15 Commits

Author	SHA1	Message	Date
Kohaku-Blueleaf	31c8cc9aa3	Correct dtype handling and better default arg	2025-06-13 12:06:12 +08:00
Kohaku-Blueleaf	218c3e3793	Use encoded latents as input	2025-06-13 11:49:49 +08:00
Kohaku-Blueleaf	23523f5894	Fix typo	2025-06-13 11:34:39 +08:00
Kohaku-Blueleaf	4dcd698e48	Update nodes_train.py	2025-06-11 00:18:23 +08:00
Kohaku-Blueleaf	1baa1bdafc	check if need to disable pbar	2025-06-07 05:34:24 +08:00
Kohaku-Blueleaf	b3b36e5a69	Use tqdm for training loop	2025-06-07 05:21:37 +08:00
Kohaku-Blueleaf	c246a1dbde	Add gradient checkpointing	2025-06-07 05:09:54 +08:00
Kohaku-Blueleaf	1870402671	Add lora model loader for onfly usage	2025-06-07 05:09:32 +08:00
Kohaku-Blueleaf	5e43ec9031	Updates of training logic ... * use seperated dtype for trainable weight * force "training module only" before training * disable gradient after training * ensure same dtype after training	2025-06-07 05:09:16 +08:00
Kohaku-Blueleaf	c8bd95a393	linting	2025-04-22 15:14:26 +08:00
Kohaku-Blueleaf	5098e9408d	Utilize weight adapter scheme in basic training node	2025-04-22 14:54:20 +08:00
Yoland Y	d58ad2dd19	Expand supported image file extensions in LoadImageSetNode	2025-03-26 17:30:42 +08:00
Yoland Yan	bfc2f177e8	Refactor import statements in nodes_train.py ... Reorganized and cleaned up import statements, removing unused imports and adding specific module imports for better clarity and organization.	2025-03-26 17:30:42 +08:00
Yoland Yan	2cd3c8a2fb	Fix ruff errors	2025-03-26 17:30:41 +08:00
Yoland Yan	225a196dae	Feat: Add basic LoRA training support ... For more details: https://github.com/Comfy-Org/rfcs/pull/26	2025-03-26 17:30:41 +08:00