Training in progress, step 100, checkpoint

checkpoint-100/adapter_config.json

@@ -20,9 +20,9 @@
   "rank_pattern": {},
   "revision": null,
   "target_modules": [
-    "v_proj",
     "k_proj",
     "q_proj",
+    "v_proj",
     "o_proj"
   ],
   "task_type": "CAUSAL_LM",

checkpoint-100/adapter_model.safetensors

@@ -1,3 +1,3 @@
 version https://git-lfs.github.com/spec/v1
-oid sha256:ad992dcdb4258d37a765425377a9b92b92fb78925955f14c74a6beec38d715c2
+oid sha256:31708829f27daf4d0a9b29f0546ba8eec73a3b1a835a02461b2817562e62db5e
 size 436242776

checkpoint-100/optimizer.pt

@@ -1,3 +1,3 @@
 version https://git-lfs.github.com/spec/v1
-oid sha256:28451a46dbd0ed11b50fd2625d02f48a1defa58eb16bf0764d57d3f9a318f66c
+oid sha256:c7e7e1c59e13d70a56ec028cbb03dc7b3b8e577a52bb9a3cca0b367450284032
 size 872631034

checkpoint-100/scheduler.pt

@@ -1,3 +1,3 @@
 version https://git-lfs.github.com/spec/v1
-oid sha256:12711e870dc2eb008cef779e781ac8fe1872c7e9d23f2bbc181148f8c2d8f981
+oid sha256:227afea00680bdcdcf19d54b572f61ab2e563bd954561db8d7fee74cde40c145
 size 1064

checkpoint-100/trainer_state.json

@@ -1,6 +1,6 @@
 {
   "best_metric": 0.47333332896232605,
-  "best_model_checkpoint": "./zephyr/08-04-24-Weni-WeniGPT-Agents-Zephyr-1.0.16-KTO_Hyperparameter search, altering lora params 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-1.0.16-KTO_Hyperparameter search, altering lora params for KTO task.-2_max_steps-145_batch_16_2024-04-09_ppid_9/checkpoint-100",
   "epoch": 0.684931506849315,
   "eval_steps": 50,
   "global_step": 100,
@@ -11,94 +11,94 @@
     {
       "epoch": 0.14,
       "grad_norm": 0.0,
-      "learning_rate": 0.00018714285714285716,
-      "loss": 0.4292,
+      "learning_rate": 0.00018,
+      "loss": 0.4713,
       "step": 20,
-      "train/kl": 4.945448875427246,
-      "train/logps/chosen": -899.4505912162163,
-      "train/logps/rejected": -1086.2307412790697,
-      "train/rewards/chosen": -61.6109058277027,
-      "train/rewards/margins": 17.706562150131596,
-      "train/rewards/rejected": -79.3174679778343
+      "train/kl": 0.0,
+      "train/logps/chosen": -1423.408907312925,
+      "train/logps/rejected": -1301.3419436416184,
+      "train/rewards/chosen": -113.72140731292517,
+      "train/rewards/margins": -11.091552724774871,
+      "train/rewards/rejected": -102.6298545881503
     },
     {
       "epoch": 0.27,
       "grad_norm": 0.0,
-      "learning_rate": 0.00015857142857142857,
-      "loss": 0.4781,
+      "learning_rate": 0.00015142857142857143,
+      "loss": 0.45,
       "step": 40,
       "train/kl": 0.0,
-      "train/logps/chosen": -2608.2385620915034,
-      "train/logps/rejected": -2442.9848428143714,
-      "train/rewards/chosen": -233.4453635620915,
-      "train/rewards/margins": -17.621004467780125,
-      "train/rewards/rejected": -215.82435909431138
+      "train/logps/chosen": -2404.9383680555557,
+      "train/logps/rejected": -2267.5051491477275,
+      "train/rewards/chosen": -211.85392252604166,
+      "train/rewards/margins": -15.253491950757564,
+      "train/rewards/rejected": -196.6004305752841
     },
     {
       "epoch": 0.34,
       "eval/kl": 0.0,
-      "eval/logps/chosen": -2513.149647887324,
-      "eval/logps/rejected": -2267.6801819620255,
-      "eval/rewards/chosen": -222.94891615316902,
-      "eval/rewards/margins": -22.617910911871547,
-      "eval/rewards/rejected": -200.33100524129748,
+      "eval/logps/chosen": -2179.756602112676,
+      "eval/logps/rejected": -1974.1839398734178,
+      "eval/rewards/chosen": -189.60969135123239,
+      "eval/rewards/margins": -18.62818423097923,
+      "eval/rewards/rejected": -170.98150712025316,
       "eval_loss": 0.47333332896232605,
-      "eval_runtime": 140.3068,
-      "eval_samples_per_second": 2.138,
-      "eval_steps_per_second": 0.535,
+      "eval_runtime": 140.7497,
+      "eval_samples_per_second": 2.131,
+      "eval_steps_per_second": 0.533,
       "step": 50
     },
     {
       "epoch": 0.41,
       "grad_norm": 0.0,
-      "learning_rate": 0.00013000000000000002,
-      "loss": 0.4594,
+      "learning_rate": 0.00012285714285714287,
+      "loss": 0.4469,
       "step": 60,
       "train/kl": 0.0,
-      "train/logps/chosen": -2768.1101190476193,
-      "train/logps/rejected": -2565.2698699421967,
-      "train/rewards/chosen": -246.74649234693877,
-      "train/rewards/margins": -19.224545092603506,
-      "train/rewards/rejected": -227.52194725433526
+      "train/logps/chosen": -2463.7814685314684,
+      "train/logps/rejected": -2104.4410310734465,
+      "train/rewards/chosen": -216.41613854895104,
+      "train/rewards/margins": -33.82629479895104,
+      "train/rewards/rejected": -182.58984375
     },
     {
       "epoch": 0.55,
       "grad_norm": 0.0,
-      "learning_rate": 0.00010142857142857143,
-      "loss": 0.4969,
+      "learning_rate": 9.428571428571429e-05,
+      "loss": 0.4281,
       "step": 80,
       "train/kl": 0.0,
-      "train/logps/chosen": -2651.341391509434,
-      "train/logps/rejected": -2455.1451863354037,
-      "train/rewards/chosen": -236.88241941823898,
-      "train/rewards/margins": -19.532070039356995,
-      "train/rewards/rejected": -217.350349378882
+      "train/logps/chosen": -2216.6425638686133,
+      "train/logps/rejected": -2146.411031420765,
+      "train/rewards/chosen": -193.77794251824818,
+      "train/rewards/margins": -6.814528925351993,
+      "train/rewards/rejected": -186.9634135928962
     },
     {
       "epoch": 0.68,
       "grad_norm": 0.0,
-      "learning_rate": 7.285714285714286e-05,
-      "loss": 0.4719,
+      "learning_rate": 6.571428571428571e-05,
+      "loss": 0.5031,
       "step": 100,
       "train/kl": 0.0,
-      "train/logps/chosen": -2469.365894039735,
-      "train/logps/rejected": -2591.453032544379,
-      "train/rewards/chosen": -219.61886899834437,
-      "train/rewards/margins": 9.505113989821325,
-      "train/rewards/rejected": -229.1239829881657
+      "train/logps/chosen": -2453.4375,
+      "train/logps/rejected": -2285.3565251572327,
+      "train/rewards/chosen": -214.367794060559,
+      "train/rewards/margins": -16.124685275181633,
+      "train/rewards/rejected": -198.24310878537736
     },
     {
       "epoch": 0.68,
       "eval/kl": 0.0,
-      "eval/logps/chosen": -2514.4253961267605,
-      "eval/logps/rejected": -2268.7335838607596,
-      "eval/rewards/chosen": -223.0765019806338,
-      "eval/rewards/margins": -22.64017632794392,
-      "eval/rewards/rejected": -200.43632565268987,
+      "eval/logps/chosen": -2181.0286091549297,
+      "eval/logps/rejected": -1975.4440268987341,
+      "eval/rewards/chosen": -189.73686454665494,
+      "eval/rewards/margins": -18.629343779249865,
+      "eval/rewards/rejected": -171.10752076740508,
       "eval_loss": 0.47333332896232605,
-      "eval_runtime": 140.3035,
-      "eval_samples_per_second": 2.138,
-      "eval_steps_per_second": 0.535,
+      "eval_runtime": 140.6959,
+      "eval_samples_per_second": 2.132,
+      "eval_steps_per_second": 0.533,
       "step": 100
     }
   ],

checkpoint-100/training_args.bin

@@ -1,3 +1,3 @@
 version https://git-lfs.github.com/spec/v1
-oid sha256:7225d407f19a33ea970c09fe3b329d792c1f46c9b0cbb091f6418b163d0104ac
+oid sha256:8fb212074cd725db22cb4056e84de5d7117e1aba74d1da2313e006cd0bf7089b
 size 5688