Query on GPU usage and training time for unitok with llamagen on ImageNet 256x256

[zgcmx]

Could you please share the GPU count and training time used separately for the tokenizer and generator when training unitok with llamagen on ImageNet 256x256?Thanks！

[machuofan]

It takes 50 hours with the equivalent computing power of 256 A100 GPUs to train the tokenizer. For the generator, the training time is roughly the same as that reported in LlamaGen.