NPU-ASLP

yufan-aslp

2023-02-16 c78474c97907c0a24c302a3952f3389360909fd9

NPU-ASLP

 egs_modelscope/asr/mfcca/speech_mfcca_asr-zh-cn-16k-alimeeting-vocab4950/RESULTS.md

@@ -1,5 +1,5 @@
# Paraformer-Large
- Model link: <https://www.modelscope.cn/models/yufan6/speech_mfcca_asr-zh-cn-16k-alimeeting-vocab4950/summary>
- Model link: <https://www.modelscope.cn/models/NPU-ASLP/speech_mfcca_asr-zh-cn-16k-alimeeting-vocab4950/summary>
- Model size: 45M

# Environments

 egs_modelscope/asr/mfcca/speech_mfcca_asr-zh-cn-16k-alimeeting-vocab4950/finetune.py

@@ -24,12 +24,12 @@

if __name__ == '__main__':
    
    params = modelscope_args(model="yufan6/speech_mfcca_asr-zh-cn-16k-alimeeting-vocab4950")
    params = modelscope_args(model="NPU-ASLP/speech_mfcca_asr-zh-cn-16k-alimeeting-vocab4950")
    params.output_dir = "./checkpoint"              # m模型保存路径
    params.data_path = "./example_data/"            # 数据路径
    params.dataset_type = "small"                   # 小数据量设置small，若数据量大于1000小时，请使用large
    params.batch_bins = 1000                       # batch size，如果dataset_type="small"，batch_bins单位为fbank特征帧数，如果dataset_type="large"，batch_bins单位为毫秒，
    params.max_epoch = 10                           # 最大训练轮数
    params.lr = 0.0001                             # 设置学习率
    params.model_revision = 'v2.0.0'
    params.model_revision = 'v1.0.0'
    modelscope_finetune(params)

 egs_modelscope/asr/mfcca/speech_mfcca_asr-zh-cn-16k-alimeeting-vocab4950/infer.py

@@ -18,8 +18,8 @@
        os.environ['CUDA_VISIBLE_DEVICES'] = str(gpu_id)
    inference_pipline = pipeline(
        task=Tasks.auto_speech_recognition,
        model='yufan6/speech_mfcca_asr-zh-cn-16k-alimeeting-vocab4950',
        model_revision='v2.0.0',
        model='NPU-ASLP/speech_mfcca_asr-zh-cn-16k-alimeeting-vocab4950',
        model_revision='v1.0.0',
        output_dir=output_dir_job,
        batch_size=1,
    )

 egs_modelscope/asr/mfcca/speech_mfcca_asr-zh-cn-16k-alimeeting-vocab4950/infer_after_finetune.py

@@ -59,7 +59,7 @@

if __name__ == '__main__':
    params = {}
    params["modelscope_model_name"] = "yufan6/speech_mfcca_asr-zh-cn-16k-alimeeting-vocab4950"
    params["modelscope_model_name"] = "NPU-ASLP/speech_mfcca_asr-zh-cn-16k-alimeeting-vocab4950"
    params["required_files"] = ["feats_stats.npz", "decoding.yaml", "configuration.json"]
    params["output_dir"] = "./checkpoint"
    params["data_dir"] = "./example_data/validation"