Update app.py

app.py

@@ -15,7 +15,7 @@ os.environ['TRANSFORMERS_CACHE'] = '/app/cache'
 os.makedirs('/app/cache', exist_ok=True)
 
 # Model konfigürasyonu
-MODEL_NAME = "ITU-TATL/gpt2-small-tr"
+MODEL_NAME = "redrussianarmy/gpt2-turkish-cased"
 
 @lru_cache(maxsize=1)
 def load_model():
@@ -30,7 +30,7 @@ def load_model():
         # CPU için float32 kullan
         model = AutoModelForCausalLM.from_pretrained(MODEL_NAME)
         model = model.to('cpu').float()  # Float32 formatına dönüştür
-        torch.set_num_threads(2)
+        torch.set_num_threads(1)
         logger.info("Model başarıyla yüklendi")
         return model, tokenizer
     except Exception as e:
@@ -66,11 +66,14 @@ def generate():
             outputs = model.generate(
                 inputs.input_ids,
                 attention_mask=inputs.attention_mask,
-                max_length=100,
+                max_length=80,
                 do_sample=True,
                 top_k=40,
                 temperature=0.7,
-                pad_token_id=tokenizer.pad_token_id
+                pad_token_id=tokenizer.pad_token_id,
+                num_return_sequences=1,
+                early_stopping=True,
+                use_cache=True
             )
         
         result = tokenizer.decode(outputs[0], skip_special_tokens=True)