Epoch 10 - Training started: 2025-11-15 21:13:58

README.md

@@ -6,7 +6,7 @@ NNUE (Efficiently Updatable Neural Network) chess evaluation model
 
 - **Model type**: NNUE Chess Evaluation
 - **Architecture**: HalfKP feature representation
-- **Uploaded**: 2025-11-15 19:27:29
+- **Uploaded**: 2025-11-15 21:32:57
 
 ## Architecture
 
@@ -29,8 +29,8 @@ Output Layer: 32 → 1 (centipawn evaluation)
 ## Training Information
 
 - **Epoch**: 10
-- **Training Loss**: 3120906.5251
-- **Validation Loss**: 2923721.9199
+- **Training Loss**: 2747954.3011
+- **Validation Loss**: 2758574.0879
 
 ## Usage
 

checkpoint_epoch_10.pt

@@ -1,3 +1,3 @@
 version https://git-lfs.github.com/spec/v1
-oid sha256:fe4f79c0dd385b53af3bb7d104ae650b21bee0e676802c5eb2fe8323c07a98f4
-size 126054987
+oid sha256:55a5ee308d6e86f8e68306c76cbf464a7ae39b68b6c2c48f8f4e824256cfbd9c
+size 126054859

config.json

@@ -34,6 +34,8 @@
     "warmup_epochs": 5,
     "warmup_start_lr": 1e-05,
     "loss_function": "mse",
+    "eval_score_min": -10000,
+    "eval_score_max": 10000,
     "log_every_n_batches": 100,
     "verbose": true,
     "seed": 42,
@@ -41,11 +43,11 @@
     "stockfish_path": "stockfish",
     "download_year": null,
     "download_month": null,
-    "download_max_games": 10000,
+    "download_max_games": 100000,
     "download_max_games_searched": null,
     "download_depth": 10,
     "download_positions_per_game": 10,
-    "download_num_workers": 4,
+    "download_num_workers": 32,
     "download_batch_size": 100,
     "download_rated_only": true,
     "download_output_dir": "data",

pytorch_model.bin

@@ -1,3 +1,3 @@
 version https://git-lfs.github.com/spec/v1
-oid sha256:d729af8233d872c00bd05b36bd5ee3d9b11a38b3d6d9abbeac94c7e0855d2bba
-size 42017953
+oid sha256:e61fc4bb9ab8e5bad96550a7bcaf1fa8df33e1e4cb7e0e0b172e7d583a369d3f
+size 42017825