cuiziteng / Illumination-Adaptive-Transformer

🌕 [BMVC 2022] You Only Need 90K Parameters to Adapt Light: A Light Weight Transformer for Image Enhancement and Exposure Correction. SOTA for low light enhancement, 0.004 seconds try this for pre-processing.
Apache License 2.0
477 stars 44 forks source link

多GPU训练 #78

Closed znmzdx-zrh closed 2 weeks ago

znmzdx-zrh commented 3 weeks ago

您好,请问一下可以使用多GPU训练吗?实验室的卡只有11G,我在训练LOL-V2时爆显存了,目前想把模型分到两张卡上。但我发现就算只指定两张卡,如5、6卡之后,它总是在0卡上还占一些显存。

cuiziteng commented 3 weeks ago

可以的,您参考一下exposure correction那里的训练代码设置,把LOL-V2的训练代码转成多GPU训练就好,然后训练时候再指定CUDA_VISIBLE_DEVICES=5,6

znmzdx-zrh commented 2 weeks ago

可以的,您参考一下exposure correction那里的训练代码设置,把LOL-V2的训练代码转成多GPU训练就好,然后训练时候再指定CUDA_VISIBLE_DEVICES=5,6

嗯嗯好滴感谢您的回复,我昨天改了一下代码可以实现多卡训练了