This repo contains functions and classes that can be used in deep learning projects.
save_root: path to the folder in which the experiment folder resides
or will reside. i.e. /home/user/project_saves/
exp_folder: this is the path to the experiment folder (the folder
that contains all model training folders for a given exp_name.)
i.e. /home/user/project_saves/my_exp_name/
model_folder: refers to the model folder without the full path. This
is the folder name that contains the checkpt files.
i.e. my_exp_name_0_lr0.001
save_folder: this is the full path to where the checkpoint files are
actually saved. the hyperparams.json and the hyperparams.txt are
both also saved to this path.
i.e. /home/user/project_saves/my_exp_name/my_exp_name_0_lr0.001
config: a dict that contains all configuration information necessary
for a training.
Clone this repo as a submodule in your project. Then install all necessary packages locally:
python3 -m pip install --user -r requirements.txtsprout_len: int- the length of the seed text in sequence models when not using teacher forcing