Update README.md

README.md

@@ -1,4 +1,3 @@
-
 ---
 base_model: meta-llama/Llama-2-8b-chat-hf
 library_name: peft
@@ -58,18 +57,32 @@ pip install torch transformers peft accelerate bitsandbytes
 ### Kullanım
 
 ```python
-from transformers import AutoTokenizer, AutoModelForCausalLM
+from transformers import AutoTokenizer, AutoModelForCausalLM, BitsAndBytesConfig
 from peft import PeftModel
 import torch
 
+# Quantization konfigürasyonu
+quantization_config = BitsAndBytesConfig(
+    load_in_4bit=True,
+    bnb_4bit_compute_dtype=torch.float16,
+    bnb_4bit_quant_type="nf4",
+    bnb_4bit_use_double_quant=True,
+)
+
 # Model ve tokenizer yükleme
 model_name = "your-username/llama-8b-mold-protection"
 tokenizer = AutoTokenizer.from_pretrained(model_name)
+
+# Pad token ayarlama
+if tokenizer.pad_token is None:
+    tokenizer.pad_token = tokenizer.eos_token
+
 model = AutoModelForCausalLM.from_pretrained(
     model_name,
     torch_dtype=torch.float16,
     device_map="auto",
-    load_in_4bit=True
+    quantization_config=quantization_config,
+    trust_remote_code=True
 )
 
 # Log analizi örneği
@@ -80,20 +93,37 @@ def analyze_log(log_entry):
 
 Açıklama ve Çözüm:"""
     
-    inputs = tokenizer(prompt, return_tensors="pt")
+    # Input'ları modelle aynı cihaza taşı
+    inputs = tokenizer(
+        prompt, 
+        return_tensors="pt", 
+        padding=True, 
+        truncation=True, 
+        max_length=512
+    )
+    
+    # Input tensor'ları GPU'ya taşı
+    inputs = {k: v.to(model.device) for k, v in inputs.items()}
     
     with torch.no_grad():
         outputs = model.generate(
             **inputs,
-            max_length=512,
+            max_new_tokens=256,  # max_length yerine max_new_tokens kullan
             temperature=0.7,
             do_sample=True,
             top_p=0.9,
-            pad_token_id=tokenizer.eos_token_id
+            top_k=50,
+            repetition_penalty=1.1,
+            pad_token_id=tokenizer.eos_token_id,
+            eos_token_id=tokenizer.eos_token_id,
+            early_stopping=True
         )
     
-    response = tokenizer.decode(outputs[0], skip_special_tokens=True)
-    return response.split("Açıklama ve Çözüm:")[-1].strip()
+    # Sadece yeni generate edilen kısmı al
+    generated_tokens = outputs[0][inputs['input_ids'].shape[1]:]
+    response = tokenizer.decode(generated_tokens, skip_special_tokens=True)
+    
+    return response.strip()
 
 # Örnek kullanım
 log_sample = "2025-01-15 14:30:22 | MoldProtection | CRITICAL | KALIP KORUMA UYARISI - Hatalı ROI'ler: ROI 1, ROI 2 | Tetikleyici: plc"