53 Commits

Author	SHA1	Message	Date
mi804	bf7b339efb	support dpo training	2025-09-22 10:14:17 +08:00
Zhongjie Duan	b0abdaffb4	Qwen image split training Bug Fix (#926) ... * bugfix	2025-09-17 20:53:46 +08:00
Artiprocher	144365b07d	merge data process to training script	2025-09-04 15:18:56 +08:00
Artiprocher	cb8de6be1b	move training code to base trainer	2025-09-03 12:03:49 +08:00
Artiprocher	b6da77e468	qwen-image splited training	2025-09-02 16:44:14 +08:00
Artiprocher	260e32217f	unified dataset	2025-09-02 13:14:08 +08:00
Zhongjie Duan	5b6d112c15	Merge pull request #843 from wuutiing/main ... add read gifs as video support	2025-08-29 12:36:24 +08:00
Artiprocher	ce0b948655	support qwen-image fp8 lora training	2025-08-25 20:32:36 +08:00
twu	f6418004bb	as numframe limit is impled in reader, add that	2025-08-22 03:00:35 +00:00
twu	e3f47a799b	make it more efficient to locate where to sample the frame	2025-08-21 09:13:45 +00:00
twu	e014cad820	add read gifs as video support	2025-08-21 09:01:48 +00:00
mi804	fff2c89360	fix bug	2025-08-19 12:38:33 +08:00
mi804	d93de98a21	fix qwen_rope	2025-08-18 17:31:18 +08:00
mi804	9f6922bba9	support qwen-image-edit	2025-08-18 16:07:45 +08:00
mi804	49f9a11eb3	lora_checkpoint & weight_decay & qwen_image_controlnet_train	2025-08-14 13:50:04 +08:00
Artiprocher	89c4e3bdb6	lora-fix	2025-08-08 18:55:13 +08:00
mi804	ef09db69cd	refactor model_logger	2025-08-06 15:47:35 +08:00
mi804	3915bc3ee6	minor fix	2025-08-06 10:58:53 +08:00
mi804	6bae70eee0	support num_workers,save_steps,find_unused_parameters	2025-08-06 10:52:59 +08:00
Zhongjie Duan	8d2f6ad32e	Merge pull request #735 from modelscope/qwen-image ... qwen-image	2025-08-04 20:40:32 +08:00
Artiprocher	c35f2d8bda	qwen-image	2025-08-04 20:24:13 +08:00
mi804	cb2caa3a36	support jsonl	2025-07-31 16:24:58 +08:00
Zhongjie Duan	db124fa6bc	Merge pull request #715 from modelscope/nexusgen-eligen ... NexusGen and EliGen	2025-07-29 20:28:07 +08:00
mi804	7df48fc2b5	remove debug out	2025-07-29 13:33:14 +08:00
mi804	8ef91b3672	support training for eligen and nexusgen	2025-07-29 13:28:42 +08:00
Artiprocher	5f68727ad3	refine code	2025-07-28 11:00:54 +08:00
Artiprocher	1363a0559f	support json dataset	2025-07-02 20:07:16 +08:00
Zhongjie Duan	3dbe271aab	Merge pull request #646 from modelscope/flux-refactor ... Flux refactor	2025-06-29 18:04:05 +08:00
Artiprocher	fcf2fbc07f	flux-refactor	2025-06-27 10:20:11 +08:00
Artiprocher	a37a83ecc3	fix videodataset to load images	2025-06-24 11:38:43 +08:00
Zhongjie Duan	c2d35a2157	update wan training (#614) ... update wan training	2025-06-16 15:48:35 +08:00
Artiprocher	830b1b7202	wan-refactor	2025-06-13 13:46:17 +08:00
Zhongjie Duan	40760ab88b	Revert "Wan refactor"	2025-06-11 17:29:27 +08:00
Artiprocher	62f6ca2b8a	new wan trainer	2025-06-06 14:58:41 +08:00
Artiprocher	8f10a9c353	training script	2025-05-19 19:02:52 +08:00
Artiprocher	675eefa07e	training framework	2025-05-12 17:48:28 +08:00
ZeYi Lin	7604be0301	output_path join swanlog	2025-03-08 13:57:08 +08:00
Artiprocher	6b67a11ad6	fix swanlab after test	2025-03-03 18:59:34 +08:00
Artiprocher	3a75026176	update wan input params	2025-02-28 19:43:18 +08:00
Zhongjie Duan	38bc785ea9	Merge branch 'main' into swanlab-dev	2025-02-28 16:16:15 +08:00
Artiprocher	a466fdca8f	del swanlab	2025-02-28 16:13:06 +08:00
ZeYi Lin	b6c3d2b74a	fix: logger	2025-02-28 12:51:58 +08:00
Artiprocher	b5c1d33e58	update swanlab log	2025-02-27 19:21:51 +08:00
ZeYi Lin	1419bec53d	feat: add swanlab logger	2025-02-26 17:12:54 +08:00
Zhongjie Duan	3b010043de	Update text_to_image.py	2025-01-02 14:23:02 +08:00
Artiprocher	919d399fdb	support resume from opensource format	2024-12-16 12:25:05 +08:00
Artiprocher	8c2671ce40	support resume training	2024-12-16 11:08:14 +08:00
Artiprocher	540c036988	add alpha to lora converter	2024-10-14 18:57:54 +08:00
Zhongjie Duan	22e4ae99e8	Flux lora update (#237) ... * update flux lora --------- Co-authored-by: tc2000731 <tc2000731@163.com>	2024-10-11 18:41:24 +08:00
Zhongjie Duan	a8cb4a21d1	align flux lora format (#204)	2024-09-12 16:01:27 +08:00