speech_recognition/FunASR
1
0
Fork 0
mirror of https://github.com/modelscope/FunASR synced 2025-09-15 14:48:36 +08:00
Code Issues Actions Packages Projects Releases Wiki Activity

Merge pull request #459 from alibaba-damo-academy/dev_sx

Browse Source 
update contextual finetune
This commit is contained in:
zhifu gao 2023-05-05 15:49:22 +08:00 committed by GitHub
commit cf96541c8c
No known key found for this signature in database
GPG Key ID: 4AEE18F83AFDEB23
1 changed files with 6 additions and 7 deletions
Show Stats Download Patch File Download Diff File Expand all files Collapse all files

13
egs_modelscope/asr/paraformer/speech_paraformer-large-contextual_asr_nat-zh-cn-16k-common-vocab8404/finetune.py
View File

@ -3,7 +3,6 @@ import os
from modelscope.metainfo import Trainers
from modelscope.trainers import build_trainer
import funasr
from funasr.datasets.ms_dataset import MsDataset
from funasr.utils.modelscope_param import modelscope_args
@ -27,11 +26,11 @@ def modelscope_finetune(params):
if __name__ == '__main__':
params = modelscope_args(model="damo/speech_paraformer-large-contextual_asr_nat-zh-cn-16k-common-vocab8404", data_path="./data")
params.output_dir = "./checkpoint" # m模型保存路径
params.output_dir = "./checkpoint" # 模型保存路径
params.data_path = "./example_data/" # 数据路径
params.dataset_type = "large" # 小数据量设置small，若数据量大于1000小时，请使用large
params.batch_bins = 2000 # 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 = 50 # 最大训练轮数
params.dataset_type = "large" # finetune contextual paraformer模型只能使用large dataset
params.batch_bins = 200000 # 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 = 20 # 最大训练轮数
params.lr = 0.00005 # 设置学习率
modelscope_finetune(params)
modelscope_finetune(params)