search

mit-han-lab / tiny-training

On-Device Training Under 256KB Memory [NeurIPS'22]
http://tinytraining.mit.edu
MIT License
429 stars 59 forks source link

how to train the quantized model on custom dataset #6

Open tonyzhangnxp opened 1 year ago

tonyzhangnxp commented 1 year ago

Hello

I 'm an engineer from NXP. I'm very interested in your incredible work. I'd read your paper "On-Device Training Under 256KB Memory", we want to reproduce your works on our platforms. but I can't find where to training the quantized model on custom dataset. I checked the mucnet repo, it only contains the model zoo, I can't find the training scripts I also checked the once-for-all repo, but it seems not support tiny subnet yet.

BR Tony

tonyzhangnxp commented 1 year ago

@songhan @zhijian-liu @Lyken17 @tonylins Sorry to loop you all, is there anybody could answer my question?

Thanks. BR Tony

loveindex commented 6 months ago

Hello

I 'm an engineer from NXP. I'm very interested in your incredible work. I'd read your paper "On-Device Training Under 256KB Memory", we want to reproduce your works on our platforms. but I can't find where to training the quantized model on custom dataset. I checked the mucnet repo, it only contains the model zoo, I can't find the training scripts I also checked the once-for-all repo, but it seems not support tiny subnet yet.

BR Tony

I also can't find the training code. In the train project, I find the train function in "main.c", it calls "invoke(label)" but "label" is not used in "invoke" function.