Train high-quality text-to-image diffusion models in a data & compute efficient manner
444
stars
31
forks
source link
Update `cc12m_256x256.yaml` with diffusion_config addition and use `null` as a correct version of `None` #13
Closed
tolgacangoz closed 1 month ago
Proposes to fix #10
This pull request updates the
cc12m_256x256.yaml
andcc12m_1024x1024.yaml
files by adding adiffusion_config
section because samplers are created withdiffusion_config.sampler_config
: https://github.com/apple/ml-mdm/blob/36959179b5c103cfe014c2ddaef91c6a24feefbe/ml_mdm/diffusion.py#L94 https://github.com/apple/ml-mdm/blob/36959179b5c103cfe014c2ddaef91c6a24feefbe/ml_mdm/diffusion.py#L287Also,
diffusion_config.no_use_residual
is needed: https://github.com/apple/ml-mdm/blob/36959179b5c103cfe014c2ddaef91c6a24feefbe/ml_mdm/diffusion.py#L262None
in the file is read by'None'
-as a string.@jlukecarlson