Junbin Xiao
Junbin Xiao
It's softmax cross-entropy.
Hi, please find the edited code for fintuning BERT on NExT-QA [here](https://drive.google.com/file/d/1Z0RMnIJrqQcFQEhhuhiRvDvyRmaKdPbg/view?usp=sharing). You can also fine-tune other datasets by using the code.
You can fork or clone this repo. It was provided in the github [folder](https://github.com/doc-doc/NExT-QA/tree/main/dataset/nextqa).
Hi, we have provided a link in the result table of readme (see BERT-FT).
Hi, please make sure you have successfully installed the following two libs: block==0.0.5 block.bootstrap.pytorch==0.1.6
Okay, are you using the right env for this repo? You can also check by vim a clean '.py' file and test: ``` from block import fusions ``` If it...
Not sure the problem here, perhaps you have to create a new env and setup the environment strictly following my guideline.
Great.
Pls try this: ``` pip install -i https://mirrors.aliyun.com/pypi/simple block==0.0.5 pip install -i https://mirrors.aliyun.com/pypi/simple block.bootstrap.pytorch==0.1.6 ```
Hi, please refer to [HCRN](https://github.com/thaolmk54/hcrn-videoqa/tree/master/preprocess) for feature extraction. I will try to add it to this repo if got time..