update data2vec_paraformer demo

speech_asr

2023-02-10 edfdc05b8c6961f7a1ba4d23f36507eb49c928c8

update data2vec_paraformer demo

 egs_modelscope/asr/data2vec/speech_data2vec_pretrain-paraformer-zh-cn-aishell2-16k/finetune.py

@@ -25,13 +25,13 @@


if __name__ == '__main__':
    params = modelscope_args(model="damo/speech_data2vec_pretrain-zh-cn-aishell2-16k-pytorch",
    params = modelscope_args(model="damo/speech_data2vec_pretrain-paraformer-zh-cn-aishell2-16k",
                             data_path="./data")
    params.output_dir = "./checkpoint"
    params.data_path = "./example_data/"
    params.dataset_type = "small"
    params.batch_bins = 16000
    params.max_epoch = 50
    params.lr = 0.00005
    params.lr = 0.00002

    modelscope_finetune(params)

 egs_modelscope/asr/data2vec/speech_data2vec_pretrain-paraformer-zh-cn-aishell2-16k/infer.py

@@ -18,7 +18,7 @@
        os.environ['CUDA_VISIBLE_DEVICES'] = str(gpu_id)
    inference_pipline = pipeline(
        task=Tasks.auto_speech_recognition,
        model="damo/speech_data2vec_pretrain-zh-cn-aishell2-16k-pytorch",
        model="damo/speech_data2vec_pretrain-paraformer-zh-cn-aishell2-16k",
        output_dir=output_dir_job,
    )
    audio_in = os.path.join(split_dir, "wav.{}.scp".format(idx))

 egs_modelscope/asr/data2vec/speech_data2vec_pretrain-paraformer-zh-cn-aishell2-16k/infer_after_finetune.py

@@ -44,7 +44,7 @@

if __name__ == '__main__':
    params = {}
    params["modelscope_model_name"] = "damo/speech_data2vec_pretrain-zh-cn-aishell2-16k-pytorch"
    params["modelscope_model_name"] = "damo/speech_data2vec_pretrain-paraformer-zh-cn-aishell2-16k"
    params["required_files"] = ["am.mvn", "decoding.yaml", "configuration.json"]
    params["output_dir"] = "./checkpoint"
    params["data_dir"] = "./data/test"