sft-6900-step

adapter_model.safetensors

@@ -1,3 +1,3 @@
 version https://git-lfs.github.com/spec/v1
-oid sha256:5afe2a027d395f551dd07fb95d989917901b4fc4d9a3f699bcf96a1a0c52045c
+oid sha256:a2ffc6552dcc71c20350838f8a219506181ba46c38f659ec852bba2bddda4dfc
 size 698419728

optimizer.pt

@@ -1,3 +1,3 @@
 version https://git-lfs.github.com/spec/v1
-oid sha256:fc2f55384d96aef8c36f040c4c76bba8c886457259f6de979d38405706b46bfd
+oid sha256:42b8045c238ff4f74f9e3fe7c94d27857f2fadb3c697f6661bb73fb9bb04a576
 size 1397136587

rng_state.pth

@@ -1,3 +1,3 @@
 version https://git-lfs.github.com/spec/v1
-oid sha256:14f4ff77bbbe606f2785ea0f04a2535a7c54901e7091c0fed26a6e7b85eda9ae
+oid sha256:be73d303d67b9e1d37ae52f58cd2c7c7c5aeb597a44b9c72b8875cd9acb7be14
 size 14645

scheduler.pt

@@ -1,3 +1,3 @@
 version https://git-lfs.github.com/spec/v1
-oid sha256:08804d0a21d8df191a267fdfad60532380afcf3b04182506e89486e7b012afc5
+oid sha256:fff36b79323a1d28015f7255a86b9e604fcdba024c3e96bcdca5e4c7054b0293
 size 1465

trainer_state.json

@@ -1,10 +1,10 @@
 {
-  "best_global_step": 6700,
-  "best_metric": 1.140816330909729,
-  "best_model_checkpoint": "/workspace/project_2026_1/checkpoints/sft/checkpoint-6700",
-  "epoch": 1.9711679905854664,
+  "best_global_step": 6800,
+  "best_metric": 1.1395292282104492,
+  "best_model_checkpoint": "/workspace/project_2026_1/checkpoints/sft/checkpoint-6800",
+  "epoch": 2.030008826125331,
   "eval_steps": 100,
-  "global_step": 6700,
+  "global_step": 6900,
   "is_hyper_param_search": false,
   "is_local_process_zero": true,
   "is_world_process_zero": true,
@@ -7445,6 +7445,228 @@
       "eval_samples_per_second": 26.029,
       "eval_steps_per_second": 3.257,
       "step": 6700
+    },
+    {
+      "entropy": 0.9509491920471191,
+      "epoch": 1.9741100323624594,
+      "grad_norm": 0.6006605625152588,
+      "learning_rate": 5.744202611276379e-05,
+      "loss": 0.9503057479858399,
+      "mean_token_accuracy": 0.7786516189575196,
+      "num_tokens": 27097949.0,
+      "step": 6710
+    },
+    {
+      "entropy": 1.0138035595417023,
+      "epoch": 1.9770520741394528,
+      "grad_norm": 0.5902991890907288,
+      "learning_rate": 5.7148775369783694e-05,
+      "loss": 1.0296749114990233,
+      "mean_token_accuracy": 0.7590757310390472,
+      "num_tokens": 27138453.0,
+      "step": 6720
+    },
+    {
+      "entropy": 0.933520519733429,
+      "epoch": 1.979994115916446,
+      "grad_norm": 0.5484936833381653,
+      "learning_rate": 5.685597532311455e-05,
+      "loss": 0.957374095916748,
+      "mean_token_accuracy": 0.7793904483318329,
+      "num_tokens": 27178805.0,
+      "step": 6730
+    },
+    {
+      "entropy": 0.9440324783325196,
+      "epoch": 1.9829361576934392,
+      "grad_norm": 0.5826029777526855,
+      "learning_rate": 5.656362905233923e-05,
+      "loss": 0.9262220382690429,
+      "mean_token_accuracy": 0.7845340669155121,
+      "num_tokens": 27219347.0,
+      "step": 6740
+    },
+    {
+      "entropy": 0.9071877419948577,
+      "epoch": 1.9858781994704324,
+      "grad_norm": 0.5721964836120605,
+      "learning_rate": 5.6271739632268094e-05,
+      "loss": 0.9060114860534668,
+      "mean_token_accuracy": 0.7890908360481262,
+      "num_tokens": 27258890.0,
+      "step": 6750
+    },
+    {
+      "entropy": 0.9562793612480164,
+      "epoch": 1.9888202412474256,
+      "grad_norm": 0.614380955696106,
+      "learning_rate": 5.598031013290631e-05,
+      "loss": 0.9876157760620117,
+      "mean_token_accuracy": 0.768429833650589,
+      "num_tokens": 27299053.0,
+      "step": 6760
+    },
+    {
+      "entropy": 0.9924969553947449,
+      "epoch": 1.991762283024419,
+      "grad_norm": 0.6030513644218445,
+      "learning_rate": 5.5689343619421906e-05,
+      "loss": 0.9977625846862793,
+      "mean_token_accuracy": 0.7658666670322418,
+      "num_tokens": 27339515.0,
+      "step": 6770
+    },
+    {
+      "entropy": 0.9534170269966126,
+      "epoch": 1.994704324801412,
+      "grad_norm": 0.5039950609207153,
+      "learning_rate": 5.539884315211321e-05,
+      "loss": 0.9545814514160156,
+      "mean_token_accuracy": 0.7779964745044708,
+      "num_tokens": 27379693.0,
+      "step": 6780
+    },
+    {
+      "entropy": 0.9789716601371765,
+      "epoch": 1.9976463665784054,
+      "grad_norm": 0.5822030305862427,
+      "learning_rate": 5.5108811786376925e-05,
+      "loss": 0.9928366661071777,
+      "mean_token_accuracy": 0.7682704031467438,
+      "num_tokens": 27419734.0,
+      "step": 6790
+    },
+    {
+      "entropy": 0.915216040611267,
+      "epoch": 2.000588408355399,
+      "grad_norm": 0.4654218554496765,
+      "learning_rate": 5.481925257267589e-05,
+      "loss": 0.8871613502502441,
+      "mean_token_accuracy": 0.7920856356620789,
+      "num_tokens": 27458303.0,
+      "step": 6800
+    },
+    {
+      "epoch": 2.000588408355399,
+      "eval_entropy": 0.9942471109663095,
+      "eval_loss": 1.1395292282104492,
+      "eval_mean_token_accuracy": 0.7511763375575148,
+      "eval_num_tokens": 27458303.0,
+      "eval_runtime": 116.8845,
+      "eval_samples_per_second": 26.051,
+      "eval_steps_per_second": 3.26,
+      "step": 6800
+    },
+    {
+      "entropy": 0.7543269693851471,
+      "epoch": 2.003530450132392,
+      "grad_norm": 0.6209985613822937,
+      "learning_rate": 5.4530168556506875e-05,
+      "loss": 0.6749869823455811,
+      "mean_token_accuracy": 0.8347735464572906,
+      "num_tokens": 27498607.0,
+      "step": 6810
+    },
+    {
+      "entropy": 0.6835850536823272,
+      "epoch": 2.0064724919093853,
+      "grad_norm": 0.781541109085083,
+      "learning_rate": 5.424156277836881e-05,
+      "loss": 0.6951170921325683,
+      "mean_token_accuracy": 0.8288436651229858,
+      "num_tokens": 27538904.0,
+      "step": 6820
+    },
+    {
+      "entropy": 0.6437631964683532,
+      "epoch": 2.0094145336863782,
+      "grad_norm": 0.8998324871063232,
+      "learning_rate": 5.395343827373053e-05,
+      "loss": 0.6296420574188233,
+      "mean_token_accuracy": 0.8461188077926636,
+      "num_tokens": 27579223.0,
+      "step": 6830
+    },
+    {
+      "entropy": 0.6127074956893921,
+      "epoch": 2.0123565754633717,
+      "grad_norm": 0.6167740225791931,
+      "learning_rate": 5.366579807299909e-05,
+      "loss": 0.5965664386749268,
+      "mean_token_accuracy": 0.850104957818985,
+      "num_tokens": 27619638.0,
+      "step": 6840
+    },
+    {
+      "entropy": 0.6964607417583466,
+      "epoch": 2.0152986172403646,
+      "grad_norm": 0.637476921081543,
+      "learning_rate": 5.337864520148768e-05,
+      "loss": 0.6968545913696289,
+      "mean_token_accuracy": 0.8300110459327698,
+      "num_tokens": 27660158.0,
+      "step": 6850
+    },
+    {
+      "entropy": 0.6738093435764313,
+      "epoch": 2.018240659017358,
+      "grad_norm": 0.7894798517227173,
+      "learning_rate": 5.309198267938402e-05,
+      "loss": 0.6670093059539794,
+      "mean_token_accuracy": 0.8377935826778412,
+      "num_tokens": 27700212.0,
+      "step": 6860
+    },
+    {
+      "entropy": 0.6280623555183411,
+      "epoch": 2.0211827007943515,
+      "grad_norm": 0.80244380235672,
+      "learning_rate": 5.280581352171836e-05,
+      "loss": 0.6267249107360839,
+      "mean_token_accuracy": 0.8437743067741394,
+      "num_tokens": 27740554.0,
+      "step": 6870
+    },
+    {
+      "entropy": 0.6882079899311065,
+      "epoch": 2.0241247425713444,
+      "grad_norm": 0.7488958835601807,
+      "learning_rate": 5.2520140738332025e-05,
+      "loss": 0.6897297382354737,
+      "mean_token_accuracy": 0.8309988558292389,
+      "num_tokens": 27781034.0,
+      "step": 6880
+    },
+    {
+      "entropy": 0.676528149843216,
+      "epoch": 2.027066784348338,
+      "grad_norm": 0.8301676511764526,
+      "learning_rate": 5.2234967333845466e-05,
+      "loss": 0.6622447490692138,
+      "mean_token_accuracy": 0.8345989942550659,
+      "num_tokens": 27821579.0,
+      "step": 6890
+    },
+    {
+      "entropy": 0.6388787865638733,
+      "epoch": 2.030008826125331,
+      "grad_norm": 0.7029614448547363,
+      "learning_rate": 5.1950296307626956e-05,
+      "loss": 0.6487605571746826,
+      "mean_token_accuracy": 0.8400563955307007,
+      "num_tokens": 27861899.0,
+      "step": 6900
+    },
+    {
+      "epoch": 2.030008826125331,
+      "eval_entropy": 0.8397969613707285,
+      "eval_loss": 1.2236672639846802,
+      "eval_mean_token_accuracy": 0.7469185830101254,
+      "eval_num_tokens": 27861899.0,
+      "eval_runtime": 116.8259,
+      "eval_samples_per_second": 26.064,
+      "eval_steps_per_second": 3.261,
+      "step": 6900
     }
   ],
   "logging_steps": 10,
@@ -7464,7 +7686,7 @@
       "attributes": {}
     }
   },
-  "total_flos": 1.2570357073262346e+18,
+  "total_flos": 1.2944070017481708e+18,
   "train_batch_size": 8,
   "trial_name": null,
   "trial_params": null