins": 4.125,
-      "rewards/rejected": -6.8125,
-      "step": 700
-    },
-    {
-      "epoch": 2.268370607028754,
-      "grad_norm": 0.09193888712214437,
-      "learning_rate": 1.3550295857988164e-07,
-      "logits/chosen": -2.765625,
-      "logits/rejected": -1.8046875,
-      "logps/chosen": -82.0,
-      "logps/rejected": -117.0,
-      "loss": 0.3288,
-      "rewards/accuracies": 0.4375,
-      "rewards/chosen": -3.65625,
-      "rewards/margins": 4.3125,
-      "rewards/rejected": -7.9375,
-      "step": 710
-    },
-    {
-      "epoch": 2.3003194888178915,
-      "grad_norm": 58.04141926321015,
-      "learning_rate": 1.2958579881656802e-07,
-      "logits/chosen": -2.0,
-      "logits/rejected": -2.125,
-      "logps/chosen": -73.0,
-      "logps/rejected": -112.0,
-      "loss": 0.351,
-      "rewards/accuracies": 0.44999998807907104,
-      "rewards/chosen": -3.546875,
-      "rewards/margins": 4.03125,
-      "rewards/rejected": -7.59375,
-      "step": 720
-    },
-    {
-      "epoch": 2.3322683706070286,
-      "grad_norm": 1.4519424377983725,
-      "learning_rate": 1.2366863905325443e-07,
-      "logits/chosen": -2.578125,
-      "logits/rejected": -2.203125,
-      "logps/chosen": -81.0,
-      "logps/rejected": -121.0,
-      "loss": 0.3597,
-      "rewards/accuracies": 0.5,
-      "rewards/chosen": -3.453125,
-      "rewards/margins": 4.59375,
-      "rewards/rejected": -8.0625,
-      "step": 730
-    },
-    {
-      "epoch": 2.364217252396166,
-      "grad_norm": 5.079228609002224,
-      "learning_rate": 1.1775147928994082e-07,
-      "logits/chosen": -2.5625,
-      "logits/rejected": -2.375,
-      "logps/chosen": -79.5,
-      "logps/rejected": -127.0,
-      "loss": 0.339,
-      "rewards/accuracies": 0.5375000238418579,
-      "rewards/chosen": -3.40625,
-      "rewards/margins": 4.8125,
-      "rewards/rejected": -8.25,
-      "step": 740
-    },
-    {
-      "epoch": 2.3961661341853033,
-      "grad_norm": 9.792346923133946,
-      "learning_rate": 1.1183431952662721e-07,
-      "logits/chosen": -2.515625,
-      "logits/rejected": -2.5625,
-      "logps/chosen": -69.5,
-      "logps/rejected": -140.0,
-      "loss": 0.367,
-      "rewards/accuracies": 0.637499988079071,
-      "rewards/chosen": -3.171875,
-      "rewards/margins": 6.21875,
-      "rewards/rejected": -9.375,
-      "step": 750
-    },
-    {
-      "epoch": 2.428115015974441,
-      "grad_norm": 1.3111870951758757,
-      "learning_rate": 1.059171597633136e-07,
-      "logits/chosen": -2.40625,
-      "logits/rejected": -1.8984375,
-      "logps/chosen": -79.0,
-      "logps/rejected": -111.5,
-      "loss": 0.3127,
-      "rewards/accuracies": 0.5,
-      "rewards/chosen": -3.34375,
-      "rewards/margins": 4.40625,
-      "rewards/rejected": -7.75,
-      "step": 760
-    },
-    {
-      "epoch": 2.460063897763578,
-      "grad_norm": 1.9510256297266442,
-      "learning_rate": 1e-07,
-      "logits/chosen": -2.09375,
-      "logits/rejected": -2.5,
-      "logps/chosen": -76.5,
-      "logps/rejected": -144.0,
-      "loss": 0.3348,
-      "rewards/accuracies": 0.574999988079071,
-      "rewards/chosen": -4.15625,
-      "rewards/margins": 5.6875,
-      "rewards/rejected": -9.875,
-      "step": 770
-    },
-    {
-      "epoch": 2.4920127795527156,
-      "grad_norm": 1.9059132414634314,
-      "learning_rate": 9.408284023668639e-08,
-      "logits/chosen": -2.421875,
-      "logits/rejected": -2.21875,
-      "logps/chosen": -80.0,
-      "logps/rejected": -126.0,
-      "loss": 0.3442,
-      "rewards/accuracies": 0.44999998807907104,
-      "rewards/chosen": -4.0625,
-      "rewards/margins": 4.375,
-      "rewards/rejected": -8.4375,
-      "step": 780
-    },
-    {
-      "epoch": 2.523961661341853,
-      "grad_norm": 0.31027943136213487,
-      "learning_rate": 8.816568047337278e-08,
-      "logits/chosen": -2.515625,
-      "logits/rejected": -2.046875,
-      "logps/chosen": -78.0,
-      "logps/rejected": -112.0,
-      "loss": 0.3619,
-      "rewards/accuracies": 0.4000000059604645,
-      "rewards/chosen": -3.71875,
-      "rewards/margins": 3.875,
-      "rewards/rejected": -7.59375,
-      "step": 790
-    },
-    {
-      "epoch": 2.5559105431309903,
-      "grad_norm": 0.2032960693026033,
-      "learning_rate": 8.224852071005916e-08,
-      "logits/chosen": -1.765625,
-      "logits/rejected": -2.40625,
-      "logps/chosen": -76.0,
-      "logps/rejected": -142.0,
-      "loss": 0.3137,
-      "rewards/accuracies": 0.5,
-      "rewards/chosen": -3.78125,
-      "rewards/margins": 5.59375,
-      "rewards/rejected": -9.375,
-      "step": 800
-    },
-    {
-      "epoch": 2.587859424920128,
-      "grad_norm": 3.385333589630319,
-      "learning_rate": 7.633136094674555e-08,
-      "logits/chosen": -2.40625,
-      "logits/rejected": -1.9765625,
-      "logps/chosen": -80.0,
-      "logps/rejected": -111.5,
-      "loss": 0.3805,
-      "rewards/accuracies": 0.48750001192092896,
-      "rewards/chosen": -3.359375,
-      "rewards/margins": 4.40625,
-      "rewards/rejected": -7.78125,
-      "step": 810
-    },
-    {
-      "epoch": 2.619808306709265,
-      "grad_norm": 13.883235922734205,
-      "learning_rate": 7.041420118343195e-08,
-      "logits/chosen": -2.515625,
-      "logits/rejected": -2.15625,
-      "logps/chosen": -80.0,
-      "logps/rejected": -129.0,
-      "loss": 0.3535,
-      "rewards/accuracies": 0.574999988079071,
-      "rewards/chosen": -2.9375,
-      "rewards/margins": 5.84375,
-      "rewards/rejected": -8.75,
-      "step": 820
-    },
-    {
-      "epoch": 2.6517571884984026,
-      "grad_norm": 0.6219296336110521,
-      "learning_rate": 6.449704142011835e-08,
-      "logits/chosen": -2.296875,
-      "logits/rejected": -2.484375,
-      "logps/chosen": -82.5,
-      "logps/rejected": -131.0,
-      "loss": 0.3589,
-      "rewards/accuracies": 0.512499988079071,
-      "rewards/chosen": -3.40625,
-      "rewards/margins": 5.0,
-      "rewards/rejected": -8.4375,
-      "step": 830
-    },
-    {
-      "epoch": 2.68370607028754,
-      "grad_norm": 0.6049056846208167,
-      "learning_rate": 5.857988165680473e-08,
-      "logits/chosen": -2.453125,
-      "logits/rejected": -1.671875,
-      "logps/chosen": -81.5,
-      "logps/rejected": -122.5,
-      "loss": 0.3279,
-      "rewards/accuracies": 0.5874999761581421,
-      "rewards/chosen": -3.28125,
-      "rewards/margins": 5.53125,
-      "rewards/rejected": -8.8125,
-      "step": 840
-    },
-    {
-      "epoch": 2.7156549520766773,
-      "grad_norm": 5.648108737262734,
-      "learning_rate": 5.266272189349112e-08,
-      "logits/chosen": -2.40625,
-      "logits/rejected": -2.453125,
-      "logps/chosen": -80.5,
-      "logps/rejected": -132.0,
-      "loss": 0.3477,
-      "rewards/accuracies": 0.5249999761581421,
-      "rewards/chosen": -3.65625,
-      "rewards/margins": 5.375,
-      "rewards/rejected": -9.0625,
-      "step": 850
-    },
-    {
-      "epoch": 2.747603833865815,
-      "grad_norm": 2.2238661378386153,
-      "learning_rate": 4.674556213017751e-08,
-      "logits/chosen": -2.265625,
-      "logits/rejected": -2.015625,
-      "logps/chosen": -82.5,
-      "logps/rejected": -117.0,
-      "loss": 0.3711,
-      "rewards/accuracies": 0.4124999940395355,
-      "rewards/chosen": -3.921875,
-      "rewards/margins": 3.875,
-      "rewards/rejected": -7.8125,
-      "step": 860
-    },
-    {
-      "epoch": 2.779552715654952,
-      "grad_norm": 3.990432381254641,
-      "learning_rate": 4.082840236686391e-08,
-      "logits/chosen": -2.21875,
-      "logits/rejected": -1.78125,
-      "logps/chosen": -81.0,
-      "logps/rejected": -119.0,
-      "loss": 0.3486,
-      "rewards/accuracies": 0.44999998807907104,
-      "rewards/chosen": -3.953125,
-      "rewards/margins": 4.53125,
-      "rewards/rejected": -8.5,
-      "step": 870
-    },
-    {
-      "epoch": 2.8115015974440896,
-      "grad_norm": 0.11041856289613622,
-      "learning_rate": 3.4911242603550294e-08,
-      "logits/chosen": -1.9609375,
-      "logits/rejected": -2.125,
-      "logps/chosen": -78.0,
-      "logps/rejected": -131.0,
-      "loss": 0.3316,
-      "rewards/accuracies": 0.5,
-      "rewards/chosen": -4.09375,
-      "rewards/margins": 4.96875,
-      "rewards/rejected": -9.0625,
-      "step": 880
-    },
-    {
-      "epoch": 2.8434504792332267,
-      "grad_norm": 0.43906962839319524,
-      "learning_rate": 2.8994082840236687e-08,
-      "logits/chosen": -2.40625,
-      "logits/rejected": -1.890625,
-      "logps/chosen": -79.0,
-      "logps/rejected": -132.0,
-      "loss": 0.3327,
-      "rewards/accuracies": 0.550000011920929,
-      "rewards/chosen": -3.453125,
-      "rewards/margins": 5.96875,
-      "rewards/rejected": -9.375,
-      "step": 890
-    },
-    {
-      "epoch": 2.8753993610223643,
-      "grad_norm": 0.08615648524755898,
-      "learning_rate": 2.3076923076923076e-08,
-      "logits/chosen": -1.734375,
-      "logits/rejected": -1.9296875,
-      "logps/chosen": -74.5,
-      "logps/rejected": -124.0,
-      "loss": 0.3596,
-      "rewards/accuracies": 0.4000000059604645,
-      "rewards/chosen": -4.21875,
-      "rewards/margins": 4.46875,
-      "rewards/rejected": -8.6875,
-      "step": 900
-    },
-    {
-      "epoch": 2.9073482428115014,
-      "grad_norm": 41.28653117321922,
-      "learning_rate": 1.7159763313609465e-08,
-      "logits/chosen": -2.421875,
-      "logits/rejected": -2.078125,
-      "logps/chosen": -76.0,
-      "logps/rejected": -119.5,
-      "loss": 0.3181,
-      "rewards/accuracies": 0.550000011920929,
-      "rewards/chosen": -3.25,
-      "rewards/margins": 4.9375,
-      "rewards/rejected": -8.1875,
-      "step": 910
-    },
-    {
-      "epoch": 2.939297124600639,
-      "grad_norm": 0.17584532058769695,
-      "learning_rate": 1.1242603550295858e-08,
-      "logits/chosen": -2.453125,
-      "logits/rejected": -2.3125,
-      "logps/chosen": -82.0,
-      "logps/rejected": -134.0,
-      "loss": 0.3534,
-      "rewards/accuracies": 0.5625,
-      "rewards/chosen": -3.4375,
-      "rewards/margins": 5.78125,
-      "rewards/rejected": -9.1875,
-      "step": 920
-    },
-    {
-      "epoch": 2.9712460063897765,
-      "grad_norm": 62.045094794045774,
-      "learning_rate": 5.325443786982248e-09,
-      "logits/chosen": -2.171875,
-      "logits/rejected": -1.8203125,
-      "logps/chosen": -82.0,
-      "logps/rejected": -103.5,
-      "loss": 0.3893,
-      "rewards/accuracies": 0.375,
-      "rewards/chosen": -4.0,
-      "rewards/margins": 3.25,
-      "rewards/rejected": -7.25,
-      "step": 930
-    },
-    {
-      "epoch": 3.0,
-      "eval_logits/chosen": -2.6875,
-      "eval_logits/rejected": -1.9296875,
-      "eval_logps/chosen": -100.5,
-      "eval_logps/rejected": -108.0,
-      "eval_loss": 0.7696093916893005,
-      "eval_rewards/accuracies": 0.4821428656578064,
-      "eval_rewards/chosen": -4.75,
-      "eval_rewards/margins": 2.265625,
-      "eval_rewards/rejected": -7.03125,
-      "eval_runtime": 15.5893,
-      "eval_samples_per_second": 12.829,
-      "eval_steps_per_second": 0.449,
-      "step": 939
     }
   ],
   "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:a328acd359f24dbd1d1d47abed6a63056da53b083fbeaa1cca55a71a6cf5453b
-size 7800
+oid sha256:6c14cd6dcdeb844089094f42b08a293cc426987779d1f95f11771d42e5d70200
+size 7864