Create app.py

app.py

@@ -0,0 +1,159 @@
+from modelscope.pipelines import pipeline
+from modelscope.utils.constant import Tasks
+import gradio as gr
+import os
+import shutil
+import os.path as osp
+from modelscope.tools import run_auto_label
+from modelscope.metainfo import Trainers
+from modelscope.trainers import build_trainer
+from modelscope.utils.audio.audio_utils import TtsTrainType
+from modelscope.models.audio.tts import SambertHifigan
+import IPython.display as ipd
+
+def launch_training_task(*audio_lst):
+    os.makedirs("/tmp/test_wavs")
+    os.makedirs("/tmp/output_training_data")
+    os.makedirs("/tmp/pretrain_work_dir")
+    for num in range(len(audio_lst)):
+        if num<10:
+            shutil.copy(audio_lst[num], "/tmp/test_wavs/01_00000"+str(num)+".wav")
+        else:
+            shutil.copy(audio_lst[num], "/tmp/test_wavs/01_0000"+str(num)+".wav")
+    
+    input_wav = "/tmp/test_wavs/"
+    output_data = "/tmp/output_training_data/"
+    ret, report = run_auto_label(input_wav=input_wav, work_dir=output_data, resource_revision="v1.0.7")
+
+    pretrained_model_id = 'damo/speech_personal_sambert-hifigan_nsf_tts_zh-cn_pretrain_16k'
+    dataset_id = "/tmp/output_training_data/"
+    pretrain_work_dir = "/tmp/pretrain_work_dir/"
+
+    train_info = {
+        TtsTrainType.TRAIN_TYPE_SAMBERT: {  # 配置训练AM（sambert）模型
+            'train_steps': 202,               # 训练多少个step 
+            'save_interval_steps': 200,       # 每训练多少个step保存一次checkpoint
+            'log_interval': 10               # 每训练多少个step打印一次训练日志
+        }
+    }
+
+    kwargs = dict(
+        model=pretrained_model_id,                  # 指定要finetune的模型
+        model_revision = "v1.0.6",
+        work_dir=pretrain_work_dir,                 # 指定临时工作目录
+        train_dataset=dataset_id,                   # 指定数据集id
+        train_type=train_info                       # 指定要训练类型及参数
+    )
+
+    trainer = build_trainer(Trainers.speech_kantts_trainer,default_args=kwargs)
+
+    trainer.train()
+
+    return "已训练完成"
+
+
+def greet(audio):
+    print(audio)
+    inference_pipeline = pipeline(task=Tasks.auto_speech_recognition, model='damo/speech_paraformer-large_asr_nat-zh-cn-16k-common-vocab8404-pytorch')
+    rec_result = inference_pipeline(audio_in=audio)
+    input_gpt3 = rec_result['text']+"|"
+
+    text_generation_zh = pipeline(Tasks.text_generation, model='damo/nlp_gpt3_text-generation_chinese-base')
+    result_gpt = text_generation_zh(input_gpt3)
+
+    input_pttp = result_gpt[result_gpt.find("|")+1:]
+
+    model_dir = os.path.abspath("/tmp/pretrain_work_dir")
+
+    custom_infer_abs = {
+        'voice_name':
+        'F7',
+        'am_ckpt':
+        os.path.join(model_dir, 'tmp_am', 'ckpt'),
+        'am_config':
+        os.path.join(model_dir, 'tmp_am', 'config.yaml'),
+        'voc_ckpt':
+        os.path.join(model_dir, 'orig_model', 'basemodel_16k', 'hifigan', 'ckpt'),
+        'voc_config':
+        os.path.join(model_dir, 'orig_model', 'basemodel_16k', 'hifigan',
+                 'config.yaml'),
+        'audio_config':
+        os.path.join(model_dir, 'data', 'audio_config.yaml'),
+        'se_file':
+        os.path.join(model_dir, 'data', 'se', 'se.npy')
+    }
+    kwargs = {'custom_ckpt': custom_infer_abs}
+
+    model_id = SambertHifigan(os.path.join(model_dir, "orig_model"), **kwargs)
+
+    inference = pipeline(task=Tasks.text_to_speech, model=model_id)
+    output = inference(input=input_pttp)
+
+    return rec_result['text'],ipd.Audio(output["output_wav"], rate=16000)
+
+with gr.Blocks() as demo:
+    with gr.Row():
+        with gr.Column(scale=1):
+            audio_lst1 = [
+                gr.Audio(label="1. 希望我们大家都能像他一样"),
+                gr.Audio(label="2. 不行, 他想了一下, 我不能这样对国王说, 这是在撒谎"),
+                gr.Audio(label="3. 但他们非常和气地问她说, 你叫什么名字"),
+                gr.Audio(label="4. 鸭子心想, 我必须去拿回我的软糖豆"),
+                gr.Audio(label="5. 小朋友, 你们不要再欺负它了"),
+            ]
+        with gr.Column(scale=1):
+            audio_lst2 = [
+                gr.Audio(label="6. 可是, 小黄鸭并不怕他们"),
+                gr.Audio(label="7. 然后, 他们一起走了很长一段时间"),
+                gr.Audio(label="8. 突然, 墙壁后面传来一阵声音"),
+                gr.Audio(label="9. 结果盘子掉在地上, 打得粉碎"),
+                gr.Audio(label="10. 四个小伙伴很开心, 一起感谢小松鼠的帮助"),
+            ]
+        with gr.Column(scale=1):
+            audio_lst3 = [
+                gr.Audio(label="11. 不过, 当他看到拇指姑娘的时候, 他马上就变得高兴起来"),
+                gr.Audio(label="12. 从此以后, 他过上了幸福的生活"),
+                gr.Audio(label="13. 老山羊最后伤心地, 哭着走了出去"),
+                gr.Audio(label="14. 而且准备一��找下去, 直到他走不动为止"),
+                gr.Audio(label="15. 海马先生轻轻游过大海"),
+            ]
+        with gr.Column(scale=1):
+            audio_lst4 = [
+                gr.Audio(label="16. 一起高高兴兴地, 回到了他们的爸爸妈妈身边"),
+                gr.Audio(label="17. 艾丽莎很小不能去上学, 但她有一个非常贵重精美的画册"),
+                gr.Audio(label="18. 狮子还是够不着, 它叫来了狐狸"),
+                gr.Audio(label="19. 姑娘坐到国王的马车上, 和国王一起回到宫中"),
+                gr.Audio(label="20. 温妮大叫了起来, 现在我们该怎么回家呀"),
+            ]
+        audio_list = audio_lst1 + audio_lst2 + audio_lst3 + audio_lst4
+
+        train_res = gr.Textbox(label="训练结果", value="当前无训练任务")
+
+        training_button = gr.Button("开始训练")
+    
+        training_button.click(launch_training_task,inputs=audio_list,outputs=train_res)
+
+
+    gr.Markdown("# 快来和模型对话吧")
+
+    with gr.Tab("麦克风"):
+        in_audio = gr.Audio(source='microphone', type='filepath')
+        button = gr.Button("开始运行", variant="primary")
+        output_txt = gr.Textbox(label='文本')
+        output_label = gr.Audio(label='音频')
+        button.click(greet,
+                     inputs=in_audio,
+                     outputs=[output_txt, output_label])
+
+    with gr.Tab("上传音频"):
+        in_audio = gr.Audio(type='filepath')
+        button = gr.Button("开始运行", variant="primary")
+        output_txt = gr.Textbox(label='文本')
+        output_label = gr.Audio(label='音频')
+        button.click(greet,
+                     inputs=in_audio,
+                     outputs=[output_txt, output_label])
+
+gr.close_all()
+demo.queue(concurrency_count=5)
+demo.launch()