Update app.py

app.py

@@ -104,8 +104,7 @@ def generate(input, instruction=DEFAULT_INPUT, sampling=False, temperature=0.7,
         omni_content = build_omni_chunks(input) + [instruction]
         sys_msg = repo.get_sys_prompt(mode="omni", language="en")
         msgs = [sys_msg, {"role": "user", "content": omni_content}]
-        params = dict(msgs=msgs, tokenizer=tokenizer, omni_input=True, **kw)
-        return repo.chat(**params)
+        print(msgs)
     elif filetype == "Audio":
         audio_np, sample_rate = librosa.load(input, sr=16000, mono=True)
         chunk_tensor = torch.from_numpy(audio_np).float().to(DEVICE)
@@ -130,7 +129,7 @@ def generate(input, instruction=DEFAULT_INPUT, sampling=False, temperature=0.7,
     inputs_payload = [{"role": "user", "content": content}]
 
     params = {
-        "msgs": inputs_payload,
+        "msgs": msgs or inputs_payload,
         "tokenizer": tokenizer,
         "sampling": sampling,
         "temperature": temperature,
@@ -138,6 +137,7 @@ def generate(input, instruction=DEFAULT_INPUT, sampling=False, temperature=0.7,
         "top_k": top_k,
         "repetition_penalty": repetition_penalty,
         "max_new_tokens": max_tokens,
+        "omni_input": filetype == "Video",
     }
 
     output = repo.chat(**params)