MDT

MDT copied to clipboard

will you release a pretrained checkpoint of B/2 model？

3

Thanks for the wonderful work again and I still wants to do some experiments on the small models like B/2 or S/2 due to my limited resources. Will you guys...

aaab8b

LT1st

Training Time

1

Dear Authors, when i use one gpu to train the model : `python -m torch.distributed.launch --nproc_per_node=1 scripts/image_train.py` and use 4 gpu to train the model: `python -m torch.distributed.launch --nproc_per_node=4 scripts/image_train.py`...

mapengsen

training problem

2

hello! when I am training with one gpu. I encountered with a problem:  and I set the script like this  I don't know which leaf Variable that requires...

ZH3nGsQ

Questions about MAE Loss

1

Great works! However, I also wonder that the reconstruction loss (MSE Loss) in MAE is lost in the code. I couldn't find it. Does the MDT computes the noising prediction...

ALEX13679173326

full and remaining unmasked latent embeddings

2

Hello, I would like to know how this sentence in the paper is achieved: "In MDT, both the full and remaining unmasked latent embeddings are fed into the diffusion model...

ZhaoLiruii

Changed hugglingface to huggingface

tanaynayak

loss m_mse of MDT-S-2 is much larger than mse and the visualization of MDT-S-2 with mask_ratio 0.3 does not work

2

From the loss of mse and m_mse, it seems that the mask branch does not work in MDT-S-2. We also visualize the generation image and find that generated image with...

ZGCTroy

batch size when training

4

Batch size is said to be 256 in the article. But why batch size in run.sh is 32? And why batch size in run_ddp_master.sh is 4?

tengjiayan20

Hello everyone. I would like to know about the memory requirements of the model. I am trying to run the original model on 2 48GB GPUs through PyTorch DDP. However,...

Bhavay-2001