search

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

confuse about the train_search/train/test #172

Open whubaichuan opened 1 year ago

whubaichuan commented 1 year ago

Hi, you mentioned in Architecture search:

Note the validation performance in this step does not indicate the final performance of the architecture. One must train the obtained genotype/architecture from scratch using full-sized models, as described in the next section.

But in Architecture search, there is also a model with weights.pt saved. Do we need to train this model again? I think the saved weights.pt in Architecture evaluation should be the almost same as the weights.pt in Architecture search?