Training in progress, step 32500, checkpoint

last-checkpoint/adapter_model.safetensors

@@ -1,3 +1,3 @@
 version https://git-lfs.github.com/spec/v1
-oid sha256:585c793fcb177f73ccea5d9e03f556c9367be9b24af17619c9336aa9784900a3
+oid sha256:dc40c7322710674ba700e7c912ed88ea4428f9c273bc8666babf624928031d55
 size 132187888

last-checkpoint/ar_diffusion_info.bin

@@ -1,3 +1,3 @@
 version https://git-lfs.github.com/spec/v1
-oid sha256:70a55c79abf48cb04d5f4a5872743d51e805cb5cf801861bd90f35f89319726c
+oid sha256:2879e5076d2053be09a0d6c73abd26b4278e5f16685fdedb148edcf4a65b4069
 size 1800

last-checkpoint/optimizer.pt

@@ -1,3 +1,3 @@
 version https://git-lfs.github.com/spec/v1
-oid sha256:0f58de48eab208cdcbb773a7a3efa7a8cc66fe9d9df03bf3d9123033730c12d8
+oid sha256:32736e80b881463272e4426f17a24ebdfddd47b5e2e3f5b99b633b357b80f7e8
 size 264665786

last-checkpoint/rng_state.pth

@@ -1,3 +1,3 @@
 version https://git-lfs.github.com/spec/v1
-oid sha256:36fd221e5d40799da39ad4f0eb8cd2d24c660462984d46b022ea86be7e548d3b
-size 14244
+oid sha256:dc5a5cccd951497473f5f616bdc9028359460cced3c38458214cef6f531dc696
+size 14308

last-checkpoint/scaler.pt

@@ -1,3 +1,3 @@
 version https://git-lfs.github.com/spec/v1
-oid sha256:7daafea84c2a1039aac52712a160f9291aa4d9b6933d9afeee1976e296f5fb3a
+oid sha256:9d6b59feadfd14316e1171dfc59fefcc9256bf9fc9745d2b33888fd1da1d95ea
 size 988

last-checkpoint/scheduler.pt

@@ -1,3 +1,3 @@
 version https://git-lfs.github.com/spec/v1
-oid sha256:2f288b947beb52c2ae256c7cb95b541005fd607b063ecb73ec691524e7e48bb2
+oid sha256:eb4a1da3914bf9e57fd6bbf0cd99b2cfb4060f68033db5609e17e853bacf6b40
 size 1064

last-checkpoint/trainer_state.json

@@ -2,9 +2,9 @@
   "best_global_step": 31000,
   "best_metric": 0.6043956279754639,
   "best_model_checkpoint": "./qwen3-4b-ar-diffusion-checkpoints/checkpoint-31000",
-  "epoch": 2.4613491269902315,
+  "epoch": 2.4998077070994538,
   "eval_steps": 250,
-  "global_step": 32000,
+  "global_step": 32500,
   "is_hyper_param_search": false,
   "is_local_process_zero": true,
   "is_world_process_zero": true,
@@ -5512,6 +5512,92 @@
       "eval_samples_per_second": 22.513,
       "eval_steps_per_second": 5.628,
       "step": 32000
+    },
+    {
+      "epoch": 2.465194985001154,
+      "grad_norm": 0.9351600408554077,
+      "learning_rate": 3.618938783990858e-05,
+      "loss": 0.6149,
+      "step": 32050
+    },
+    {
+      "epoch": 2.469040843012076,
+      "grad_norm": 1.4392043352127075,
+      "learning_rate": 3.59296678180921e-05,
+      "loss": 0.6133,
+      "step": 32100
+    },
+    {
+      "epoch": 2.4728867010229982,
+      "grad_norm": 1.143561601638794,
+      "learning_rate": 3.566994779627561e-05,
+      "loss": 0.6284,
+      "step": 32150
+    },
+    {
+      "epoch": 2.4767325590339206,
+      "grad_norm": 0.5455169081687927,
+      "learning_rate": 3.541022777445914e-05,
+      "loss": 0.6546,
+      "step": 32200
+    },
+    {
+      "epoch": 2.4805784170448426,
+      "grad_norm": 0.9387330412864685,
+      "learning_rate": 3.515050775264265e-05,
+      "loss": 0.6164,
+      "step": 32250
+    },
+    {
+      "epoch": 2.4805784170448426,
+      "eval_loss": 0.6368865966796875,
+      "eval_runtime": 21.3563,
+      "eval_samples_per_second": 23.412,
+      "eval_steps_per_second": 5.853,
+      "step": 32250
+    },
+    {
+      "epoch": 2.484424275055765,
+      "grad_norm": 0.7866175770759583,
+      "learning_rate": 3.489078773082617e-05,
+      "loss": 0.6219,
+      "step": 32300
+    },
+    {
+      "epoch": 2.488270133066687,
+      "grad_norm": 0.870444655418396,
+      "learning_rate": 3.463106770900969e-05,
+      "loss": 0.5986,
+      "step": 32350
+    },
+    {
+      "epoch": 2.4921159910776094,
+      "grad_norm": 1.548135757446289,
+      "learning_rate": 3.4371347687193205e-05,
+      "loss": 0.6284,
+      "step": 32400
+    },
+    {
+      "epoch": 2.495961849088532,
+      "grad_norm": 1.0319703817367554,
+      "learning_rate": 3.4111627665376724e-05,
+      "loss": 0.6055,
+      "step": 32450
+    },
+    {
+      "epoch": 2.4998077070994538,
+      "grad_norm": 1.011017918586731,
+      "learning_rate": 3.3851907643560244e-05,
+      "loss": 0.6341,
+      "step": 32500
+    },
+    {
+      "epoch": 2.4998077070994538,
+      "eval_loss": 0.6336821913719177,
+      "eval_runtime": 22.1686,
+      "eval_samples_per_second": 22.554,
+      "eval_steps_per_second": 5.639,
+      "step": 32500
     }
   ],
   "logging_steps": 50,