miluELK/ddpm-ema-pokemon-128

Training data

huggan/few-shot-pokemon

Training hyperparameters

The following hyperparameters were used during training: --checkpointingsteps=1000 \ --datasetname="huggan/few-shot-pokemon" \ --resolution=128 \ --outputdir="ddpm-ema-pokemon-128" \ --trainbatchsize=16 \ --evalbatchsize=16 \ --numepochs=100 \ --gradientaccumulationsteps=1 \ --learningrate=1e-4 \ --lrwarmupsteps=800 \ --mixedprecision="fp16" \ --pushtohub

Training results

📈 [TensorBoard logs]()