what is the mean of the full model during training?

[xushilin1]

Does it mean that you have trained all visual encoders during the fine-tuning and what are the specific training settings