Training in progress, step 100, checkpoint

checkpoint-100/adapter_config.json

@@ -20,10 +20,10 @@
   "rank_pattern": {},
   "revision": null,
   "target_modules": [
-    "v_proj",
-    "o_proj",
     "k_proj",
-    "q_proj"
+    "v_proj",
+    "q_proj",
+    "o_proj"
   ],
   "task_type": "CAUSAL_LM",
   "use_dora": false,

checkpoint-100/adapter_model.safetensors

@@ -1,3 +1,3 @@
 version https://git-lfs.github.com/spec/v1
-oid sha256:b522f4d26fb3f5c7f67fb6d1e57c69d93d012b49ae040b33d304619d5e700627
+oid sha256:66298e88cce2631c468d9284935e93b4d249f9dfa01080d4519bdf0479322d10
 size 54560368

checkpoint-100/optimizer.pt

@@ -1,3 +1,3 @@
 version https://git-lfs.github.com/spec/v1
-oid sha256:be7f37e8a7a420f9026fe26771af86c4119f4a70b6df31a422e59a46e283f0b9
+oid sha256:a75f97964133629dde1484c53985c085332965a689a7d706870c0f1e35d92853
 size 109267450

checkpoint-100/scheduler.pt

@@ -1,3 +1,3 @@
 version https://git-lfs.github.com/spec/v1
-oid sha256:99cdf514d195ace594a247b30bc61c0b0ccb08b6431f56f75e4d6b9fc7392b92
+oid sha256:d18870198c31c5821df8b9e9d648f47112d84cad0aef4ef7508fe2f35542f854
 size 1064

checkpoint-100/trainer_state.json

@@ -1,5 +1,5 @@
 {
-  "best_metric": 0.3754348158836365,
+  "best_metric": 0.47333332896232605,
   "best_model_checkpoint": "./zephyr/05-04-24-Weni-WeniGPT-Agents-Zephyr-1.0.1-KTO_testing kto dataset during training-3_max_steps-145_batch_16_2024-04-05_ppid_9/checkpoint-100",
   "epoch": 0.684931506849315,
   "eval_steps": 50,
@@ -10,95 +10,95 @@
   "log_history": [
     {
       "epoch": 0.14,
-      "grad_norm": 7.029934406280518,
-      "kl": 2.2542672157287598,
-      "learning_rate": 0.00018,
-      "logps/chosen": -356.6376037597656,
-      "logps/rejected": -393.9584655761719,
-      "loss": 0.4417,
-      "rewards/chosen": -3.5771310329437256,
-      "rewards/margins": 2.5301473140716553,
-      "rewards/rejected": -6.391368865966797,
+      "grad_norm": 32.5521125793457,
+      "kl": 6.568634033203125,
+      "learning_rate": 0.00018285714285714286,
+      "logps/chosen": -309.11578369140625,
+      "logps/rejected": -330.8851013183594,
+      "loss": 0.4463,
+      "rewards/chosen": -1.5149470567703247,
+      "rewards/margins": 0.730856716632843,
+      "rewards/rejected": -2.1397347450256348,
       "step": 20
     },
     {
       "epoch": 0.27,
-      "grad_norm": 19.954599380493164,
-      "kl": 6.632220268249512,
-      "learning_rate": 0.0001542857142857143,
-      "logps/chosen": -319.3048400878906,
-      "logps/rejected": -348.5643615722656,
-      "loss": 0.4315,
-      "rewards/chosen": -1.7924585342407227,
-      "rewards/margins": 1.6441251039505005,
-      "rewards/rejected": -3.656928539276123,
+      "grad_norm": 5.194253444671631,
+      "kl": 9.558149337768555,
+      "learning_rate": 0.00015571428571428572,
+      "logps/chosen": -317.7120666503906,
+      "logps/rejected": -336.9992980957031,
+      "loss": 0.4204,
+      "rewards/chosen": -0.900913655757904,
+      "rewards/margins": 1.870524287223816,
+      "rewards/rejected": -2.7778050899505615,
       "step": 40
     },
     {
       "epoch": 0.34,
-      "eval_kl": 32.934268951416016,
-      "eval_logps/chosen": -264.7897033691406,
-      "eval_logps/rejected": -267.6116027832031,
-      "eval_loss": 0.4092177450656891,
-      "eval_rewards/chosen": 5.2044806480407715,
-      "eval_rewards/margins": 2.720277786254883,
-      "eval_rewards/rejected": 2.5654892921447754,
-      "eval_runtime": 139.3761,
-      "eval_samples_per_second": 2.152,
-      "eval_steps_per_second": 0.538,
+      "eval_kl": 0.0,
+      "eval_logps/chosen": -471.27984619140625,
+      "eval_logps/rejected": -470.8447570800781,
+      "eval_loss": 0.47356343269348145,
+      "eval_rewards/chosen": -17.10463523864746,
+      "eval_rewards/margins": 1.9298614263534546,
+      "eval_rewards/rejected": -19.51471710205078,
+      "eval_runtime": 137.637,
+      "eval_samples_per_second": 2.18,
+      "eval_steps_per_second": 0.545,
       "step": 50
     },
     {
       "epoch": 0.41,
-      "grad_norm": 4.9014763832092285,
-      "kl": 22.45343017578125,
-      "learning_rate": 0.00012571428571428572,
-      "logps/chosen": -290.8168029785156,
-      "logps/rejected": -297.7256774902344,
-      "loss": 0.4154,
-      "rewards/chosen": 3.169417381286621,
-      "rewards/margins": 1.9798123836517334,
-      "rewards/rejected": 1.3037060499191284,
+      "grad_norm": 0.0,
+      "kl": 0.2518434524536133,
+      "learning_rate": 0.00012714285714285714,
+      "logps/chosen": -638.1253051757812,
+      "logps/rejected": -641.34716796875,
+      "loss": 0.402,
+      "rewards/chosen": -33.93111038208008,
+      "rewards/margins": 1.04580819606781,
+      "rewards/rejected": -34.0406379699707,
       "step": 60
     },
     {
       "epoch": 0.55,
-      "grad_norm": 2.188873291015625,
-      "kl": 15.003771781921387,
-      "learning_rate": 9.714285714285715e-05,
-      "logps/chosen": -279.8699951171875,
-      "logps/rejected": -332.38848876953125,
-      "loss": 0.3492,
-      "rewards/chosen": 1.9723680019378662,
-      "rewards/margins": 4.276235103607178,
-      "rewards/rejected": -2.454475164413452,
+      "grad_norm": 0.0,
+      "kl": 0.0,
+      "learning_rate": 9.857142857142858e-05,
+      "logps/chosen": -1467.325927734375,
+      "logps/rejected": -1537.5704345703125,
+      "loss": 0.45,
+      "rewards/chosen": -117.26831817626953,
+      "rewards/margins": 9.93100357055664,
+      "rewards/rejected": -124.3635025024414,
       "step": 80
     },
     {
       "epoch": 0.68,
-      "grad_norm": 4.462972640991211,
-      "kl": 4.606294631958008,
-      "learning_rate": 6.857142857142858e-05,
-      "logps/chosen": -278.8154602050781,
-      "logps/rejected": -356.94451904296875,
-      "loss": 0.3678,
-      "rewards/chosen": 1.00811767578125,
-      "rewards/margins": 5.261899471282959,
-      "rewards/rejected": -4.226025581359863,
+      "grad_norm": 0.0,
+      "kl": 0.0,
+      "learning_rate": 7e-05,
+      "logps/chosen": -2025.484375,
+      "logps/rejected": -1882.913818359375,
+      "loss": 0.4813,
+      "rewards/chosen": -169.48184204101562,
+      "rewards/margins": -10.249415397644043,
+      "rewards/rejected": -157.5347900390625,
       "step": 100
     },
     {
       "epoch": 0.68,
-      "eval_kl": 1.3235526084899902,
-      "eval_logps/chosen": -330.0657653808594,
-      "eval_logps/rejected": -357.2244873046875,
-      "eval_loss": 0.3754348158836365,
-      "eval_rewards/chosen": -1.3231240510940552,
-      "eval_rewards/margins": 5.1890459060668945,
-      "eval_rewards/rejected": -6.395801067352295,
-      "eval_runtime": 139.0506,
-      "eval_samples_per_second": 2.157,
-      "eval_steps_per_second": 0.539,
+      "eval_kl": 0.0,
+      "eval_logps/chosen": -1703.4486083984375,
+      "eval_logps/rejected": -1636.100341796875,
+      "eval_loss": 0.47333332896232605,
+      "eval_rewards/chosen": -140.32151794433594,
+      "eval_rewards/margins": -6.890669822692871,
+      "eval_rewards/rejected": -136.040283203125,
+      "eval_runtime": 137.631,
+      "eval_samples_per_second": 2.18,
+      "eval_steps_per_second": 0.545,
       "step": 100
     }
   ],

checkpoint-100/training_args.bin

@@ -1,3 +1,3 @@
 version https://git-lfs.github.com/spec/v1
-oid sha256:918924d56c60c496a10afc622ed24a5105bed8686a07c0f35eeb9917bbadd344
+oid sha256:2537b1649ac52556ff9ceed8db71e4fc060d1ffc901556fa0b90e1d0d70234d5
 size 5624