Config loader

Functions to load and store models and datasets.

Code using omegaconf to handle IO.

IO

---

source

class_to_str

 class_to_str (cls)

---

source

load_config

 load_config (file_path)

---

source

config_to_dict

 config_to_dict (config)

---

source

save_dataclass_yaml

 save_dataclass_yaml (data_obj, file_path)

---

source

save_dict_yaml

 save_dict_yaml (dict_obj, file_path)

Test

@dataclass
class MyConfig:    
    target:str = class_to_str(OmegaConf)
    clr_dim: int = 80
    features: list[int]=None
    
c = MyConfig()
c.features = [1,2,3]

OmegaConf.structured(c)

{'target': 'omegaconf.omegaconf.OmegaConf', 'clr_dim': 80, 'features': [1, 2, 3]}

Object config load

Adapted from: https://github.com/Stability-AI/generative-models

---

source

get_obj_from_str

 get_obj_from_str (string, reload=False, invalidate_cache=True)

---

source

instantiate_from_config

 instantiate_from_config (config)

Models

---

source

store_model_state_dict

 store_model_state_dict (state_dict, save_path)

---

source

load_model_state_dict

 load_model_state_dict (save_path, device)

Tensors and numpy

torch.serialization.DEFAULT_PROTOCOL

2

---

source

store_tensor

 store_tensor (tensor, save_path, type='tensor')

---

source

load_tensor

 load_tensor (save_path, device, type='tensor')