app.py

# Import necessary libraries
import gradio as gr
import torch
import torch.nn as nn
from torch.utils.data import Dataset, DataLoader
from transformers import WhisperModel, WhisperFeatureExtractor
import datasets
from datasets import load_dataset, DatasetDict, Audio
from huggingface_hub import PyTorchModelHubMixin

# Define data class
class SpeechInferenceDataset(Dataset):
    def __init__(self, audio_data, text_processor):
        self.audio_data = audio_data
        self.text_processor = text_processor

    def __len__(self):
        return len(self.audio_data)

    def __getitem__(self, index):
        inputs = self.text_processor(self.audio_data[index]["audio"]["array"],
                                     return_tensors="pt",
                                     sampling_rate=self.audio_data[index]["audio"]["sampling_rate"])
        input_features = inputs.input_features
        # Assuming 'encoder' is defined or available in the scope
        decoder_input_ids = torch.tensor([[1, 1]]) * encoder.config.decoder_start_token_id
        return input_features, decoder_input_ids

# Define model class
class SpeechClassifier(nn.Module, PyTorchModelHubMixin):
    def __init__(self, config):
        super(SpeechClassifier, self).__init__()
        self.encoder = WhisperModel.from_pretrained(config["encoder"])
        self.classifier = nn.Sequential(
            nn.Linear(self.encoder.config.hidden_size, 4096),
            nn.ReLU(),
            nn.Linear(4096, 2048),
            nn.ReLU(),
            nn.Linear(2048, 1024),
            nn.ReLU(),
            nn.Linear(1024, 512),
            nn.ReLU(),
            nn.Linear(512, config["num_labels"])
        )

    def forward(self, input_features, decoder_input_ids):
        outputs = self.encoder(input_features, decoder_input_ids=decoder_input_ids)
        pooled_output = outputs['last_hidden_state'][:, 0, :]
        logits = self.classifier(pooled_output)
        return logits

# Prepare data function
def prepare_data(audio_file_path, model_checkpoint="openai/whisper-base"):
    feature_extractor = WhisperFeatureExtractor.from_pretrained(model_checkpoint)
    inference_data = datasets.Dataset.from_dict({"path": [audio_file_path], "audio": [audio_file_path]}).cast_column("audio", Audio(sampling_rate=16_000))
    inference_dataset = SpeechInferenceDataset(inference_data, feature_extractor)
    inference_loader = DataLoader(inference_dataset, batch_size=1, shuffle=False)
    input_features, decoder_input_ids = next(iter(inference_loader))
    # Ensure 'device' is defined or replace with 'torch.device("cpu")' if GPU is not available
    input_features = input_features.squeeze(1).to(device)
    decoder_input_ids = decoder_input_ids.squeeze(1).to(device)
    return input_features, decoder_input_ids

# Prediction function
def predict(audio_file_path, config={"encoder": "openai/whisper-base", "num_labels": 2}):
    input_features, decoder_input_ids = prepare_data(audio_file_path)
    model = SpeechClassifier(config)
    model.eval()
    with torch.no_grad():
        logits = model(input_features, decoder_input_ids)
        predicted_ids = int(torch.argmax(logits, dim=-1))
    return predicted_ids

# Gradio Interface function
def gradio_interface(uploaded_file):
    with open(uploaded_file.name, "wb") as f:
        f.write(uploaded_file.read())
    prediction = predict(uploaded_file.name)
    label = "Hypernasality Detected" if prediction == 1 else "No Hypernasality Detected"
    return label

# Create and launch Gradio Interface with File upload input
iface = gr.Interface(fn=gradio_interface, 
                     inputs=gr.inputs.File(label="Upload Audio File"), 
                     outputs="text")
iface.launch()