End of training

adapter_config.json

@@ -24,13 +24,13 @@
   "rank_pattern": {},
   "revision": null,
   "target_modules": [
-    "gate_proj",
-    "down_proj",
     "v_proj",
-    "q_proj",
     "o_proj",
     "k_proj",
-    "up_proj"
+    "gate_proj",
+    "up_proj",
+    "q_proj",
+    "down_proj"
   ],
   "task_type": "CAUSAL_LM",
   "trainable_token_indices": null,

adapter_model.safetensors

@@ -1,3 +1,3 @@
 version https://git-lfs.github.com/spec/v1
-oid sha256:232e7e8a88bc8d6a7ad1c9a02dd8ae4c482b7dd03897fe558551bb106f1ff028
+oid sha256:8bad1efde990dbc16a8a08a683016a42d4e555d1877ccfe293dd47291f130a9c
 size 35237104

all_results.json

@@ -1,8 +1,8 @@
 {
     "total_flos": 0.0,
-    "train_loss": 0.0013348124106414617,
-    "train_runtime": 223.6386,
+    "train_loss": 2.7043432623941042e-06,
+    "train_runtime": 967.2338,
     "train_samples": 14,
-    "train_samples_per_second": 1.431,
-    "train_steps_per_second": 0.089
+    "train_samples_per_second": 0.331,
+    "train_steps_per_second": 0.021
 }

train_results.json

@@ -1,8 +1,8 @@
 {
     "total_flos": 0.0,
-    "train_loss": 0.0013348124106414617,
-    "train_runtime": 223.6386,
+    "train_loss": 2.7043432623941042e-06,
+    "train_runtime": 967.2338,
     "train_samples": 14,
-    "train_samples_per_second": 1.431,
-    "train_steps_per_second": 0.089
+    "train_samples_per_second": 0.331,
+    "train_steps_per_second": 0.021
 }

trainer_state.json

@@ -9,203 +9,203 @@
   "is_world_process_zero": true,
   "log_history": [
     {
-      "completion_length": 224.375,
+      "completion_length": 357.59375,
       "epoch": 0.5714285714285714,
-      "grad_norm": 4.6907958984375,
-      "kl": 0.7289290763437748,
+      "grad_norm": 1.550470232963562,
+      "kl": 0.0067819759860867634,
       "learning_rate": 4.965903258506806e-07,
-      "loss": 0.0007,
-      "reward": 2.210346505045891,
-      "reward_std": 1.094197740778327,
-      "rewards/concensus_correctness_reward_func": 0.23987499997019768,
-      "rewards/consensus_reward_func": 0.25,
+      "loss": 0.0,
+      "reward": 0.359556189738214,
+      "reward_std": 0.5430519804358482,
+      "rewards/concensus_correctness_reward_func": 0.015124999918043613,
+      "rewards/consensus_reward_func": 0.0,
       "rewards/cumulative_reward_2": 0.0,
-      "rewards/final_correctness_reward_func": 0.1875,
-      "rewards/question_recreation_reward_func": 0.37503394670784473,
+      "rewards/final_correctness_reward_func": 0.0,
+      "rewards/question_recreation_reward_func": 0.20552494376897812,
       "rewards/soft_format_reward_func": 0.0,
-      "rewards/strict_format_reward_func": 0.21875,
-      "rewards/xmlcount_reward_func": 0.9391875043511391,
+      "rewards/strict_format_reward_func": 0.0,
+      "rewards/xmlcount_reward_func": 0.13890625163912773,
       "step": 2
     },
     {
-      "completion_length": 106.0,
+      "completion_length": 385.2083333333333,
       "epoch": 1.0,
-      "grad_norm": 3.8641750812530518,
-      "kl": 0.3693273241321246,
+      "grad_norm": 1.0296989679336548,
+      "kl": 0.0013993951433803886,
       "learning_rate": 4.698684378016222e-07,
-      "loss": 0.0003,
-      "reward": 3.373856246471405,
-      "reward_std": 2.5780710726976395,
-      "rewards/concensus_correctness_reward_func": 0.992999995748202,
-      "rewards/consensus_reward_func": 0.16666666666666666,
+      "loss": 0.0,
+      "reward": 0.7451420103510221,
+      "reward_std": 0.9848504811525345,
+      "rewards/concensus_correctness_reward_func": 0.010416666666666666,
+      "rewards/consensus_reward_func": 0.0,
       "rewards/cumulative_reward_2": 0.0,
-      "rewards/final_correctness_reward_func": 0.3333333333333333,
-      "rewards/question_recreation_reward_func": 0.5558145766456922,
+      "rewards/final_correctness_reward_func": 0.4166666666666667,
+      "rewards/question_recreation_reward_func": 0.3348920363932848,
       "rewards/soft_format_reward_func": 0.0,
-      "rewards/strict_format_reward_func": 0.25,
-      "rewards/xmlcount_reward_func": 1.075041671593984,
+      "rewards/strict_format_reward_func": 0.0,
+      "rewards/xmlcount_reward_func": -0.01683332274357478,
       "step": 4
     },
     {
-      "completion_length": 190.5,
+      "completion_length": 387.0,
       "epoch": 1.5714285714285714,
-      "grad_norm": 3.4989094734191895,
-      "kl": 1.3430762365460396,
+      "grad_norm": 2.4776289463043213,
+      "kl": 0.004400491248816252,
       "learning_rate": 4.193203929064353e-07,
-      "loss": 0.0013,
-      "reward": 1.9955407828092575,
-      "reward_std": 1.9218887463212013,
-      "rewards/concensus_correctness_reward_func": 0.3619999960064888,
-      "rewards/consensus_reward_func": 0.3125,
+      "loss": 0.0,
+      "reward": 0.3268889347091317,
+      "reward_std": 0.5764277145499364,
+      "rewards/concensus_correctness_reward_func": 0.0,
+      "rewards/consensus_reward_func": 0.0,
       "rewards/cumulative_reward_2": 0.0,
-      "rewards/final_correctness_reward_func": 0.25,
-      "rewards/question_recreation_reward_func": 0.22154081566259265,
-      "rewards/soft_format_reward_func": 0.0,
-      "rewards/strict_format_reward_func": 0.203125,
-      "rewards/xmlcount_reward_func": 0.6463750079274178,
+      "rewards/final_correctness_reward_func": 0.0,
+      "rewards/question_recreation_reward_func": 0.24620142811909318,
+      "rewards/soft_format_reward_func": 0.015625,
+      "rewards/strict_format_reward_func": 0.0,
+      "rewards/xmlcount_reward_func": 0.06506249727681279,
       "step": 6
     },
     {
-      "completion_length": 197.29166666666666,
+      "completion_length": 301.625,
       "epoch": 2.0,
-      "grad_norm": 1.6900211572647095,
-      "kl": 10.67092294494311,
+      "grad_norm": 4.630437850952148,
+      "kl": 0.008956898062024266,
       "learning_rate": 3.5042385616324236e-07,
-      "loss": 0.008,
-      "reward": 2.3348856767018638,
-      "reward_std": 0.928510844707489,
-      "rewards/concensus_correctness_reward_func": 0.15933332840601602,
-      "rewards/consensus_reward_func": 0.16666666666666666,
+      "loss": 0.0,
+      "reward": 0.669816846648852,
+      "reward_std": 0.9645149211088816,
+      "rewards/concensus_correctness_reward_func": 0.0,
+      "rewards/consensus_reward_func": 0.0,
       "rewards/cumulative_reward_2": 0.0,
-      "rewards/final_correctness_reward_func": 0.3333333333333333,
-      "rewards/question_recreation_reward_func": 0.5513440370559692,
+      "rewards/final_correctness_reward_func": 0.16666666666666666,
+      "rewards/question_recreation_reward_func": 0.28952519533534843,
       "rewards/soft_format_reward_func": 0.0,
-      "rewards/strict_format_reward_func": 0.1875,
-      "rewards/xmlcount_reward_func": 0.9367083311080933,
+      "rewards/strict_format_reward_func": 0.0,
+      "rewards/xmlcount_reward_func": 0.21362500016887984,
       "step": 8
     },
     {
-      "completion_length": 176.40625,
+      "completion_length": 360.84375,
       "epoch": 2.571428571428571,
-      "grad_norm": 3.9968183040618896,
-      "kl": 0.845702612772584,
+      "grad_norm": 3.974666118621826,
+      "kl": 0.0014987692411523312,
       "learning_rate": 2.706448363680831e-07,
-      "loss": 0.0008,
-      "reward": 2.57753723859787,
-      "reward_std": 1.3181768357753754,
-      "rewards/concensus_correctness_reward_func": 0.3593749888241291,
-      "rewards/consensus_reward_func": 0.375,
+      "loss": 0.0,
+      "reward": 1.445620215497911,
+      "reward_std": 1.892067939043045,
+      "rewards/concensus_correctness_reward_func": 0.625,
+      "rewards/consensus_reward_func": 0.0625,
       "rewards/cumulative_reward_2": 0.0,
       "rewards/final_correctness_reward_func": 0.25,
-      "rewards/question_recreation_reward_func": 0.40291227400302887,
+      "rewards/question_recreation_reward_func": 0.2774638633709401,
       "rewards/soft_format_reward_func": 0.0,
-      "rewards/strict_format_reward_func": 0.1875,
-      "rewards/xmlcount_reward_func": 1.0027499943971634,
+      "rewards/strict_format_reward_func": 0.0,
+      "rewards/xmlcount_reward_func": 0.23065625689923763,
       "step": 10
     },
     {
-      "completion_length": 157.41666666666666,
+      "completion_length": 430.0833333333333,
       "epoch": 3.0,
-      "grad_norm": 2.858097553253174,
-      "kl": 0.4163093989094098,
+      "grad_norm": 1.2759487628936768,
+      "kl": 0.0015422078043532868,
       "learning_rate": 1.886286282148002e-07,
-      "loss": 0.0003,
-      "reward": 2.235823631286621,
-      "reward_std": 1.7302586088577907,
-      "rewards/concensus_correctness_reward_func": 0.23999999463558197,
-      "rewards/consensus_reward_func": 0.25,
+      "loss": 0.0,
+      "reward": 0.15630086387197176,
+      "reward_std": 0.9415244202439984,
+      "rewards/concensus_correctness_reward_func": 0.0,
+      "rewards/consensus_reward_func": 0.0,
       "rewards/cumulative_reward_2": 0.0,
-      "rewards/final_correctness_reward_func": 0.3333333333333333,
-      "rewards/question_recreation_reward_func": 0.4821569509804249,
+      "rewards/final_correctness_reward_func": 0.08333333333333333,
+      "rewards/question_recreation_reward_func": 0.2368425317108631,
       "rewards/soft_format_reward_func": 0.0,
-      "rewards/strict_format_reward_func": 0.16666666666666666,
-      "rewards/xmlcount_reward_func": 0.7636666595935822,
+      "rewards/strict_format_reward_func": 0.0,
+      "rewards/xmlcount_reward_func": -0.16387500117222467,
       "step": 12
     },
     {
-      "completion_length": 119.46875,
+      "completion_length": 353.65625,
       "epoch": 3.571428571428571,
-      "grad_norm": 3.6504130363464355,
-      "kl": 0.5300018824636936,
+      "grad_norm": 3.5632550716400146,
+      "kl": 0.0016433208002126776,
       "learning_rate": 1.1326296046939333e-07,
-      "loss": 0.0005,
-      "reward": 2.6175443530082703,
-      "reward_std": 1.588691521435976,
-      "rewards/concensus_correctness_reward_func": 0.3619999960064888,
-      "rewards/consensus_reward_func": 0.3125,
+      "loss": 0.0,
+      "reward": 0.6132600959390402,
+      "reward_std": 0.8666674289852381,
+      "rewards/concensus_correctness_reward_func": 0.0,
+      "rewards/consensus_reward_func": 0.0,
       "rewards/cumulative_reward_2": 0.0,
-      "rewards/final_correctness_reward_func": 0.1875,
-      "rewards/question_recreation_reward_func": 0.49226302560418844,
+      "rewards/final_correctness_reward_func": 0.25,
+      "rewards/question_recreation_reward_func": 0.2559788469225168,
       "rewards/soft_format_reward_func": 0.0,
-      "rewards/strict_format_reward_func": 0.265625,
-      "rewards/xmlcount_reward_func": 0.9976562485098839,
+      "rewards/strict_format_reward_func": 0.0,
+      "rewards/xmlcount_reward_func": 0.10728124715387821,
       "step": 14
     },
     {
-      "completion_length": 152.16666666666666,
+      "completion_length": 477.875,
       "epoch": 4.0,
-      "grad_norm": 1.7039144039154053,
-      "kl": 0.3265587662657102,
+      "grad_norm": 2.0089707374572754,
+      "kl": 0.0014073919446673244,
       "learning_rate": 5.271487265090163e-08,
-      "loss": 0.0002,
-      "reward": 2.018751641114553,
-      "reward_std": 1.0230716715256374,
-      "rewards/concensus_correctness_reward_func": 0.07999999821186066,
-      "rewards/consensus_reward_func": 0.08333333333333333,
+      "loss": 0.0,
+      "reward": 0.03820791778465112,
+      "reward_std": 0.6295135350277027,
+      "rewards/concensus_correctness_reward_func": 0.020166666557391483,
+      "rewards/consensus_reward_func": 0.0,
       "rewards/cumulative_reward_2": 0.0,
-      "rewards/final_correctness_reward_func": 0.3333333333333333,
-      "rewards/question_recreation_reward_func": 0.2857933019598325,
+      "rewards/final_correctness_reward_func": 0.0,
+      "rewards/question_recreation_reward_func": 0.166707926740249,
       "rewards/soft_format_reward_func": 0.0,
-      "rewards/strict_format_reward_func": 0.25,
-      "rewards/xmlcount_reward_func": 0.9862916668256124,
+      "rewards/strict_format_reward_func": 0.0,
+      "rewards/xmlcount_reward_func": -0.14866666992505392,
       "step": 16
     },
     {
-      "completion_length": 166.28125,
+      "completion_length": 394.34375,
       "epoch": 4.571428571428571,
-      "grad_norm": 3.133758544921875,
-      "kl": 0.742509638890624,
+      "grad_norm": 2.704482316970825,
+      "kl": 0.0014581629366148263,
       "learning_rate": 1.3545689574841341e-08,
-      "loss": 0.0007,
-      "reward": 2.7243216931819916,
-      "reward_std": 2.6493867076933384,
-      "rewards/concensus_correctness_reward_func": 0.804749995470047,
-      "rewards/consensus_reward_func": 0.1875,
+      "loss": 0.0,
+      "reward": 0.3895467920228839,
+      "reward_std": 0.7705125007778406,
+      "rewards/concensus_correctness_reward_func": 0.014937499538064003,
+      "rewards/consensus_reward_func": 0.0,
       "rewards/cumulative_reward_2": 0.0,
-      "rewards/final_correctness_reward_func": 0.25,
-      "rewards/question_recreation_reward_func": 0.4959466829895973,
+      "rewards/final_correctness_reward_func": 0.125,
+      "rewards/question_recreation_reward_func": 0.30085928039625287,
       "rewards/soft_format_reward_func": 0.0,
-      "rewards/strict_format_reward_func": 0.25,
-      "rewards/xmlcount_reward_func": 0.7361249998211861,
+      "rewards/strict_format_reward_func": 0.0,
+      "rewards/xmlcount_reward_func": -0.051249995827674866,
       "step": 18
     },
     {
-      "completion_length": 182.70833333333334,
+      "completion_length": 338.4583333333333,
       "epoch": 5.0,
-      "grad_norm": 1.9381123781204224,
-      "kl": 0.42725201696157455,
+      "grad_norm": 1.2185847759246826,
+      "kl": 0.0017583037454945345,
       "learning_rate": 0.0,
-      "loss": 0.0003,
-      "reward": 2.9027516742547355,
-      "reward_std": 2.645435392856598,
-      "rewards/concensus_correctness_reward_func": 0.9129999975363413,
-      "rewards/consensus_reward_func": 0.08333333333333333,
+      "loss": 0.0,
+      "reward": 2.36764890452226,
+      "reward_std": 2.5429103871186576,
+      "rewards/concensus_correctness_reward_func": 1.6666666666666667,
+      "rewards/consensus_reward_func": 0.16666666666666666,
       "rewards/cumulative_reward_2": 0.0,
-      "rewards/final_correctness_reward_func": 0.3333333333333333,
-      "rewards/question_recreation_reward_func": 0.3405850703517596,
+      "rewards/final_correctness_reward_func": 0.16666666666666666,
+      "rewards/question_recreation_reward_func": 0.21544045334060988,
       "rewards/soft_format_reward_func": 0.0,
-      "rewards/strict_format_reward_func": 0.25,
-      "rewards/xmlcount_reward_func": 0.9824999968210856,
+      "rewards/strict_format_reward_func": 0.0,
+      "rewards/xmlcount_reward_func": 0.15220833321412405,
       "step": 20
     },
     {
       "epoch": 5.0,
       "step": 20,
       "total_flos": 0.0,
-      "train_loss": 0.0013348124106414617,
-      "train_runtime": 223.6386,
-      "train_samples_per_second": 1.431,
-      "train_steps_per_second": 0.089
+      "train_loss": 2.7043432623941042e-06,
+      "train_runtime": 967.2338,
+      "train_samples_per_second": 0.331,
+      "train_steps_per_second": 0.021
     }
   ],
   "logging_steps": 2,