thuiar / MMSA

MMSA is a unified framework for Multimodal Sentiment Analysis.
MIT License
634 stars 104 forks source link

如何基于单模态进行训练 #86

Open bie33333 opened 6 months ago

bie33333 commented 6 months ago

比方说我只需要基于SIMS数据集训练视频模态,该如何设置config?

bie33333 commented 6 months ago

是修改下面两个参数吗? 'seq_lens': [39, 400, 55], 'feature_dims': [768, 33, 709]

Columbine21 commented 6 months ago

只需要使用原来的借口获得三个模态的序列,但只将获得的 video 模态序列输入到单模态模型中即可。