[Experimental] Add cache mechanism for dataset groups to avoid long waiting time for initilization (#1178)

* support meta cached dataset

* add cache meta scripts

* random ip_noise_gamma strength

* random noise_offset strength

* use correct settings for parser

* cache path/caption/size only

* revert mess up commit

* revert mess up commit

* Update requirements.txt

* Add arguments for meta cache.

* remove pickle implementation

* Return sizes when enable cache

---------

Co-authored-by: Kohya S <52813779+kohya-ss@users.noreply.github.com>

train_network.py

@@ -6,6 +6,7 @@ import sys
 import random
 import time
 import json
+import pickle
 from multiprocessing import Value
 import toml
 
@@ -23,7 +24,7 @@ from library import model_util
 
 import library.train_util as train_util
 from library.train_util import (
-    DreamBoothDataset,
+    DreamBoothDataset, DatasetGroup
 )
 import library.config_util as config_util
 from library.config_util import (