bdsqlsz

/

HunYuanDiT-V1.2-fp16-pruned

like 6

Hi, this repository is only used for kohya sd-scripts training so there is no configuration of model_index.json

Yes, I trained on Kohya, but the script says, 'Model is not found as a file or in Hugging Face.', perhaps file name is wrong?” and “OSError: Error no file named model_index.json found in directory”
May I ask what the problem is?

Yes, I trained on Kohya, but the script says, 'Model is not found as a file or in Hugging Face.', perhaps file name is wrong?” and “OSError: Error no file named model_index.json found in directory”
May I ask what the problem is?

Can you provide scripts where you downloaded?

The sh script I used:
pretrained_model="./Stable-diffusion/HunYuanDiT_fp16"

The hunyuan model directory is located in the script root directory:
/data/train/lora-scripts/Stable-diffusion/HunYuanDiT_fp16
clip
denoiser
mt5
vae

python -m accelerate.commands.launch --num_processes=2 ${launchArgs[@]} "./sd-scripts/$train_script.py"
--pretrained_model_name_or_path=$pretrained_model
--train_data_dir=$train_data_dir
--output_dir="/sd-train/images2/logs"
--logging_dir="/sd-train/images2/logs"
--resolution=$resolution
--max_train_epochs=$max_train_epoches
--learning_rate=$lr
--output_name=$output_name
--train_batch_size=$batch_size
--save_every_n_epochs=$save_every_n_epochs
--save_precision=$save_precision
--seed=$seed
--max_token_length=225
--caption_extension=".txt"
--save_model_as=$save_model_as
${extArgs[@]}

The sh script I used:
pretrained_model="./Stable-diffusion/HunYuanDiT_fp16"

The hunyuan model directory is located in the script root directory:
/data/train/lora-scripts/Stable-diffusion/HunYuanDiT_fp16
clip
denoiser
mt5
vae

python -m accelerate.commands.launch --num_processes=2 ${launchArgs[@]} "./sd-scripts/$train_script.py"
--pretrained_model_name_or_path=$pretrained_model
--train_data_dir=$train_data_dir
--output_dir="/sd-train/images2/logs"
--logging_dir="/sd-train/images2/logs"
--resolution=$resolution
--max_train_epochs=$max_train_epoches
--learning_rate=$lr
--output_name=$output_name
--train_batch_size=$batch_size
--save_every_n_epochs=$save_every_n_epochs
--save_precision=$save_precision
--seed=$seed
--max_token_length=225
--caption_extension=".txt"
--save_model_as=$save_model_as
${extArgs[@]}

I mean that giving where you downloaded the script from, it doesn't look like it's the version I released, so it may not be supported.

The sh script I used:
pretrained_model="./Stable-diffusion/HunYuanDiT_fp16"

The hunyuan model directory is located in the script root directory:
/data/train/lora-scripts/Stable-diffusion/HunYuanDiT_fp16
clip
denoiser
mt5
vae

python -m accelerate.commands.launch --num_processes=2 ${launchArgs[@]} "./sd-scripts/$train_script.py"
--pretrained_model_name_or_path=$pretrained_model
--train_data_dir=$train_data_dir
--output_dir="/sd-train/images2/logs"
--logging_dir="/sd-train/images2/logs"
--resolution=$resolution
--max_train_epochs=$max_train_epoches
--learning_rate=$lr
--output_name=$output_name
--train_batch_size=$batch_size
--save_every_n_epochs=$save_every_n_epochs
--save_precision=$save_precision
--seed=$seed
--max_token_length=225
--caption_extension=".txt"
--save_model_as=$save_model_as
${extArgs[@]}

Update V19.1 fix this bug.