added prompt debug

app.py

@@ -119,11 +119,11 @@ def generate_beam(model, tokenizer, beam_size: int = 5, prompt=None, embed=None,
     with torch.no_grad():
         if embed is not None:
             generated = embed
-        else:
-            if tokens is None:
-                tokens = torch.tensor(tokenizer.encode(prompt))
-                tokens = tokens.unsqueeze(0).to(device)
-                generated = model.gpt.transformer.wte(tokens)
+        if prompt is not 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)
         for i in range(entry_length):
             outputs = model.gpt(inputs_embeds=generated)
             logits = outputs.logits