app fix

app.py

@@ -15,25 +15,8 @@ def get_model(model_name, model_path):
     return model, tokenizer
 
 
-def predict(text, model, tokenizer, n_beams=5, temperature=2.5, top_p=0.8, length_of_generated=300):
-    # text += '\n'
-    input_ids = tokenizer.encode(text, return_tensors="pt")
-    length_of_prompt = len(input_ids[0])
-    with torch.no_grad():
-        out = model.generate(input_ids,
-                             do_sample=True,
-                             num_beams=n_beams,
-                             temperature=temperature,
-                             top_p=top_p,
-                             max_length=length_of_prompt + length_of_generated,
-                             eos_token_id=tokenizer.eos_token_id
-                             )
-                          
-    generated = list(map(tokenizer.decode, out))[0]
-    return generated.replace('\n[EOS]\n', '')
-
 def predict_gpt(text, model, tokenizer,):
-    input_ids = tokenizer.encode(text, return_tensors="pt").to(device)
+    input_ids = tokenizer.encode(text, return_tensors="pt")
     with torch.no_grad():
         out = model.generate(input_ids, 
                             do_sample=True,
@@ -53,7 +36,7 @@ def predict_gpt(text, model, tokenizer,):
     return generated_text
 
 def predict_t5(text, model, tokenizer,):
-    input_ids = tokenizer.encode(text, return_tensors="pt").to(device)
+    input_ids = tokenizer.encode(text, return_tensors="pt")
     with torch.no_grad():
         out = model.generate(input_ids, 
                         do_sample=True,