Model save

README.md

@@ -2,10 +2,7 @@
 license: mit
 base_model: microsoft/Phi-3-small-8k-instruct
 tags:
-- alignment-handbook
 - generated_from_trainer
-datasets:
-- AmberYifan/spin-v
 model-index:
 - name: phi3-spin-zephyr-data
   results: []
@@ -16,15 +13,15 @@ should probably proofread and complete it, then remove this comment. -->
 
 # phi3-spin-zephyr-data
 
-This model is a fine-tuned version of [microsoft/Phi-3-small-8k-instruct](https://huggingface.co/microsoft/Phi-3-small-8k-instruct) on the AmberYifan/spin-v dataset.
+This model is a fine-tuned version of [microsoft/Phi-3-small-8k-instruct](https://huggingface.co/microsoft/Phi-3-small-8k-instruct) on the None dataset.
 It achieves the following results on the evaluation set:
-- Loss: 0.1680
-- Rewards/real: -6.9251
-- Rewards/generated: -22.7311
+- Loss: 0.1748
+- Rewards/real: -6.0641
+- Rewards/generated: -21.7894
 - Rewards/accuracies: 0.9443
-- Rewards/margins: 15.8060
-- Logps/generated: -518.7460
-- Logps/real: -321.6379
+- Rewards/margins: 15.7252
+- Logps/generated: -509.3286
+- Logps/real: -313.0280
 - Logits/generated: -inf
 - Logits/real: -inf
 
@@ -62,7 +59,7 @@ The following hyperparameters were used during training:
 
 | Training Loss | Epoch | Step | Validation Loss | Rewards/real | Rewards/generated | Rewards/accuracies | Rewards/margins | Logps/generated | Logps/real | Logits/generated | Logits/real |
 |:-------------:|:-----:|:----:|:---------------:|:------------:|:-----------------:|:------------------:|:---------------:|:---------------:|:----------:|:----------------:|:-----------:|
-| 0.2807        | 0.64  | 500  | 0.1680          | -6.9251      | -22.7311          | 0.9443             | 15.8060         | -518.7460       | -321.6379  | -inf             | -inf        |
+| 0.2945        | 0.64  | 500  | 0.1748          | -6.0641      | -21.7894          | 0.9443             | 15.7252         | -509.3286       | -313.0280  | -inf             | -inf        |
 
 
 ### Framework versions

all_results.json

@@ -1,8 +1,8 @@
 {
     "epoch": 1.0,
-    "train_loss": 0.6284302195911938,
-    "train_runtime": 5860.6637,
+    "train_loss": 0.6284351689954493,
+    "train_runtime": 6560.8804,
     "train_samples": 25000,
-    "train_samples_per_second": 4.266,
-    "train_steps_per_second": 0.133
+    "train_samples_per_second": 3.81,
+    "train_steps_per_second": 0.119
 }

model-00001-of-00004.safetensors

@@ -1,3 +1,3 @@
 version https://git-lfs.github.com/spec/v1
-oid sha256:e61a098b9291287a13df8ae4fcdb00643a5ec324bf65af655801e5cb47a21618
+oid sha256:1efa10e64d5859b0237937a454e8d0e674af6517f71086f2e7e98bc3147fe71c
 size 4832943104

model-00002-of-00004.safetensors

@@ -1,3 +1,3 @@
 version https://git-lfs.github.com/spec/v1
-oid sha256:1b112110dccae5c1237da32a3ef6bd78b31a00f092836ea5d6725a775764041d
+oid sha256:1ac4af61d8a50b24b2a0451b0d0d8d681f445298adb99946728417f8ea8a5f8c
 size 4799608224

model-00003-of-00004.safetensors

@@ -1,3 +1,3 @@
 version https://git-lfs.github.com/spec/v1
-oid sha256:6b8607487f2f6919b39039edda7f2d6f3a681042b5433ed53b46c2b0e1484738
+oid sha256:4cbc2426e9fb18df7c282f39a9fd4f5523abf48457b84c80ecdc7b84bf172b63
 size 4799608240

model-00004-of-00004.safetensors

@@ -1,3 +1,3 @@
 version https://git-lfs.github.com/spec/v1
-oid sha256:713883d1a11f48447b2142edf25e4ea6b77b0ed0851ce9d404c3ca3b854a8975
+oid sha256:856cab213b70e9a43d567a410c8964012fdc3434091381c99187734430660556
 size 352437304

runs/Jul24_14-12-34_gilbreth-j001.rcac.purdue.edu/events.out.tfevents.1721844898.gilbreth-j001.rcac.purdue.edu.5513.0

@@ -1,3 +1,3 @@
 version https://git-lfs.github.com/spec/v1
-oid sha256:a9c9931847440eb80ad54887342cd5c41ed40869da6dc121426354a790bd2778
-size 52207
+oid sha256:0ff6e5db5f8fbe5faa8099ace021ff1818e37a6f9aa6825947a27551f6fd0333
+size 57609

train_results.json

@@ -1,8 +1,8 @@
 {
     "epoch": 1.0,
-    "train_loss": 0.6284302195911938,
-    "train_runtime": 5860.6637,
+    "train_loss": 0.6284351689954493,
+    "train_runtime": 6560.8804,
     "train_samples": 25000,
-    "train_samples_per_second": 4.266,
-    "train_steps_per_second": 0.133
+    "train_samples_per_second": 3.81,
+    "train_steps_per_second": 0.119
 }

trainer_state.json

@@ -27,13 +27,13 @@
       "learning_rate": 6.329113924050633e-08,
       "logits/generated": -Infinity,
       "logits/real": -Infinity,
-      "logps/generated": -487.2403869628906,
-      "logps/real": -384.84637451171875,
-      "loss": 3.4512,
-      "rewards/accuracies": 0.6666666865348816,
-      "rewards/generated": -19.510425567626953,
-      "rewards/margins": 6.074762344360352,
-      "rewards/real": -13.435664176940918,
+      "logps/generated": -487.2853088378906,
+      "logps/real": -384.8033142089844,
+      "loss": 3.4516,
+      "rewards/accuracies": 0.6111111044883728,
+      "rewards/generated": -19.51491928100586,
+      "rewards/margins": 6.083561420440674,
+      "rewards/real": -13.431358337402344,
       "step": 10
     },
     {
@@ -41,13 +41,13 @@
       "learning_rate": 1.2658227848101266e-07,
       "logits/generated": -Infinity,
       "logits/real": -Infinity,
-      "logps/generated": -473.37353515625,
-      "logps/real": -413.13916015625,
-      "loss": 3.9704,
+      "logps/generated": -473.446044921875,
+      "logps/real": -413.171630859375,
+      "loss": 3.9705,
       "rewards/accuracies": 0.637499988079071,
-      "rewards/generated": -18.36065101623535,
-      "rewards/margins": 3.8984241485595703,
-      "rewards/real": -14.462226867675781,
+      "rewards/generated": -18.367900848388672,
+      "rewards/margins": 3.9024269580841064,
+      "rewards/real": -14.465472221374512,
       "step": 20
     },
     {
@@ -55,13 +55,13 @@
       "learning_rate": 1.89873417721519e-07,
       "logits/generated": -Infinity,
       "logits/real": -Infinity,
-      "logps/generated": -510.23455810546875,
-      "logps/real": -408.78350830078125,
-      "loss": 3.6848,
+      "logps/generated": -510.4585876464844,
+      "logps/real": -408.74212646484375,
+      "loss": 3.6955,
       "rewards/accuracies": 0.6875,
-      "rewards/generated": -21.037378311157227,
-      "rewards/margins": 7.7357025146484375,
-      "rewards/real": -13.301675796508789,
+      "rewards/generated": -21.059778213500977,
+      "rewards/margins": 7.762242317199707,
+      "rewards/real": -13.297533988952637,
       "step": 30
     },
     {
@@ -69,13 +69,13 @@
       "learning_rate": 2.5316455696202533e-07,
       "logits/generated": -Infinity,
       "logits/real": -Infinity,
-      "logps/generated": -451.669189453125,
-      "logps/real": -368.1625061035156,
-      "loss": 3.9079,
+      "logps/generated": -451.734130859375,
+      "logps/real": -368.16998291015625,
+      "loss": 3.9056,
       "rewards/accuracies": 0.625,
-      "rewards/generated": -16.798627853393555,
-      "rewards/margins": 4.454809665679932,
-      "rewards/real": -12.343819618225098,
+      "rewards/generated": -16.805126190185547,
+      "rewards/margins": 4.4605584144592285,
+      "rewards/real": -12.344568252563477,
       "step": 40
     },
     {
@@ -83,13 +83,13 @@
       "learning_rate": 3.1645569620253163e-07,
       "logits/generated": -Infinity,
       "logits/real": -Infinity,
-      "logps/generated": -552.6080322265625,
-      "logps/real": -397.91064453125,
-      "loss": 2.9787,
+      "logps/generated": -552.9141845703125,
+      "logps/real": -397.95318603515625,
+      "loss": 2.9742,
       "rewards/accuracies": 0.737500011920929,
-      "rewards/generated": -25.66506004333496,
-      "rewards/margins": 10.62718391418457,
-      "rewards/real": -15.037874221801758,
+      "rewards/generated": -25.695674896240234,
+      "rewards/margins": 10.653547286987305,
+      "rewards/real": -15.042126655578613,
       "step": 50
     },
     {
@@ -97,13 +97,13 @@
       "learning_rate": 3.79746835443038e-07,
       "logits/generated": -Infinity,
       "logits/real": -Infinity,
-      "logps/generated": -509.36578369140625,
-      "logps/real": -390.8323059082031,
-      "loss": 2.5737,
+      "logps/generated": -509.5240783691406,
+      "logps/real": -390.6533203125,
+      "loss": 2.5682,
       "rewards/accuracies": 0.737500011920929,
-      "rewards/generated": -22.55277442932129,
-      "rewards/margins": 7.870760440826416,
-      "rewards/real": -14.682014465332031,
+      "rewards/generated": -22.568603515625,
+      "rewards/margins": 7.904494285583496,
+      "rewards/real": -14.664111137390137,
       "step": 60
     },
     {
@@ -111,13 +111,13 @@
       "learning_rate": 4.4303797468354424e-07,
       "logits/generated": -Infinity,
       "logits/real": -Infinity,
-      "logps/generated": -554.7327270507812,
-      "logps/real": -382.78350830078125,
-      "loss": 1.7978,
+      "logps/generated": -554.5638427734375,
+      "logps/real": -382.84832763671875,
+      "loss": 1.7793,
       "rewards/accuracies": 0.8374999761581421,
-      "rewards/generated": -28.232168197631836,
-      "rewards/margins": 14.969076156616211,
-      "rewards/real": -13.263090133666992,
+      "rewards/generated": -28.215290069580078,
+      "rewards/margins": 14.945714950561523,
+      "rewards/real": -13.269571304321289,
       "step": 70
     },
     {
@@ -125,13 +125,13 @@
       "learning_rate": 4.992887624466572e-07,
       "logits/generated": -Infinity,
       "logits/real": -Infinity,
-      "logps/generated": -573.7469482421875,
-      "logps/real": -402.5927429199219,
-      "loss": 1.4178,
+      "logps/generated": -574.03759765625,
+      "logps/real": -402.69232177734375,
+      "loss": 1.4194,
       "rewards/accuracies": 0.800000011920929,
-      "rewards/generated": -28.089534759521484,
-      "rewards/margins": 12.5045166015625,
-      "rewards/real": -15.5850191116333,
+      "rewards/generated": -28.118602752685547,
+      "rewards/margins": 12.523630142211914,
+      "rewards/real": -15.594972610473633,
       "step": 80
     },
     {
@@ -139,13 +139,13 @@
       "learning_rate": 4.92176386913229e-07,
       "logits/generated": -Infinity,
       "logits/real": -Infinity,
-      "logps/generated": -576.7484741210938,
-      "logps/real": -400.97186279296875,
-      "loss": 1.2467,
-      "rewards/accuracies": 0.7875000238418579,
-      "rewards/generated": -30.259912490844727,
-      "rewards/margins": 14.735580444335938,
-      "rewards/real": -15.524327278137207,
+      "logps/generated": -577.7903442382812,
+      "logps/real": -401.81561279296875,
+      "loss": 1.2635,
+      "rewards/accuracies": 0.7749999761581421,
+      "rewards/generated": -30.364093780517578,
+      "rewards/margins": 14.755389213562012,
+      "rewards/real": -15.608701705932617,
       "step": 90
     },
     {
@@ -153,13 +153,13 @@
       "learning_rate": 4.850640113798008e-07,
       "logits/generated": -Infinity,
       "logits/real": -Infinity,
-      "logps/generated": -654.0052490234375,
-      "logps/real": -442.6397399902344,
-      "loss": 0.9329,
+      "logps/generated": -654.4276733398438,
+      "logps/real": -442.893798828125,
+      "loss": 0.9213,
       "rewards/accuracies": 0.8999999761581421,
-      "rewards/generated": -35.535160064697266,
-      "rewards/margins": 18.725465774536133,
-      "rewards/real": -16.8096923828125,
+      "rewards/generated": -35.57740020751953,
+      "rewards/margins": 18.7423038482666,
+      "rewards/real": -16.835100173950195,
       "step": 100
     },
     {
@@ -167,13 +167,13 @@
       "learning_rate": 4.779516358463727e-07,
       "logits/generated": -Infinity,
       "logits/real": -Infinity,
-      "logps/generated": -638.0022583007812,
-      "logps/real": -427.4214782714844,
-      "loss": 1.1736,
+      "logps/generated": -637.9244384765625,
+      "logps/real": -426.7471618652344,
+      "loss": 1.0798,
       "rewards/accuracies": 0.875,
-      "rewards/generated": -35.03214645385742,
-      "rewards/margins": 18.788606643676758,
-      "rewards/real": -16.24354362487793,
+      "rewards/generated": -35.024356842041016,
+      "rewards/margins": 18.848251342773438,
+      "rewards/real": -16.17610740661621,
       "step": 110
     },
     {
@@ -181,13 +181,13 @@
       "learning_rate": 4.7083926031294454e-07,
       "logits/generated": -Infinity,
       "logits/real": -Infinity,
-      "logps/generated": -647.80859375,
-      "logps/real": -420.6051330566406,
-      "loss": 0.7488,
+      "logps/generated": -647.3193969726562,
+      "logps/real": -419.98095703125,
+      "loss": 0.7428,
       "rewards/accuracies": 0.9375,
-      "rewards/generated": -35.87686538696289,
-      "rewards/margins": 21.418781280517578,
-      "rewards/real": -14.458081245422363,
+      "rewards/generated": -35.827938079833984,
+      "rewards/margins": 21.43227767944336,
+      "rewards/real": -14.395665168762207,
       "step": 120
     },
     {
@@ -195,13 +195,13 @@
       "learning_rate": 4.6372688477951633e-07,
       "logits/generated": -Infinity,
       "logits/real": -Infinity,
-      "logps/generated": -647.1038818359375,
-      "logps/real": -412.8779296875,
-      "loss": 0.6447,
+      "logps/generated": -647.6121826171875,
+      "logps/real": -413.42401123046875,
+      "loss": 0.6566,
       "rewards/accuracies": 0.875,
-      "rewards/generated": -35.35771560668945,
-      "rewards/margins": 20.0076904296875,
-      "rewards/real": -15.350027084350586,
+      "rewards/generated": -35.408538818359375,
+      "rewards/margins": 20.003904342651367,
+      "rewards/real": -15.404635429382324,
       "step": 130
     },
     {
@@ -209,13 +209,13 @@
       "learning_rate": 4.5661450924608817e-07,
       "logits/generated": -Infinity,
       "logits/real": -Infinity,
-      "logps/generated": -699.005126953125,
-      "logps/real": -401.8106384277344,
-      "loss": 0.9455,
+      "logps/generated": -697.9991455078125,
+      "logps/real": -401.5552673339844,
+      "loss": 0.9421,
       "rewards/accuracies": 0.9375,
-      "rewards/generated": -38.48512268066406,
-      "rewards/margins": 23.671005249023438,
-      "rewards/real": -14.814119338989258,
+      "rewards/generated": -38.38452911376953,
+      "rewards/margins": 23.595943450927734,
+      "rewards/real": -14.788581848144531,
       "step": 140
     },
     {
@@ -223,13 +223,13 @@
       "learning_rate": 4.4950213371266e-07,
       "logits/generated": -Infinity,
       "logits/real": -Infinity,
-      "logps/generated": -673.933349609375,
-      "logps/real": -404.8531494140625,
-      "loss": 0.6848,
-      "rewards/accuracies": 0.9125000238418579,
-      "rewards/generated": -38.79594039916992,
-      "rewards/margins": 22.858118057250977,
-      "rewards/real": -15.937822341918945,
+      "logps/generated": -672.8363037109375,
+      "logps/real": -401.61590576171875,
+      "loss": 0.6587,
+      "rewards/accuracies": 0.8999999761581421,
+      "rewards/generated": -38.686241149902344,
+      "rewards/margins": 23.072139739990234,
+      "rewards/real": -15.614102363586426,
       "step": 150
     },
     {
@@ -237,13 +237,13 @@
       "learning_rate": 4.4238975817923186e-07,
       "logits/generated": -Infinity,
       "logits/real": -Infinity,
-      "logps/generated": -650.0132446289062,
-      "logps/real": -380.55413818359375,
-      "loss": 0.5912,
+      "logps/generated": -641.0380859375,
+      "logps/real": -377.9221496582031,
+      "loss": 0.6261,
       "rewards/accuracies": 0.8999999761581421,
-      "rewards/generated": -36.62251663208008,
-      "rewards/margins": 22.439044952392578,
-      "rewards/real": -14.183469772338867,
+      "rewards/generated": -35.72499084472656,
+      "rewards/margins": 21.804719924926758,
+      "rewards/real": -13.920272827148438,
       "step": 160
     },
     {
@@ -251,13 +251,13 @@
       "learning_rate": 4.3527738264580364e-07,
       "logits/generated": -Infinity,
       "logits/real": -Infinity,
-      "logps/generated": -676.5908203125,
-      "logps/real": -413.38287353515625,
-      "loss": 0.6238,
+      "logps/generated": -676.1116943359375,
+      "logps/real": -410.857421875,
+      "loss": 0.6189,
       "rewards/accuracies": 0.9125000238418579,
-      "rewards/generated": -37.932899475097656,
-      "rewards/margins": 22.803815841674805,
-      "rewards/real": -15.129079818725586,
+      "rewards/generated": -37.884979248046875,
+      "rewards/margins": 23.008445739746094,
+      "rewards/real": -14.876535415649414,
       "step": 170
     },
     {
@@ -265,13 +265,13 @@
       "learning_rate": 4.2816500711237554e-07,
       "logits/generated": -Infinity,
       "logits/real": -Infinity,
-      "logps/generated": -660.9660034179688,
-      "logps/real": -372.2694396972656,
-      "loss": 0.3412,
+      "logps/generated": -654.6505737304688,
+      "logps/real": -370.24224853515625,
+      "loss": 0.3247,
       "rewards/accuracies": 0.9375,
-      "rewards/generated": -37.31585693359375,
-      "rewards/margins": 23.456205368041992,
-      "rewards/real": -13.859651565551758,
+      "rewards/generated": -36.68430709838867,
+      "rewards/margins": 23.027376174926758,
+      "rewards/real": -13.656933784484863,
       "step": 180
     },
     {
@@ -279,13 +279,13 @@
       "learning_rate": 4.2105263157894733e-07,
       "logits/generated": -Infinity,
       "logits/real": -Infinity,
-      "logps/generated": -668.5888671875,
-      "logps/real": -366.01507568359375,
-      "loss": 0.5799,
+      "logps/generated": -665.4993286132812,
+      "logps/real": -364.58880615234375,
+      "loss": 0.5709,
       "rewards/accuracies": 0.9624999761581421,
-      "rewards/generated": -37.69170379638672,
-      "rewards/margins": 25.54647445678711,
-      "rewards/real": -12.145231246948242,
+      "rewards/generated": -37.38275146484375,
+      "rewards/margins": 25.380136489868164,
+      "rewards/real": -12.002609252929688,
       "step": 190
     },
     {
@@ -293,13 +293,13 @@
       "learning_rate": 4.1394025604551917e-07,
       "logits/generated": -Infinity,
       "logits/real": -Infinity,
-      "logps/generated": -681.8294677734375,
-      "logps/real": -392.57513427734375,
-      "loss": 0.6742,
+      "logps/generated": -684.380615234375,
+      "logps/real": -392.02288818359375,
+      "loss": 0.6823,
       "rewards/accuracies": 0.8999999761581421,
-      "rewards/generated": -38.56768035888672,
-      "rewards/margins": 24.48987579345703,
-      "rewards/real": -14.077804565429688,
+      "rewards/generated": -38.82280349731445,
+      "rewards/margins": 24.800228118896484,
+      "rewards/real": -14.02257251739502,
       "step": 200
     },
     {
@@ -307,13 +307,13 @@
       "learning_rate": 4.06827880512091e-07,
       "logits/generated": -Infinity,
       "logits/real": -Infinity,
-      "logps/generated": -754.3670654296875,
-      "logps/real": -383.51702880859375,
-      "loss": 0.519,
+      "logps/generated": -741.0132446289062,
+      "logps/real": -383.6390075683594,
+      "loss": 0.5369,
       "rewards/accuracies": 0.925000011920929,
-      "rewards/generated": -44.508426666259766,
-      "rewards/margins": 30.150854110717773,
-      "rewards/real": -14.357576370239258,
+      "rewards/generated": -43.17304229736328,
+      "rewards/margins": 28.80326271057129,
+      "rewards/real": -14.369776725769043,
       "step": 210
     },
     {
@@ -321,13 +321,13 @@
       "learning_rate": 3.9971550497866285e-07,
       "logits/generated": -Infinity,
       "logits/real": -Infinity,
-      "logps/generated": -669.1099243164062,
-      "logps/real": -404.3431091308594,
-      "loss": 0.3679,
+      "logps/generated": -665.7184448242188,
+      "logps/real": -401.6400146484375,
+      "loss": 0.3696,
       "rewards/accuracies": 0.9375,
-      "rewards/generated": -39.06100845336914,
-      "rewards/margins": 24.223697662353516,
-      "rewards/real": -14.837308883666992,
+      "rewards/generated": -38.721866607666016,
+      "rewards/margins": 24.154865264892578,
+      "rewards/real": -14.567001342773438,
       "step": 220
     },
     {
@@ -335,13 +335,13 @@
       "learning_rate": 3.926031294452347e-07,
       "logits/generated": -Infinity,
       "logits/real": -Infinity,
-      "logps/generated": -688.25048828125,
-      "logps/real": -396.00018310546875,
-      "loss": 0.2677,
+      "logps/generated": -681.5037231445312,
+      "logps/real": -394.03509521484375,
+      "loss": 0.2532,
       "rewards/accuracies": 0.949999988079071,
-      "rewards/generated": -39.11343765258789,
-      "rewards/margins": 24.355987548828125,
-      "rewards/real": -14.757448196411133,
+      "rewards/generated": -38.438758850097656,
+      "rewards/margins": 23.877817153930664,
+      "rewards/real": -14.560938835144043,
       "step": 230
     },
     {
@@ -349,13 +349,13 @@
       "learning_rate": 3.8549075391180653e-07,
       "logits/generated": -Infinity,
       "logits/real": -Infinity,
-      "logps/generated": -714.4971923828125,
-      "logps/real": -415.25262451171875,
-      "loss": 0.4572,
+      "logps/generated": -705.4946899414062,
+      "logps/real": -414.25146484375,
+      "loss": 0.4719,
       "rewards/accuracies": 0.925000011920929,
-      "rewards/generated": -40.9525032043457,
-      "rewards/margins": 25.465791702270508,
-      "rewards/real": -15.486714363098145,
+      "rewards/generated": -40.05225372314453,
+      "rewards/margins": 24.665658950805664,
+      "rewards/real": -15.386594772338867,
       "step": 240
     },
     {
@@ -363,13 +363,13 @@
       "learning_rate": 3.783783783783784e-07,
       "logits/generated": -Infinity,
       "logits/real": -Infinity,
-      "logps/generated": -729.0126342773438,
-      "logps/real": -394.86724853515625,
-      "loss": 0.345,
+      "logps/generated": -718.469970703125,
+      "logps/real": -393.9696350097656,
+      "loss": 0.326,
       "rewards/accuracies": 0.9624999761581421,
-      "rewards/generated": -44.133033752441406,
-      "rewards/margins": 29.94510841369629,
-      "rewards/real": -14.187917709350586,
+      "rewards/generated": -43.078765869140625,
+      "rewards/margins": 28.980609893798828,
+      "rewards/real": -14.09815788269043,
       "step": 250
     },
     {
@@ -377,13 +377,13 @@
       "learning_rate": 3.7126600284495016e-07,
       "logits/generated": -Infinity,
       "logits/real": -Infinity,
-      "logps/generated": -722.02490234375,
-      "logps/real": -403.8961181640625,
-      "loss": 0.4853,
+      "logps/generated": -721.9268798828125,
+      "logps/real": -403.2998046875,
+      "loss": 0.4804,
       "rewards/accuracies": 0.9375,
-      "rewards/generated": -41.30921173095703,
-      "rewards/margins": 26.302785873413086,
-      "rewards/real": -15.006426811218262,
+      "rewards/generated": -41.29940414428711,
+      "rewards/margins": 26.352609634399414,
+      "rewards/real": -14.946797370910645,
       "step": 260
     },
     {
@@ -391,13 +391,13 @@
       "learning_rate": 3.6415362731152206e-07,
       "logits/generated": -Infinity,
       "logits/real": -Infinity,
-      "logps/generated": -695.8910522460938,
-      "logps/real": -374.00433349609375,
-      "loss": 0.3413,
+      "logps/generated": -690.9553833007812,
+      "logps/real": -374.5037536621094,
+      "loss": 0.3827,
       "rewards/accuracies": 0.949999988079071,
-      "rewards/generated": -41.190391540527344,
-      "rewards/margins": 27.258630752563477,
-      "rewards/real": -13.931764602661133,
+      "rewards/generated": -40.69682693481445,
+      "rewards/margins": 26.715112686157227,
+      "rewards/real": -13.981710433959961,
       "step": 270
     },
     {
@@ -405,13 +405,13 @@
       "learning_rate": 3.5704125177809385e-07,
       "logits/generated": -Infinity,
       "logits/real": -Infinity,
-      "logps/generated": -734.4241943359375,
-      "logps/real": -404.4153137207031,
-      "loss": 0.4178,
+      "logps/generated": -725.0104370117188,
+      "logps/real": -406.488037109375,
+      "loss": 0.4101,
       "rewards/accuracies": 0.9624999761581421,
-      "rewards/generated": -43.61100387573242,
-      "rewards/margins": 28.391124725341797,
-      "rewards/real": -15.219881057739258,
+      "rewards/generated": -42.66962432861328,
+      "rewards/margins": 27.242467880249023,
+      "rewards/real": -15.427154541015625,
       "step": 280
     },
     {
@@ -419,13 +419,13 @@
       "learning_rate": 3.4992887624466574e-07,
       "logits/generated": -Infinity,
       "logits/real": -Infinity,
-      "logps/generated": -703.3042602539062,
-      "logps/real": -400.789306640625,
-      "loss": 0.358,
+      "logps/generated": -697.3206787109375,
+      "logps/real": -400.65753173828125,
+      "loss": 0.3629,
       "rewards/accuracies": 0.9624999761581421,
-      "rewards/generated": -40.19432830810547,
-      "rewards/margins": 25.275157928466797,
-      "rewards/real": -14.919171333312988,
+      "rewards/generated": -39.595970153808594,
+      "rewards/margins": 24.689983367919922,
+      "rewards/real": -14.90599250793457,
       "step": 290
     },
     {
@@ -433,13 +433,13 @@
       "learning_rate": 3.4281650071123753e-07,
       "logits/generated": -Infinity,
       "logits/real": -Infinity,
-      "logps/generated": -717.597900390625,
-      "logps/real": -387.6546325683594,
-      "loss": 0.2285,
+      "logps/generated": -711.7612915039062,
+      "logps/real": -388.0057678222656,
+      "loss": 0.2122,
       "rewards/accuracies": 0.949999988079071,
-      "rewards/generated": -42.98313522338867,
-      "rewards/margins": 27.72798728942871,
-      "rewards/real": -15.255144119262695,
+      "rewards/generated": -42.39947509765625,
+      "rewards/margins": 27.109222412109375,
+      "rewards/real": -15.290254592895508,
       "step": 300
     },
     {
@@ -447,13 +447,13 @@
       "learning_rate": 3.3570412517780937e-07,
       "logits/generated": -Infinity,
       "logits/real": -Infinity,
-      "logps/generated": -714.7513427734375,
-      "logps/real": -424.5567932128906,
-      "loss": 0.362,
-      "rewards/accuracies": 0.8999999761581421,
-      "rewards/generated": -42.02880859375,
-      "rewards/margins": 26.301555633544922,
-      "rewards/real": -15.727252006530762,
+      "logps/generated": -713.0751953125,
+      "logps/real": -423.7640686035156,
+      "loss": 0.3657,
+      "rewards/accuracies": 0.9125000238418579,
+      "rewards/generated": -41.861183166503906,
+      "rewards/margins": 26.213199615478516,
+      "rewards/real": -15.647982597351074,
       "step": 310
     },
     {
@@ -461,13 +461,13 @@
       "learning_rate": 3.285917496443812e-07,
       "logits/generated": -Infinity,
       "logits/real": -Infinity,
-      "logps/generated": -704.7682495117188,
-      "logps/real": -389.437744140625,
-      "loss": 0.3054,
+      "logps/generated": -695.4622192382812,
+      "logps/real": -386.4324645996094,
+      "loss": 0.3005,
       "rewards/accuracies": 0.925000011920929,
-      "rewards/generated": -40.38294982910156,
-      "rewards/margins": 26.326675415039062,
-      "rewards/real": -14.05627727508545,
+      "rewards/generated": -39.45234298706055,
+      "rewards/margins": 25.696590423583984,
+      "rewards/real": -13.755752563476562,
       "step": 320
     },
     {
@@ -475,13 +475,13 @@
       "learning_rate": 3.2147937411095305e-07,
       "logits/generated": -Infinity,
       "logits/real": -Infinity,
-      "logps/generated": -788.1136474609375,
-      "logps/real": -426.383544921875,
-      "loss": 0.2255,
-      "rewards/accuracies": 0.9750000238418579,
-      "rewards/generated": -49.49579620361328,
-      "rewards/margins": 32.605934143066406,
-      "rewards/real": -16.889863967895508,
+      "logps/generated": -774.9369506835938,
+      "logps/real": -426.2724609375,
+      "loss": 0.2325,
+      "rewards/accuracies": 0.9624999761581421,
+      "rewards/generated": -48.17813491821289,
+      "rewards/margins": 31.29937171936035,
+      "rewards/real": -16.87876319885254,
       "step": 330
     },
     {
@@ -489,13 +489,13 @@
       "learning_rate": 3.1436699857752484e-07,
       "logits/generated": -Infinity,
       "logits/real": -Infinity,
-      "logps/generated": -720.3298950195312,
-      "logps/real": -392.3839111328125,
-      "loss": 0.3118,
+      "logps/generated": -714.9234008789062,
+      "logps/real": -395.0069885253906,
+      "loss": 0.2981,
       "rewards/accuracies": 0.9624999761581421,
-      "rewards/generated": -42.088287353515625,
-      "rewards/margins": 27.807758331298828,
-      "rewards/real": -14.28053092956543,
+      "rewards/generated": -41.547645568847656,
+      "rewards/margins": 27.00480079650879,
+      "rewards/real": -14.54284381866455,
       "step": 340
     },
     {
@@ -503,13 +503,13 @@
       "learning_rate": 3.0725462304409674e-07,
       "logits/generated": -Infinity,
       "logits/real": -Infinity,
-      "logps/generated": -721.5670776367188,
-      "logps/real": -399.3511047363281,
-      "loss": 0.2108,
+      "logps/generated": -721.8656005859375,
+      "logps/real": -400.1606750488281,
+      "loss": 0.2019,
       "rewards/accuracies": 0.9375,
-      "rewards/generated": -44.001930236816406,
-      "rewards/margins": 28.8182430267334,
-      "rewards/real": -15.183688163757324,
+      "rewards/generated": -44.031803131103516,
+      "rewards/margins": 28.76715087890625,
+      "rewards/real": -15.264646530151367,
       "step": 350
     },
     {
@@ -517,13 +517,13 @@
       "learning_rate": 3.001422475106685e-07,
       "logits/generated": -Infinity,
       "logits/real": -Infinity,
-      "logps/generated": -734.0841064453125,
-      "logps/real": -473.2635192871094,
-      "loss": 0.6012,
-      "rewards/accuracies": 0.925000011920929,
-      "rewards/generated": -44.092620849609375,
-      "rewards/margins": 25.192630767822266,
-      "rewards/real": -18.89999008178711,
+      "logps/generated": -728.89599609375,
+      "logps/real": -470.1971130371094,
+      "loss": 0.5931,
+      "rewards/accuracies": 0.9375,
+      "rewards/generated": -43.573814392089844,
+      "rewards/margins": 24.9804630279541,
+      "rewards/real": -18.593351364135742,
       "step": 360
     },
     {
@@ -531,13 +531,13 @@
       "learning_rate": 2.9302987197724037e-07,
       "logits/generated": -Infinity,
       "logits/real": -Infinity,
-      "logps/generated": -666.3681030273438,
-      "logps/real": -369.57232666015625,
-      "loss": 0.3811,
+      "logps/generated": -667.4353637695312,
+      "logps/real": -369.68328857421875,
+      "loss": 0.4031,
       "rewards/accuracies": 0.9624999761581421,
-      "rewards/generated": -38.6079216003418,
-      "rewards/margins": 23.917612075805664,
-      "rewards/real": -14.690305709838867,
+      "rewards/generated": -38.71464920043945,
+      "rewards/margins": 24.013246536254883,
+      "rewards/real": -14.701400756835938,
       "step": 370
     },
     {
@@ -545,13 +545,13 @@
       "learning_rate": 2.8591749644381226e-07,
       "logits/generated": -Infinity,
       "logits/real": -Infinity,
-      "logps/generated": -709.2445068359375,
-      "logps/real": -394.5869140625,
-      "loss": 0.3208,
-      "rewards/accuracies": 0.949999988079071,
-      "rewards/generated": -42.30268096923828,
-      "rewards/margins": 26.500690460205078,
-      "rewards/real": -15.80199146270752,
+      "logps/generated": -708.6209716796875,
+      "logps/real": -395.7604675292969,
+      "loss": 0.3691,
+      "rewards/accuracies": 0.9375,
+      "rewards/generated": -42.240318298339844,
+      "rewards/margins": 26.32097816467285,
+      "rewards/real": -15.919347763061523,
       "step": 380
     },
     {
@@ -559,13 +559,13 @@
       "learning_rate": 2.7880512091038405e-07,
       "logits/generated": -Infinity,
       "logits/real": -Infinity,
-      "logps/generated": -775.5260009765625,
-      "logps/real": -426.09869384765625,
-      "loss": 0.2884,
+      "logps/generated": -766.4454345703125,
+      "logps/real": -423.30755615234375,
+      "loss": 0.2758,
       "rewards/accuracies": 0.9624999761581421,
-      "rewards/generated": -48.343849182128906,
-      "rewards/margins": 31.10616683959961,
-      "rewards/real": -17.237682342529297,
+      "rewards/generated": -47.435791015625,
+      "rewards/margins": 30.477214813232422,
+      "rewards/real": -16.958572387695312,
       "step": 390
     },
     {
@@ -573,13 +573,13 @@
       "learning_rate": 2.716927453769559e-07,
       "logits/generated": -Infinity,
       "logits/real": -Infinity,
-      "logps/generated": -773.4555053710938,
-      "logps/real": -450.7288513183594,
-      "loss": 0.3376,
+      "logps/generated": -768.6455078125,
+      "logps/real": -449.0997009277344,
+      "loss": 0.3455,
       "rewards/accuracies": 0.949999988079071,
-      "rewards/generated": -47.49498748779297,
-      "rewards/margins": 28.620986938476562,
-      "rewards/real": -18.873998641967773,
+      "rewards/generated": -47.01398849487305,
+      "rewards/margins": 28.302906036376953,
+      "rewards/real": -18.711084365844727,
       "step": 400
     },
     {
@@ -587,13 +587,13 @@
       "learning_rate": 2.6458036984352773e-07,
       "logits/generated": -Infinity,
       "logits/real": -Infinity,
-      "logps/generated": -741.7351684570312,
-      "logps/real": -449.17816162109375,
-      "loss": 0.3754,
-      "rewards/accuracies": 0.925000011920929,
-      "rewards/generated": -46.99356460571289,
-      "rewards/margins": 27.995086669921875,
-      "rewards/real": -18.998476028442383,
+      "logps/generated": -734.2765502929688,
+      "logps/real": -447.5357971191406,
+      "loss": 0.375,
+      "rewards/accuracies": 0.9375,
+      "rewards/generated": -46.2476921081543,
+      "rewards/margins": 27.413455963134766,
+      "rewards/real": -18.834239959716797,
       "step": 410
     },
     {
@@ -601,13 +601,13 @@
       "learning_rate": 2.574679943100996e-07,
       "logits/generated": -Infinity,
       "logits/real": -Infinity,
-      "logps/generated": -738.3614501953125,
-      "logps/real": -456.7314453125,
-      "loss": 0.0844,
-      "rewards/accuracies": 0.9375,
-      "rewards/generated": -46.36874008178711,
-      "rewards/margins": 27.84047508239746,
-      "rewards/real": -18.528268814086914,
+      "logps/generated": -730.5106201171875,
+      "logps/real": -453.9371643066406,
+      "loss": 0.0816,
+      "rewards/accuracies": 0.925000011920929,
+      "rewards/generated": -45.583648681640625,
+      "rewards/margins": 27.334802627563477,
+      "rewards/real": -18.248844146728516,
       "step": 420
     },
     {
@@ -615,13 +615,13 @@
       "learning_rate": 2.5035561877667136e-07,
       "logits/generated": -Infinity,
       "logits/real": -Infinity,
-      "logps/generated": -768.1492309570312,
-      "logps/real": -441.343017578125,
-      "loss": 0.1898,
+      "logps/generated": -766.5603637695312,
+      "logps/real": -438.89788818359375,
+      "loss": 0.1735,
       "rewards/accuracies": 0.9750000238418579,
-      "rewards/generated": -47.066619873046875,
-      "rewards/margins": 29.435678482055664,
-      "rewards/real": -17.63094139099121,
+      "rewards/generated": -46.9077262878418,
+      "rewards/margins": 29.52130126953125,
+      "rewards/real": -17.386432647705078,
       "step": 430
     },
     {
@@ -629,13 +629,13 @@
       "learning_rate": 2.4324324324324326e-07,
       "logits/generated": -Infinity,
       "logits/real": -Infinity,
-      "logps/generated": -787.4987182617188,
-      "logps/real": -408.2841796875,
-      "loss": 0.3922,
+      "logps/generated": -781.421630859375,
+      "logps/real": -405.1501770019531,
+      "loss": 0.3965,
       "rewards/accuracies": 0.987500011920929,
-      "rewards/generated": -50.27423095703125,
-      "rewards/margins": 33.921939849853516,
-      "rewards/real": -16.352293014526367,
+      "rewards/generated": -49.6665153503418,
+      "rewards/margins": 33.62762451171875,
+      "rewards/real": -16.03889274597168,
       "step": 440
     },
     {
@@ -643,13 +643,13 @@
       "learning_rate": 2.3613086770981507e-07,
       "logits/generated": -Infinity,
       "logits/real": -Infinity,
-      "logps/generated": -771.3111572265625,
-      "logps/real": -408.37896728515625,
-      "loss": 0.3608,
+      "logps/generated": -760.2469482421875,
+      "logps/real": -404.4407653808594,
+      "loss": 0.4128,
       "rewards/accuracies": 0.949999988079071,
-      "rewards/generated": -48.568450927734375,
-      "rewards/margins": 31.147785186767578,
-      "rewards/real": -17.420665740966797,
+      "rewards/generated": -47.4620361328125,
+      "rewards/margins": 30.4351863861084,
+      "rewards/real": -17.026851654052734,
       "step": 450
     },
     {
@@ -657,13 +657,13 @@
       "learning_rate": 2.290184921763869e-07,
       "logits/generated": -Infinity,
       "logits/real": -Infinity,
-      "logps/generated": -827.6937255859375,
-      "logps/real": -423.6630859375,
-      "loss": 0.1253,
+      "logps/generated": -820.1843872070312,
+      "logps/real": -420.02587890625,
+      "loss": 0.1306,
       "rewards/accuracies": 1.0,
-      "rewards/generated": -51.94307327270508,
-      "rewards/margins": 35.77669143676758,
-      "rewards/real": -16.166385650634766,
+      "rewards/generated": -51.192138671875,
+      "rewards/margins": 35.38946533203125,
+      "rewards/real": -15.8026704788208,
       "step": 460
     },
     {
@@ -671,13 +671,13 @@
       "learning_rate": 2.2190611664295875e-07,
       "logits/generated": -Infinity,
       "logits/real": -Infinity,
-      "logps/generated": -790.8399658203125,
-      "logps/real": -440.3358459472656,
-      "loss": 0.2949,
+      "logps/generated": -759.2930908203125,
+      "logps/real": -438.888916015625,
+      "loss": 0.2487,
       "rewards/accuracies": 0.925000011920929,
-      "rewards/generated": -48.574867248535156,
-      "rewards/margins": 30.353759765625,
-      "rewards/real": -18.22110939025879,
+      "rewards/generated": -45.420188903808594,
+      "rewards/margins": 27.34377670288086,
+      "rewards/real": -18.0764102935791,
       "step": 470
     },
     {
@@ -685,13 +685,13 @@
       "learning_rate": 2.1479374110953057e-07,
       "logits/generated": -Infinity,
       "logits/real": -Infinity,
-      "logps/generated": -781.35791015625,
-      "logps/real": -390.54168701171875,
-      "loss": 0.3877,
-      "rewards/accuracies": 0.9624999761581421,
-      "rewards/generated": -49.05020523071289,
-      "rewards/margins": 32.30176544189453,
-      "rewards/real": -16.748441696166992,
+      "logps/generated": -777.7584838867188,
+      "logps/real": -388.6899108886719,
+      "loss": 0.3922,
+      "rewards/accuracies": 0.949999988079071,
+      "rewards/generated": -48.69025421142578,
+      "rewards/margins": 32.126991271972656,
+      "rewards/real": -16.563264846801758,
       "step": 480
     },
     {
@@ -699,13 +699,13 @@
       "learning_rate": 2.076813655761024e-07,
       "logits/generated": -Infinity,
       "logits/real": -Infinity,
-      "logps/generated": -810.448486328125,
-      "logps/real": -441.8717346191406,
-      "loss": 0.2217,
+      "logps/generated": -804.2822875976562,
+      "logps/real": -438.558837890625,
+      "loss": 0.2558,
       "rewards/accuracies": 0.987500011920929,
-      "rewards/generated": -51.156005859375,
-      "rewards/margins": 31.813852310180664,
-      "rewards/real": -19.342153549194336,
+      "rewards/generated": -50.539390563964844,
+      "rewards/margins": 31.528533935546875,
+      "rewards/real": -19.0108585357666,
       "step": 490
     },
     {
@@ -713,29 +713,29 @@
       "learning_rate": 2.0056899004267425e-07,
       "logits/generated": -Infinity,
       "logits/real": -Infinity,
-      "logps/generated": -802.3043212890625,
-      "logps/real": -487.0044860839844,
-      "loss": 0.2807,
+      "logps/generated": -799.6202392578125,
+      "logps/real": -483.55096435546875,
+      "loss": 0.2945,
       "rewards/accuracies": 0.9624999761581421,
-      "rewards/generated": -51.0858039855957,
-      "rewards/margins": 30.591796875,
-      "rewards/real": -20.494003295898438,
+      "rewards/generated": -50.81739807128906,
+      "rewards/margins": 30.668746948242188,
+      "rewards/real": -20.14865493774414,
       "step": 500
     },
     {
       "epoch": 0.64,
       "eval_logits/generated": -Infinity,
       "eval_logits/real": -Infinity,
-      "eval_logps/generated": -518.7460327148438,
-      "eval_logps/real": -321.637939453125,
-      "eval_loss": 0.16799330711364746,
+      "eval_logps/generated": -509.32861328125,
+      "eval_logps/real": -313.0279541015625,
+      "eval_loss": 0.1748354285955429,
       "eval_rewards/accuracies": 0.9442675113677979,
-      "eval_rewards/generated": -22.73108673095703,
-      "eval_rewards/margins": 15.805987358093262,
-      "eval_rewards/real": -6.9250993728637695,
-      "eval_runtime": 514.9684,
-      "eval_samples_per_second": 9.709,
-      "eval_steps_per_second": 0.305,
+      "eval_rewards/generated": -21.789350509643555,
+      "eval_rewards/margins": 15.725247383117676,
+      "eval_rewards/real": -6.0641021728515625,
+      "eval_runtime": 590.026,
+      "eval_samples_per_second": 8.474,
+      "eval_steps_per_second": 0.266,
       "step": 500
     },
     {
@@ -743,13 +743,13 @@
       "learning_rate": 1.9345661450924607e-07,
       "logits/generated": -Infinity,
       "logits/real": -Infinity,
-      "logps/generated": -764.2227783203125,
-      "logps/real": -417.8741149902344,
-      "loss": 0.2566,
-      "rewards/accuracies": 0.9624999761581421,
-      "rewards/generated": -46.21670150756836,
-      "rewards/margins": 29.2624568939209,
-      "rewards/real": -16.954248428344727,
+      "logps/generated": -757.3658447265625,
+      "logps/real": -412.8812561035156,
+      "loss": 0.2523,
+      "rewards/accuracies": 0.9750000238418579,
+      "rewards/generated": -45.53099822998047,
+      "rewards/margins": 29.076038360595703,
+      "rewards/real": -16.454959869384766,
       "step": 510
     },
     {
@@ -757,13 +757,13 @@
       "learning_rate": 1.863442389758179e-07,
       "logits/generated": -Infinity,
       "logits/real": -Infinity,
-      "logps/generated": -785.6708984375,
-      "logps/real": -436.8841857910156,
-      "loss": 0.2175,
+      "logps/generated": -780.4212036132812,
+      "logps/real": -437.5557556152344,
+      "loss": 0.2147,
       "rewards/accuracies": 0.987500011920929,
-      "rewards/generated": -50.03806686401367,
-      "rewards/margins": 33.52396011352539,
-      "rewards/real": -16.514102935791016,
+      "rewards/generated": -49.51309585571289,
+      "rewards/margins": 32.93183517456055,
+      "rewards/real": -16.581256866455078,
       "step": 520
     },
     {
@@ -771,13 +771,13 @@
       "learning_rate": 1.7923186344238975e-07,
       "logits/generated": -Infinity,
       "logits/real": -Infinity,
-      "logps/generated": -741.9736328125,
-      "logps/real": -408.57763671875,
-      "loss": 0.3024,
-      "rewards/accuracies": 0.9750000238418579,
-      "rewards/generated": -45.586387634277344,
-      "rewards/margins": 29.933719635009766,
-      "rewards/real": -15.652669906616211,
+      "logps/generated": -735.9564208984375,
+      "logps/real": -407.6651306152344,
+      "loss": 0.2975,
+      "rewards/accuracies": 0.9624999761581421,
+      "rewards/generated": -44.98466491699219,
+      "rewards/margins": 29.423254013061523,
+      "rewards/real": -15.561413764953613,
       "step": 530
     },
     {
@@ -785,13 +785,13 @@
       "learning_rate": 1.721194879089616e-07,
       "logits/generated": -Infinity,
       "logits/real": -Infinity,
-      "logps/generated": -813.8468017578125,
-      "logps/real": -430.96026611328125,
-      "loss": 0.2112,
+      "logps/generated": -807.4033203125,
+      "logps/real": -429.4000549316406,
+      "loss": 0.2254,
       "rewards/accuracies": 0.949999988079071,
-      "rewards/generated": -52.11174392700195,
-      "rewards/margins": 33.244110107421875,
-      "rewards/real": -18.867637634277344,
+      "rewards/generated": -51.46739959716797,
+      "rewards/margins": 32.75578689575195,
+      "rewards/real": -18.711612701416016,
       "step": 540
     },
     {
@@ -799,13 +799,13 @@
       "learning_rate": 1.650071123755334e-07,
       "logits/generated": -Infinity,
       "logits/real": -Infinity,
-      "logps/generated": -813.7432250976562,
-      "logps/real": -425.29754638671875,
-      "loss": 0.3095,
+      "logps/generated": -811.77099609375,
+      "logps/real": -425.84649658203125,
+      "loss": 0.3105,
       "rewards/accuracies": 0.949999988079071,
-      "rewards/generated": -52.5365104675293,
-      "rewards/margins": 34.71119689941406,
-      "rewards/real": -17.825315475463867,
+      "rewards/generated": -52.33929443359375,
+      "rewards/margins": 34.459083557128906,
+      "rewards/real": -17.88020896911621,
       "step": 550
     },
     {
@@ -813,13 +813,13 @@
       "learning_rate": 1.5789473684210525e-07,
       "logits/generated": -Infinity,
       "logits/real": -Infinity,
-      "logps/generated": -790.2058715820312,
-      "logps/real": -440.711669921875,
-      "loss": 0.1864,
+      "logps/generated": -788.242919921875,
+      "logps/real": -440.695556640625,
+      "loss": 0.1702,
       "rewards/accuracies": 0.9624999761581421,
-      "rewards/generated": -49.65260314941406,
-      "rewards/margins": 31.609088897705078,
-      "rewards/real": -18.04351234436035,
+      "rewards/generated": -49.456298828125,
+      "rewards/margins": 31.41439437866211,
+      "rewards/real": -18.041906356811523,
       "step": 560
     },
     {
@@ -827,13 +827,13 @@
       "learning_rate": 1.507823613086771e-07,
       "logits/generated": -Infinity,
       "logits/real": -Infinity,
-      "logps/generated": -752.94677734375,
-      "logps/real": -427.7891540527344,
-      "loss": 0.3443,
-      "rewards/accuracies": 0.925000011920929,
-      "rewards/generated": -47.858802795410156,
-      "rewards/margins": 30.949520111083984,
-      "rewards/real": -16.90927505493164,
+      "logps/generated": -752.26611328125,
+      "logps/real": -427.10113525390625,
+      "loss": 0.3281,
+      "rewards/accuracies": 0.9375,
+      "rewards/generated": -47.79071807861328,
+      "rewards/margins": 30.950246810913086,
+      "rewards/real": -16.84047508239746,
       "step": 570
     },
     {
@@ -841,13 +841,13 @@
       "learning_rate": 1.436699857752489e-07,
       "logits/generated": -Infinity,
       "logits/real": -Infinity,
-      "logps/generated": -843.44140625,
-      "logps/real": -425.5953063964844,
-      "loss": 0.3088,
+      "logps/generated": -826.0186767578125,
+      "logps/real": -424.66680908203125,
+      "loss": 0.3042,
       "rewards/accuracies": 0.9750000238418579,
-      "rewards/generated": -53.65172576904297,
-      "rewards/margins": 36.3214111328125,
-      "rewards/real": -17.330326080322266,
+      "rewards/generated": -51.909454345703125,
+      "rewards/margins": 34.6719856262207,
+      "rewards/real": -17.237468719482422,
       "step": 580
     },
     {
@@ -855,13 +855,13 @@
       "learning_rate": 1.3655761024182077e-07,
       "logits/generated": -Infinity,
       "logits/real": -Infinity,
-      "logps/generated": -816.5244750976562,
-      "logps/real": -419.6240234375,
-      "loss": 0.2445,
+      "logps/generated": -816.61962890625,
+      "logps/real": -420.100830078125,
+      "loss": 0.2635,
       "rewards/accuracies": 0.9624999761581421,
-      "rewards/generated": -52.9045524597168,
-      "rewards/margins": 36.242149353027344,
-      "rewards/real": -16.662403106689453,
+      "rewards/generated": -52.91407012939453,
+      "rewards/margins": 36.203983306884766,
+      "rewards/real": -16.710086822509766,
       "step": 590
     },
     {
@@ -869,13 +869,13 @@
       "learning_rate": 1.2944523470839261e-07,
       "logits/generated": -Infinity,
       "logits/real": -Infinity,
-      "logps/generated": -824.7449340820312,
-      "logps/real": -429.56414794921875,
-      "loss": 0.3351,
-      "rewards/accuracies": 0.949999988079071,
-      "rewards/generated": -51.94647979736328,
-      "rewards/margins": 34.617332458496094,
-      "rewards/real": -17.329151153564453,
+      "logps/generated": -820.4493408203125,
+      "logps/real": -428.4378967285156,
+      "loss": 0.347,
+      "rewards/accuracies": 0.9375,
+      "rewards/generated": -51.51692581176758,
+      "rewards/margins": 34.30039596557617,
+      "rewards/real": -17.216527938842773,
       "step": 600
     },
     {
@@ -883,13 +883,13 @@
       "learning_rate": 1.2233285917496443e-07,
       "logits/generated": -Infinity,
       "logits/real": -Infinity,
-      "logps/generated": -775.5467529296875,
-      "logps/real": -422.3451232910156,
-      "loss": 0.07,
+      "logps/generated": -768.8092041015625,
+      "logps/real": -420.62200927734375,
+      "loss": 0.0588,
       "rewards/accuracies": 0.987500011920929,
-      "rewards/generated": -48.54518508911133,
-      "rewards/margins": 32.839900970458984,
-      "rewards/real": -15.705281257629395,
+      "rewards/generated": -47.871437072753906,
+      "rewards/margins": 32.33845901489258,
+      "rewards/real": -15.532976150512695,
       "step": 610
     },
     {
@@ -897,13 +897,13 @@
       "learning_rate": 1.1522048364153626e-07,
       "logits/generated": -Infinity,
       "logits/real": -Infinity,
-      "logps/generated": -779.6900634765625,
-      "logps/real": -441.49365234375,
-      "loss": 0.2883,
+      "logps/generated": -786.0013427734375,
+      "logps/real": -442.2757873535156,
+      "loss": 0.2483,
       "rewards/accuracies": 0.9624999761581421,
-      "rewards/generated": -49.64753341674805,
-      "rewards/margins": 32.548099517822266,
-      "rewards/real": -17.099435806274414,
+      "rewards/generated": -50.278663635253906,
+      "rewards/margins": 33.101016998291016,
+      "rewards/real": -17.17764663696289,
       "step": 620
     },
     {
@@ -911,13 +911,13 @@
       "learning_rate": 1.0810810810810811e-07,
       "logits/generated": -Infinity,
       "logits/real": -Infinity,
-      "logps/generated": -785.1602172851562,
-      "logps/real": -403.9580993652344,
-      "loss": 0.1209,
+      "logps/generated": -787.63818359375,
+      "logps/real": -404.19122314453125,
+      "loss": 0.1336,
       "rewards/accuracies": 0.9624999761581421,
-      "rewards/generated": -49.906761169433594,
-      "rewards/margins": 33.77416229248047,
-      "rewards/real": -16.13260269165039,
+      "rewards/generated": -50.15456008911133,
+      "rewards/margins": 33.99864959716797,
+      "rewards/real": -16.15591049194336,
       "step": 630
     },
     {
@@ -925,13 +925,13 @@
       "learning_rate": 1.0099573257467994e-07,
       "logits/generated": -Infinity,
       "logits/real": -Infinity,
-      "logps/generated": -830.7962036132812,
-      "logps/real": -457.5830993652344,
-      "loss": 0.2802,
+      "logps/generated": -830.99462890625,
+      "logps/real": -457.3229064941406,
+      "loss": 0.2925,
       "rewards/accuracies": 0.949999988079071,
-      "rewards/generated": -53.04398727416992,
-      "rewards/margins": 34.22355270385742,
-      "rewards/real": -18.820430755615234,
+      "rewards/generated": -53.06385040283203,
+      "rewards/margins": 34.269439697265625,
+      "rewards/real": -18.794404983520508,
       "step": 640
     },
     {
@@ -939,13 +939,13 @@
       "learning_rate": 9.388335704125178e-08,
       "logits/generated": -Infinity,
       "logits/real": -Infinity,
-      "logps/generated": -774.2494506835938,
-      "logps/real": -407.559326171875,
-      "loss": 0.1348,
+      "logps/generated": -777.55517578125,
+      "logps/real": -405.3326110839844,
+      "loss": 0.137,
       "rewards/accuracies": 1.0,
-      "rewards/generated": -49.61153030395508,
-      "rewards/margins": 32.225975036621094,
-      "rewards/real": -17.385555267333984,
+      "rewards/generated": -49.94211196899414,
+      "rewards/margins": 32.77922821044922,
+      "rewards/real": -17.16288948059082,
       "step": 650
     },
     {
@@ -953,13 +953,13 @@
       "learning_rate": 8.677098150782361e-08,
       "logits/generated": -Infinity,
       "logits/real": -Infinity,
-      "logps/generated": -772.5739135742188,
-      "logps/real": -409.3276672363281,
-      "loss": 0.0997,
+      "logps/generated": -775.7921142578125,
+      "logps/real": -408.35302734375,
+      "loss": 0.0992,
       "rewards/accuracies": 0.987500011920929,
-      "rewards/generated": -48.17709732055664,
-      "rewards/margins": 32.27601623535156,
-      "rewards/real": -15.901082038879395,
+      "rewards/generated": -48.498905181884766,
+      "rewards/margins": 32.695289611816406,
+      "rewards/real": -15.803617477416992,
       "step": 660
     },
     {
@@ -967,13 +967,13 @@
       "learning_rate": 7.965860597439544e-08,
       "logits/generated": -Infinity,
       "logits/real": -Infinity,
-      "logps/generated": -764.9151000976562,
-      "logps/real": -425.33709716796875,
-      "loss": 0.2364,
+      "logps/generated": -764.4813232421875,
+      "logps/real": -427.140625,
+      "loss": 0.2396,
       "rewards/accuracies": 0.925000011920929,
-      "rewards/generated": -48.48712921142578,
-      "rewards/margins": 31.54388427734375,
-      "rewards/real": -16.9432430267334,
+      "rewards/generated": -48.443748474121094,
+      "rewards/margins": 31.320148468017578,
+      "rewards/real": -17.123600006103516,
       "step": 670
     },
     {
@@ -981,13 +981,13 @@
       "learning_rate": 7.254623044096728e-08,
       "logits/generated": -Infinity,
       "logits/real": -Infinity,
-      "logps/generated": -822.6394653320312,
-      "logps/real": -434.67877197265625,
-      "loss": 0.242,
+      "logps/generated": -803.4662475585938,
+      "logps/real": -435.402099609375,
+      "loss": 0.2362,
       "rewards/accuracies": 0.987500011920929,
-      "rewards/generated": -53.5615348815918,
-      "rewards/margins": 35.22587585449219,
-      "rewards/real": -18.33565902709961,
+      "rewards/generated": -51.644203186035156,
+      "rewards/margins": 33.23621368408203,
+      "rewards/real": -18.40799331665039,
       "step": 680
     },
     {
@@ -995,13 +995,13 @@
       "learning_rate": 6.543385490753911e-08,
       "logits/generated": -Infinity,
       "logits/real": -Infinity,
-      "logps/generated": -868.9513549804688,
-      "logps/real": -439.777587890625,
-      "loss": 0.203,
-      "rewards/accuracies": 0.9624999761581421,
-      "rewards/generated": -56.48514938354492,
-      "rewards/margins": 37.33259201049805,
-      "rewards/real": -19.152555465698242,
+      "logps/generated": -863.4392700195312,
+      "logps/real": -439.1971130371094,
+      "loss": 0.1949,
+      "rewards/accuracies": 0.9750000238418579,
+      "rewards/generated": -55.93394088745117,
+      "rewards/margins": 36.83943176269531,
+      "rewards/real": -19.094507217407227,
       "step": 690
     },
     {
@@ -1009,13 +1009,13 @@
       "learning_rate": 5.832147937411095e-08,
       "logits/generated": -Infinity,
       "logits/real": -Infinity,
-      "logps/generated": -784.9397583007812,
-      "logps/real": -413.0848083496094,
-      "loss": 0.2787,
-      "rewards/accuracies": 0.925000011920929,
-      "rewards/generated": -47.916893005371094,
-      "rewards/margins": 29.88480567932129,
-      "rewards/real": -18.032087326049805,
+      "logps/generated": -779.9309692382812,
+      "logps/real": -411.91455078125,
+      "loss": 0.2666,
+      "rewards/accuracies": 0.9375,
+      "rewards/generated": -47.416011810302734,
+      "rewards/margins": 29.500951766967773,
+      "rewards/real": -17.915063858032227,
       "step": 700
     },
     {
@@ -1023,13 +1023,13 @@
       "learning_rate": 5.120910384068278e-08,
       "logits/generated": -Infinity,
       "logits/real": -Infinity,
-      "logps/generated": -824.5567626953125,
-      "logps/real": -448.226318359375,
-      "loss": 0.2584,
+      "logps/generated": -826.6921997070312,
+      "logps/real": -447.87762451171875,
+      "loss": 0.258,
       "rewards/accuracies": 0.949999988079071,
-      "rewards/generated": -53.5278205871582,
-      "rewards/margins": 35.21510696411133,
-      "rewards/real": -18.312713623046875,
+      "rewards/generated": -53.74137496948242,
+      "rewards/margins": 35.463531494140625,
+      "rewards/real": -18.277841567993164,
       "step": 710
     },
     {
@@ -1037,13 +1037,13 @@
       "learning_rate": 4.4096728307254624e-08,
       "logits/generated": -Infinity,
       "logits/real": -Infinity,
-      "logps/generated": -828.8753662109375,
-      "logps/real": -427.9056091308594,
-      "loss": 0.1088,
+      "logps/generated": -834.2398681640625,
+      "logps/real": -427.90966796875,
+      "loss": 0.1272,
       "rewards/accuracies": 0.987500011920929,
-      "rewards/generated": -53.2839469909668,
-      "rewards/margins": 35.713260650634766,
-      "rewards/real": -17.57068634033203,
+      "rewards/generated": -53.82038497924805,
+      "rewards/margins": 36.249298095703125,
+      "rewards/real": -17.571086883544922,
       "step": 720
     },
     {
@@ -1051,13 +1051,13 @@
       "learning_rate": 3.698435277382646e-08,
       "logits/generated": -Infinity,
       "logits/real": -Infinity,
-      "logps/generated": -839.0155029296875,
-      "logps/real": -435.33428955078125,
-      "loss": 0.2107,
-      "rewards/accuracies": 1.0,
-      "rewards/generated": -55.168609619140625,
-      "rewards/margins": 37.164310455322266,
-      "rewards/real": -18.004301071166992,
+      "logps/generated": -838.7642822265625,
+      "logps/real": -436.14642333984375,
+      "loss": 0.2078,
+      "rewards/accuracies": 0.9750000238418579,
+      "rewards/generated": -55.14348602294922,
+      "rewards/margins": 37.05797576904297,
+      "rewards/real": -18.08551597595215,
       "step": 730
     },
     {
@@ -1065,13 +1065,13 @@
       "learning_rate": 2.9871977240398294e-08,
       "logits/generated": -Infinity,
       "logits/real": -Infinity,
-      "logps/generated": -805.4227905273438,
-      "logps/real": -428.77923583984375,
-      "loss": 0.2453,
+      "logps/generated": -798.5349731445312,
+      "logps/real": -427.6378479003906,
+      "loss": 0.2621,
       "rewards/accuracies": 0.9624999761581421,
-      "rewards/generated": -52.77501678466797,
-      "rewards/margins": 34.709877014160156,
-      "rewards/real": -18.065141677856445,
+      "rewards/generated": -52.086219787597656,
+      "rewards/margins": 34.135223388671875,
+      "rewards/real": -17.95099449157715,
       "step": 740
     },
     {
@@ -1079,13 +1079,13 @@
       "learning_rate": 2.275960170697013e-08,
       "logits/generated": -Infinity,
       "logits/real": -Infinity,
-      "logps/generated": -834.6575927734375,
-      "logps/real": -415.8843688964844,
-      "loss": 0.1762,
+      "logps/generated": -835.6724853515625,
+      "logps/real": -414.00152587890625,
+      "loss": 0.1564,
       "rewards/accuracies": 0.9750000238418579,
-      "rewards/generated": -52.49644088745117,
-      "rewards/margins": 36.04848098754883,
-      "rewards/real": -16.447967529296875,
+      "rewards/generated": -52.597930908203125,
+      "rewards/margins": 36.33824920654297,
+      "rewards/real": -16.259681701660156,
       "step": 750
     },
     {
@@ -1093,13 +1093,13 @@
       "learning_rate": 1.564722617354196e-08,
       "logits/generated": -Infinity,
       "logits/real": -Infinity,
-      "logps/generated": -850.2274169921875,
-      "logps/real": -395.8170471191406,
-      "loss": 0.1395,
+      "logps/generated": -850.904296875,
+      "logps/real": -392.68310546875,
+      "loss": 0.1515,
       "rewards/accuracies": 0.9750000238418579,
-      "rewards/generated": -55.13444900512695,
-      "rewards/margins": 38.56990432739258,
-      "rewards/real": -16.564544677734375,
+      "rewards/generated": -55.202125549316406,
+      "rewards/margins": 38.95097351074219,
+      "rewards/real": -16.25115394592285,
       "step": 760
     },
     {
@@ -1107,13 +1107,13 @@
       "learning_rate": 8.534850640113798e-09,
       "logits/generated": -Infinity,
       "logits/real": -Infinity,
-      "logps/generated": -807.279296875,
-      "logps/real": -415.69586181640625,
-      "loss": 0.1756,
-      "rewards/accuracies": 0.9750000238418579,
-      "rewards/generated": -52.691368103027344,
-      "rewards/margins": 36.51746368408203,
-      "rewards/real": -16.173904418945312,
+      "logps/generated": -814.2432250976562,
+      "logps/real": -415.373779296875,
+      "loss": 0.1867,
+      "rewards/accuracies": 0.9624999761581421,
+      "rewards/generated": -53.38776397705078,
+      "rewards/margins": 37.246055603027344,
+      "rewards/real": -16.141704559326172,
       "step": 770
     },
     {
@@ -1121,23 +1121,23 @@
       "learning_rate": 1.422475106685633e-09,
       "logits/generated": -Infinity,
       "logits/real": -Infinity,
-      "logps/generated": -819.2388916015625,
-      "logps/real": -390.07391357421875,
-      "loss": 0.2143,
+      "logps/generated": -811.1026611328125,
+      "logps/real": -388.99835205078125,
+      "loss": 0.2403,
       "rewards/accuracies": 0.9750000238418579,
-      "rewards/generated": -52.7367057800293,
-      "rewards/margins": 36.12593460083008,
-      "rewards/real": -16.610774993896484,
+      "rewards/generated": -51.92308807373047,
+      "rewards/margins": 35.41986846923828,
+      "rewards/real": -16.503215789794922,
       "step": 780
     },
     {
       "epoch": 1.0,
       "step": 782,
       "total_flos": 0.0,
-      "train_loss": 0.6284302195911938,
-      "train_runtime": 5860.6637,
-      "train_samples_per_second": 4.266,
-      "train_steps_per_second": 0.133
+      "train_loss": 0.6284351689954493,
+      "train_runtime": 6560.8804,
+      "train_samples_per_second": 3.81,
+      "train_steps_per_second": 0.119
     }
   ],
   "logging_steps": 10,