Update app.py

app.py

@@ -48,7 +48,21 @@ user_role = "user"
 tts_model = TTS(language="EN_NEWEST", device="auto")
 speaker_id = tts_model.hps.data.spk2id["EN-Newest"]
 blocksize = 512
+tts_model.tts_to_file("text", speaker_id, quiet=True)
+dummy_input = torch.randn(
+        (1, 80, 3000),
+        dtype="float16",
+        device="cuda",
+)
 transcriber = pipeline("automatic-speech-recognition", model="openai/whisper-tiny.en", device="cuda")
+start_event = torch.cuda.Event(enable_timing=True)
+end_event = torch.cuda.Event(enable_timing=True)
+torch.cuda.synchronize()
+start_event.record()
+transcriber({"sampling_rate": sr, "raw": dummy_input})["text"]
+end_event.record()
+torch.cuda.synchronize()
+
 def int2float(sound):
     """
     Taken from https://github.com/snakers4/silero-vad
@@ -78,7 +92,21 @@ LM_model = AutoModelForCausalLM.from_pretrained(
 LM_pipe = pipeline(
     "text-generation", model=LM_model, tokenizer=LM_tokenizer, device="cuda"
 )
-
+dummy_input_text = "Write me a poem about Machine Learning."
+dummy_chat = [{"role": "user", "content": dummy_input_text}]
+start_event = torch.cuda.Event(enable_timing=True)
+end_event = torch.cuda.Event(enable_timing=True)
+torch.cuda.synchronize()
+start_event.record()
+LM_pipe(
+    dummy_chat,
+    max_new_tokens=32,
+    min_new_tokens=0,
+    temperature=0.0,
+    do_sample=False,
+)
+end_event.record()
+torch.cuda.synchronize()
 vad_model, _ = torch.hub.load("snakers4/silero-vad:v4.0", "silero_vad")
 vad_iterator = VADIterator(
     vad_model,