Peize Sun

icon indicating a website link https://peizesun.github.io/ icon indicating an email [email protected]

icon company The University of Hong Kong icon location Hong Kong icon location PhD student, The University of Hong Kong, Computer Vision

Results 64 comments of Peize Sun

FID results of GPT-L and GPT-1B on 256*256 images

Hi~ I don’t understand what is reproducing the result on 224x224. The expected FID is in 256x256.

KeyError: 'optimizer'

Hi~ This is not the mistake in your side. It is because our released vq_ds16_t2i.pt doesn’t include optimizer parameters.

can i train in colab free?

Hi~ We will try to see whether Colab works. If it goes well, we will update here.

Training Results

Hi~ This is not normal. In general, you can see meaningful images after 10 epochs training. Can you first try to check whether the image code is correct? You can...

Training Results

Did you change the batch size, learning rate or other training parameters ?

RuntimeError: shape '[1, 512, 1, 32, 2]' is invalid for input of size 16448

Hi~ Can you provide more descriptions, like - what is the running common you use? - What basic things do you change?

Question about why not try using image tokenizer and a ready made llama3 etc LLM model with lora?

Hi~ Lora fine-tuning an existing LLM to enable image generation ability is a very promising direction, but how to **NOT** harm LLM text ability is quite challenging. In my opinion,...

Question about why not try using image tokenizer and a ready made llama3 etc LLM model with lora?

To evaluate an tokenizers is a open problem. It really depends on which tasks you focus on. For a universal tokenizer, I think we need to: - Image reconstrcution. -...

Question about why not try using image tokenizer and a ready made llama3 etc LLM model with lora?

I am not aware of details about Omnitokenizer, maybe you can connect its first author?

Will this line of code blur the image during evaluation? tokenizer/tokenizer_image/reconstruction_vq_ddp.py: transforms.Normalize(mean=[0.5, 0.5, 0.5], std=[0.5, 0.5, 0.5], inplace=True)

Hi~ Normalizing an image before feeding it into a network is a widely used operation, which is not usually considered to have the effect of blurring the image.