Training in progress, step 100, checkpoint

checkpoint-100/adapter_config.json

@@ -20,10 +20,10 @@
   "rank_pattern": {},
   "revision": null,
   "target_modules": [
-    "q_proj",
     "v_proj",
     "o_proj",
-    "k_proj"
+    "k_proj",
+    "q_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 sha256:0a09efd490b1230a03de342cc81ebd092ee71bdf105cf41f55fd564ccfb7fe67
+oid sha256:b522f4d26fb3f5c7f67fb6d1e57c69d93d012b49ae040b33d304619d5e700627
 size 54560368

checkpoint-100/optimizer.pt

@@ -1,3 +1,3 @@
 version https://git-lfs.github.com/spec/v1
-oid sha256:31ec931eae12d877a7421bb24a3cd866d5cb82e68cb89575eac37dd6eb7bf739
+oid sha256:be7f37e8a7a420f9026fe26771af86c4119f4a70b6df31a422e59a46e283f0b9
 size 109267450

checkpoint-100/trainer_state.json

@@ -1,5 +1,5 @@
 {
-  "best_metric": 0.47333332896232605,
+  "best_metric": 0.3754348158836365,
   "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": 4.650723457336426,
-      "kl": 4.865365505218506,
-      "learning_rate": 0.00018285714285714286,
-      "logps/chosen": -310.4614562988281,
-      "logps/rejected": -364.2945556640625,
-      "loss": 0.45,
-      "rewards/chosen": -2.890953779220581,
-      "rewards/margins": 3.259232521057129,
-      "rewards/rejected": -5.586672306060791,
+      "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,
       "step": 20
     },
     {
       "epoch": 0.27,
-      "grad_norm": 0.0,
-      "kl": 1.256190538406372,
+      "grad_norm": 19.954599380493164,
+      "kl": 6.632220268249512,
       "learning_rate": 0.0001542857142857143,
-      "logps/chosen": -1522.9342041015625,
-      "logps/rejected": -1577.1060791015625,
-      "loss": 0.4463,
-      "rewards/chosen": -123.79159545898438,
-      "rewards/margins": 4.876855373382568,
-      "rewards/rejected": -128.13319396972656,
+      "logps/chosen": -319.3048400878906,
+      "logps/rejected": -348.5643615722656,
+      "loss": 0.4315,
+      "rewards/chosen": -1.7924585342407227,
+      "rewards/margins": 1.6441251039505005,
+      "rewards/rejected": -3.656928539276123,
       "step": 40
     },
     {
       "epoch": 0.34,
-      "eval_kl": 0.0,
-      "eval_logps/chosen": -1422.4874267578125,
-      "eval_logps/rejected": -1508.553955078125,
-      "eval_loss": 0.473336786031723,
-      "eval_rewards/chosen": -114.58648681640625,
-      "eval_rewards/margins": 10.318593978881836,
-      "eval_rewards/rejected": -124.07293701171875,
-      "eval_runtime": 140.0926,
-      "eval_samples_per_second": 2.141,
-      "eval_steps_per_second": 0.535,
+      "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,
       "step": 50
     },
     {
       "epoch": 0.41,
-      "grad_norm": 3.258587184973294e-06,
-      "kl": 0.0,
+      "grad_norm": 4.9014763832092285,
+      "kl": 22.45343017578125,
       "learning_rate": 0.00012571428571428572,
-      "logps/chosen": -1605.4840087890625,
-      "logps/rejected": -1731.5294189453125,
-      "loss": 0.4812,
-      "rewards/chosen": -133.641357421875,
-      "rewards/margins": 11.06646728515625,
-      "rewards/rejected": -144.48353576660156,
+      "logps/chosen": -290.8168029785156,
+      "logps/rejected": -297.7256774902344,
+      "loss": 0.4154,
+      "rewards/chosen": 3.169417381286621,
+      "rewards/margins": 1.9798123836517334,
+      "rewards/rejected": 1.3037060499191284,
       "step": 60
     },
     {
       "epoch": 0.55,
-      "grad_norm": 0.0,
-      "kl": 0.0,
+      "grad_norm": 2.188873291015625,
+      "kl": 15.003771781921387,
       "learning_rate": 9.714285714285715e-05,
-      "logps/chosen": -2286.116455078125,
-      "logps/rejected": -2312.602294921875,
-      "loss": 0.4688,
-      "rewards/chosen": -200.4071044921875,
-      "rewards/margins": 2.4775140285491943,
-      "rewards/rejected": -200.7866973876953,
+      "logps/chosen": -279.8699951171875,
+      "logps/rejected": -332.38848876953125,
+      "loss": 0.3492,
+      "rewards/chosen": 1.9723680019378662,
+      "rewards/margins": 4.276235103607178,
+      "rewards/rejected": -2.454475164413452,
       "step": 80
     },
     {
       "epoch": 0.68,
-      "grad_norm": 0.0,
-      "kl": 0.0,
+      "grad_norm": 4.462972640991211,
+      "kl": 4.606294631958008,
       "learning_rate": 6.857142857142858e-05,
-      "logps/chosen": -2744.5634765625,
-      "logps/rejected": -2576.698974609375,
-      "loss": 0.4719,
-      "rewards/chosen": -246.3837890625,
-      "rewards/margins": -13.795512199401855,
-      "rewards/rejected": -228.7157440185547,
+      "logps/chosen": -278.8154602050781,
+      "logps/rejected": -356.94451904296875,
+      "loss": 0.3678,
+      "rewards/chosen": 1.00811767578125,
+      "rewards/margins": 5.261899471282959,
+      "rewards/rejected": -4.226025581359863,
       "step": 100
     },
     {
       "epoch": 0.68,
-      "eval_kl": 0.0,
-      "eval_logps/chosen": -2469.165771484375,
-      "eval_logps/rejected": -2225.683349609375,
-      "eval_loss": 0.47333332896232605,
-      "eval_rewards/chosen": -219.25433349609375,
-      "eval_rewards/margins": -26.235870361328125,
-      "eval_rewards/rejected": -195.78585815429688,
-      "eval_runtime": 139.9824,
-      "eval_samples_per_second": 2.143,
-      "eval_steps_per_second": 0.536,
+      "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,
       "step": 100
     }
   ],

checkpoint-100/training_args.bin

@@ -1,3 +1,3 @@
 version https://git-lfs.github.com/spec/v1
-oid sha256:91658cc872c39fb432cb322f2450e9b0c776e35ec752fc4911f1c95a922ea14a
+oid sha256:918924d56c60c496a10afc622ed24a5105bed8686a07c0f35eeb9917bbadd344
 size 5624