yhd-123

How to use 2 GPUs?

2

 python train.py --cfg_path ${cfgs/anet_c3d_pdvc.yml} But only one GPU is used!

三种模态的特征序列长度都不同，怎么修改Cross_VQEmbeddingEMA_AVT模块

4

如果audio_feat，video_feat，text_feat的特征序列长度都不同，AVT_VQVAE_Encoder中的self.Cross_quantizer = Cross_VQEmbeddingEMA_AVT(n_embeddings, self.hidden_dim)传播会出错。 v_ph = torch.reshape(v_ph, ((B, T, M))) # [BxT, M] -> [B, T, M] RuntimeError: shape '[16, 99, 400]' is invalid for input of size 236800 Cross_VQEmbeddingEMA_AVT部分怎么修改代码，我想直接用audio_feat，video_feat，text_feat通过AVT_VQVAE_Encoder获取量化后语义对齐的特征表示audio_vq，video_vq，text_vq，进行下游任务。

With the increase of the number of epoch, the inference sentences obtained from each epoch occupy the CPU memory, how to clear the CPU memory occupied by the inference sentences...