search

RoyWheels / stylegan2-ada

StyleGAN2 with adaptive discriminator augmentation (ADA) - Official TensorFlow implementation
https://arxiv.org/abs/2006.06676
Other
76 stars 13 forks source link

result and epoch #1

Open Adithia99 opened 4 years ago

Adithia99 commented 4 years ago

the training was succesfull running.. but i dont know where the result , the model pkl, and how to define epoch ? its runing forever

RoyWheels commented 4 years ago

The results folder is set when you run the training command.

!python train.py --gpus=1 --data="/content/datasets/atari_384x256_dataset_raw" --min_h=2 --min_w=3 --res_log2=7 --snap=6 --cfg="auto" --mirror=1 --resume="latest" --metrics="none" --outdir="/content/drive/My Drive/sg2ada_atari_results"

You can see in the above training run using colab that I'm putting the results in my google drive. If the colab session expires , my latest PKL files will be safe.

By default the training runs for 25000 kimg. This is set in config, within the train.py file. I've never got to an end point myself. I'm producing models for creative uses so finishing training isn't really a thing I worry about.

Adithia99 commented 4 years ago

Yeah thanks but i got little confuse in here , can u make the codes run forever (25k kimg) , and got result per image in content result in gcollab?

Please i really need that give me the ipynb file :)