app.py

import gradio as gr
import torch
import torchaudio
from torchaudio.transforms import Resample
import importlib.util

# Function to dynamically import wav2vec2 module and avoid duplicate registration
def import_wav2vec2():
    if 'wav2vec2' not in sys.modules:
        spec = importlib.util.spec_from_file_location("wav2vec2", "wav2vec2.py")
        wav2vec2 = importlib.util.module_from_spec(spec)
        sys.modules['wav2vec2'] = wav2vec2
        spec.loader.exec_module(wav2vec2)
    else:
        wav2vec2 = sys.modules['wav2vec2']
    
    Wav2Vec2Model = wav2vec2.Wav2Vec2Model
    Wav2Vec2Config = wav2vec2.Wav2Vec2Config
    return Wav2Vec2Model, Wav2Vec2Config

Wav2Vec2Model, Wav2Vec2Config = import_wav2vec2()

# 定义模型路径
model_path = "https://huggingface.co/Tele-AI/TeleSpeech-ASR1.0/resolve/main/finetune_large_kespeech.pt"

# 下载模型文件
print("Downloading model file...")
torch.hub.download_url_to_file(model_path, 'large.pt')
print("Model file downloaded.")

# 加载模型配置和初始化模型
config = Wav2Vec2Config()
model = Wav2Vec2Model.build_model(config)

# 加载模型参数
print("Loading model checkpoint...")
checkpoint = torch.load('large.pt', map_location=torch.device('cpu'))
print("Checkpoint keys:", checkpoint.keys())

# 打印模型参数中的键
if 'model' in checkpoint:
    state_dict = checkpoint['model']
    print("Model state_dict keys:", state_dict.keys())
else:
    print("Key 'model' not found in checkpoint.")
    state_dict = checkpoint

# 加载模型状态字典
try:
    model.load_state_dict(state_dict)
    print("Model state_dict loaded successfully.")
except Exception as e:
    print("Error loading model state_dict:", str(e))

model.eval()

# 定义处理函数
def transcribe(audio):
    print("Transcribing audio...")
    waveform, sample_rate = torchaudio.load(audio)
    if sample_rate != 16000:
        resample = Resample(orig_freq=sample_rate, new_freq=16000)
        waveform = resample(waveform).squeeze()
    else:
        waveform = waveform.squeeze()

    # 将输入数据转换为符合模型预期的形状
    input_values = waveform.unsqueeze(0)  # (batch_size, seq_len)

    with torch.no_grad():
        outputs = model.extract_features(input_values, padding_mask=None)
        logits = outputs["x"]
    predicted_ids = torch.argmax(logits, dim=-1)
    transcription = ''.join([chr(i) for i in predicted_ids[0].tolist()])  # 解码预测到字符
    print("Transcription:", transcription)
    return transcription

# 创建 Gradio 界面
iface = gr.Interface(
    fn=transcribe,
    inputs=gr.Audio(type="filepath"),
    outputs="text",
    title="TeleSpeech ASR",
    description="Upload an audio file or record your voice to transcribe speech to text using the TeleSpeech ASR model."
)

print("Launching Gradio interface...")
iface.launch()