Xueyan Zou

> > > > > > I got the same question. > > > > > > > > > > > > > > > Friend, have you resolved...

All the packages including Detectorn, Whisper, einops need to be installed from source code. It would be really appreciated if you could provide more details for installation errors. Note, the...

Yes, you can do so, however you may need to follow mpi settings, here is what I am using as a job template for job submission: ``` job_template: name: train_seem_v1_focalt_enc6_fpn_dec10_lang_bs{batch_size_train}_ep{epoch}_scw{spatial_class_weight}_sdw{spatial_dice_weight}_smw{spatial_mask_weight}_nsm{num_spatial_memories}_lr{lr}_ts{top_spatial_layers}_fb{fix_backbone}_fl{fix_lang}_fe{fix_encoder}_spa{spatial_on}_grd{grounding_on}_iterbase_pn_v2_maxi{max_iter}_qsn{spatial_query_num}_mc{max_candidate}...

The problem with that is probably caused by we only support evaluation of one image at one GPU, because of padding problem. You could change the code for that part...

semseg = torch.einsum("qc,qhw->chw", mask_cls, mask_pred), this step actually do a weighted sum of all the object queries, instead of selecting one object query, it actually weighted sum all the queries...

I think this may caused by you need to first: sudo apt-get install openmpi-bin libopenmpi-dev

What python version you are using? If you are using python3.6 please try: pip uninstall -y enum34

Thanks a lot that would be a very great suggestions, would provide detailed instructors for Windows and Mac Users.

Interesting, I am surprised about this too! Could you please share the demo image? As what I am asking, we do not have any specific design : )

I see, thanks for submitting the image! This is actually caused by these two categories are actually in coco training dataset, not because of zero-shot generalization.