storytold/FineTrainers-Conditioning
2
0
Fork 0
mirror of https://github.com/storytold/FineTrainers-Conditioning.git synced 2026-10-19 07:43:28 +00:00
Code Issues Packages Projects Releases Wiki Activity

update (#198)

Browse Source 
Co-authored-by: Sayak Paul <spsayakpaul@gmail.com>
This commit is contained in:
Aryan
2025-01-09 11:06:34 +05:30
committed by GitHub
parent dbffc80e7c
commit 6716ccf8c2
2 changed files with 8 additions and 6 deletions
Download Patch File Download Diff File Expand all files Collapse all files
README.md
+5 -3
View File
@@ -43,6 +43,8 @@ Then launch LoRA fine-tuning. Below we provide an example for LTX-Video. We refe
<details> <details>
<summary>Training command</summary> <summary>Training command</summary>
TODO: LTX does not do too well with the disney dataset. We will update this to use a better example soon.
```bash ```bash
#!/bin/bash #!/bin/bash
export WANDB_MODE="offline" export WANDB_MODE="offline"
@@ -75,18 +77,18 @@ dataset_cmd="--data_root $DATA_ROOT \
dataloader_cmd="--dataloader_num_workers 0" dataloader_cmd="--dataloader_num_workers 0"
# Diffusion arguments # Diffusion arguments
diffusion_cmd="--flow_resolution_shifting" diffusion_cmd="--flow_weighting_scheme logit_normal"
# Training arguments # Training arguments
training_cmd="--training_type lora \ training_cmd="--training_type lora \
--seed 42 \ --seed 42 \
--mixed_precision bf16 \ --mixed_precision bf16 \
--batch_size 1 \ --batch_size 1 \
--train_steps 1200 \ --train_steps 3000 \
--rank 128 \ --rank 128 \
--lora_alpha 128 \ --lora_alpha 128 \
--target_modules to_q to_k to_v to_out.0 \ --target_modules to_q to_k to_v to_out.0 \
--gradient_accumulation_steps 1 \ --gradient_accumulation_steps 4 \
--gradient_checkpointing \ --gradient_checkpointing \
--checkpointing_steps 500 \ --checkpointing_steps 500 \
--checkpointing_limit 2 \ --checkpointing_limit 2 \
docs/training/ltx_video.md
+3 -3
View File
@@ -36,18 +36,18 @@ dataset_cmd="--data_root $DATA_ROOT \
dataloader_cmd="--dataloader_num_workers 0" dataloader_cmd="--dataloader_num_workers 0"
# Diffusion arguments # Diffusion arguments
diffusion_cmd="--flow_resolution_shifting" diffusion_cmd="--flow_weighting_scheme logit_normal"
# Training arguments # Training arguments
training_cmd="--training_type lora \ training_cmd="--training_type lora \
--seed 42 \ --seed 42 \
--mixed_precision bf16 \ --mixed_precision bf16 \
--batch_size 1 \ --batch_size 1 \
--train_steps 1200 \ --train_steps 3000 \
--rank 128 \ --rank 128 \
--lora_alpha 128 \ --lora_alpha 128 \
--target_modules to_q to_k to_v to_out.0 \ --target_modules to_q to_k to_v to_out.0 \
--gradient_accumulation_steps 1 \ --gradient_accumulation_steps 4 \
--gradient_checkpointing \ --gradient_checkpointing \
--checkpointing_steps 500 \ --checkpointing_steps 500 \
--checkpointing_limit 2 \ --checkpointing_limit 2 \