training time for similar performance with pre-trained checkpoint

JackAILab / ConsistentID

Customized ID Consistent for human

MIT License

825 stars 75 forks source link

training time for similar performance with pre-trained checkpoint #50

Open hwopark opened 1 month ago

hwopark commented 1 month ago

In the paper, the model was optimized using Adam on 8 NVIDIA 3090 GPUs with a batch size of 16. How many epochs are needed to achieve performance comparable to the released pre-trained checkpoint? and could you share how much time it will take?

JackAILab commented 1 month ago

Hi, @hwopark , We trained for about 15 epochs on all the data, and it took about 15 days to achieve the desired results.