step27036: step 27036, 1730304 examples, eval {'val_loss': 1.3545, 'val_wp_mae': 0.0855, 'top_move_acc': 0.25333333333333335, 'step': 26000}

README.md

@@ -9,7 +9,7 @@ tags: [chess, laya, decision-model]
 for each legal move, a Laya `score` question predicts the side to move's win chance over 10 levels.
 The engine plays the move with the highest expected win chance (no search).
 
-- checkpoint: `step25985` (step 25985, 1,663,040 training examples)
-- latest eval: `{'val_loss': 1.3659, 'val_wp_mae': 0.0888, 'top_move_acc': 0.26666666666666666, 'step': 24000}`
+- checkpoint: `step27036` (step 27036, 1,730,304 training examples)
+- latest eval: `{'val_loss': 1.3545, 'val_wp_mae': 0.0855, 'top_move_acc': 0.25333333333333335, 'step': 26000}`
 - encoding / levels / prompt template: see `chess_meta.json`
 - code: github.com/devroopsaha744/LayaChess

chess_meta.json

@@ -1,8 +1,8 @@
 {
- "tag": "step25985",
+ "tag": "step27036",
  "final": false,
- "step": 25985,
- "consumed": 1663040,
+ "step": 27036,
+ "consumed": 1730304,
  "n_levels": 10,
  "crit": [
   "0-10%",
@@ -45,96 +45,6 @@
   "SEED": 0
  },
  "history": [
-  {
-   "step": 23550,
-   "loss": 1.3889
-  },
-  {
-   "step": 23600,
-   "loss": 1.3918
-  },
-  {
-   "step": 23650,
-   "loss": 1.3792
-  },
-  {
-   "step": 23700,
-   "loss": 1.3689
-  },
-  {
-   "step": 23750,
-   "loss": 1.3709
-  },
-  {
-   "step": 23800,
-   "loss": 1.3425
-  },
-  {
-   "step": 23850,
-   "loss": 1.4021
-  },
-  {
-   "step": 23900,
-   "loss": 1.3816
-  },
-  {
-   "step": 23950,
-   "loss": 1.3642
-  },
-  {
-   "step": 24000,
-   "loss": 1.3775
-  },
-  {
-   "val_loss": 1.3659,
-   "val_wp_mae": 0.0888,
-   "top_move_acc": 0.26666666666666666,
-   "step": 24000
-  },
-  {
-   "step": 24050,
-   "loss": 1.3651
-  },
-  {
-   "step": 24100,
-   "loss": 1.3638
-  },
-  {
-   "step": 24150,
-   "loss": 1.3713
-  },
-  {
-   "step": 24200,
-   "loss": 1.3664
-  },
-  {
-   "step": 24250,
-   "loss": 1.3665
-  },
-  {
-   "step": 24300,
-   "loss": 1.3642
-  },
-  {
-   "step": 24350,
-   "loss": 1.3734
-  },
-  {
-   "step": 24400,
-   "loss": 1.3731
-  },
-  {
-   "step": 24450,
-   "loss": 1.3679
-  },
-  {
-   "step": 24500,
-   "loss": 1.3862
-  },
-  {
-   "step": 24550,
-   "loss": 1.386
-  },
   {
    "step": 24600,
    "loss": 1.3717
@@ -246,6 +156,96 @@
   {
    "step": 25950,
    "loss": 1.3739
+  },
+  {
+   "step": 26000,
+   "loss": 1.3455
+  },
+  {
+   "val_loss": 1.3545,
+   "val_wp_mae": 0.0855,
+   "top_move_acc": 0.25333333333333335,
+   "step": 26000
+  },
+  {
+   "step": 26050,
+   "loss": 1.3481
+  },
+  {
+   "step": 26100,
+   "loss": 1.3556
+  },
+  {
+   "step": 26150,
+   "loss": 1.3564
+  },
+  {
+   "step": 26200,
+   "loss": 1.3745
+  },
+  {
+   "step": 26250,
+   "loss": 1.3554
+  },
+  {
+   "step": 26300,
+   "loss": 1.3706
+  },
+  {
+   "step": 26350,
+   "loss": 1.3639
+  },
+  {
+   "step": 26400,
+   "loss": 1.3593
+  },
+  {
+   "step": 26450,
+   "loss": 1.355
+  },
+  {
+   "step": 26500,
+   "loss": 1.3479
+  },
+  {
+   "step": 26550,
+   "loss": 1.3517
+  },
+  {
+   "step": 26600,
+   "loss": 1.3729
+  },
+  {
+   "step": 26650,
+   "loss": 1.3756
+  },
+  {
+   "step": 26700,
+   "loss": 1.3366
+  },
+  {
+   "step": 26750,
+   "loss": 1.3626
+  },
+  {
+   "step": 26800,
+   "loss": 1.3675
+  },
+  {
+   "step": 26850,
+   "loss": 1.3547
+  },
+  {
+   "step": 26900,
+   "loss": 1.3472
+  },
+  {
+   "step": 26950,
+   "loss": 1.3692
+  },
+  {
+   "step": 27000,
+   "loss": 1.3456
   }
  ]
 }

model.safetensors

@@ -1,3 +1,3 @@
 version https://git-lfs.github.com/spec/v1
-oid sha256:ffdaa44e31bbf2b98f6cd12fc26a3da5f8b3c0424a3488d44b18f6dd0d14c884
+oid sha256:987587b480af038b8fbb43ae64550458bbd08e8dba54aae247ef94e0f51500d2
 size 842609220