theluyuan/DiffSynth-Studio
1
0
Fork 0
mirror of https://github.com/modelscope/DiffSynth-Studio.git synced 2026-03-23 17:38:10 +00:00
Code Issues Packages Projects Releases Wiki Activity

support ltx2 train -2

Browse Source
This commit is contained in:
mi804
2026-02-25 18:06:02 +08:00
parent 586ac9d8a6
commit 8e15dcd289
32 changed files with 175 additions and 39 deletions
Download Patch File Download Diff File Expand all files Collapse all files

8
examples/ltx2/model_training/lora/LTX-2-T2AV-noaudio.sh
View File

@@ -22,8 +22,8 @@
accelerate launch examples/ltx2/model_training/train.py \
--dataset_base_path data/example_video_dataset/ltx2 \
--dataset_metadata_path data/example_video_dataset/ltx2_t2av.csv \
--height 256 \
--width 384 \
--height 512 \
--width 768 \
--num_frames 49\
--dataset_repeat 1 \
--model_id_with_origin_paths "DiffSynth-Studio/LTX-2-Repackage:text_encoder_post_modules.safetensors,DiffSynth-Studio/LTX-2-Repackage:video_vae_encoder.safetensors,DiffSynth-Studio/LTX-2-Repackage:audio_vae_encoder.safetensors,google/gemma-3-12b-it-qat-q4_0-unquantized:model-*.safetensors" \
@@ -40,8 +40,8 @@ accelerate launch examples/ltx2/model_training/train.py \
accelerate launch examples/ltx2/model_training/train.py \
--dataset_base_path ./models/train/LTX2-T2AV-noaudio_lora-splited-cache \
--height 256 \
--width 384 \
--height 512 \
--width 768 \
--num_frames 49\
--dataset_repeat 100 \
--model_id_with_origin_paths "DiffSynth-Studio/LTX-2-Repackage:transformer.safetensors" \

22
examples/ltx2/model_training/lora/LTX-2-T2AV-splited.sh
View File

@@ -1,3 +1,24 @@
# Single Stage Training not recommended for T2AV due to the large memory consumption. Please use the Splited Training instead.
# accelerate launch examples/ltx2/model_training/train.py \
# --dataset_base_path data/example_video_dataset/ltx2 \
# --dataset_metadata_path data/example_video_dataset/ltx2_t2av.csv \
# --data_file_keys "video,input_audio" \
# --extra_inputs "input_audio" \
# --height 256 \
# --width 384 \
# --num_frames 25\
# --dataset_repeat 100 \
# --model_id_with_origin_paths "DiffSynth-Studio/LTX-2-Repackage:transformer.safetensors,DiffSynth-Studio/LTX-2-Repackage:text_encoder_post_modules.safetensors,DiffSynth-Studio/LTX-2-Repackage:video_vae_encoder.safetensors,DiffSynth-Studio/LTX-2-Repackage:audio_vae_encoder.safetensors,google/gemma-3-12b-it-qat-q4_0-unquantized:model-*.safetensors" \
# --learning_rate 1e-4 \
# --num_epochs 5 \
# --remove_prefix_in_ckpt "pipe.dit." \
# --output_path "./models/train/LTX2-T2AV_lora" \
# --lora_base_model "dit" \
# --lora_target_modules "to_k,to_q,to_v,to_out.0" \
# --lora_rank 32 \
# --use_gradient_checkpointing \
# --find_unused_parameters
# Splited Training
accelerate launch examples/ltx2/model_training/train.py \
--dataset_base_path data/example_video_dataset/ltx2 \
@@ -19,7 +40,6 @@ accelerate launch examples/ltx2/model_training/train.py \
--use_gradient_checkpointing \
--task "sft:data_process"
accelerate launch examples/ltx2/model_training/train.py \
--dataset_base_path ./models/train/LTX2-T2AV_lora-splited-cache \
--data_file_keys "video,input_audio" \

19
examples/ltx2/model_training/lora/LTX-2-T2AV.sh
View File

@@ -1,19 +0,0 @@
accelerate launch examples/ltx2/model_training/train.py \
--dataset_base_path data/example_video_dataset/ltx2 \
--dataset_metadata_path data/example_video_dataset/ltx2_t2av.csv \
--data_file_keys "video,input_audio" \
--extra_inputs "input_audio" \
--height 256 \
--width 384 \
--num_frames 25\
--dataset_repeat 100 \
--model_id_with_origin_paths "DiffSynth-Studio/LTX-2-Repackage:transformer.safetensors,DiffSynth-Studio/LTX-2-Repackage:text_encoder_post_modules.safetensors,DiffSynth-Studio/LTX-2-Repackage:video_vae_encoder.safetensors,DiffSynth-Studio/LTX-2-Repackage:audio_vae_encoder.safetensors,google/gemma-3-12b-it-qat-q4_0-unquantized:model-*.safetensors" \
--learning_rate 1e-4 \
--num_epochs 5 \
--remove_prefix_in_ckpt "pipe.dit." \
--output_path "./models/train/LTX2-T2AV_lora" \
--lora_base_model "dit" \
--lora_target_modules "to_k,to_q,to_v,to_out.0" \
--lora_rank 32 \
--use_gradient_checkpointing \
--find_unused_parameters