fixed prompt

app.py

@@ -123,7 +123,8 @@ def generate_beam(model, tokenizer, beam_size: int = 5, prompt=None, embed=None,
             tokens = torch.tensor(tokenizer.encode(prompt))
             tokens = tokens.unsqueeze(0).to(device)
             prompt_tokens = model.gpt.transformer.wte(tokens)
-        print(">>>>", generated.shape, prompt_tokens.shape)
+        generated = torch.cat((generated, prompt_tokens), dim=1)
+
         for i in range(entry_length):
             outputs = model.gpt(inputs_embeds=generated)
             logits = outputs.logits