Open fxnnxc opened 3 years ago
Because the return size of mu and var is [BATCH x Dimesion], I had to use mean value here. (I tried sum --> grad inf error!) Should I find a better way to generate one vector Z?
https://github.com/fxnnxc/text_summarization/blob/b7a879b1ebbd6e79b590ba9a266772cd5c10e59a/experiments/experiment3/vae_bart_attn_src/models/vae_bart.py#L73-L84
@ntnguyen1234
Because the return size of mu and var is [BATCH x Dimesion], I had to use mean value here. (I tried sum --> grad inf error!) Should I find a better way to generate one vector Z?
https://github.com/fxnnxc/text_summarization/blob/b7a879b1ebbd6e79b590ba9a266772cd5c10e59a/experiments/experiment3/vae_bart_attn_src/models/vae_bart.py#L73-L84