step28087: step 28087, 1797568 examples, eval {'val_loss': 1.3418, 'val_wp_mae': 0.0845, 'top_move_acc': 0.29333333333333333, 'step': 28000}

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: `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}`
+- checkpoint: `step28087` (step 28087, 1,797,568 training examples)
+- latest eval: `{'val_loss': 1.3418, 'val_wp_mae': 0.0845, 'top_move_acc': 0.29333333333333333, 'step': 28000}`
 - encoding / levels / prompt template: see `chess_meta.json`
 - code: github.com/devroopsaha744/LayaChess

chess_meta.json

@@ -1,8 +1,8 @@
 {
- "tag": "step27036",
+ "tag": "step28087",
  "final": false,
- "step": 27036,
- "consumed": 1730304,
+ "step": 28087,
+ "consumed": 1797568,
  "n_levels": 10,
  "crit": [
   "0-10%",
@@ -45,94 +45,6 @@
   "SEED": 0
  },
  "history": [
-  {
-   "step": 24600,
-   "loss": 1.3717
-  },
-  {
-   "step": 24650,
-   "loss": 1.3684
-  },
-  {
-   "step": 24700,
-   "loss": 1.3761
-  },
-  {
-   "step": 24750,
-   "loss": 1.3618
-  },
-  {
-   "step": 24800,
-   "loss": 1.3795
-  },
-  {
-   "step": 24850,
-   "loss": 1.3916
-  },
-  {
-   "step": 24900,
-   "loss": 1.3475
-  },
-  {
-   "step": 24950,
-   "loss": 1.373
-  },
-  {
-   "step": 25000,
-   "loss": 1.359
-  },
-  {
-   "step": 25050,
-   "loss": 1.337
-  },
-  {
-   "step": 25100,
-   "loss": 1.3572
-  },
-  {
-   "step": 25150,
-   "loss": 1.3517
-  },
-  {
-   "step": 25200,
-   "loss": 1.3712
-  },
-  {
-   "step": 25250,
-   "loss": 1.3447
-  },
-  {
-   "step": 25300,
-   "loss": 1.3592
-  },
-  {
-   "step": 25350,
-   "loss": 1.3427
-  },
-  {
-   "step": 25400,
-   "loss": 1.3306
-  },
-  {
-   "step": 25450,
-   "loss": 1.3534
-  },
-  {
-   "step": 25500,
-   "loss": 1.3726
-  },
-  {
-   "step": 25550,
-   "loss": 1.3724
-  },
-  {
-   "step": 25600,
-   "loss": 1.3529
-  },
-  {
-   "step": 25650,
-   "loss": 1.363
-  },
   {
    "step": 25700,
    "loss": 1.3591
@@ -246,6 +158,96 @@
   {
    "step": 27000,
    "loss": 1.3456
+  },
+  {
+   "step": 27050,
+   "loss": 1.3589
+  },
+  {
+   "step": 27100,
+   "loss": 1.3419
+  },
+  {
+   "step": 27150,
+   "loss": 1.3537
+  },
+  {
+   "step": 27200,
+   "loss": 1.3614
+  },
+  {
+   "step": 27250,
+   "loss": 1.3599
+  },
+  {
+   "step": 27300,
+   "loss": 1.3478
+  },
+  {
+   "step": 27350,
+   "loss": 1.3533
+  },
+  {
+   "step": 27400,
+   "loss": 1.3597
+  },
+  {
+   "step": 27450,
+   "loss": 1.3687
+  },
+  {
+   "step": 27500,
+   "loss": 1.3475
+  },
+  {
+   "step": 27550,
+   "loss": 1.3591
+  },
+  {
+   "step": 27600,
+   "loss": 1.3554
+  },
+  {
+   "step": 27650,
+   "loss": 1.3452
+  },
+  {
+   "step": 27700,
+   "loss": 1.3307
+  },
+  {
+   "step": 27750,
+   "loss": 1.3563
+  },
+  {
+   "step": 27800,
+   "loss": 1.3645
+  },
+  {
+   "step": 27850,
+   "loss": 1.3637
+  },
+  {
+   "step": 27900,
+   "loss": 1.3384
+  },
+  {
+   "step": 27950,
+   "loss": 1.3549
+  },
+  {
+   "step": 28000,
+   "loss": 1.3128
+  },
+  {
+   "val_loss": 1.3418,
+   "val_wp_mae": 0.0845,
+   "top_move_acc": 0.29333333333333333,
+   "step": 28000
+  },
+  {
+   "step": 28050,
+   "loss": 1.3337
   }
  ]
 }

model.safetensors

@@ -1,3 +1,3 @@
 version https://git-lfs.github.com/spec/v1
-oid sha256:987587b480af038b8fbb43ae64550458bbd08e8dba54aae247ef94e0f51500d2
+oid sha256:2f2dd82e5911e06eb7a192e0a6d8bf79d1839041c8c9844bbdc360d2a0066533
 size 842609220