import torch
import torch.nn as nn
import torch.nn.functional as F
import torch.optim as optim
import torchaudio
import sys

import matplotlib.pyplot as plt
import IPython.display as ipd

from tqdm import tqdm
import gradio as gr

device = torch.device("cuda" if torch.cuda.is_available() else "cpu")
sample_rate = 16000
new_sample_rate = 8000
transform = torchaudio.transforms.Resample(orig_freq=sample_rate, new_freq=new_sample_rate)

class M5(nn.Module):
    def __init__(self, n_input=1, n_output=35, stride=16, n_channel=32):
        super().__init__()
        self.conv1 = nn.Conv1d(n_input, n_channel, kernel_size=80, stride=stride)
        self.bn1 = nn.BatchNorm1d(n_channel)
        self.pool1 = nn.MaxPool1d(4)
        self.conv2 = nn.Conv1d(n_channel, n_channel, kernel_size=3)
        self.bn2 = nn.BatchNorm1d(n_channel)
        self.pool2 = nn.MaxPool1d(4)
        self.conv3 = nn.Conv1d(n_channel, 2 * n_channel, kernel_size=3)
        self.bn3 = nn.BatchNorm1d(2 * n_channel)
        self.pool3 = nn.MaxPool1d(4)
        self.conv4 = nn.Conv1d(2 * n_channel, 2 * n_channel, kernel_size=3)
        self.bn4 = nn.BatchNorm1d(2 * n_channel)
        self.pool4 = nn.MaxPool1d(4)
        self.fc1 = nn.Linear(2 * n_channel, n_output)

    def forward(self, x):
        x = self.conv1(x)
        x = F.relu(self.bn1(x))
        x = self.pool1(x)
        x = self.conv2(x)
        x = F.relu(self.bn2(x))
        x = self.pool2(x)
        x = self.conv3(x)
        x = F.relu(self.bn3(x))
        x = self.pool3(x)
        x = self.conv4(x)
        x = F.relu(self.bn4(x))
        x = self.pool4(x)
        x = F.avg_pool1d(x, x.shape[-1])
        x = x.permute(0, 2, 1)
        x = self.fc1(x)
        return F.log_softmax(x, dim=2)

def get_likely_index(tensor):
    # find most likely label index for each element in the batch
    return tensor.argmax(dim=-1)
 
def index_to_label(index):
    # Return the word corresponding to the index in labels
    # This is the inverse of label_to_index
    return labels[index]
    
def predict(filepath):
    tensor=(torchaudio.load(filepath))[0]
    # Use the model to predict the label of the waveform
    tensor = tensor.to(device)
    tensor = transform(tensor)
    tensor = model(tensor.unsqueeze(0))
    tensor = get_likely_index(tensor)
    tensor = index_to_label(tensor.squeeze())
    return tensor
    
def record(seconds=1):

    from google.colab import output as colab_output
    from base64 import b64decode
    from io import BytesIO
    from pydub import AudioSegment

    RECORD = (
        b"const sleep  = time => new Promise(resolve => setTimeout(resolve, time))\n"
        b"const b2text = blob => new Promise(resolve => {\n"
        b"  const reader = new FileReader()\n"
        b"  reader.onloadend = e => resolve(e.srcElement.result)\n"
        b"  reader.readAsDataURL(blob)\n"
        b"})\n"
        b"var record = time => new Promise(async resolve => {\n"
        b"  stream = await navigator.mediaDevices.getUserMedia({ audio: true })\n"
        b"  recorder = new MediaRecorder(stream)\n"
        b"  chunks = []\n"
        b"  recorder.ondataavailable = e => chunks.push(e.data)\n"
        b"  recorder.start()\n"
        b"  await sleep(time)\n"
        b"  recorder.onstop = async ()=>{\n"
        b"    blob = new Blob(chunks)\n"
        b"    text = await b2text(blob)\n"
        b"    resolve(text)\n"
        b"  }\n"
        b"  recorder.stop()\n"
        b"})"
    )
    RECORD = RECORD.decode("ascii")

    print(f"Recording started for {seconds} seconds.")
    display(ipd.Javascript(RECORD))
    s = colab_output.eval_js("record(%d)" % (seconds * 1000))
    print("Recording ended.")
    b = b64decode(s.split(",")[1])

    fileformat = "wav"
    filename = f"_audio.{fileformat}"
    AudioSegment.from_file(BytesIO(b)).export(filename, format=fileformat)
    return torchaudio.load(filename)

    
model = torch.load('export.pkl',map_location=torch.device('cpu'))

  
      
gr.Interface(fn=predict, inputs=gr.inputs.Audio(source=record()[0]), outputs=gr.outputs.Label(num_top_classes=3)).launch(share=True)