mod gpt

gpt.py

@@ -16,10 +16,7 @@ def prepare_simple_dataset():
 
 def main():
     # Inicjalizacja tokenizera
-    tokenizer = AutoTokenizer.from_pretrained(
-        MODEL_NAME,
-        mean_resizing=False
-    )
+    tokenizer = AutoTokenizer.from_pretrained(MODEL_NAME)
     tokenizer.add_special_tokens({"additional_special_tokens": SPECIAL_TOKENS})
     tokenizer.pad_token = tokenizer.eos_token
 
@@ -42,7 +39,10 @@ def main():
     tokenized_dataset = dataset.map(tokenize_function, batched=True)
 
     # Model i data collator
-    model = AutoModelForCausalLM.from_pretrained(MODEL_NAME)
+    model = AutoModelForCausalLM.from_pretrained(
+        MODEL_NAME,
+        mean_resizing=False
+    )
     model.resize_token_embeddings(len(tokenizer))
     
     data_collator = DataCollatorForLanguageModeling(