step8904: step 8904, 569856 examples, eval {'val_loss': 1.4704, 'val_wp_mae': 0.1043, 'top_move_acc': 0.21666666666666667, 'step': 8000}

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: `step7827` (step 7827, 500,928 training examples)
-- latest eval: `{'val_loss': 1.5209, 'val_wp_mae': 0.1171, 'top_move_acc': 0.19666666666666666, 'step': 6000}`
+- checkpoint: `step8904` (step 8904, 569,856 training examples)
+- latest eval: `{'val_loss': 1.4704, 'val_wp_mae': 0.1043, 'top_move_acc': 0.21666666666666667, 'step': 8000}`
 - encoding / levels / prompt template: see `chess_meta.json`
 - code: github.com/devroopsaha744/LayaChess

chess_meta.json

@@ -1,8 +1,8 @@
 {
- "tag": "step7827",
+ "tag": "step8904",
  "final": false,
- "step": 7827,
- "consumed": 500928,
+ "step": 8904,
+ "consumed": 569856,
  "n_levels": 10,
  "crit": [
   "0-10%",
@@ -45,100 +45,6 @@
   "SEED": 0
  },
  "history": [
-  {
-   "step": 5400,
-   "loss": 1.5546
-  },
-  {
-   "step": 5450,
-   "loss": 1.5622
-  },
-  {
-   "step": 5500,
-   "loss": 1.5426
-  },
-  {
-   "step": 5550,
-   "loss": 1.5501
-  },
-  {
-   "step": 5600,
-   "loss": 1.5601
-  },
-  {
-   "step": 5650,
-   "loss": 1.5411
-  },
-  {
-   "step": 5700,
-   "loss": 1.5534
-  },
-  {
-   "step": 5750,
-   "loss": 1.5509
-  },
-  {
-   "step": 5800,
-   "loss": 1.5414
-  },
-  {
-   "step": 5850,
-   "loss": 1.5512
-  },
-  {
-   "step": 5900,
-   "loss": 1.522
-  },
-  {
-   "step": 5950,
-   "loss": 1.5126
-  },
-  {
-   "step": 6000,
-   "loss": 1.5183
-  },
-  {
-   "val_loss": 1.5209,
-   "val_wp_mae": 0.1171,
-   "top_move_acc": 0.19666666666666666,
-   "step": 6000
-  },
-  {
-   "step": 6050,
-   "loss": 1.5171
-  },
-  {
-   "step": 6100,
-   "loss": 1.5371
-  },
-  {
-   "step": 6150,
-   "loss": 1.5258
-  },
-  {
-   "step": 6200,
-   "loss": 1.5161
-  },
-  {
-   "step": 6250,
-   "loss": 1.517
-  },
-  {
-   "step": 6300,
-   "loss": 1.5615
-  },
-  {
-   "step": 6350,
-   "loss": 1.5175
-  },
-  {
-   "step": 6400,
-   "loss": 1.5343
-  },
-  {
-   "step": 6450,
-   "loss": 1.5301
-  },
   {
    "step": 6500,
    "loss": 1.5242
@@ -246,6 +152,100 @@
   {
    "step": 7800,
    "loss": 1.4677
+  },
+  {
+   "step": 7850,
+   "loss": 1.4853
+  },
+  {
+   "step": 7900,
+   "loss": 1.4717
+  },
+  {
+   "step": 7950,
+   "loss": 1.4859
+  },
+  {
+   "step": 8000,
+   "loss": 1.4934
+  },
+  {
+   "val_loss": 1.4704,
+   "val_wp_mae": 0.1043,
+   "top_move_acc": 0.21666666666666667,
+   "step": 8000
+  },
+  {
+   "step": 8050,
+   "loss": 1.4751
+  },
+  {
+   "step": 8100,
+   "loss": 1.4735
+  },
+  {
+   "step": 8150,
+   "loss": 1.4506
+  },
+  {
+   "step": 8200,
+   "loss": 1.4816
+  },
+  {
+   "step": 8250,
+   "loss": 1.4646
+  },
+  {
+   "step": 8300,
+   "loss": 1.4727
+  },
+  {
+   "step": 8350,
+   "loss": 1.4546
+  },
+  {
+   "step": 8400,
+   "loss": 1.4819
+  },
+  {
+   "step": 8450,
+   "loss": 1.4764
+  },
+  {
+   "step": 8500,
+   "loss": 1.4726
+  },
+  {
+   "step": 8550,
+   "loss": 1.4564
+  },
+  {
+   "step": 8600,
+   "loss": 1.4769
+  },
+  {
+   "step": 8650,
+   "loss": 1.4726
+  },
+  {
+   "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
   }
  ]
 }

model.safetensors

@@ -1,3 +1,3 @@
 version https://git-lfs.github.com/spec/v1
-oid sha256:bd94f798940f4d0f94c368a7c307f0ce7659d8a234cb27fff6cc9632c46cce6d
+oid sha256:ea9d548376ed8146bf5b304895cfae9f8ac824820df2fb6e33593a70b0f53398
 size 842609220