Add information about training

[Andrey36652]

Hello, thank you for the research. Please share more info about pre-training process.

Data:

• language data (total number of text tokens model have seen during pretraining)
• images (total amount of images for pretraining)
• videos (total amount of videos for pretraining)

Hardware: what kind of hardware were used, for how long, and maybe pretraining cost estimation.

[nunbuzor]

and why not use natural language synthetic captioning for images?

[Molly-3000]

Please also include the source of the videos. Additionally, provide the composition ratio of different categories as mentioned in the paper.

[yinsong1986]

+1 for this question. Thanks!