param_size of searched model is too heavy

quark0 / darts

Differentiable architecture search for convolutional and recurrent networks

https://arxiv.org/abs/1806.09055

Apache License 2.0

3.92k stars 843 forks source link

param_size of searched model is too heavy #118

Open xingyueye opened 5 years ago

xingyueye commented 5 years ago

@quark0 hi, firstly, thanks for your work , I'm really interested in Darts. But I have some confusion about my reproduce result. I found that the network I searched was too big compared to the network released in Darts. The param_size of my result model are generally 8M， while released mode‘s is nearly 3M. Could you tell me the reason why the results are different. Or I just should continue to search for more networks. Thanks

epriwahyu commented 5 years ago

@xingyueye : can you teach me how to run DART from the bigining?