FoundationVision / OmniTokenizer

OmniTokenizer: one model and one weight for image-video joint tokenization.
https://www.wangjunke.info/OmniTokenizer/
MIT License
234 stars 5 forks source link

Inquiry about the version of pytorch lightning #7

Closed hyc9 closed 3 months ago

hyc9 commented 3 months ago

Thank you for your excellent work! I've tried switching between a few versions of pytorch lightning, but I'm encountering conflicts here or there. I wonder what version is being used in the code?

hyc9 commented 3 months ago

I tried the version in requirements.txt( i.e. 1.5.4) too and there are still some conflicts

hyc9 commented 3 months ago

Here are some error lines encountered in different versions, for example : `parser = pl.Trainer.add_argparse_args(parser) #with pl 2.x.x from pytorch_lightning.utilities.rank_zero import rank_zero_only #with pl 1.5.4

with 1.8.1 / 1.9.5

return [opt_ae, opt_disc], [{"scheduler": sch_ae, "interval": "step"}, {"scheduler": sch_disc, "interval": "step"}]

or [{"optimizer": opt_ae, "scheduler": sch_ae, "interval": "step"}, {"optimizer": opt_disc, "scheduler": sch_disc, "interval": "step"}]

sch1, sch2 = self.lr_schedulers() `

wdrink commented 3 months ago
截屏2024-06-26 09 20 22

I use pl == 1.5.4, it seems to work well for me.