Han Cai
Han Cai
Thanks for your interest in our work. Releasing the training code and scripts on segmentation is on the schedule. Please stay tuned. Best regards, Han
Hello Ross, Thank you for sharing your findings! I also have similar findings that q/k/v Matmul and the division need to be float32 during training to avoid NaN loss. We...
That's a good point. @chenjy2003, we should add the command to train DiT-XL on ImageNet 512x512.
Thanks for your interest in our work! VQ is one direction we are working on. We will push our updates to this repo.
Thank you for the question. I have not trained efficientvit classification models on the imagenet-22k dataset. I am not sure what the best practice is for setting training hyperparameters for...
You can find the model checkpoints here: https://huggingface.co/collections/mit-han-lab/efficientvit
Hi RunMarshal, Can you give more details about the issue? Which model is missing? Thank you, Han