search

isl-org / lang-seg

Language-Driven Semantic Segmentation
MIT License
691 stars 84 forks source link

Training configuration #7

Closed XAVILLA closed 2 years ago

XAVILLA commented 2 years ago

Hi! Can you please let me know what is the correct training configuration to reproduce the performance reported in the paper?

In the paper you mentioned that 6 GPUs were used and batch size was set to 6. Does this mean that I should just launch train.sh with 6 GPUs available? And can you please let me know what is the approximate time for training? Thanks!

XAVILLA commented 2 years ago

When using 6gpus it seems to take 1h per epoch. Is this expected?

Boyiliee commented 2 years ago

Hi, @XAVILLA ,

1) We have provided a training example in README.MD. Please check it for details.
2) It depends on your machine. Any number of GPUs should work. It also depends on the task (dataset) for each epoch. 3) If you train ADE20k, 1-2days for all in general, but it totally depends on what kind of machine you use.

Hope this helps!