Dongzhi Jiang comments

Results 8 comments of


                                            Dongzhi Jiang

When finetune the 7b model, I get runtime error: The size of tensor a (65537024) must match the size of tensor b (262148096)

In my case, you have to mask sure one epoch cannot be finished in a single step, note that here one step is counted as one step with backward pass...

typeError: Unsupported format:color

Hi @justinday123 , thanks for your feedback. We fix the bug. It should work fine now. Please refer to #8 for more details.

typeError: Unsupported format:color

Hi, the problem seems to be related to your environment. Maybe this [link](https://github.com/facebookresearch/pytorch3d/issues/1399) can be of help.

Can Hop used in PETR model?

Hi @FDInSky . HoP cannot directly be used in PETR model since it does not explicitly construct BEV feature.

Some Issues with Replication

Hi @wuwangzhuanwan , thanks for your feedback. We fix the bug for loading data by disk. It should work fine now. Specifically, I find that `mmcv.imread` is not equal to...

Some Issues with Replication

Hi, the performance in the log file is based on epoch_24.pth. However, we test the model with epoch_24_ema.pth. The ema checkpoint is empirically better than the original model. This should...

Ablation study of the bev feature at timestamp t-k

We conducted the experiment on the model below: and we found that the performance of HoP branch is: .

Ablation study of the bev feature at timestamp t-k

Hi, @jinhuan-hit . I am afraid that my last answer may mislead you. Let me clarify it. The performance of HoP branch, which is shown in the second picture, means...