Training in progress, step 100, checkpoint

checkpoint-100/adapter_config.json

@@ -20,10 +20,10 @@
   "rank_pattern": {},
   "revision": null,
   "target_modules": [
-    "o_proj",
     "q_proj",
+    "k_proj",
     "v_proj",
-    "k_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 sha256:6e3a00b2723924481e99ca7420904a3063b73b668f14806f7769bf6279ec38f7
+oid sha256:d1d26bc61e21d0b93fd9b446c555c662c0dced2ed04a45c6e79b26eeb3ee52e9
 size 6849416

checkpoint-100/optimizer.pt

@@ -1,3 +1,3 @@
 version https://git-lfs.github.com/spec/v1
-oid sha256:101b9b6f741ef23eec4c6ec86585ed72f0eb201356b3c71f133178a65ad02838
+oid sha256:9a3d921e02a11aa9aed101e1624589a67a8ae88f444b6f2746ea10128ca55dc8
 size 13846266

checkpoint-100/scheduler.pt

@@ -1,3 +1,3 @@
 version https://git-lfs.github.com/spec/v1
-oid sha256:227afea00680bdcdcf19d54b572f61ab2e563bd954561db8d7fee74cde40c145
+oid sha256:363bcb1976d3d8f69d575a3bb74fad2f79e9d75da57793c889da5e2ae17ef801
 size 1064

checkpoint-100/trainer_state.json

@@ -1,5 +1,5 @@
 {
-  "best_metric": 0.47333332896232605,
+  "best_metric": 0.3468475937843323,
   "best_model_checkpoint": "./zephyr/08-04-24-Weni-WeniGPT-Agents-Zephyr-1.0.18-KTO_Hyperparameter search, altering lora params for KTO task.-2_max_steps-145_batch_16_2024-04-08_ppid_9/checkpoint-100",
   "epoch": 0.684931506849315,
   "eval_steps": 50,
@@ -10,95 +10,95 @@
   "log_history": [
     {
       "epoch": 0.14,
-      "grad_norm": 1.8155961925003794e-06,
-      "learning_rate": 0.00018,
-      "loss": 0.4616,
+      "grad_norm": 3.3949286937713623,
+      "learning_rate": 0.00018142857142857142,
+      "loss": 0.4993,
       "step": 20,
-      "train/kl": 0.0,
-      "train/logps/chosen": -384.02016395246477,
-      "train/logps/rejected": -362.7105863764045,
-      "train/rewards/chosen": -9.318169392330546,
-      "train/rewards/margins": -0.6738980560858128,
-      "train/rewards/rejected": -8.644271336244733
+      "train/kl": 2.0773630142211914,
+      "train/logps/chosen": -280.7225560897436,
+      "train/logps/rejected": -280.629668445122,
+      "train/rewards/chosen": -0.4658302894005409,
+      "train/rewards/margins": 0.008156592134686835,
+      "train/rewards/rejected": -0.4739868815352277
     },
     {
       "epoch": 0.27,
-      "grad_norm": 0.0,
-      "learning_rate": 0.00015142857142857143,
-      "loss": 0.4875,
+      "grad_norm": 2.1768083572387695,
+      "learning_rate": 0.00015285714285714287,
+      "loss": 0.412,
       "step": 40,
-      "train/kl": 0.0,
-      "train/logps/chosen": -517.3250200320513,
-      "train/logps/rejected": -543.1232374237804,
-      "train/rewards/chosen": -23.67786583533654,
-      "train/rewards/margins": 1.115628412472148,
-      "train/rewards/rejected": -24.793494247808688
+      "train/kl": 11.56242561340332,
+      "train/logps/chosen": -279.2820556640625,
+      "train/logps/rejected": -285.233056640625,
+      "train/rewards/chosen": 1.6047859191894531,
+      "train/rewards/margins": 1.591878777742386,
+      "train/rewards/rejected": 0.01290714144706726
     },
     {
       "epoch": 0.34,
-      "eval/kl": 0.0,
-      "eval/logps/chosen": -559.2342099471831,
-      "eval/logps/rejected": -549.5850474683544,
-      "eval/rewards/chosen": -27.557097958846832,
-      "eval/rewards/margins": 0.9651281091911414,
-      "eval/rewards/rejected": -28.522226068037973,
-      "eval_loss": 0.47333332896232605,
-      "eval_runtime": 140.5206,
-      "eval_samples_per_second": 2.135,
-      "eval_steps_per_second": 0.534,
+      "eval/kl": 34.189842224121094,
+      "eval/logps/chosen": -240.1980633802817,
+      "eval/logps/rejected": -241.65251681170886,
+      "eval/rewards/chosen": 4.346213757152289,
+      "eval/rewards/margins": 2.0745202015376725,
+      "eval/rewards/rejected": 2.2716935556146165,
+      "eval_loss": 0.3987465500831604,
+      "eval_runtime": 139.8252,
+      "eval_samples_per_second": 2.146,
+      "eval_steps_per_second": 0.536,
       "step": 50
     },
     {
       "epoch": 0.41,
-      "grad_norm": 0.0,
-      "learning_rate": 0.00012285714285714287,
-      "loss": 0.4531,
+      "grad_norm": 3.7959017753601074,
+      "learning_rate": 0.00012428571428571428,
+      "loss": 0.409,
       "step": 60,
-      "train/kl": 0.0,
-      "train/logps/chosen": -566.4386314655172,
-      "train/logps/rejected": -569.3920535714286,
-      "train/rewards/chosen": -27.128678946659484,
-      "train/rewards/margins": 0.9857937095905172,
-      "train/rewards/rejected": -28.11447265625
+      "train/kl": 25.960163116455078,
+      "train/logps/chosen": -228.58444552951389,
+      "train/logps/rejected": -259.24209872159093,
+      "train/rewards/chosen": 3.7380718655056424,
+      "train/rewards/margins": 2.2719342395512747,
+      "train/rewards/rejected": 1.466137625954368
     },
     {
       "epoch": 0.55,
-      "grad_norm": 0.0,
-      "learning_rate": 9.428571428571429e-05,
-      "loss": 0.475,
+      "grad_norm": 5.033076763153076,
+      "learning_rate": 9.571428571428573e-05,
+      "loss": 0.3576,
       "step": 80,
-      "train/kl": 0.0,
-      "train/logps/chosen": -543.1592824835526,
-      "train/logps/rejected": -573.1459728422619,
-      "train/rewards/chosen": -26.98720189144737,
-      "train/rewards/margins": 1.380697685375548,
-      "train/rewards/rejected": -28.367899576822918
+      "train/kl": 5.953394412994385,
+      "train/logps/chosen": -279.7611678685897,
+      "train/logps/rejected": -306.2880144817073,
+      "train/rewards/chosen": 1.7610896183894231,
+      "train/rewards/margins": 3.6376326169126703,
+      "train/rewards/rejected": -1.876542998523247
     },
     {
       "epoch": 0.68,
-      "grad_norm": 0.0,
-      "learning_rate": 6.571428571428571e-05,
-      "loss": 0.4562,
+      "grad_norm": 6.184478759765625,
+      "learning_rate": 6.714285714285714e-05,
+      "loss": 0.3304,
       "step": 100,
-      "train/kl": 0.0,
-      "train/logps/chosen": -549.855522260274,
-      "train/logps/rejected": -568.3944863505748,
-      "train/rewards/chosen": -27.00509685359589,
-      "train/rewards/margins": 2.2099236878911768,
-      "train/rewards/rejected": -29.215020541487068
+      "train/kl": 0.5991002321243286,
+      "train/logps/chosen": -307.0083233173077,
+      "train/logps/rejected": -345.8705592105263,
+      "train/rewards/chosen": -1.1401430570162259,
+      "train/rewards/margins": 4.161937278098906,
+      "train/rewards/rejected": -5.302080335115131
     },
     {
       "epoch": 0.68,
-      "eval/kl": 0.0,
-      "eval/logps/chosen": -562.8151408450705,
-      "eval/logps/rejected": -553.574119857595,
-      "eval/rewards/chosen": -27.915193799515844,
-      "eval/rewards/margins": 1.0059395074461825,
-      "eval/rewards/rejected": -28.921133306962027,
-      "eval_loss": 0.47333332896232605,
-      "eval_runtime": 140.4904,
-      "eval_samples_per_second": 2.135,
-      "eval_steps_per_second": 0.534,
+      "eval/kl": 3.7238929271698,
+      "eval/logps/chosen": -265.10247029049293,
+      "eval/logps/rejected": -277.03555181962025,
+      "eval/rewards/chosen": 1.855774355606294,
+      "eval/rewards/margins": 3.1223836440754393,
+      "eval/rewards/rejected": -1.2666092884691456,
+      "eval_loss": 0.3468475937843323,
+      "eval_runtime": 139.8048,
+      "eval_samples_per_second": 2.146,
+      "eval_steps_per_second": 0.536,
       "step": 100
     }
   ],

checkpoint-100/training_args.bin

@@ -1,3 +1,3 @@
 version https://git-lfs.github.com/spec/v1
-oid sha256:7226114c5e2b217740bdcf70fa54f9c5ff75648b1c63ef0cf288d14e4f73b13c
+oid sha256:38ae2e87f29eb2d5b20b8a1633c876370c1a196bac5b23847645ad024eea5fb3
 size 5688