Haisong Ding issues

Results 6 issues of


                                            Haisong Ding

Potential overflow when calculating exp

https://github.com/srvk/eesen/blob/c64c152aa60729e6b754c9274dc2bd4b386f45d6/asr_egs/wsj/utils/nnet_notf.py#L228 One simple fix would be: ` mat_array = np.asarray(mat)` ` mat_array = (mat_array.transpose() - mat_array.max(axis=1)).transpose()` ` scoreMatExp = np.exp(mat_array)`

COCO performance after finetuning

Have you tried to finetune the stage 2 model on COCO dataset and see the limit of the COCO performance?

Train/val split of HCCDoc for text recognition experiments

I wonder if there is an official train/validation split of the dataset for the text recognition task.

Eval function is not wrapped with torch.no_grad()

https://github.com/tianrun-chen/SAM-Adapter-PyTorch/blob/39c339cf481c068405e64ba48122cf3dda782260/train.py#L79 Eval is not wrapped with torch.no_grad(), there could be a data (grad) leak.

Support for GS-LRM

Does this repo have any plan to re-produce the [GS-LRM](https://sai-bi.github.io/project/gs-lrm/) paper from Adobe?

Very confused about this line of code

https://github.com/3DTopia/LGM/blob/fe8d12cff8c827df7bb77a3c8e8b37408cb6fe4c/core/provider_objaverse.py#L161 In this code, given a camera pose, you want to convert it from opengl to colmap format by reverting the up and forward direction. According to https://github.com/ashawkey/kiuikit/blob/73e34203fd3bd4b5979dc3a38057dff117b39521/kiui/cam.py#L31, it is...