Merge pull request #473 from alibaba-damo-academy/dev_smohan

jmwang66

2023-05-09 8dab6d184a034ca86eafa644ea0d2100aadfe27d

 funasr/bin/modelscope_infer.py

@@ -15,6 +15,10 @@
                        type=str,
                        default="speech_paraformer-large_asr_nat-zh-cn-16k-common-vocab8404-pytorch",
                        help="model name in modelscope")
    parser.add_argument("--model_revision",
                        type=str,
                        default="v1.0.4",
                        help="model revision in modelscope")
    parser.add_argument("--local_model_path",
                        type=str,
                        default=None,
@@ -62,7 +66,8 @@
    if args.local_model_path is None:
        inference_pipeline = pipeline(
            task=Tasks.auto_speech_recognition,
            model="damo/{}".format(args.model_name))
            model="damo/{}".format(args.model_name),
            model_revision=args.model_revision)
    else:
        inference_pipeline = pipeline(
            task=Tasks.auto_speech_recognition,
@@ -77,6 +82,9 @@
            wav_id, wav_path = line.strip().split()
            logging.info("decoding, utt_id: ['{}']".format(wav_id))
            rec_result = inference_pipeline(audio_in=wav_path)
            text = rec_result["text"]
            if 'text' in rec_result:
                text = rec_result["text"]
            else:
                text = ''
            f_out.write(wav_id + " " + text + "\n")
            logging.info("best hypo: {} \n".format(text))