Training in progress, step 12, checkpoint

checkpoint-12/adapter_config.json

@@ -16,24 +16,24 @@
   "layers_pattern": null,
   "layers_to_transform": null,
   "loftq_config": {},
-  "lora_alpha": 8,
+  "lora_alpha": 16,
   "lora_bias": false,
   "lora_dropout": 0.0,
   "megatron_config": null,
   "megatron_core": "megatron.core",
   "modules_to_save": null,
   "peft_type": "LORA",
-  "r": 8,
+  "r": 16,
   "rank_pattern": {},
   "revision": null,
   "target_modules": [
-    "down_proj",
+    "up_proj",
+    "k_proj",
     "o_proj",
-    "v_proj",
     "gate_proj",
-    "q_proj",
-    "up_proj",
-    "k_proj"
+    "v_proj",
+    "down_proj",
+    "q_proj"
   ],
   "task_type": null,
   "trainable_token_indices": null,

checkpoint-12/adapter_model.safetensors

@@ -1,3 +1,3 @@
 version https://git-lfs.github.com/spec/v1
-oid sha256:851f4b0fe956eac509642f70fd14815493e76066844d87f487c1d0ec98575524
-size 8841928
+oid sha256:b49989a6af1fef377ce007ebbb3081e857efdd9ac5ed32f0572b399f54638528
+size 17640808

checkpoint-12/global_step12/bf16_zero_pp_rank_0_mp_rank_00_optim_states.pt

@@ -1,3 +1,3 @@
 version https://git-lfs.github.com/spec/v1
-oid sha256:78a75e2ea5a78d36924f5dad07d8c356999b5449bd6dce79c735da1d4208e705
-size 52792816
+oid sha256:deb87890f0d8546ed6dd84ca0c41ac09657443cf10153ed506ca3b47dd93d8f1
+size 105582064

checkpoint-12/global_step12/zero_pp_rank_0_mp_rank_00_model_states.pt

@@ -1,3 +1,3 @@
 version https://git-lfs.github.com/spec/v1
-oid sha256:ba10126db40fa1da3e0640e6c7a9549056758e44fb6c2aa6c123abc0b8ef91d7
+oid sha256:d588e76fb3c7b7eedb51521f9c49f616d9d8ac5c2c9cfc7d8e22297a7c040817
 size 332955

checkpoint-12/trainer_state.json

@@ -10,13 +10,13 @@
   "log_history": [
     {
       "clip_ratio": 0.0,
-      "completion_length": 598.6875,
+      "completion_length": 576.9375,
       "epoch": 0.0002457304337142155,
       "grad_norm": 0.0,
       "kl": 0.0,
       "learning_rate": 5e-07,
       "loss": 0.0,
-      "num_tokens": 12883.0,
+      "num_tokens": 12535.0,
       "reward": 0.0,
       "reward_std": 0.0,
       "rewards/reward_func": 0.0,
@@ -24,13 +24,13 @@
     },
     {
       "clip_ratio": 0.0,
-      "completion_length": 579.75,
+      "completion_length": 879.0625,
       "epoch": 0.000491460867428431,
-      "grad_norm": 0.057158572334860595,
+      "grad_norm": 0.11151622953359844,
       "kl": 0.0,
       "learning_rate": 5e-07,
-      "loss": 0.0991,
-      "num_tokens": 23183.0,
+      "loss": 0.0776,
+      "num_tokens": 27624.0,
       "reward": 0.125,
       "reward_std": 0.2314550280570984,
       "rewards/reward_func": 0.125,
@@ -38,13 +38,13 @@
     },
     {
       "clip_ratio": 0.0,
-      "completion_length": 712.5625,
+      "completion_length": 923.75,
       "epoch": 0.0007371913011426465,
-      "grad_norm": 9.369531689325602e-07,
-      "kl": 0.00038623809814453125,
+      "grad_norm": 1.7342617962764825e-06,
+      "kl": 0.0003814697265625,
       "learning_rate": 5e-07,
       "loss": 0.0,
-      "num_tokens": 36072.0,
+      "num_tokens": 43892.0,
       "reward": 0.0,
       "reward_std": 0.0,
       "rewards/reward_func": 0.0,
@@ -52,69 +52,69 @@
     },
     {
       "clip_ratio": 0.0,
-      "completion_length": 578.1875,
+      "completion_length": 682.9375,
       "epoch": 0.000982921734856862,
-      "grad_norm": 0.1134692235444953,
-      "kl": 0.00043392181396484375,
+      "grad_norm": 0.1605198642986145,
+      "kl": 0.00042629241943359375,
       "learning_rate": 5e-07,
-      "loss": -0.0204,
-      "num_tokens": 46523.0,
-      "reward": 0.0625,
-      "reward_std": 0.1767766922712326,
-      "rewards/reward_func": 0.0625,
+      "loss": -0.1736,
+      "num_tokens": 56019.0,
+      "reward": 0.1875,
+      "reward_std": 0.408231720328331,
+      "rewards/reward_func": 0.1875,
       "step": 4
     },
     {
       "clip_ratio": 0.0,
-      "completion_length": 708.0625,
+      "completion_length": 729.5,
       "epoch": 0.0012286521685710775,
-      "grad_norm": 0.13726250703666534,
-      "kl": 0.0003414154052734375,
+      "grad_norm": 0.15191784640809727,
+      "kl": 0.00035190582275390625,
       "learning_rate": 5e-07,
-      "loss": -0.0446,
-      "num_tokens": 59748.0,
-      "reward": 0.125,
-      "reward_std": 0.2314550280570984,
-      "rewards/reward_func": 0.125,
+      "loss": 0.1212,
+      "num_tokens": 69587.0,
+      "reward": 0.0625,
+      "reward_std": 0.1767766922712326,
+      "rewards/reward_func": 0.0625,
       "step": 5
     },
     {
       "clip_ratio": 0.0,
-      "completion_length": 351.9375,
+      "completion_length": 680.3125,
       "epoch": 0.001474382602285293,
-      "grad_norm": 0.13704685404012662,
-      "kl": 0.00045680999755859375,
+      "grad_norm": 0.00010216089929885942,
+      "kl": 0.0011739730834960938,
       "learning_rate": 5e-07,
-      "loss": -0.0854,
-      "num_tokens": 66971.0,
-      "reward": 0.125,
-      "reward_std": 0.2314550280570984,
-      "rewards/reward_func": 0.125,
+      "loss": 0.0,
+      "num_tokens": 82064.0,
+      "reward": 0.0,
+      "reward_std": 0.0,
+      "rewards/reward_func": 0.0,
       "step": 6
     },
     {
       "clip_ratio": 0.0,
-      "completion_length": 670.375,
+      "completion_length": 723.1875,
       "epoch": 0.0017201130359995086,
-      "grad_norm": 0.05810748919014082,
-      "kl": 0.000301361083984375,
+      "grad_norm": 0.053916157377856225,
+      "kl": 0.00029850006103515625,
       "learning_rate": 5e-07,
-      "loss": 0.0394,
-      "num_tokens": 80449.0,
-      "reward": 0.125,
-      "reward_std": 0.2314550280570984,
-      "rewards/reward_func": 0.125,
+      "loss": 0.0843,
+      "num_tokens": 96387.0,
+      "reward": 0.0625,
+      "reward_std": 0.1767766922712326,
+      "rewards/reward_func": 0.0625,
       "step": 7
     },
     {
       "clip_ratio": 0.0,
-      "completion_length": 797.375,
+      "completion_length": 646.1875,
       "epoch": 0.001965843469713724,
-      "grad_norm": 0.09117945345657646,
-      "kl": 0.0003681182861328125,
+      "grad_norm": 0.16676206881832856,
+      "kl": 0.0003566741943359375,
       "learning_rate": 5e-07,
-      "loss": 0.0619,
-      "num_tokens": 95399.0,
+      "loss": 0.0941,
+      "num_tokens": 108918.0,
       "reward": 0.25,
       "reward_std": 0.4355512708425522,
       "rewards/reward_func": 0.25,
@@ -122,41 +122,41 @@
     },
     {
       "clip_ratio": 0.0,
-      "completion_length": 455.1875,
+      "completion_length": 605.875,
       "epoch": 0.0022115739034279398,
-      "grad_norm": 0.059279116418829485,
-      "kl": 0.0003509521484375,
+      "grad_norm": 1.5988321121918276e-06,
+      "kl": 0.0003833770751953125,
       "learning_rate": 5e-07,
-      "loss": 0.0553,
-      "num_tokens": 104498.0,
-      "reward": 0.125,
-      "reward_std": 0.2314550280570984,
-      "rewards/reward_func": 0.125,
+      "loss": 0.0,
+      "num_tokens": 120428.0,
+      "reward": 0.0,
+      "reward_std": 0.0,
+      "rewards/reward_func": 0.0,
       "step": 9
     },
     {
       "clip_ratio": 0.0,
-      "completion_length": 569.25,
+      "completion_length": 551.4375,
       "epoch": 0.002457304337142155,
-      "grad_norm": 0.11258647371214518,
-      "kl": 0.0002951622009277344,
+      "grad_norm": 0.16315342243657902,
+      "kl": 0.00034236907958984375,
       "learning_rate": 5e-07,
-      "loss": -0.118,
-      "num_tokens": 114646.0,
-      "reward": 0.1875,
-      "reward_std": 0.2587745785713196,
-      "rewards/reward_func": 0.1875,
+      "loss": -0.0335,
+      "num_tokens": 130291.0,
+      "reward": 0.0625,
+      "reward_std": 0.1767766922712326,
+      "rewards/reward_func": 0.0625,
       "step": 10
     },
     {
       "clip_ratio": 0.0,
-      "completion_length": 461.0625,
+      "completion_length": 757.3125,
       "epoch": 0.0027030347708563705,
-      "grad_norm": 1.7798072854936817e-06,
-      "kl": 0.0003662109375,
+      "grad_norm": 1.5633163558139792e-06,
+      "kl": 0.00029850006103515625,
       "learning_rate": 5e-07,
       "loss": 0.0,
-      "num_tokens": 128663.0,
+      "num_tokens": 149048.0,
       "reward": 0.0,
       "reward_std": 0.0,
       "rewards/reward_func": 0.0,
@@ -164,16 +164,16 @@
     },
     {
       "clip_ratio": 0.0,
-      "completion_length": 378.375,
+      "completion_length": 766.375,
       "epoch": 0.002948765204570586,
-      "grad_norm": 0.09280689232901017,
-      "kl": 0.0002918243408203125,
+      "grad_norm": 0.08173214068073498,
+      "kl": 0.00022268295288085938,
       "learning_rate": 5e-07,
-      "loss": -0.1028,
-      "num_tokens": 135957.0,
-      "reward": 0.1875,
-      "reward_std": 0.2587745785713196,
-      "rewards/reward_func": 0.1875,
+      "loss": 0.1401,
+      "num_tokens": 162550.0,
+      "reward": 0.125,
+      "reward_std": 0.2314550280570984,
+      "rewards/reward_func": 0.125,
       "step": 12
     }
   ],

checkpoint-12/training_args.bin

@@ -1,3 +1,3 @@
 version https://git-lfs.github.com/spec/v1
-oid sha256:fff5d866f3f58eded13df4e5d0f17d86b3f86cb9253959e5d718e586d0617838
+oid sha256:af3974b5d650bad10616ac85d5be43a43be3a632e9e6ca9838e55c33bd2fac4e
 size 7544