step25985: step 25985, 1663040 examples, eval {'val_loss': 1.3659, 'val_wp_mae': 0.0888, 'top_move_acc': 0.26666666666666666, 'step': 24000}

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: `step24859` (step 24859, 1,590,976 training examples)
+- checkpoint: `step25985` (step 25985, 1,663,040 training examples)
 - latest eval: `{'val_loss': 1.3659, 'val_wp_mae': 0.0888, 'top_move_acc': 0.26666666666666666, 'step': 24000}`
 - encoding / levels / prompt template: see `chess_meta.json`
 - code: github.com/devroopsaha744/LayaChess

chess_meta.json

@@ -1,8 +1,8 @@
 {
- "tag": "step24859",
+ "tag": "step25985",
  "final": false,
- "step": 24859,
- "consumed": 1590976,
+ "step": 25985,
+ "consumed": 1663040,
  "n_levels": 10,
  "crit": [
   "0-10%",
@@ -45,94 +45,6 @@
   "SEED": 0
  },
  "history": [
-  {
-   "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
-  },
-  {
-   "step": 22700,
-   "loss": 1.3735
-  },
-  {
-   "step": 22750,
-   "loss": 1.3734
-  },
-  {
-   "step": 22800,
-   "loss": 1.3657
-  },
-  {
-   "step": 22850,
-   "loss": 1.4064
-  },
-  {
-   "step": 22900,
-   "loss": 1.3548
-  },
-  {
-   "step": 22950,
-   "loss": 1.3686
-  },
-  {
-   "step": 23000,
-   "loss": 1.3758
-  },
-  {
-   "step": 23050,
-   "loss": 1.3828
-  },
-  {
-   "step": 23100,
-   "loss": 1.3893
-  },
-  {
-   "step": 23150,
-   "loss": 1.3928
-  },
-  {
-   "step": 23200,
-   "loss": 1.3711
-  },
-  {
-   "step": 23250,
-   "loss": 1.3611
-  },
-  {
-   "step": 23300,
-   "loss": 1.3704
-  },
-  {
-   "step": 23350,
-   "loss": 1.3356
-  },
-  {
-   "step": 23400,
-   "loss": 1.3683
-  },
-  {
-   "step": 23450,
-   "loss": 1.4022
-  },
-  {
-   "step": 23500,
-   "loss": 1.3709
-  },
   {
    "step": 23550,
    "loss": 1.3889
@@ -246,6 +158,94 @@
   {
    "step": 24850,
    "loss": 1.3916
+  },
+  {
+   "step": 24900,
+   "loss": 1.3475
+  },
+  {
+   "step": 24950,
+   "loss": 1.373
+  },
+  {
+   "step": 25000,
+   "loss": 1.359
+  },
+  {
+   "step": 25050,
+   "loss": 1.337
+  },
+  {
+   "step": 25100,
+   "loss": 1.3572
+  },
+  {
+   "step": 25150,
+   "loss": 1.3517
+  },
+  {
+   "step": 25200,
+   "loss": 1.3712
+  },
+  {
+   "step": 25250,
+   "loss": 1.3447
+  },
+  {
+   "step": 25300,
+   "loss": 1.3592
+  },
+  {
+   "step": 25350,
+   "loss": 1.3427
+  },
+  {
+   "step": 25400,
+   "loss": 1.3306
+  },
+  {
+   "step": 25450,
+   "loss": 1.3534
+  },
+  {
+   "step": 25500,
+   "loss": 1.3726
+  },
+  {
+   "step": 25550,
+   "loss": 1.3724
+  },
+  {
+   "step": 25600,
+   "loss": 1.3529
+  },
+  {
+   "step": 25650,
+   "loss": 1.363
+  },
+  {
+   "step": 25700,
+   "loss": 1.3591
+  },
+  {
+   "step": 25750,
+   "loss": 1.3538
+  },
+  {
+   "step": 25800,
+   "loss": 1.3595
+  },
+  {
+   "step": 25850,
+   "loss": 1.3606
+  },
+  {
+   "step": 25900,
+   "loss": 1.3513
+  },
+  {
+   "step": 25950,
+   "loss": 1.3739
   }
  ]
 }

model.safetensors

@@ -1,3 +1,3 @@
 version https://git-lfs.github.com/spec/v1
-oid sha256:12306a9b0b52efab2ea4a9b2fa9bfadf6dde034cbd053bb85a44a541bdce774b
+oid sha256:ffdaa44e31bbf2b98f6cd12fc26a3da5f8b3c0424a3488d44b18f6dd0d14c884
 size 842609220