test

app.py

@@ -14,15 +14,15 @@ tokenizer = AutoTokenizer.from_pretrained(tokenizer_name, trust_remote_code=True
 device = "cuda" if torch.cuda.is_available() else "cpu"
 model.to(device)
 
-# Define generation configuration with a maximum length
+# Define generation configuration
 generation_config = GenerationConfig(
-    max_length=150,               # Set a maximum length for output
+    max_length=100,               # Adjust this based on your translation requirements
     max_new_tokens=50,            # Ensure sufficient tokens for your translations
     num_beams=5,                  # Moderate number of beams for a balance between speed and quality
     do_sample=False,              # Disable sampling to make output deterministic
     temperature=1.0,              # Neutral temperature since sampling is off
-    top_k=None,                   # Set to None for deterministic generation
-    top_p=None,                   # Set to None for deterministic generation
+    top_k=0,                      # Disable top-k sampling (since sampling is off)
+    top_p=0,                      # Disable top-p (nucleus) sampling (since sampling is off)
     repetition_penalty=4.0,       # Neutral repetition penalty for translation
     length_penalty=3.0,           # No penalty for sequence length; modify if your translations tend to be too short/long
     early_stopping=True            # Stop early when all beams finish to speed up generation