step3370: step 3370, 215680 examples, eval {'val_loss': 1.5918, 'val_wp_mae': 0.1285, 'top_move_acc': 0.16333333333333333, 'step': 2000}

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: `step2220` (step 2220, 142,080 training examples)
+- checkpoint: `step3370` (step 3370, 215,680 training examples)
 - latest eval: `{'val_loss': 1.5918, 'val_wp_mae': 0.1285, 'top_move_acc': 0.16333333333333333, 'step': 2000}`
 - encoding / levels / prompt template: see `chess_meta.json`
 - code: github.com/devroopsaha744/LayaChess

chess_meta.json

@@ -1,8 +1,8 @@
 {
- "tag": "step2220",
+ "tag": "step3370",
  "final": false,
- "step": 2220,
- "consumed": 142080,
+ "step": 3370,
+ "consumed": 215680,
  "n_levels": 10,
  "crit": [
   "0-10%",
@@ -45,84 +45,6 @@
   "SEED": 0
  },
  "history": [
-  {
-   "val_loss": 2.4097,
-   "val_wp_mae": 0.2865,
-   "top_move_acc": 0.06,
-   "step": 0
-  },
-  {
-   "step": 50,
-   "loss": 2.1504
-  },
-  {
-   "step": 100,
-   "loss": 2.0519
-  },
-  {
-   "step": 150,
-   "loss": 1.9668
-  },
-  {
-   "step": 200,
-   "loss": 1.8753
-  },
-  {
-   "step": 250,
-   "loss": 1.8255
-  },
-  {
-   "step": 300,
-   "loss": 1.7603
-  },
-  {
-   "step": 350,
-   "loss": 1.7439
-  },
-  {
-   "step": 400,
-   "loss": 1.7451
-  },
-  {
-   "step": 450,
-   "loss": 1.7204
-  },
-  {
-   "step": 500,
-   "loss": 1.7132
-  },
-  {
-   "step": 550,
-   "loss": 1.7237
-  },
-  {
-   "step": 600,
-   "loss": 1.7013
-  },
-  {
-   "step": 650,
-   "loss": 1.6971
-  },
-  {
-   "step": 700,
-   "loss": 1.6698
-  },
-  {
-   "step": 750,
-   "loss": 1.6473
-  },
-  {
-   "step": 800,
-   "loss": 1.6473
-  },
-  {
-   "step": 850,
-   "loss": 1.6712
-  },
-  {
-   "step": 900,
-   "loss": 1.6612
-  },
   {
    "step": 950,
    "loss": 1.642
@@ -232,6 +154,98 @@
   {
    "step": 2200,
    "loss": 1.5952
+  },
+  {
+   "step": 2250,
+   "loss": 1.5796
+  },
+  {
+   "step": 2300,
+   "loss": 1.5788
+  },
+  {
+   "step": 2350,
+   "loss": 1.5687
+  },
+  {
+   "step": 2400,
+   "loss": 1.6003
+  },
+  {
+   "step": 2450,
+   "loss": 1.5733
+  },
+  {
+   "step": 2500,
+   "loss": 1.5897
+  },
+  {
+   "step": 2550,
+   "loss": 1.5945
+  },
+  {
+   "step": 2600,
+   "loss": 1.5713
+  },
+  {
+   "step": 2650,
+   "loss": 1.5836
+  },
+  {
+   "step": 2700,
+   "loss": 1.5618
+  },
+  {
+   "step": 2750,
+   "loss": 1.5639
+  },
+  {
+   "step": 2800,
+   "loss": 1.5866
+  },
+  {
+   "step": 2850,
+   "loss": 1.6121
+  },
+  {
+   "step": 2900,
+   "loss": 1.5878
+  },
+  {
+   "step": 2950,
+   "loss": 1.5798
+  },
+  {
+   "step": 3000,
+   "loss": 1.5747
+  },
+  {
+   "step": 3050,
+   "loss": 1.5969
+  },
+  {
+   "step": 3100,
+   "loss": 1.5824
+  },
+  {
+   "step": 3150,
+   "loss": 1.5848
+  },
+  {
+   "step": 3200,
+   "loss": 1.5735
+  },
+  {
+   "step": 3250,
+   "loss": 1.5577
+  },
+  {
+   "step": 3300,
+   "loss": 1.5671
+  },
+  {
+   "step": 3350,
+   "loss": 1.5548
   }
  ]
 }

model.safetensors

@@ -1,3 +1,3 @@
 version https://git-lfs.github.com/spec/v1
-oid sha256:137650d99cc30af206e7d2055a45fa419b0d0cd629ff993319897a3e2af58466
+oid sha256:7d136467af3e26da6adb542f7c95151f0b262db96a767cc1ea921bcf7a1873a7
 size 842609220