Rongjie Li

Same problem, can not inference where batch size greater than 1.

I think it may be caused by a torch.DDP, I adpot the training and evaluation loop of blip2-opt for instruct-blip vicuna and cause this error. @LiJunnan1992 Would you have plan...

> add `outputs[outputs == -1] = 1` to > > https://github.com/salesforce/LAVIS/blob/59273f651b9bffb193d1b12a235e909e9f826dda/lavis/models/blip2_models/blip2_vicuna_instruct.py#L372 > > > You can have a try Does this modification will work? The error is occur within llm_model.generate()

Hi, Is VGG16 backbone ready? Seems that [the VGG backbone are not in the dev plan of maskrcnn-benchmark](https://github.com/facebookresearch/maskrcnn-benchmark/issues/214), but VGG is the most important backbone for Scene graph generation. It...

@aryanmangal769 An error occurred while calling self.get_statistics(). Remove the try-except, and you will see the error caused by the dtype conversion of numpy. https://github.com/ubc-vision/IterativeSG/blob/7abd162cb8e510cfbcededbeb18b36b54f381189/data/datasets/visual_genome.py#L49

Same question; I'm wondering about the implementation of how the author achieves M=3.

Please refer to this issue. You may load the config incorrectly. https://github.com/SHTUPLUS/PySGG/issues/5

Hi, It is a unpublished model proposed in Scene-Graph-Benchmark. It seems like using the Transformer for context modeling instead of LSTM.

Hi, For vctree, I simply adopt the implementation from Scene-Graph-Benchmark. It does slow than other model. You can try to comment the torch.autograd.set_detect_anomaly(True) in line 58 of tools/relation_train_net.py. It may...

Hi， I‘m wondering which predictor is the baseline you refered to？RelDN or MSDN？