step22680: step 22680, 1451520 examples, eval {'val_loss': 1.3755, 'val_wp_mae': 0.0906, 'top_move_acc': 0.23666666666666666, 'step': 22000}

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: `step21628` (step 21628, 1,384,192 training examples)
-- latest eval: `{'val_loss': 1.3962, 'val_wp_mae': 0.0942, 'top_move_acc': 0.20666666666666667, 'step': 20000}`
+- checkpoint: `step22680` (step 22680, 1,451,520 training examples)
+- latest eval: `{'val_loss': 1.3755, 'val_wp_mae': 0.0906, 'top_move_acc': 0.23666666666666666, 'step': 22000}`
 - encoding / levels / prompt template: see `chess_meta.json`
 - code: github.com/devroopsaha744/LayaChess

chess_meta.json

@@ -1,8 +1,8 @@
 {
- "tag": "step21628",
+ "tag": "step22680",
  "final": false,
- "step": 21628,
- "consumed": 1384192,
+ "step": 22680,
+ "consumed": 1451520,
  "n_levels": 10,
  "crit": [
   "0-10%",
@@ -45,96 +45,6 @@
   "SEED": 0
  },
  "history": [
-  {
-   "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
-  },
-  {
-   "step": 19500,
-   "loss": 1.4003
-  },
-  {
-   "step": 19550,
-   "loss": 1.4134
-  },
-  {
-   "step": 19600,
-   "loss": 1.4052
-  },
-  {
-   "step": 19650,
-   "loss": 1.4031
-  },
-  {
-   "step": 19700,
-   "loss": 1.3919
-  },
-  {
-   "step": 19750,
-   "loss": 1.4166
-  },
-  {
-   "step": 19800,
-   "loss": 1.3965
-  },
-  {
-   "step": 19850,
-   "loss": 1.4093
-  },
-  {
-   "step": 19900,
-   "loss": 1.3845
-  },
-  {
-   "step": 19950,
-   "loss": 1.4027
-  },
-  {
-   "step": 20000,
-   "loss": 1.4191
-  },
-  {
-   "val_loss": 1.3962,
-   "val_wp_mae": 0.0942,
-   "top_move_acc": 0.20666666666666667,
-   "step": 20000
-  },
-  {
-   "step": 20050,
-   "loss": 1.4214
-  },
-  {
-   "step": 20100,
-   "loss": 1.4163
-  },
-  {
-   "step": 20150,
-   "loss": 1.3903
-  },
-  {
-   "step": 20200,
-   "loss": 1.4205
-  },
   {
    "step": 20250,
    "loss": 1.4059
@@ -246,6 +156,96 @@
   {
    "step": 21600,
    "loss": 1.3959
+  },
+  {
+   "step": 21650,
+   "loss": 1.3747
+  },
+  {
+   "step": 21700,
+   "loss": 1.3837
+  },
+  {
+   "step": 21750,
+   "loss": 1.3837
+  },
+  {
+   "step": 21800,
+   "loss": 1.4037
+  },
+  {
+   "step": 21850,
+   "loss": 1.3647
+  },
+  {
+   "step": 21900,
+   "loss": 1.389
+  },
+  {
+   "step": 21950,
+   "loss": 1.4138
+  },
+  {
+   "step": 22000,
+   "loss": 1.3847
+  },
+  {
+   "val_loss": 1.3755,
+   "val_wp_mae": 0.0906,
+   "top_move_acc": 0.23666666666666666,
+   "step": 22000
+  },
+  {
+   "step": 22050,
+   "loss": 1.3796
+  },
+  {
+   "step": 22100,
+   "loss": 1.3865
+  },
+  {
+   "step": 22150,
+   "loss": 1.3848
+  },
+  {
+   "step": 22200,
+   "loss": 1.3707
+  },
+  {
+   "step": 22250,
+   "loss": 1.3894
+  },
+  {
+   "step": 22300,
+   "loss": 1.3708
+  },
+  {
+   "step": 22350,
+   "loss": 1.3802
+  },
+  {
+   "step": 22400,
+   "loss": 1.3836
+  },
+  {
+   "step": 22450,
+   "loss": 1.3872
+  },
+  {
+   "step": 22500,
+   "loss": 1.3895
+  },
+  {
+   "step": 22550,
+   "loss": 1.3755
+  },
+  {
+   "step": 22600,
+   "loss": 1.3604
+  },
+  {
+   "step": 22650,
+   "loss": 1.3829
   }
  ]
 }

model.safetensors

@@ -1,3 +1,3 @@
 version https://git-lfs.github.com/spec/v1
-oid sha256:e18d84d2ab921fbff155f15e26d62b2154239888fd0d3aa24f23ae0b20847dea
+oid sha256:0dd60a9d8f3c888feaf50359f0e41c72b2f16ff70bc37be0579d3c1c10242e7d
 size 842609220