step12221: step 12221, 782144 examples, eval {'val_loss': 1.4059, 'val_wp_mae': 0.0964, 'top_move_acc': 0.19666666666666666, 'step': 12000}

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: `step11149` (step 11149, 713,536 training examples)
-- latest eval: `{'val_loss': 1.4349, 'val_wp_mae': 0.0981, 'top_move_acc': 0.19666666666666666, 'step': 10000}`
+- checkpoint: `step12221` (step 12221, 782,144 training examples)
+- latest eval: `{'val_loss': 1.4059, 'val_wp_mae': 0.0964, 'top_move_acc': 0.19666666666666666, 'step': 12000}`
 - encoding / levels / prompt template: see `chess_meta.json`
 - code: github.com/devroopsaha744/LayaChess

chess_meta.json

@@ -1,8 +1,8 @@
 {
- "tag": "step11149",
+ "tag": "step12221",
  "final": false,
- "step": 11149,
- "consumed": 713536,
+ "step": 12221,
+ "consumed": 782144,
  "n_levels": 10,
  "crit": [
   "0-10%",
@@ -45,98 +45,6 @@
   "SEED": 0
  },
  "history": [
-  {
-   "step": 8700,
-   "loss": 1.4775
-  },
-  {
-   "step": 8750,
-   "loss": 1.4575
-  },
-  {
-   "step": 8800,
-   "loss": 1.4616
-  },
-  {
-   "step": 8850,
-   "loss": 1.4477
-  },
-  {
-   "step": 8900,
-   "loss": 1.4476
-  },
-  {
-   "step": 8950,
-   "loss": 1.4855
-  },
-  {
-   "step": 9000,
-   "loss": 1.4658
-  },
-  {
-   "step": 9050,
-   "loss": 1.4446
-  },
-  {
-   "step": 9100,
-   "loss": 1.463
-  },
-  {
-   "step": 9150,
-   "loss": 1.4703
-  },
-  {
-   "step": 9200,
-   "loss": 1.4479
-  },
-  {
-   "step": 9250,
-   "loss": 1.4321
-  },
-  {
-   "step": 9300,
-   "loss": 1.45
-  },
-  {
-   "step": 9350,
-   "loss": 1.4566
-  },
-  {
-   "step": 9400,
-   "loss": 1.4516
-  },
-  {
-   "step": 9450,
-   "loss": 1.4543
-  },
-  {
-   "step": 9500,
-   "loss": 1.459
-  },
-  {
-   "step": 9550,
-   "loss": 1.4481
-  },
-  {
-   "step": 9600,
-   "loss": 1.4224
-  },
-  {
-   "step": 9650,
-   "loss": 1.4161
-  },
-  {
-   "step": 9700,
-   "loss": 1.4238
-  },
-  {
-   "step": 9750,
-   "loss": 1.4439
-  },
-  {
-   "step": 9800,
-   "loss": 1.4393
-  },
   {
    "step": 9850,
    "loss": 1.4532
@@ -246,6 +154,100 @@
   {
    "step": 11100,
    "loss": 1.426
+  },
+  {
+   "step": 11150,
+   "loss": 1.4295
+  },
+  {
+   "step": 11200,
+   "loss": 1.4177
+  },
+  {
+   "step": 11250,
+   "loss": 1.4365
+  },
+  {
+   "step": 11300,
+   "loss": 1.4089
+  },
+  {
+   "step": 11350,
+   "loss": 1.4332
+  },
+  {
+   "step": 11400,
+   "loss": 1.4171
+  },
+  {
+   "step": 11450,
+   "loss": 1.4313
+  },
+  {
+   "step": 11500,
+   "loss": 1.4352
+  },
+  {
+   "step": 11550,
+   "loss": 1.4189
+  },
+  {
+   "step": 11600,
+   "loss": 1.4116
+  },
+  {
+   "step": 11650,
+   "loss": 1.3983
+  },
+  {
+   "step": 11700,
+   "loss": 1.4193
+  },
+  {
+   "step": 11750,
+   "loss": 1.4245
+  },
+  {
+   "step": 11800,
+   "loss": 1.4241
+  },
+  {
+   "step": 11850,
+   "loss": 1.4153
+  },
+  {
+   "step": 11900,
+   "loss": 1.4077
+  },
+  {
+   "step": 11950,
+   "loss": 1.4323
+  },
+  {
+   "step": 12000,
+   "loss": 1.4093
+  },
+  {
+   "val_loss": 1.4059,
+   "val_wp_mae": 0.0964,
+   "top_move_acc": 0.19666666666666666,
+   "step": 12000
+  },
+  {
+   "step": 12050,
+   "loss": 1.4258
+  },
+  {
+   "step": 12100,
+   "loss": 1.3939
+  },
+  {
+   "step": 12150,
+   "loss": 1.4145
+  },
+  {
+   "step": 12200,
+   "loss": 1.4209
   }
  ]
 }

model.safetensors

@@ -1,3 +1,3 @@
 version https://git-lfs.github.com/spec/v1
-oid sha256:23be7dbd5150706a9b07bbae1339c9770ea8e79e5128f9a1ccfc75271f254ad4
+oid sha256:8ba78ea307ca5fdb0be0c6779aeeffde94bdf8ed6124162242eb1def2efac9ad
 size 842609220