Hi @Adrien987k /V-JEPA 2.1 Team,
Congratulations on the release and your new model. I am trying to understand how to interpret the number of epochs and ipe in vit-large training config
V-JEPA 2.1 paper mentions that the pertaining is done for 135_000 steps using AdamW. However, I see the config has 1000 epochs \times 300 ipe set in it which leads to 300_000 steps. Could you please let me know what I might be missing in my understanding of the information in config vs the paper?
Hi @Adrien987k /V-JEPA 2.1 Team,
Congratulations on the release and your new model. I am trying to understand how to interpret the number of epochs and ipe in vit-large training config
V-JEPA 2.1 paper mentions that the pertaining is done for 135_000 steps using AdamW. However, I see the config has 1000 epochs \times 300 ipe set in it which leads to 300_000 steps. Could you please let me know what I might be missing in my understanding of the information in config vs the paper?