Yingzi Ma

Why can't I sync files?

1

This error occurs every time when I click **sync local -> remote** `[07-10 15:23:31] [error] Error: please select one of the assets at 101 (c:\Users\gray\.vscode\extensions\natizyskunk.sftp-1.15.14\node_modules\ssh2\lib\protocol\SFTP.js:2618:19) at SFTP.push (c:\Users\gray\.vscode\extensions\natizyskunk.sftp-1.15.14\node_modules\ssh2\lib\protocol\SFTP.js:278:11) at CHANNEL_DATA...

Hello, I've been trying to reproduce BEVFormerV2 recently, but according to the code, it seems that BEVFormerV2 doesn't support Batch Size > 1 during training, is this the case or...

Question about training loss

1

Hi, I'm very interested in your project, but during the training, I found that the training loss will be very big, more than 30, is it normal?

For the evaluation of classification tasks, such as ImageNet1k, I would like to know if using kv_cathe for evaluation would have any impact on the results? My personal understanding is...

I would like to know if [otter-mpt7b](https://huggingface.co/luodian/OTTER-Image-MPT7B) also does in-context instruction tuning and if I want to test otter's ability in in-context learning, which model should I choose, please!

Question about 2D bounding box

2

I want to transform the 3d_bbs to 2D bounding boxes of the each frame, but I can't get extrinsic matrix of the sensor. Can you offer some suggestions for me...

Question about PPO Training (Qwen2.5)

1

I am using 8 A100 40GB GPUs. The following script for PPO training works successfully. However, when I switch the actor pretrained model to "qwen2.5-7B-instruct", it runs out of VRAM...

Some Suggestion/Tips for Reproducing

7

Thanks for the great work in the AV domain! It's an impressive project. However, when I attempted to reproduce the results, I encountered some issues and have a few tips...