stage3_hybrid_v2: epoch 10 (loss=0.4964)

stage3_hybrid_v2/latest/model.safetensors

@@ -1,3 +1,3 @@
 version https://git-lfs.github.com/spec/v1
-oid sha256:4baa521b13b1572e2e72698bf2b469dbd131dc2b0bf90933f68d96b8801ababd
+oid sha256:01570f587802a6df6d56749a819b7f239d848f105a13c6be27f4ad77b0e2714f
 size 1192585656

stage3_hybrid_v2/latest/optimizer.pt

@@ -1,3 +1,3 @@
 version https://git-lfs.github.com/spec/v1
-oid sha256:1e6a6105cfa76c5c80dc51a842d1e32c028c43b66e4f5b23d8ff3aa9b7a204d6
+oid sha256:194906cbd2185be267ef1e4238f311c2a7d04d0560104f5a48b15c7da9fe15d6
 size 2385361355

stage3_hybrid_v2/latest/scheduler.pt

@@ -1,3 +1,3 @@
 version https://git-lfs.github.com/spec/v1
-oid sha256:277a9dfa56b4fd8442853a893751a8ecacb6bd5e91257187416dceba7cca9c86
+oid sha256:92a7ebab7534079d1a6ffe1dbc7c1edafd8bc018aebc08978b18933369821282
 size 1465

stage3_hybrid_v2/latest/training_state.json

@@ -1,9 +1,9 @@
 {
   "stage": "stage3_hybrid_v2",
-  "epoch_completed": 8,
+  "epoch_completed": 10,
   "total_epochs": 20,
-  "avg_loss": 0.4968204984052607,
-  "global_step": 20988,
-  "curriculum_max_len": 148,
-  "saved_at": "2026-01-22T12:20:14.880321Z"
+  "avg_loss": 0.4963871867867316,
+  "global_step": 26852,
+  "curriculum_max_len": 169,
+  "saved_at": "2026-01-22T13:02:48.797119Z"
 }