徐铭远

Thank you for your code, I change the code of pytorch to paddlepaddle, and deploy it on Baidu free server to run online. I have tested both your network and...

I didn't find the ablation experiment results of the number of experts in your paper. When I have a total of 9 tasks, how many experts do I need to...

大佬，请问有测试384和vit或者swin对比的结果嘛。

作者的工作很不错，但是我测试同义词替换之类的基本用的繁体字或者很奇怪的东西换的... 自己写了一般非常简单的几个增强，代码也比较简单⬇️ https://github.com/xmy0916/nlp_aug

[maybe a bug] loss nan

4

https://github.com/yitu-opensource/T2T-ViT/blob/main/models/token_performer.py#L18 My code has turned on fp16, so the 1e-8 on this line to prevent division by 0 is not enough for my code... the loss of the network calculation...

msmt17的训练脚本没有嘛？

3

直接改了tt参数每次评估都报错 ValueError: Input contains NaN, infinity or a value too large for dtype('float32').

I'm also training this... I haven't downloaded webvid2.5m yet and then I found that you have done everything I want to do, hahahaha

1

I'm also training this... I haven't downloaded webvid2.5m yet and then I found that you have done everything I want to do, hahahaha

There seems to be some discrepancy between the cogvlm indicators posted in the paper.

1

indicators fron cogvlm's paper.  indicators fron this paper. 

Features will release training code of VL?

1

### Reminder - [X] I have searched the Github Discussion and issues and have not found anything similar to this. ### Motivation will release training code of VL? ### Solution...

https://github.com/DAMO-NLP-SG/Video-LLaMA/blob/main/train_configs/visionbranch_stage1_pretrain.yaml#L73 这一行设置一个epoch跑2500个iters，数据集下载完是2253277个，batchsize 32 8张卡应该是8801个iters，是不是不太对呀