Create app.py

app.py

@@ -0,0 +1,84 @@
+import torch
+import spaces
+import numpy as np
+import gradio as gr
+from gtts import gTTS
+from transformers import pipeline
+from huggingface_hub import InferenceClient
+
+
+ASR_MODEL_NAME = "openai/whisper-small"
+NLP_MODEL_NAME = "mistralai/Mistral-7B-Instruct-v0.2"
+system_prompt = """"<s> [INST] You are Friday a helpful and conversational assistant. [/INST]"""
+
+client = InferenceClient(NLP_MODEL_NAME)
+
+device = 0 if torch.cuda.is_available() else "cpu"
+
+pipe = pipeline(
+    task="automatic-speech-recognition",
+    model=ASR_MODEL_NAME,
+    device=device,
+)
+
+
+def generate(prompt, temperature=0.1, max_new_tokens=64, top_p=0.95, repetition_penalty=1.0):
+    temperature = float(temperature)
+    if temperature < 1e-2:
+        temperature = 1e-2
+    top_p = float(top_p)
+
+    generate_kwargs = dict(
+        temperature=temperature,
+        max_new_tokens=max_new_tokens,
+        top_p=top_p,
+        repetition_penalty=repetition_penalty,
+        do_sample=True,
+        seed=42,
+    )
+
+    formatted_prompt = system_prompt + f""" {prompt} </s>"""
+
+    output = client.text_generation(
+        formatted_prompt, **generate_kwargs, stream=False, details=False, return_full_text=False)
+
+    print(output)
+    return output
+
+
+@spaces.GPU(duration=60)
+def transcribe(audio):
+    sr, y = audio
+    y = y.astype(np.float32)
+    y /= np.max(np.abs(y))
+
+    inputs = pipe({"sampling_rate": sr, "raw": y})["text"]
+
+    print("User transcription: ", inputs)
+
+    response = generate(inputs)
+    audio_response = gTTS(response)
+    audio_response.save("response.mp3")
+
+    print(audio_response)
+
+    return "response.mp3"
+
+
+with gr.Blocks() as demo:
+    gr.HTML("<center><h1>Friday: AI Virtual Assistant<h1><center>")
+
+    with gr.Row():
+        audio_input = gr.Audio(label="Human", sources="microphone")
+        output_audio = gr.Audio(label="Friday", type="filepath",
+                                interactive=False,
+                                autoplay=True,
+                                elem_classes="audio")
+
+    transcribe_btn = gr.Button("Transcribe")
+    transcribe_btn.click(fn=transcribe, inputs=audio_input,
+                         outputs=output_audio)
+
+
+demo.queue()
+demo.launch()