Training in progress, step 33000, checkpoint

last-checkpoint/adapter_model.safetensors

@@ -1,3 +1,3 @@
 version https://git-lfs.github.com/spec/v1
-oid sha256:dc40c7322710674ba700e7c912ed88ea4428f9c273bc8666babf624928031d55
+oid sha256:c0017e49ae1179cf7a55f9321475e16cc1d09f3892ce98421f7979982523ebb9
 size 132187888

last-checkpoint/ar_diffusion_info.bin

@@ -1,3 +1,3 @@
 version https://git-lfs.github.com/spec/v1
-oid sha256:2879e5076d2053be09a0d6c73abd26b4278e5f16685fdedb148edcf4a65b4069
+oid sha256:bbd1ae1aad603802c82ed073dbd51ff3ad9656e5ab94f3bf3f7d0f8825c93935
 size 1800

last-checkpoint/optimizer.pt

@@ -1,3 +1,3 @@
 version https://git-lfs.github.com/spec/v1
-oid sha256:32736e80b881463272e4426f17a24ebdfddd47b5e2e3f5b99b633b357b80f7e8
+oid sha256:30ff43ec8181901deebb427ddf6e8551d899dcdbad73bc946f675627d7969b43
 size 264665786

last-checkpoint/rng_state.pth

@@ -1,3 +1,3 @@
 version https://git-lfs.github.com/spec/v1
-oid sha256:dc5a5cccd951497473f5f616bdc9028359460cced3c38458214cef6f531dc696
-size 14308
+oid sha256:e489367499db5673a7d65f2da50fa2212ff1d663a956b395bb78fec25dc26db1
+size 14244

last-checkpoint/scaler.pt

@@ -1,3 +1,3 @@
 version https://git-lfs.github.com/spec/v1
-oid sha256:9d6b59feadfd14316e1171dfc59fefcc9256bf9fc9745d2b33888fd1da1d95ea
+oid sha256:79aee1eff14611f1a5c55aaad748b38158c5fb7de7271250f2ef9599cd3ffe75
 size 988

last-checkpoint/scheduler.pt

@@ -1,3 +1,3 @@
 version https://git-lfs.github.com/spec/v1
-oid sha256:eb4a1da3914bf9e57fd6bbf0cd99b2cfb4060f68033db5609e17e853bacf6b40
+oid sha256:99a95e7d3fa1b2f1e9c1ee3dba6fdbbd55025c6a74911523945c9d063de75830
 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": 2.4998077070994538,
+  "epoch": 2.5382662872086765,
   "eval_steps": 250,
-  "global_step": 32500,
+  "global_step": 33000,
   "is_hyper_param_search": false,
   "is_local_process_zero": true,
   "is_world_process_zero": true,
@@ -5598,6 +5598,92 @@
       "eval_samples_per_second": 22.554,
       "eval_steps_per_second": 5.639,
       "step": 32500
+    },
+    {
+      "epoch": 2.503653565110376,
+      "grad_norm": 0.832127034664154,
+      "learning_rate": 3.359218762174376e-05,
+      "loss": 0.6311,
+      "step": 32550
+    },
+    {
+      "epoch": 2.507499423121298,
+      "grad_norm": 1.1985678672790527,
+      "learning_rate": 3.333246759992728e-05,
+      "loss": 0.6464,
+      "step": 32600
+    },
+    {
+      "epoch": 2.5113452811322206,
+      "grad_norm": 0.9630841612815857,
+      "learning_rate": 3.3072747578110796e-05,
+      "loss": 0.6098,
+      "step": 32650
+    },
+    {
+      "epoch": 2.515191139143143,
+      "grad_norm": 1.0516365766525269,
+      "learning_rate": 3.2813027556294315e-05,
+      "loss": 0.6099,
+      "step": 32700
+    },
+    {
+      "epoch": 2.519036997154065,
+      "grad_norm": 1.056236982345581,
+      "learning_rate": 3.2553307534477835e-05,
+      "loss": 0.6225,
+      "step": 32750
+    },
+    {
+      "epoch": 2.519036997154065,
+      "eval_loss": 0.6316623687744141,
+      "eval_runtime": 21.245,
+      "eval_samples_per_second": 23.535,
+      "eval_steps_per_second": 5.884,
+      "step": 32750
+    },
+    {
+      "epoch": 2.5228828551649873,
+      "grad_norm": 0.5901683568954468,
+      "learning_rate": 3.2293587512661355e-05,
+      "loss": 0.6281,
+      "step": 32800
+    },
+    {
+      "epoch": 2.5267287131759097,
+      "grad_norm": 1.385683536529541,
+      "learning_rate": 3.2033867490844874e-05,
+      "loss": 0.6392,
+      "step": 32850
+    },
+    {
+      "epoch": 2.5305745711868317,
+      "grad_norm": 1.0138803720474243,
+      "learning_rate": 3.177414746902839e-05,
+      "loss": 0.6125,
+      "step": 32900
+    },
+    {
+      "epoch": 2.534420429197754,
+      "grad_norm": 0.7247025370597839,
+      "learning_rate": 3.151442744721191e-05,
+      "loss": 0.6315,
+      "step": 32950
+    },
+    {
+      "epoch": 2.5382662872086765,
+      "grad_norm": 0.9154751896858215,
+      "learning_rate": 3.1254707425395426e-05,
+      "loss": 0.6231,
+      "step": 33000
+    },
+    {
+      "epoch": 2.5382662872086765,
+      "eval_loss": 0.6276845932006836,
+      "eval_runtime": 22.2178,
+      "eval_samples_per_second": 22.504,
+      "eval_steps_per_second": 5.626,
+      "step": 33000
     }
   ],
   "logging_steps": 50,