Update compounds GPT-2 small (retokenized, fix #12)

config.json

@@ -1,43 +1,34 @@
 {
-  "_name_or_path": "learning_source_20260316/compounds/gpt2-output/compounds-small/checkpoint-6000",
-  "_riken_bias": false,
-  "_riken_model_args": {
-    "bias": false,
-    "block_size": 1024,
-    "dropout": 0.0,
-    "n_embd": 768,
-    "n_head": 12,
-    "n_layer": 12,
-    "vocab_size": 612
-  },
-  "activation_function": "gelu_new",
   "architectures": [
     "GPT2LMHeadModel"
   ],
-  "attn_pdrop": 0.0,
-  "bos_token_id": 0,
-  "embd_pdrop": 0.0,
-  "eos_token_id": 0,
-  "initializer_range": 0.02,
-  "layer_norm_epsilon": 1e-05,
   "model_type": "gpt2",
+  "vocab_size": 612,
+  "n_positions": 1024,
   "n_ctx": 1024,
   "n_embd": 768,
+  "n_layer": 12,
   "n_head": 12,
   "n_inner": 3072,
-  "n_layer": 12,
-  "n_positions": 1024,
-  "reorder_and_upcast_attn": false,
+  "activation_function": "gelu_new",
   "resid_pdrop": 0.0,
-  "scale_attn_by_inverse_layer_idx": false,
-  "scale_attn_weights": true,
-  "summary_activation": null,
-  "summary_first_dropout": 0.1,
-  "summary_proj_to_labels": true,
-  "summary_type": "cls_index",
-  "summary_use_proj": true,
-  "torch_dtype": "float32",
-  "transformers_version": "4.45.1",
+  "embd_pdrop": 0.0,
+  "attn_pdrop": 0.0,
+  "layer_norm_epsilon": 1e-05,
+  "initializer_range": 0.02,
   "use_cache": true,
-  "vocab_size": 612
-}
+  "bos_token_id": 0,
+  "eos_token_id": 0,
+  "transformers_version": "4.0.0",
+  "_name_or_path": "riken-gpt2",
+  "_riken_model_args": {
+    "n_layer": 12,
+    "n_head": 12,
+    "n_embd": 768,
+    "block_size": 1024,
+    "bias": false,
+    "vocab_size": 612,
+    "dropout": 0.0
+  },
+  "_riken_bias": false
+}

pytorch_model.bin

@@ -1,3 +1,3 @@
 version https://git-lfs.github.com/spec/v1
-oid sha256:3d56393096242d67d18968e31823147a9494aed1d383b2dce0535ccbbcf5ff55
+oid sha256:6672efa2908532b33521405cf67139238c56f00800366d0b67facaf5974768d6
 size 344866536

training_args.json

@@ -1,7 +1,7 @@
 {
   "iteration": 6000,
-  "best_val_loss": 0.030465448275208473,
-  "early_stopping_counter": 1,
+  "best_val_loss": 0.030528374016284943,
+  "early_stopping_counter": 10,
   "learning_rate": 6e-06,
   "batch_size": 8,
   "block_size": 1024,