Training in progress, step 30500, checkpoint

last-checkpoint/adapter_model.safetensors

@@ -1,3 +1,3 @@
 version https://git-lfs.github.com/spec/v1
-oid sha256:c51c95b62fe64237a3070b38732becb17c727e262f2a166cf9b540a95a07ffa4
+oid sha256:8f72bd5a3c71ee7c0378f6c13844dc64c89949a7e7e1e07f8933751b8df932c7
 size 132187888

last-checkpoint/ar_diffusion_info.bin

@@ -1,3 +1,3 @@
 version https://git-lfs.github.com/spec/v1
-oid sha256:c956420b50cd1beed5b84ba3d1b039f88d0f55421c32ed697563eb48dbe3e467
+oid sha256:e85124ad974e9aa6591318b973e27a8c6a21355c73a2b688604b07888ca77bfd
 size 1800

last-checkpoint/optimizer.pt

@@ -1,3 +1,3 @@
 version https://git-lfs.github.com/spec/v1
-oid sha256:7a6ad13ae7222893577b884130d3ce0bc936a3a1c24d4768f253e6c8a17784ab
+oid sha256:1bfbd83026836fb7e11711c7d54e9b00d7e03835d945121611d08eb2713a80bd
 size 264665786

last-checkpoint/rng_state.pth

@@ -1,3 +1,3 @@
 version https://git-lfs.github.com/spec/v1
-oid sha256:072a1f78d2c87d7bd95194d5360ec258a354ea618d79e75e77fefc7224373d1f
+oid sha256:7e898de494648db9b26425188387425acd19b2117f233464d12d3be9a3ed8c13
 size 14244

last-checkpoint/scaler.pt

@@ -1,3 +1,3 @@
 version https://git-lfs.github.com/spec/v1
-oid sha256:db92c81e1d1adfede0ee8a02a6941b37c6dd7a52f54da9d0ece7ad64c5439504
+oid sha256:851ff73959206b5bf637351b456e6acc61cd20020d42c449aed8671d33f868f2
 size 988

last-checkpoint/scheduler.pt

@@ -1,3 +1,3 @@
 version https://git-lfs.github.com/spec/v1
-oid sha256:961ccdacdfed72cc3f6e549530202fa2bdca81a4b4bb1d566331261ea797f82d
+oid sha256:52513f6a22920e0a6a384b951f3cbc4c30d39da66d57f762b33f885b2ce6ee0e
 size 1064

last-checkpoint/trainer_state.json

@@ -1,10 +1,10 @@
 {
-  "best_global_step": 30000,
-  "best_metric": 0.6152763962745667,
-  "best_model_checkpoint": "./qwen3-4b-ar-diffusion-checkpoints/checkpoint-30000",
-  "epoch": 2.307514806553342,
+  "best_global_step": 30500,
+  "best_metric": 0.6086920499801636,
+  "best_model_checkpoint": "./qwen3-4b-ar-diffusion-checkpoints/checkpoint-30500",
+  "epoch": 2.3459733866625645,
   "eval_steps": 250,
-  "global_step": 30000,
+  "global_step": 30500,
   "is_hyper_param_search": false,
   "is_local_process_zero": true,
   "is_world_process_zero": true,
@@ -5168,6 +5168,92 @@
       "eval_samples_per_second": 22.296,
       "eval_steps_per_second": 5.574,
       "step": 30000
+    },
+    {
+      "epoch": 2.311360664564264,
+      "grad_norm": 0.9155976176261902,
+      "learning_rate": 4.6572994312131526e-05,
+      "loss": 0.6241,
+      "step": 30050
+    },
+    {
+      "epoch": 2.3152065225751866,
+      "grad_norm": 0.861916720867157,
+      "learning_rate": 4.6313274290315045e-05,
+      "loss": 0.6359,
+      "step": 30100
+    },
+    {
+      "epoch": 2.3190523805861085,
+      "grad_norm": 0.9024108052253723,
+      "learning_rate": 4.605355426849856e-05,
+      "loss": 0.6343,
+      "step": 30150
+    },
+    {
+      "epoch": 2.322898238597031,
+      "grad_norm": 1.2562297582626343,
+      "learning_rate": 4.579383424668208e-05,
+      "loss": 0.5956,
+      "step": 30200
+    },
+    {
+      "epoch": 2.3267440966079533,
+      "grad_norm": 0.7998838424682617,
+      "learning_rate": 4.55341142248656e-05,
+      "loss": 0.6162,
+      "step": 30250
+    },
+    {
+      "epoch": 2.3267440966079533,
+      "eval_loss": 0.6139496564865112,
+      "eval_runtime": 21.4503,
+      "eval_samples_per_second": 23.31,
+      "eval_steps_per_second": 5.827,
+      "step": 30250
+    },
+    {
+      "epoch": 2.3305899546188753,
+      "grad_norm": 0.6242274641990662,
+      "learning_rate": 4.527439420304912e-05,
+      "loss": 0.5912,
+      "step": 30300
+    },
+    {
+      "epoch": 2.3344358126297977,
+      "grad_norm": 1.1920520067214966,
+      "learning_rate": 4.5014674181232637e-05,
+      "loss": 0.6258,
+      "step": 30350
+    },
+    {
+      "epoch": 2.33828167064072,
+      "grad_norm": 1.370483160018921,
+      "learning_rate": 4.475495415941615e-05,
+      "loss": 0.6229,
+      "step": 30400
+    },
+    {
+      "epoch": 2.342127528651642,
+      "grad_norm": 1.0605512857437134,
+      "learning_rate": 4.449523413759967e-05,
+      "loss": 0.6621,
+      "step": 30450
+    },
+    {
+      "epoch": 2.3459733866625645,
+      "grad_norm": 0.874275803565979,
+      "learning_rate": 4.423551411578319e-05,
+      "loss": 0.6004,
+      "step": 30500
+    },
+    {
+      "epoch": 2.3459733866625645,
+      "eval_loss": 0.6086920499801636,
+      "eval_runtime": 22.3525,
+      "eval_samples_per_second": 22.369,
+      "eval_steps_per_second": 5.592,
+      "step": 30500
     }
   ],
   "logging_steps": 50,