Training in progress, step 28500, checkpoint

last-checkpoint/adapter_model.safetensors

@@ -1,3 +1,3 @@
 version https://git-lfs.github.com/spec/v1
-oid sha256:9780097950800686c9322dd13a3faf02ad28fadad548e324be714511b2714201
+oid sha256:0bfc389ecdd34314f2f612c53571c8e1b47d7411169c05b6e5e9ef64751f9336
 size 132187888

last-checkpoint/ar_diffusion_info.bin

@@ -1,3 +1,3 @@
 version https://git-lfs.github.com/spec/v1
-oid sha256:f0eeadcb7061b328806de65f859dc63fe68b92f70ad10c9992993064e4e91786
+oid sha256:e03d21f59c7390814d3683840e8e05a5e87623d8d2c21d59a6e20c5b10100a19
 size 1800

last-checkpoint/optimizer.pt

@@ -1,3 +1,3 @@
 version https://git-lfs.github.com/spec/v1
-oid sha256:897b5043f12ab5b080ec026763f5e58f9947bfeb9d3b5e68e21c2b6493f9e676
+oid sha256:8c90e918c468147f036c1471580e464a3a1b2bd0dab65ebc2790466b8b13eec4
 size 264665786

last-checkpoint/rng_state.pth

@@ -1,3 +1,3 @@
 version https://git-lfs.github.com/spec/v1
-oid sha256:91254c943660a3203eec1d026396102f3e0b80f3a8c66cc2f6add4b242607689
-size 14180
+oid sha256:4e97ab479ccef726fdca432f369bd24340a85bd69660c81814e1117757ffe24e
+size 14308

last-checkpoint/scaler.pt

@@ -1,3 +1,3 @@
 version https://git-lfs.github.com/spec/v1
-oid sha256:f013f71783c0dc60fa63541cd9fac53da38b82998b3731a80d7c115cba7f0bf1
+oid sha256:4d31641ba54273e3788b2cfec72f598fb159706e78b7db0851c1ea541532286b
 size 988

last-checkpoint/scheduler.pt

@@ -1,3 +1,3 @@
 version https://git-lfs.github.com/spec/v1
-oid sha256:ff477fcb74f9478dd012f656769bb7c4aa62a4bc43e62fb29e6d1b12821b2b70
+oid sha256:50474d23052aaa3aaeaaa97d7b77d240e068c7314f12dd1aceb1c97f3be499a2
 size 1064

last-checkpoint/trainer_state.json

@@ -1,10 +1,10 @@
 {
-  "best_global_step": 28000,
-  "best_metric": 0.6302720904350281,
-  "best_model_checkpoint": "./qwen3-4b-ar-diffusion-checkpoints/checkpoint-28000",
-  "epoch": 2.1536804861164525,
+  "best_global_step": 28500,
+  "best_metric": 0.6302415132522583,
+  "best_model_checkpoint": "./qwen3-4b-ar-diffusion-checkpoints/checkpoint-28500",
+  "epoch": 2.192139066225675,
   "eval_steps": 250,
-  "global_step": 28000,
+  "global_step": 28500,
   "is_hyper_param_search": false,
   "is_local_process_zero": true,
   "is_world_process_zero": true,
@@ -4824,6 +4824,92 @@
       "eval_samples_per_second": 22.513,
       "eval_steps_per_second": 5.628,
       "step": 28000
+    },
+    {
+      "epoch": 2.157526344127375,
+      "grad_norm": 1.0334737300872803,
+      "learning_rate": 5.69617951847908e-05,
+      "loss": 0.6715,
+      "step": 28050
+    },
+    {
+      "epoch": 2.1613722021382973,
+      "grad_norm": 0.9481180310249329,
+      "learning_rate": 5.670207516297432e-05,
+      "loss": 0.6529,
+      "step": 28100
+    },
+    {
+      "epoch": 2.1652180601492192,
+      "grad_norm": 0.7230287790298462,
+      "learning_rate": 5.6442355141157834e-05,
+      "loss": 0.6703,
+      "step": 28150
+    },
+    {
+      "epoch": 2.1690639181601417,
+      "grad_norm": 0.9549979567527771,
+      "learning_rate": 5.6182635119341354e-05,
+      "loss": 0.6266,
+      "step": 28200
+    },
+    {
+      "epoch": 2.1729097761710636,
+      "grad_norm": 1.2103570699691772,
+      "learning_rate": 5.5922915097524874e-05,
+      "loss": 0.6111,
+      "step": 28250
+    },
+    {
+      "epoch": 2.1729097761710636,
+      "eval_loss": 0.6317173838615417,
+      "eval_runtime": 21.5185,
+      "eval_samples_per_second": 23.236,
+      "eval_steps_per_second": 5.809,
+      "step": 28250
+    },
+    {
+      "epoch": 2.176755634181986,
+      "grad_norm": 0.84665846824646,
+      "learning_rate": 5.5663195075708387e-05,
+      "loss": 0.6481,
+      "step": 28300
+    },
+    {
+      "epoch": 2.1806014921929084,
+      "grad_norm": 0.6889792084693909,
+      "learning_rate": 5.5403475053891906e-05,
+      "loss": 0.6393,
+      "step": 28350
+    },
+    {
+      "epoch": 2.1844473502038304,
+      "grad_norm": 1.1163616180419922,
+      "learning_rate": 5.5143755032075426e-05,
+      "loss": 0.6209,
+      "step": 28400
+    },
+    {
+      "epoch": 2.188293208214753,
+      "grad_norm": 0.8563538789749146,
+      "learning_rate": 5.488403501025894e-05,
+      "loss": 0.6071,
+      "step": 28450
+    },
+    {
+      "epoch": 2.192139066225675,
+      "grad_norm": 0.8409940600395203,
+      "learning_rate": 5.462431498844246e-05,
+      "loss": 0.6398,
+      "step": 28500
+    },
+    {
+      "epoch": 2.192139066225675,
+      "eval_loss": 0.6302415132522583,
+      "eval_runtime": 22.229,
+      "eval_samples_per_second": 22.493,
+      "eval_steps_per_second": 5.623,
+      "step": 28500
     }
   ],
   "logging_steps": 50,