Yuren Cong comments

Results 46 comments of


                                            Yuren Cong

checkpoint should be updated with enhanced version

Hi, I remember that the checkpoint is the same but we add a post-processing. Please check (https://github.com/yrcong/RelTR/issues/34)

Inquiry about code behavior in relation to constraints of evaluation metrics

Hi, I am very happy if our work can help you! However I forgot where(what) the variable is (for). Could you give me a link for it? I haven't understand...

Model not getting trained on single GPU

We train the model for 150 epochs. 38th epoch might be just warm-up. Maybe you can try to load some pretrained weights to accelerate the training?

Model not getting trained on single GPU

It is not related to the port. Make --nproc_per_node=1 pls

annotation files

We create the .json files in COCO format. You can refer to the code we used: ``` from dataloaders.visual_genome_coco import VGDataLoader, VG import numpy as np from torch import optim...

I can only get the relationships between entities, how can I get the scene graph?

Actually we draw it manually. You can visualize the graph automatically with GraphViz, although it is not very elegant.

NaN Loss

When do you meet NaN? Training Stage C or Stage B?

Hi, thank you for your work. I have no idea why this error occurs

Hi, it seems something wrong with your SD. I am not sure where you downloaded the ```stable diffusion's .pt```. Please directly download SD2.1 from huggingface: https://huggingface.co/stabilityai/stable-diffusion-2-1-base with ``` git lfs...

Implementation details

> We use the same set of hyperparameters used in the original papers for each model. The dataset size is changed (for example CUB-Color should have fewer training images than...

Implementation details

I also encountered a problem when I was trying to reproduce DAMSM R-precision (for C-CUB-Color). In the original AttnGAN, there are 5450 tokens. I downloaded the pretrained text encoder provided...