Training in progress, step 100, checkpoint

checkpoint-100/adapter_config.json

@@ -20,10 +20,10 @@
   "rank_pattern": {},
   "revision": null,
   "target_modules": [
-    "o_proj",
     "v_proj",
-    "q_proj",
-    "k_proj"
+    "o_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:e5de099bbf6d137eaf417dc4d3bde76b542612bb9c07e6fd415dfcd2f2cf14e7
+oid sha256:6072b6642115fd4383ff47af572051e28f499a366efad2d6af9e21576673d0c0
 size 54560368

checkpoint-100/optimizer.pt

@@ -1,3 +1,3 @@
 version https://git-lfs.github.com/spec/v1
-oid sha256:ea4410718a328f195e3f750d20e13b9b7c1d8b1b31bdb5f3c47cce06b4b10824
+oid sha256:8677aaa3491cf25746d02e89ba51b5f2cb7001c8cc9e27de8cdb551d46e668aa
 size 109267450

checkpoint-100/trainer_state.json

@@ -1,6 +1,6 @@
 {
   "best_metric": 0.4496666491031647,
-  "best_model_checkpoint": "./zephyr/08-04-24-Weni-WeniGPT-Agents-Zephyr-1.0.11-KTO_Hyperparameter search, altering desired and undesired weights for KTO task.-2_max_steps-145_batch_16_2024-04-08_ppid_9/checkpoint-100",
+  "best_model_checkpoint": "./zephyr/09-04-24-Weni-WeniGPT-Agents-Zephyr-1.0.11-KTO_Hyperparameter search, altering desired and undesired weights for KTO task.-2_max_steps-145_batch_16_2024-04-09_ppid_10/checkpoint-100",
   "epoch": 0.684931506849315,
   "eval_steps": 50,
   "global_step": 100,
@@ -13,12 +13,12 @@
       "grad_norm": 0.0,
       "kl": 0.0,
       "learning_rate": 0.00018142857142857142,
-      "logps/chosen": -497.94134521484375,
-      "logps/rejected": -483.5335998535156,
-      "loss": 0.4555,
-      "rewards/chosen": -20.328125,
-      "rewards/margins": 0.06232962757349014,
-      "rewards/rejected": -19.38505744934082,
+      "logps/chosen": -1014.4324340820312,
+      "logps/rejected": -961.4172973632812,
+      "loss": 0.4278,
+      "rewards/chosen": -71.9658432006836,
+      "rewards/margins": -3.0392403602600098,
+      "rewards/rejected": -67.96768188476562,
       "step": 20
     },
     {
@@ -26,26 +26,26 @@
       "grad_norm": 0.0,
       "kl": 0.0,
       "learning_rate": 0.00015285714285714287,
-      "logps/chosen": -894.6141967773438,
-      "logps/rejected": -926.7986450195312,
-      "loss": 0.4602,
-      "rewards/chosen": -60.24372482299805,
-      "rewards/margins": 5.149778842926025,
-      "rewards/rejected": -66.03770446777344,
+      "logps/chosen": -2804.0458984375,
+      "logps/rejected": -2825.398193359375,
+      "loss": 0.4513,
+      "rewards/chosen": -251.50927734375,
+      "rewards/margins": -4.426294326782227,
+      "rewards/rejected": -251.947265625,
       "step": 40
     },
     {
       "epoch": 0.34,
       "eval_kl": 0.0,
-      "eval_logps/chosen": -1047.365234375,
-      "eval_logps/rejected": -1037.4864501953125,
+      "eval_logps/chosen": -2748.7060546875,
+      "eval_logps/rejected": -2395.84228515625,
       "eval_loss": 0.4496666491031647,
-      "eval_rewards/chosen": -75.97087860107422,
-      "eval_rewards/margins": 1.050919532775879,
-      "eval_rewards/rejected": -76.91150665283203,
-      "eval_runtime": 138.5255,
-      "eval_samples_per_second": 2.166,
-      "eval_steps_per_second": 0.541,
+      "eval_rewards/chosen": -246.0056915283203,
+      "eval_rewards/margins": -31.47684669494629,
+      "eval_rewards/rejected": -213.32154846191406,
+      "eval_runtime": 140.7571,
+      "eval_samples_per_second": 2.131,
+      "eval_steps_per_second": 0.533,
       "step": 50
     },
     {
@@ -53,12 +53,12 @@
       "grad_norm": 0.0,
       "kl": 0.0,
       "learning_rate": 0.00012428571428571428,
-      "logps/chosen": -1031.908447265625,
-      "logps/rejected": -1113.5894775390625,
-      "loss": 0.4394,
-      "rewards/chosen": -75.6056900024414,
-      "rewards/margins": 5.431799411773682,
-      "rewards/rejected": -82.01710510253906,
+      "logps/chosen": -2966.9404296875,
+      "logps/rejected": -2732.59423828125,
+      "loss": 0.4483,
+      "rewards/chosen": -268.0592041015625,
+      "rewards/margins": -19.579919815063477,
+      "rewards/rejected": -244.43467712402344,
       "step": 60
     },
     {
@@ -66,12 +66,12 @@
       "grad_norm": 0.0,
       "kl": 0.0,
       "learning_rate": 9.571428571428573e-05,
-      "logps/chosen": -1048.306640625,
-      "logps/rejected": -1061.301025390625,
-      "loss": 0.4869,
-      "rewards/chosen": -76.67771911621094,
-      "rewards/margins": 0.8430765271186829,
-      "rewards/rejected": -78.49474334716797,
+      "logps/chosen": -2559.466064453125,
+      "logps/rejected": -2662.142578125,
+      "loss": 0.4572,
+      "rewards/chosen": -229.73390197753906,
+      "rewards/margins": 8.422286987304688,
+      "rewards/rejected": -237.15330505371094,
       "step": 80
     },
     {
@@ -79,26 +79,26 @@
       "grad_norm": 0.0,
       "kl": 0.0,
       "learning_rate": 6.714285714285714e-05,
-      "logps/chosen": -1074.357177734375,
-      "logps/rejected": -1137.3873291015625,
-      "loss": 0.4483,
-      "rewards/chosen": -77.28292083740234,
-      "rewards/margins": 9.020448684692383,
-      "rewards/rejected": -85.49586486816406,
+      "logps/chosen": -2944.9951171875,
+      "logps/rejected": -2686.48046875,
+      "loss": 0.475,
+      "rewards/chosen": -264.62896728515625,
+      "rewards/margins": -24.418039321899414,
+      "rewards/rejected": -238.73057556152344,
       "step": 100
     },
     {
       "epoch": 0.68,
       "eval_kl": 0.0,
-      "eval_logps/chosen": -1065.047607421875,
-      "eval_logps/rejected": -1055.3685302734375,
+      "eval_logps/chosen": -2736.132568359375,
+      "eval_logps/rejected": -2383.06005859375,
       "eval_loss": 0.4496666491031647,
-      "eval_rewards/chosen": -77.73910522460938,
-      "eval_rewards/margins": 1.0638010501861572,
-      "eval_rewards/rejected": -78.69970703125,
-      "eval_runtime": 138.4826,
-      "eval_samples_per_second": 2.166,
-      "eval_steps_per_second": 0.542,
+      "eval_rewards/chosen": -244.74835205078125,
+      "eval_rewards/margins": -31.5240421295166,
+      "eval_rewards/rejected": -212.04331970214844,
+      "eval_runtime": 140.895,
+      "eval_samples_per_second": 2.129,
+      "eval_steps_per_second": 0.532,
       "step": 100
     }
   ],

checkpoint-100/training_args.bin

@@ -1,3 +1,3 @@
 version https://git-lfs.github.com/spec/v1
-oid sha256:33b50e60be62ce43760f9c0e318a7d543aed98ed5967b04c00f11575781e0c9c
+oid sha256:3e692c85e2382863d47509bc470768505f535684608d322bb3d14aa5f9ed78ae
 size 5688