Training time on colab - Githubissues

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

Training time on colab #162

Open pizhn opened 3 years ago

pizhn commented 3 years ago

When running train_search in colab pro using GPU, every epoch takes around 3100 secs. Then total 50 epochs takes around 1.8 days. Anybody trying colab how much training time do you need? Can you reproduce the 1 day searching time in paper?

pizhn commented 3 years ago

I realized I didn't copy my data in google drive to local env. I just added that in my workflow, but it only accelerates a little, seems still take 31 hrs or so

myl11045 commented 2 years ago

When running MNIST ,i caculate each epoch consumes about 3000secs when i set batch size=16, which runs in 2080 10G gpu,and intel E5 2650 V4 CPU.I find that GPU utilize rate are very low ,around 8-25%,which makes me annoyed