search

UX-Decoder / Semantic-SAM

[ECCV 2024] Official implementation of the paper "Semantic-SAM: Segment and Recognize Anything at Any Granularity"
2.26k stars 107 forks source link

about training config #75

Open Shi-qingyu opened 11 months ago

Shi-qingyu commented 11 months ago

What was the duration of your model's training, and could you provide details about your training configuration?

xushilin1 commented 11 months ago

@FengLi-ust Hi, Thanks for your open-source code. I am re-training your model using the SAM dataset with SWIN-L backbone. Could you please provide me with information about the duration of your training time, the specifications of the machine used, and the batch size? Thanks a lot.

FengLi-ust commented 10 months ago

Yes, we train the whole SA-1B dataset for around 2 epochs. We use 64 V100 machines with batch size 64.

JosephKJ commented 10 months ago

Thank you for the response. How much time did it take?