28 Commits

Author SHA1 Message Date
OleehyO
60f6a3d7ee feat: add base trainer implementation and training script
- Add Trainer base class with core training loop functionality
- Implement distributed training setup with Accelerate
- Add training script with model/trainer initialization
- Support LoRA fine-tuning with checkpointing and validation
2025-01-01 15:10:41 +00:00
OleehyO
a505f2e312 Add constants.py 2025-01-01 15:10:40 +00:00
OleehyO
78f655a9a4 Add utils 2025-01-01 15:10:40 +00:00
OleehyO
85e00a1082 feat(models): add scaffolding 2025-01-01 15:10:40 +00:00
OleehyO
918ebb5a54 feat(datasets): implement video dataset modules
- Add dataset implementations for text-to-video and image-to-video
- Include bucket sampler for efficient batch processing
- Add utility functions for data processing
- Create dataset package structure with proper initialization
2025-01-01 15:10:40 +00:00
OleehyO
e3f6def234 feat: add video frame extraction tool
Add utility script to extract first frames from videos, helping users convert T2V datasets to I2V format
2025-01-01 15:10:39 +00:00
OleehyO
7b282246dd chore: remove unused configuration files after refactoring
Delete accelerate configs, deepspeed config and host file that are no longer needed
2025-01-01 15:10:39 +00:00
Gforky
48ac9c1066 [fix]fix typo in train_cogvideox_image_to_video_lora.py 2025-01-01 15:10:30 +00:00
Zheng Guang Cong
21693ca770 fix bugs of image-to-video without image-condition 2025-01-01 15:10:30 +00:00
OleehyO
7b4c9db6d9 Fix for CogVideoX-{2B,5B}
When loading CogVideX-{2B,5B}, `patch_size_t` is None,
which results in the `prepare_rotary_position_embeddings` function.
2024-12-13 04:02:27 +00:00
OleehyO
36f1333788 Fix for deepspeed training 2024-12-13 04:02:26 +00:00
OleehyO
4d1b9fd166 Fix for Disney video dataset 2024-12-13 04:02:21 +00:00
Yuxuan.Zhang
5aa6d3a9ee
Merge pull request #515 from Gforky/fix_finetune_demo
[fix]fix deepspeed initialization issue in finetune examples
2024-12-02 11:29:42 +08:00
spacegoing
2fb763d25f [Fix] fix rope temporal patch size 2024-11-21 16:26:45 +00:00
luwen.miao
ac2f2c78f7 [fix]fix deepspeed initialization issue in finetune examples 2024-11-18 09:49:31 +00:00
zR
e6ee283d0e Merge branch 'CogVideoX_dev' of github.com:THUDM/CogVideo into CogVideoX_dev 2024-10-14 11:34:40 +08:00
zR
e169e7b045 Update train_cogvideox_image_to_video_lora.py 2024-10-06 22:50:56 +08:00
Yuxuan.Zhang
532f246d7c
Merge pull request #389 from THUDM/CogVideoX_dev
I2V Finetune of CogVIdeoX-5B-I2V
2024-10-05 22:14:52 +08:00
zR
f28708d845 Update train_cogvideox_image_to_video_lora.py 2024-10-05 22:12:22 +08:00
zR
4339f65660 update 2024-10-05 01:05:27 +08:00
LittleNyima
a59ed84b52
fix deprecation of clear_objs_and_retain_memory 2024-09-30 20:45:53 +08:00
zR
d9e75ce3f5 horse 2024-09-28 16:36:40 +08:00
zR
fbfad9c361 readme change 2024-09-20 17:18:31 +08:00
zR
1f6d9032cc test finetune 2024-09-20 16:22:58 +08:00
zR
2db0453b96 finetune requirement change 2024-09-20 13:35:03 +08:00
zR
66369a90aa update of readme and hostfile 2024-09-17 23:42:35 +08:00
zR
db309f3242 update llm_cogvideox_flux demo test 2024-09-17 23:15:19 +08:00
zR
6e64359524 finetune and infer upload 2024-09-16 12:02:27 +08:00