| | |
|---|
| | | |
|---|
| | | > **Note**: |
|---|
| | | > The modelscope pipeline supports all the models in |
|---|
| | | [model zoo](https://alibaba-damo-academy.github.io/FunASR/en/modelscope_models.html#pretrained-models-on-modelscope) |
|---|
| | | [model zoo](https://alibaba-damo-academy.github.io/FunASR/en/model_zoo/modelscope_models.html#pretrained-models-on-modelscope) |
|---|
| | | to inference and finetine. Here we take the model of xvector_sv as example to demonstrate the usage. |
|---|
| | | |
|---|
| | | ## Inference with pipeline |
|---|
| | |
|---|
| | | ### API-reference |
|---|
| | | #### Define pipeline |
|---|
| | | - `task`: `Tasks.speaker_diarization` |
|---|
| | | - `model`: model name in [model zoo](https://alibaba-damo-academy.github.io/FunASR/en/modelscope_models.html#pretrained-models-on-modelscope), or model path in local disk |
|---|
| | | - `model`: model name in [model zoo](https://alibaba-damo-academy.github.io/FunASR/en/model_zoo/modelscope_models.html#pretrained-models-on-modelscope), or model path in local disk |
|---|
| | | - `ngpu`: `1` (Default), decoding on GPU. If ngpu=0, decoding on CPU |
|---|
| | | - `output_dir`: `None` (Default), the output path of results if set |
|---|
| | | - `batch_size`: `1` (Default), batch size when decoding |
|---|
