step17271: step 17271, 1105344 examples, eval {'val_loss': 1.3803, 'val_wp_mae': 0.091, 'top_move_acc': 0.23666666666666666, 'step': 16145}

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: `final` (step 16145, 1,033,280 training examples)
+- checkpoint: `step17271` (step 17271, 1,105,344 training examples)
 - latest eval: `{'val_loss': 1.3803, 'val_wp_mae': 0.091, 'top_move_acc': 0.23666666666666666, 'step': 16145}`
 - encoding / levels / prompt template: see `chess_meta.json`
 - code: github.com/devroopsaha744/LayaChess

chess_meta.json

@@ -1,8 +1,8 @@
 {
- "tag": "final",
- "final": true,
- "step": 16145,
- "consumed": 1033280,
+ "tag": "step17271",
+ "final": false,
+ "step": 17271,
+ "consumed": 1105344,
  "n_levels": 10,
  "crit": [
   "0-10%",
@@ -20,8 +20,8 @@
  "state_format": "piece lists v2",
  "cfg": {
   "MODEL_ID": "convaiinnovations/laya",
-  "N_TRAIN_SHARDS": 1,
-  "MAX_TRAIN_RECORDS": 10000000,
+  "N_TRAIN_SHARDS": 2,
+  "MAX_TRAIN_RECORDS": null,
   "N_VAL_RECORDS": 4000,
   "N_VAL_POSITIONS": 300,
   "N_LEVELS": 10,
@@ -39,217 +39,109 @@
   "EVAL_EVERY_STEPS": 2000,
   "LOG_EVERY_STEPS": 50,
   "SAVE_OPTIMIZER": true,
-  "RESUME_FROM": null,
+  "RESUME_FROM": "/tmp/resume_ckpt",
   "RESUME_LR_SCALE": 0.5,
   "HF_REPO_NAME": "laya-chess",
   "SEED": 0
  },
  "history": [
   {
-   "step": 13800,
-   "loss": 1.4002
-  },
-  {
-   "step": 13850,
-   "loss": 1.413
-  },
-  {
-   "step": 13900,
-   "loss": 1.4033
-  },
-  {
-   "step": 13950,
-   "loss": 1.3729
-  },
-  {
-   "step": 14000,
-   "loss": 1.4235
-  },
-  {
-   "val_loss": 1.3893,
-   "val_wp_mae": 0.0943,
+   "val_loss": 1.3803,
+   "val_wp_mae": 0.091,
    "top_move_acc": 0.23666666666666666,
-   "step": 14000
-  },
-  {
-   "step": 14050,
-   "loss": 1.3853
-  },
-  {
-   "step": 14100,
-   "loss": 1.4006
-  },
-  {
-   "step": 14150,
-   "loss": 1.39
-  },
-  {
-   "step": 14200,
-   "loss": 1.4043
-  },
-  {
-   "step": 14250,
-   "loss": 1.3979
-  },
-  {
-   "step": 14300,
-   "loss": 1.4103
-  },
-  {
-   "step": 14350,
-   "loss": 1.4165
-  },
-  {
-   "step": 14400,
-   "loss": 1.403
-  },
-  {
-   "step": 14450,
-   "loss": 1.3911
-  },
-  {
-   "step": 14500,
-   "loss": 1.4061
-  },
-  {
-   "step": 14550,
-   "loss": 1.3982
-  },
-  {
-   "step": 14600,
-   "loss": 1.4163
-  },
-  {
-   "step": 14650,
-   "loss": 1.3909
-  },
-  {
-   "step": 14700,
-   "loss": 1.399
-  },
-  {
-   "step": 14750,
-   "loss": 1.4142
-  },
-  {
-   "step": 14800,
-   "loss": 1.4087
-  },
-  {
-   "step": 14850,
-   "loss": 1.4004
-  },
-  {
-   "step": 14900,
-   "loss": 1.3804
-  },
-  {
-   "step": 14950,
-   "loss": 1.3818
-  },
-  {
-   "step": 15000,
-   "loss": 1.3935
-  },
-  {
-   "step": 15050,
-   "loss": 1.4219
+   "step": 16145
   },
   {
-   "step": 15100,
-   "loss": 1.3838
+   "step": 16150,
+   "loss": 0.1361
   },
   {
-   "step": 15150,
-   "loss": 1.392
+   "step": 16200,
+   "loss": 1.3798
   },
   {
-   "step": 15200,
-   "loss": 1.3909
+   "step": 16250,
+   "loss": 1.4033
   },
   {
-   "step": 15250,
-   "loss": 1.4165
+   "step": 16300,
+   "loss": 1.3921
   },
   {
-   "step": 15300,
-   "loss": 1.3905
+   "step": 16350,
+   "loss": 1.4022
   },
   {
-   "step": 15350,
-   "loss": 1.4124
+   "step": 16400,
+   "loss": 1.4241
   },
   {
-   "step": 15400,
-   "loss": 1.3939
+   "step": 16450,
+   "loss": 1.4322
   },
   {
-   "step": 15450,
-   "loss": 1.3821
+   "step": 16500,
+   "loss": 1.4343
   },
   {
-   "step": 15500,
-   "loss": 1.395
+   "step": 16550,
+   "loss": 1.4183
   },
   {
-   "step": 15550,
-   "loss": 1.3914
+   "step": 16600,
+   "loss": 1.4138
   },
   {
-   "step": 15600,
-   "loss": 1.3927
+   "step": 16650,
+   "loss": 1.4078
   },
   {
-   "step": 15650,
-   "loss": 1.3722
+   "step": 16700,
+   "loss": 1.4313
   },
   {
-   "step": 15700,
-   "loss": 1.3652
+   "step": 16750,
+   "loss": 1.4149
   },
   {
-   "step": 15750,
-   "loss": 1.4063
+   "step": 16800,
+   "loss": 1.4067
   },
   {
-   "step": 15800,
-   "loss": 1.3709
+   "step": 16850,
+   "loss": 1.4214
   },
   {
-   "step": 15850,
-   "loss": 1.3887
+   "step": 16900,
+   "loss": 1.4348
   },
   {
-   "step": 15900,
-   "loss": 1.4116
+   "step": 16950,
+   "loss": 1.4161
   },
   {
-   "step": 15950,
-   "loss": 1.3723
+   "step": 17000,
+   "loss": 1.4159
   },
   {
-   "step": 16000,
-   "loss": 1.4105
+   "step": 17050,
+   "loss": 1.4399
   },
   {
-   "val_loss": 1.3828,
-   "val_wp_mae": 0.0915,
-   "top_move_acc": 0.23333333333333334,
-   "step": 16000
+   "step": 17100,
+   "loss": 1.414
   },
   {
-   "step": 16050,
-   "loss": 1.3975
+   "step": 17150,
+   "loss": 1.4278
   },
   {
-   "step": 16100,
-   "loss": 1.3831
+   "step": 17200,
+   "loss": 1.4079
   },
   {
-   "val_loss": 1.3803,
-   "val_wp_mae": 0.091,
-   "top_move_acc": 0.23666666666666666,
-   "step": 16145
+   "step": 17250,
+   "loss": 1.42
   }
  ]
 }

model.safetensors

@@ -1,3 +1,3 @@
 version https://git-lfs.github.com/spec/v1
-oid sha256:8ec7a035dc1cd8457aa3f39cfe6d8b30478146d27e9ce5091532bb3bd449a81b
+oid sha256:3f259b32bf55d57bb8de4dc4e845cc668dcbe9b235f56139f442b27833927d0c
 size 842609220