Kunchang Li

Pretrained model for sth-sth v1 and kinetics got lower accuracy

2

Hello, thanks for your great job!!! Today when I tried to fine-tuning your model and test the accuracy of your model, something strange happened. I used uniform sampling, but it...

Why using ReLU before split-attention?

1

I have read your codes about split-attention and I found that you use ReLU before split-attention. https://github.com/zhanghang1989/ResNeSt/blob/76debaa9b9444742599d104609b8ee984b207332/resnest/torch/splat.py#L48-L76 **However, in MobileNetV3,CBAM-ResNet or other models which use attention mechanism, activation is usually...

Good job! My question is that why to use different class tokens for each stage but **only the final class token is used for classification**? https://github.com/microsoft/CvT/blob/34d1af94c95442b19fb9470e0c9dd5ee11be2024/lib/models/cls_cvt.py#L607

Thanks for your great job! I notice you create an interesting toy dataset in Section 4. As I'm not familiar with creating a dataset, do you plan to make your...

Some question

2

In the paper, EViT with oracle can obtain higher accuracy when training longer epochs. Similar results are also shown in the DeiT paper. Thus I think the comparison is not...

Ask anything in video

4

Hi! We have simply extended MiniGPT-4 for video question answering in our project [Ask-Anything](https://github.com/OpenGVLab/Ask-Anything). Without extra instruction fine-tuning, current results are not satisfactory.  In our other try, we simply...

Thanks for your great job! I have checked the code of v0 and v1, and found some differences between the prompts: ```python # Vicuna-v0 sep = "###" train_prompt = "system.###...

Will the multi-choice version be released

1

Hi! Thanks for the exciting jobs! I have noticed another multi-choice version dataset but can not find the annotation files. Will it be released? 😁

Should input_embeddings and out_embeddings be updated in Stage2?

1

Hi! Thanks for your interesting job! I just find that when using LoRA, the `input_embeddings` and `out_embeddings` are not updated with the following code. https://github.com/eric-ai-lab/MiniGPT-5/blob/2121c745b2cb2d7e842e03b4bcaa89c63f2ee6c1/minigpt4/models/mini_gpt5.py#L115-L116 Considering the LoRA is used...