Update app.py

app.py

@@ -15,11 +15,11 @@ def load_model(name: str) -> Any:
 def load_tokenizer(name: str) -> Any:
     return AutoTokenizer.from_pretrained(name)
 
-def create_generator():
+def create_generator(temperature, top_p, num_beams):
     return GenerationConfig(
-    temperature=1.0,
-    top_p=0.75,
-    num_beams=4,
+    temperature=temperature,
+    top_p=top_p,
+    num_beams=num_beams,
 )
     
 def generate_prompt(instruction, input=None):
@@ -46,7 +46,7 @@ def generate_prompt(instruction, input=None):
 generation_config = create_generator()
 
 
-def evaluate(instruction, input, model, tokenizer):
+def evaluate(instruction, input, model, tokenizer, generation_config):
     prompt = generate_prompt(instruction, input)
     inputs = tokenizer(prompt, return_tensors="pt")
     input_ids = inputs["input_ids"]
@@ -64,9 +64,10 @@ def evaluate(instruction, input, model, tokenizer):
     return ' '.join(el for el in result)            
 
 def inference(model_name, text, input, temperature, top_p, num_beams):
+    generation_config = create_generator(temperature, top_p, num_beams)
     model = load_model(model_name)
     tokenizer = load_tokenizer(model_name)
-    output = evaluate(instruction = text, input = input, model = model, tokenizer = tokenizer)
+    output = evaluate(instruction = text, input = input, model = model, tokenizer = tokenizer, generation_config = generation_config)
     return output
 
 def choose_model(name):