Open sh0416 opened 1 year ago
I want to finetune the 16B scale codegen checkpoint using TPU.
In the config directory, there is no configuration for that.
Could you share about the configuration? or some documentation for scaling model parameter?
FYI, I am planning to use TPUv3-256 or more core for that.
I want to finetune the 16B scale codegen checkpoint using TPU.
In the config directory, there is no configuration for that.
Could you share about the configuration? or some documentation for scaling model parameter?