step19451: step 19451, 1244864 examples, eval {'val_loss': 1.4023, 'val_wp_mae': 0.0945, 'top_move_acc': 0.21333333333333335, 'step': 18000}

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: `step18324` (step 18324, 1,172,736 training examples)
+- checkpoint: `step19451` (step 19451, 1,244,864 training examples)
 - latest eval: `{'val_loss': 1.4023, 'val_wp_mae': 0.0945, 'top_move_acc': 0.21333333333333335, 'step': 18000}`
 - encoding / levels / prompt template: see `chess_meta.json`
 - code: github.com/devroopsaha744/LayaChess

chess_meta.json

@@ -1,8 +1,8 @@
 {
- "tag": "step18324",
+ "tag": "step19451",
  "final": false,
- "step": 18324,
- "consumed": 1172736,
+ "step": 19451,
+ "consumed": 1244864,
  "n_levels": 10,
  "crit": [
   "0-10%",
@@ -45,84 +45,6 @@
   "SEED": 0
  },
  "history": [
-  {
-   "val_loss": 1.3803,
-   "val_wp_mae": 0.091,
-   "top_move_acc": 0.23666666666666666,
-   "step": 16145
-  },
-  {
-   "step": 16150,
-   "loss": 0.1361
-  },
-  {
-   "step": 16200,
-   "loss": 1.3798
-  },
-  {
-   "step": 16250,
-   "loss": 1.4033
-  },
-  {
-   "step": 16300,
-   "loss": 1.3921
-  },
-  {
-   "step": 16350,
-   "loss": 1.4022
-  },
-  {
-   "step": 16400,
-   "loss": 1.4241
-  },
-  {
-   "step": 16450,
-   "loss": 1.4322
-  },
-  {
-   "step": 16500,
-   "loss": 1.4343
-  },
-  {
-   "step": 16550,
-   "loss": 1.4183
-  },
-  {
-   "step": 16600,
-   "loss": 1.4138
-  },
-  {
-   "step": 16650,
-   "loss": 1.4078
-  },
-  {
-   "step": 16700,
-   "loss": 1.4313
-  },
-  {
-   "step": 16750,
-   "loss": 1.4149
-  },
-  {
-   "step": 16800,
-   "loss": 1.4067
-  },
-  {
-   "step": 16850,
-   "loss": 1.4214
-  },
-  {
-   "step": 16900,
-   "loss": 1.4348
-  },
-  {
-   "step": 16950,
-   "loss": 1.4161
-  },
-  {
-   "step": 17000,
-   "loss": 1.4159
-  },
   {
    "step": 17050,
    "loss": 1.4399
@@ -232,6 +154,98 @@
   {
    "step": 18300,
    "loss": 1.4302
+  },
+  {
+   "step": 18350,
+   "loss": 1.4108
+  },
+  {
+   "step": 18400,
+   "loss": 1.4365
+  },
+  {
+   "step": 18450,
+   "loss": 1.4038
+  },
+  {
+   "step": 18500,
+   "loss": 1.3829
+  },
+  {
+   "step": 18550,
+   "loss": 1.3745
+  },
+  {
+   "step": 18600,
+   "loss": 1.391
+  },
+  {
+   "step": 18650,
+   "loss": 1.3841
+  },
+  {
+   "step": 18700,
+   "loss": 1.4053
+  },
+  {
+   "step": 18750,
+   "loss": 1.4357
+  },
+  {
+   "step": 18800,
+   "loss": 1.4191
+  },
+  {
+   "step": 18850,
+   "loss": 1.4105
+  },
+  {
+   "step": 18900,
+   "loss": 1.4052
+  },
+  {
+   "step": 18950,
+   "loss": 1.3759
+  },
+  {
+   "step": 19000,
+   "loss": 1.4155
+  },
+  {
+   "step": 19050,
+   "loss": 1.4118
+  },
+  {
+   "step": 19100,
+   "loss": 1.4039
+  },
+  {
+   "step": 19150,
+   "loss": 1.4244
+  },
+  {
+   "step": 19200,
+   "loss": 1.3875
+  },
+  {
+   "step": 19250,
+   "loss": 1.3887
+  },
+  {
+   "step": 19300,
+   "loss": 1.4035
+  },
+  {
+   "step": 19350,
+   "loss": 1.4129
+  },
+  {
+   "step": 19400,
+   "loss": 1.4091
+  },
+  {
+   "step": 19450,
+   "loss": 1.4036
   }
  ]
 }

model.safetensors

@@ -1,3 +1,3 @@
 version https://git-lfs.github.com/spec/v1
-oid sha256:820b70b5f4fe7078357af40c29d8bd4e3635efee178a6dfd1f065e2050b77a2d
+oid sha256:d9c36919078e90f1467759c0c2e53d0e28158043b6ecf36ece5b9324e2d529e7
 size 842609220