MILVLG / bottom-up-attention.pytorch

A PyTorch reimplementation of bottom-up-attention models
Apache License 2.0
294 stars 76 forks source link

extract 2048-D features? [Help ASAP] #112

Open DLrook opened 1 year ago

DLrook commented 1 year ago

Thank you for the excellent code, I have some problems : 1:How to extract 2048-D features during feature extraction in this repository? I am getting this error during test phase of 'mcan-vqa': RuntimeError: mat1 and mat2 shapes cannot be multiplied (100x100 and 2048x512)

2:RuntimeError: stack expects each tensor to be equal size, but got [100, 100] at entry 0 and [100, 95] at entry 56

@JimLee4530 @Zoroaster97 @nbgao @LLipsky @jessiSYJ