Initial kimetsu-style tiny GPT

README.md

@@ -13,8 +13,8 @@ tags:
 
 # Kimetsu-Style GPT (Tiny)
 
-**“鬼滅風”のセリフを生成する超小型デモモデル**（教育/勉強会向け）。  
-このノートで、**自作コーパス → トークナイズ → 事前学習 → 生成 → 公開**までを約1時間で再現しています。
+**“鬼滅風”のセリフを生成する超小型デモモデル**（教育/勉強会向け）。
+このノートで、**自作コーパス → トークナイズ → 事前学習 → 生成 → 公開**まで実施しています。
 
 ## Training
 - Tokenizer: ByteLevel BPE (~5000)

config.json

@@ -4,7 +4,7 @@
     "GPT2LMHeadModel"
   ],
   "attn_pdrop": 0.1,
-  "bos_token_id": 0,
+  "bos_token_id": 1,
   "dtype": "float32",
   "embd_pdrop": 0.1,
   "eos_token_id": 2,
@@ -28,5 +28,5 @@
   "summary_use_proj": true,
   "transformers_version": "4.56.1",
   "use_cache": true,
-  "vocab_size": 412
+  "vocab_size": 32000
 }

generation_config.json

@@ -1,6 +1,6 @@
 {
   "_from_model_config": true,
-  "bos_token_id": 0,
+  "bos_token_id": 1,
   "eos_token_id": 2,
   "transformers_version": "4.56.1"
 }

model.safetensors

@@ -1,3 +1,3 @@
 version https://git-lfs.github.com/spec/v1
-oid sha256:c0afc1c59b629af6065545d201338189d736b716f1a78d5c878209a8abad3c3f
-size 6875816
+oid sha256:504d915f248c64adad7ad6677af48d648944d1d93c33fb4e44c650a1a5e3aed8
+size 39221936

special_tokens_map.json

@@ -1,11 +1,6 @@
 {
-  "bos_token": {
-    "content": "<s>",
-    "lstrip": false,
-    "normalized": false,
-    "rstrip": false,
-    "single_word": false
-  },
+  "bos_token": "<s>",
+  "cls_token": "[CLS]",
   "eos_token": {
     "content": "</s>",
     "lstrip": false,
@@ -13,13 +8,15 @@
     "rstrip": false,
     "single_word": false
   },
+  "mask_token": "[MASK]",
   "pad_token": {
-    "content": "<pad>",
+    "content": "[PAD]",
     "lstrip": false,
     "normalized": false,
     "rstrip": false,
     "single_word": false
   },
+  "sep_token": "[SEP]",
   "unk_token": {
     "content": "<unk>",
     "lstrip": false,

spiece.model

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:b5cbdfa8aa7c54c8c5af85b78c309c54a5f2749a20468bf6f60eee007fe6fec1
+size 805634

tokenizer_config.json

@@ -1,7 +1,8 @@
 {
+  "add_prefix_space": null,
   "added_tokens_decoder": {
     "0": {
-      "content": "<s>",
+      "content": "<unk>",
       "lstrip": false,
       "normalized": false,
       "rstrip": false,
@@ -9,7 +10,7 @@
       "special": true
     },
     "1": {
-      "content": "<pad>",
+      "content": "<s>",
       "lstrip": false,
       "normalized": false,
       "rstrip": false,
@@ -25,7 +26,31 @@
       "special": true
     },
     "3": {
-      "content": "<unk>",
+      "content": "[PAD]",
+      "lstrip": false,
+      "normalized": false,
+      "rstrip": false,
+      "single_word": false,
+      "special": true
+    },
+    "4": {
+      "content": "[CLS]",
+      "lstrip": false,
+      "normalized": false,
+      "rstrip": false,
+      "single_word": false,
+      "special": true
+    },
+    "5": {
+      "content": "[SEP]",
+      "lstrip": false,
+      "normalized": false,
+      "rstrip": false,
+      "single_word": false,
+      "special": true
+    },
+    "6": {
+      "content": "[MASK]",
       "lstrip": false,
       "normalized": false,
       "rstrip": false,
@@ -33,12 +58,20 @@
       "special": true
     }
   },
+  "additional_special_tokens": [],
   "bos_token": "<s>",
   "clean_up_tokenization_spaces": false,
+  "cls_token": "[CLS]",
+  "do_lower_case": true,
   "eos_token": "</s>",
+  "extra_ids": 0,
   "extra_special_tokens": {},
+  "legacy": true,
+  "mask_token": "[MASK]",
   "model_max_length": 1000000000000000019884624838656,
-  "pad_token": "<pad>",
-  "tokenizer_class": "PreTrainedTokenizerFast",
+  "pad_token": "[PAD]",
+  "sep_token": "[SEP]",
+  "sp_model_kwargs": {},
+  "tokenizer_class": "T5Tokenizer",
   "unk_token": "<unk>"
 }

training_args.bin

@@ -1,3 +1,3 @@
 version https://git-lfs.github.com/spec/v1
-oid sha256:7b761e97981f33bde7c70dfe5c9b438eac49034f1648cbdce0a9a315b71b86e5
+oid sha256:83477d2c8dfea6010fdbc937e2b76bfb8940d94a0bd43cdb4777bf525745985d
 size 5777