mirror of https://github.com/bmaltais/kohya_ss
Updates
bmaltais
parent
4104599942
commit
fc5b0b1798
|
|
@ -511,6 +511,7 @@ If you come across a `FileNotFoundError`, it is likely due to an installation is
|
|||
|
||||
* 2023/08/05 (v21.8.8)
|
||||
- Fix issue with aiofiles: https://github.com/bmaltais/kohya_ss/issues/1359
|
||||
- Merge sd-scripts updates as of Aug 11 2023
|
||||
* 2023/08/05 (v21.8.7)
|
||||
- Add manual captioning option. Thanks to https://github.com/channelcat for this great contribution. (https://github.com/bmaltais/kohya_ss/pull/1352)
|
||||
- Added support for `v_pred_like_loss` to the advanced training tab
|
||||
|
|
|
|||
|
|
@ -1,5 +1,5 @@
|
|||
import gradio as gr
|
||||
from library.merge_lora_gui import gradio_merge_lora_tab
|
||||
from library.merge_lora_gui import GradioMergeLoRaTab
|
||||
from library.svd_merge_lora_gui import gradio_svd_merge_lora_tab
|
||||
from library.verify_lora_gui import gradio_verify_lora_tab
|
||||
from library.resize_lora_gui import gradio_resize_lora_tab
|
||||
|
|
@ -25,7 +25,7 @@ class LoRATools:
|
|||
gradio_extract_dylora_tab(headless=headless)
|
||||
gradio_extract_lora_tab(headless=headless)
|
||||
gradio_extract_lycoris_locon_tab(headless=headless)
|
||||
gradio_merge_lora_tab(headless=headless)
|
||||
gradio_merge_lora_tab = GradioMergeLoRaTab()
|
||||
gradio_merge_lycoris_tab(headless=headless)
|
||||
gradio_svd_merge_lora_tab(headless=headless)
|
||||
gradio_resize_lora_tab(headless=headless)
|
||||
|
|
|
|||
|
|
@ -37,261 +37,257 @@ def verify_conditions(sd_model, lora_models):
|
|||
return True
|
||||
return False
|
||||
|
||||
class GradioMergeLoRaTab:
|
||||
def __init__(self, headless=False):
|
||||
self.headless = headless
|
||||
self.build_tab()
|
||||
|
||||
def save_inputs_to_json(self, file_path, inputs):
|
||||
with open(file_path, 'w') as file:
|
||||
json.dump(inputs, file)
|
||||
log.info(f'Saved inputs to {file_path}')
|
||||
|
||||
def merge_lora(
|
||||
sd_model,
|
||||
sdxl_model,
|
||||
lora_a_model,
|
||||
lora_b_model,
|
||||
lora_c_model,
|
||||
lora_d_model,
|
||||
ratio_a,
|
||||
ratio_b,
|
||||
ratio_c,
|
||||
ratio_d,
|
||||
save_to,
|
||||
precision,
|
||||
save_precision,
|
||||
):
|
||||
log.info('Merge model...')
|
||||
models = [sd_model, lora_a_model, lora_b_model, lora_c_model, lora_d_model]
|
||||
lora_models = models[1:]
|
||||
ratios = [ratio_a, ratio_b, ratio_c, ratio_d]
|
||||
def load_inputs_from_json(self, file_path):
|
||||
with open(file_path, 'r') as file:
|
||||
inputs = json.load(file)
|
||||
log.info(f'Loaded inputs from {file_path}')
|
||||
return inputs
|
||||
|
||||
if not verify_conditions(sd_model, lora_models):
|
||||
log.info(
|
||||
'Warning: Either provide at least one LoRa model along with the sd_model or at least two LoRa models if no sd_model is provided.'
|
||||
)
|
||||
return
|
||||
def build_tab(self):
|
||||
with gr.Tab('Merge LoRA'):
|
||||
gr.Markdown(
|
||||
'This utility can merge up to 4 LoRA together or alternatively merge up to 4 LoRA into a SD checkpoint.'
|
||||
)
|
||||
|
||||
for model in models:
|
||||
if not check_model(model):
|
||||
lora_ext = gr.Textbox(value='*.safetensors *.pt', visible=False)
|
||||
lora_ext_name = gr.Textbox(value='LoRA model types', visible=False)
|
||||
ckpt_ext = gr.Textbox(value='*.safetensors *.ckpt', visible=False)
|
||||
ckpt_ext_name = gr.Textbox(value='SD model types', visible=False)
|
||||
|
||||
with gr.Row():
|
||||
sd_model = gr.Textbox(
|
||||
label='SD Model',
|
||||
placeholder='(Optional) Stable Diffusion model',
|
||||
interactive=True,
|
||||
info='Provide a SD file path IF you want to merge it with LoRA files',
|
||||
)
|
||||
sd_model_file = gr.Button(
|
||||
folder_symbol,
|
||||
elem_id='open_folder_small',
|
||||
visible=(not self.headless),
|
||||
)
|
||||
sd_model_file.click(
|
||||
get_file_path,
|
||||
inputs=[sd_model, ckpt_ext, ckpt_ext_name],
|
||||
outputs=sd_model,
|
||||
show_progress=False,
|
||||
)
|
||||
sdxl_model = gr.Checkbox(label='SDXL model', value=False)
|
||||
|
||||
with gr.Row():
|
||||
lora_a_model = gr.Textbox(
|
||||
label='LoRA model "A"',
|
||||
placeholder='Path to the LoRA A model',
|
||||
interactive=True,
|
||||
)
|
||||
button_lora_a_model_file = gr.Button(
|
||||
folder_symbol,
|
||||
elem_id='open_folder_small',
|
||||
visible=(not self.headless),
|
||||
)
|
||||
button_lora_a_model_file.click(
|
||||
get_file_path,
|
||||
inputs=[lora_a_model, lora_ext, lora_ext_name],
|
||||
outputs=lora_a_model,
|
||||
show_progress=False,
|
||||
)
|
||||
|
||||
lora_b_model = gr.Textbox(
|
||||
label='LoRA model "B"',
|
||||
placeholder='Path to the LoRA B model',
|
||||
interactive=True,
|
||||
)
|
||||
button_lora_b_model_file = gr.Button(
|
||||
folder_symbol,
|
||||
elem_id='open_folder_small',
|
||||
visible=(not self.headless),
|
||||
)
|
||||
button_lora_b_model_file.click(
|
||||
get_file_path,
|
||||
inputs=[lora_b_model, lora_ext, lora_ext_name],
|
||||
outputs=lora_b_model,
|
||||
show_progress=False,
|
||||
)
|
||||
|
||||
with gr.Row():
|
||||
ratio_a = gr.Slider(
|
||||
label='Model A merge ratio (eg: 0.5 mean 50%)',
|
||||
minimum=0,
|
||||
maximum=1,
|
||||
step=0.01,
|
||||
value=0.0,
|
||||
interactive=True,
|
||||
)
|
||||
|
||||
ratio_b = gr.Slider(
|
||||
label='Model B merge ratio (eg: 0.5 mean 50%)',
|
||||
minimum=0,
|
||||
maximum=1,
|
||||
step=0.01,
|
||||
value=0.0,
|
||||
interactive=True,
|
||||
)
|
||||
|
||||
with gr.Row():
|
||||
lora_c_model = gr.Textbox(
|
||||
label='LoRA model "C"',
|
||||
placeholder='Path to the LoRA C model',
|
||||
interactive=True,
|
||||
)
|
||||
button_lora_c_model_file = gr.Button(
|
||||
folder_symbol,
|
||||
elem_id='open_folder_small',
|
||||
visible=(not self.headless),
|
||||
)
|
||||
button_lora_c_model_file.click(
|
||||
get_file_path,
|
||||
inputs=[lora_c_model, lora_ext, lora_ext_name],
|
||||
outputs=lora_c_model,
|
||||
show_progress=False,
|
||||
)
|
||||
|
||||
lora_d_model = gr.Textbox(
|
||||
label='LoRA model "D"',
|
||||
placeholder='Path to the LoRA D model',
|
||||
interactive=True,
|
||||
)
|
||||
button_lora_d_model_file = gr.Button(
|
||||
folder_symbol,
|
||||
elem_id='open_folder_small',
|
||||
visible=(not self.headless),
|
||||
)
|
||||
button_lora_d_model_file.click(
|
||||
get_file_path,
|
||||
inputs=[lora_d_model, lora_ext, lora_ext_name],
|
||||
outputs=lora_d_model,
|
||||
show_progress=False,
|
||||
)
|
||||
|
||||
with gr.Row():
|
||||
ratio_c = gr.Slider(
|
||||
label='Model C merge ratio (eg: 0.5 mean 50%)',
|
||||
minimum=0,
|
||||
maximum=1,
|
||||
step=0.01,
|
||||
value=0.0,
|
||||
interactive=True,
|
||||
)
|
||||
|
||||
ratio_d = gr.Slider(
|
||||
label='Model D merge ratio (eg: 0.5 mean 50%)',
|
||||
minimum=0,
|
||||
maximum=1,
|
||||
step=0.01,
|
||||
value=0.0,
|
||||
interactive=True,
|
||||
)
|
||||
|
||||
with gr.Row():
|
||||
save_to = gr.Textbox(
|
||||
label='Save to',
|
||||
placeholder='path for the file to save...',
|
||||
interactive=True,
|
||||
)
|
||||
button_save_to = gr.Button(
|
||||
folder_symbol,
|
||||
elem_id='open_folder_small',
|
||||
visible=(not self.headless),
|
||||
)
|
||||
button_save_to.click(
|
||||
get_saveasfilename_path,
|
||||
inputs=[save_to, lora_ext, lora_ext_name],
|
||||
outputs=save_to,
|
||||
show_progress=False,
|
||||
)
|
||||
precision = gr.Dropdown(
|
||||
label='Merge precision',
|
||||
choices=['fp16', 'bf16', 'float'],
|
||||
value='float',
|
||||
interactive=True,
|
||||
)
|
||||
save_precision = gr.Dropdown(
|
||||
label='Save precision',
|
||||
choices=['fp16', 'bf16', 'float'],
|
||||
value='fp16',
|
||||
interactive=True,
|
||||
)
|
||||
|
||||
merge_button = gr.Button('Merge model')
|
||||
|
||||
merge_button.click(
|
||||
self.merge_lora,
|
||||
inputs=[
|
||||
sd_model,
|
||||
sdxl_model,
|
||||
lora_a_model,
|
||||
lora_b_model,
|
||||
lora_c_model,
|
||||
lora_d_model,
|
||||
ratio_a,
|
||||
ratio_b,
|
||||
ratio_c,
|
||||
ratio_d,
|
||||
save_to,
|
||||
precision,
|
||||
save_precision,
|
||||
],
|
||||
show_progress=False,
|
||||
)
|
||||
|
||||
def merge_lora(self, sd_model, sdxl_model, lora_a_model, lora_b_model, lora_c_model, lora_d_model, ratio_a, ratio_b, ratio_c, ratio_d, save_to, precision, save_precision):
|
||||
|
||||
log.info('Merge model...')
|
||||
models = [sd_model, lora_a_model, lora_b_model, lora_c_model, lora_d_model]
|
||||
lora_models = models[1:]
|
||||
ratios = [ratio_a, ratio_b, ratio_c, ratio_d]
|
||||
|
||||
if not verify_conditions(sd_model, lora_models):
|
||||
log.info(
|
||||
'Warning: Either provide at least one LoRa model along with the sd_model or at least two LoRa models if no sd_model is provided.'
|
||||
)
|
||||
return
|
||||
|
||||
if not sdxl_model:
|
||||
run_cmd = f'{PYTHON} "{os.path.join("networks","merge_lora.py")}"'
|
||||
else:
|
||||
run_cmd = f'{PYTHON} "{os.path.join("networks","sdxl_merge_lora.py")}"'
|
||||
if sd_model:
|
||||
run_cmd += f' --sd_model "{sd_model}"'
|
||||
run_cmd += f' --save_precision {save_precision}'
|
||||
run_cmd += f' --precision {precision}'
|
||||
run_cmd += f' --save_to "{save_to}"'
|
||||
for model in models:
|
||||
if not check_model(model):
|
||||
return
|
||||
|
||||
# Create a space-separated string of non-empty models (from the second element onwards), enclosed in double quotes
|
||||
models_cmd = ' '.join([f'"{model}"' for model in lora_models if model])
|
||||
if not sdxl_model:
|
||||
run_cmd = f'{PYTHON} "{os.path.join("networks","merge_lora.py")}"'
|
||||
else:
|
||||
run_cmd = f'{PYTHON} "{os.path.join("networks","sdxl_merge_lora.py")}"'
|
||||
if sd_model:
|
||||
run_cmd += f' --sd_model "{sd_model}"'
|
||||
run_cmd += f' --save_precision {save_precision}'
|
||||
run_cmd += f' --precision {precision}'
|
||||
run_cmd += f' --save_to "{save_to}"'
|
||||
|
||||
# Create a space-separated string of non-zero ratios corresponding to non-empty LoRa models
|
||||
valid_ratios = [ratios[i] for i, model in enumerate(lora_models) if model]
|
||||
ratios_cmd = ' '.join([str(ratio) for ratio in valid_ratios])
|
||||
# Create a space-separated string of non-empty models (from the second element onwards), enclosed in double quotes
|
||||
models_cmd = ' '.join([f'"{model}"' for model in lora_models if model])
|
||||
|
||||
if models_cmd:
|
||||
run_cmd += f' --models {models_cmd}'
|
||||
run_cmd += f' --ratios {ratios_cmd}'
|
||||
# Create a space-separated string of non-zero ratios corresponding to non-empty LoRa models
|
||||
valid_ratios = [ratios[i] for i, model in enumerate(lora_models) if model]
|
||||
ratios_cmd = ' '.join([str(ratio) for ratio in valid_ratios])
|
||||
|
||||
log.info(run_cmd)
|
||||
if models_cmd:
|
||||
run_cmd += f' --models {models_cmd}'
|
||||
run_cmd += f' --ratios {ratios_cmd}'
|
||||
|
||||
# Run the command
|
||||
if os.name == 'posix':
|
||||
os.system(run_cmd)
|
||||
else:
|
||||
subprocess.run(run_cmd)
|
||||
log.info(run_cmd)
|
||||
|
||||
log.info('Done merging...')
|
||||
# Run the command
|
||||
if os.name == 'posix':
|
||||
os.system(run_cmd)
|
||||
else:
|
||||
subprocess.run(run_cmd)
|
||||
|
||||
|
||||
###
|
||||
# Gradio UI
|
||||
###
|
||||
|
||||
|
||||
def gradio_merge_lora_tab(headless=False):
|
||||
with gr.Tab('Merge LoRA'):
|
||||
gr.Markdown(
|
||||
'This utility can merge up to 4 LoRA together or alternatively merge up to 4 LoRA into a SD checkpoint.'
|
||||
)
|
||||
|
||||
lora_ext = gr.Textbox(value='*.safetensors *.pt', visible=False)
|
||||
lora_ext_name = gr.Textbox(value='LoRA model types', visible=False)
|
||||
ckpt_ext = gr.Textbox(value='*.safetensors *.ckpt', visible=False)
|
||||
ckpt_ext_name = gr.Textbox(value='SD model types', visible=False)
|
||||
|
||||
with gr.Row():
|
||||
sd_model = gr.Textbox(
|
||||
label='SD Model',
|
||||
placeholder='(Optional) Stable Diffusion model',
|
||||
interactive=True,
|
||||
info='Provide a SD file path IF you want to merge it with LoRA files',
|
||||
)
|
||||
sd_model_file = gr.Button(
|
||||
folder_symbol,
|
||||
elem_id='open_folder_small',
|
||||
visible=(not headless),
|
||||
)
|
||||
sd_model_file.click(
|
||||
get_file_path,
|
||||
inputs=[sd_model, ckpt_ext, ckpt_ext_name],
|
||||
outputs=sd_model,
|
||||
show_progress=False,
|
||||
)
|
||||
sdxl_model = gr.Checkbox(label='SDXL model', value=False)
|
||||
|
||||
with gr.Row():
|
||||
lora_a_model = gr.Textbox(
|
||||
label='LoRA model "A"',
|
||||
placeholder='Path to the LoRA A model',
|
||||
interactive=True,
|
||||
)
|
||||
button_lora_a_model_file = gr.Button(
|
||||
folder_symbol,
|
||||
elem_id='open_folder_small',
|
||||
visible=(not headless),
|
||||
)
|
||||
button_lora_a_model_file.click(
|
||||
get_file_path,
|
||||
inputs=[lora_a_model, lora_ext, lora_ext_name],
|
||||
outputs=lora_a_model,
|
||||
show_progress=False,
|
||||
)
|
||||
|
||||
lora_b_model = gr.Textbox(
|
||||
label='LoRA model "B"',
|
||||
placeholder='Path to the LoRA B model',
|
||||
interactive=True,
|
||||
)
|
||||
button_lora_b_model_file = gr.Button(
|
||||
folder_symbol,
|
||||
elem_id='open_folder_small',
|
||||
visible=(not headless),
|
||||
)
|
||||
button_lora_b_model_file.click(
|
||||
get_file_path,
|
||||
inputs=[lora_b_model, lora_ext, lora_ext_name],
|
||||
outputs=lora_b_model,
|
||||
show_progress=False,
|
||||
)
|
||||
|
||||
with gr.Row():
|
||||
ratio_a = gr.Slider(
|
||||
label='Model A merge ratio (eg: 0.5 mean 50%)',
|
||||
minimum=0,
|
||||
maximum=1,
|
||||
step=0.01,
|
||||
value=0.0,
|
||||
interactive=True,
|
||||
)
|
||||
|
||||
ratio_b = gr.Slider(
|
||||
label='Model B merge ratio (eg: 0.5 mean 50%)',
|
||||
minimum=0,
|
||||
maximum=1,
|
||||
step=0.01,
|
||||
value=0.0,
|
||||
interactive=True,
|
||||
)
|
||||
|
||||
with gr.Row():
|
||||
lora_c_model = gr.Textbox(
|
||||
label='LoRA model "C"',
|
||||
placeholder='Path to the LoRA C model',
|
||||
interactive=True,
|
||||
)
|
||||
button_lora_c_model_file = gr.Button(
|
||||
folder_symbol,
|
||||
elem_id='open_folder_small',
|
||||
visible=(not headless),
|
||||
)
|
||||
button_lora_c_model_file.click(
|
||||
get_file_path,
|
||||
inputs=[lora_c_model, lora_ext, lora_ext_name],
|
||||
outputs=lora_c_model,
|
||||
show_progress=False,
|
||||
)
|
||||
|
||||
lora_d_model = gr.Textbox(
|
||||
label='LoRA model "D"',
|
||||
placeholder='Path to the LoRA D model',
|
||||
interactive=True,
|
||||
)
|
||||
button_lora_d_model_file = gr.Button(
|
||||
folder_symbol,
|
||||
elem_id='open_folder_small',
|
||||
visible=(not headless),
|
||||
)
|
||||
button_lora_d_model_file.click(
|
||||
get_file_path,
|
||||
inputs=[lora_d_model, lora_ext, lora_ext_name],
|
||||
outputs=lora_d_model,
|
||||
show_progress=False,
|
||||
)
|
||||
|
||||
with gr.Row():
|
||||
ratio_c = gr.Slider(
|
||||
label='Model C merge ratio (eg: 0.5 mean 50%)',
|
||||
minimum=0,
|
||||
maximum=1,
|
||||
step=0.01,
|
||||
value=0.0,
|
||||
interactive=True,
|
||||
)
|
||||
|
||||
ratio_d = gr.Slider(
|
||||
label='Model D merge ratio (eg: 0.5 mean 50%)',
|
||||
minimum=0,
|
||||
maximum=1,
|
||||
step=0.01,
|
||||
value=0.0,
|
||||
interactive=True,
|
||||
)
|
||||
|
||||
with gr.Row():
|
||||
save_to = gr.Textbox(
|
||||
label='Save to',
|
||||
placeholder='path for the file to save...',
|
||||
interactive=True,
|
||||
)
|
||||
button_save_to = gr.Button(
|
||||
folder_symbol,
|
||||
elem_id='open_folder_small',
|
||||
visible=(not headless),
|
||||
)
|
||||
button_save_to.click(
|
||||
get_saveasfilename_path,
|
||||
inputs=[save_to, lora_ext, lora_ext_name],
|
||||
outputs=save_to,
|
||||
show_progress=False,
|
||||
)
|
||||
precision = gr.Dropdown(
|
||||
label='Merge precision',
|
||||
choices=['fp16', 'bf16', 'float'],
|
||||
value='float',
|
||||
interactive=True,
|
||||
)
|
||||
save_precision = gr.Dropdown(
|
||||
label='Save precision',
|
||||
choices=['fp16', 'bf16', 'float'],
|
||||
value='fp16',
|
||||
interactive=True,
|
||||
)
|
||||
|
||||
merge_button = gr.Button('Merge model')
|
||||
|
||||
merge_button.click(
|
||||
merge_lora,
|
||||
inputs=[
|
||||
sd_model,
|
||||
sdxl_model,
|
||||
lora_a_model,
|
||||
lora_b_model,
|
||||
lora_c_model,
|
||||
lora_d_model,
|
||||
ratio_a,
|
||||
ratio_b,
|
||||
ratio_c,
|
||||
ratio_d,
|
||||
save_to,
|
||||
precision,
|
||||
save_precision,
|
||||
],
|
||||
show_progress=False,
|
||||
)
|
||||
log.info('Done merging...')
|
||||
|
|
|
|||
|
|
@ -0,0 +1,95 @@
|
|||
{
|
||||
"LoRA_type": "Standard",
|
||||
"adaptive_noise_scale": 0,
|
||||
"additional_parameters": "--max_grad_norm=1",
|
||||
"block_alphas": "",
|
||||
"block_dims": "",
|
||||
"block_lr_zero_threshold": "",
|
||||
"bucket_no_upscale": true,
|
||||
"bucket_reso_steps": 32,
|
||||
"cache_latents": true,
|
||||
"cache_latents_to_disk": true,
|
||||
"caption_dropout_every_n_epochs": 0.0,
|
||||
"caption_dropout_rate": 0,
|
||||
"caption_extension": ".txt",
|
||||
"clip_skip": "1",
|
||||
"color_aug": false,
|
||||
"conv_alpha": 4,
|
||||
"conv_block_alphas": "",
|
||||
"conv_block_dims": "",
|
||||
"conv_dim": 8,
|
||||
"decompose_both": false,
|
||||
"dim_from_weights": false,
|
||||
"down_lr_weight": "",
|
||||
"enable_bucket": true,
|
||||
"epoch": 160,
|
||||
"factor": -1,
|
||||
"flip_aug": false,
|
||||
"full_bf16": false,
|
||||
"full_fp16": false,
|
||||
"gradient_accumulation_steps": 1.0,
|
||||
"gradient_checkpointing": true,
|
||||
"keep_tokens": "0",
|
||||
"learning_rate": 1.0,
|
||||
"lora_network_weights": "",
|
||||
"lr_scheduler": "cosine",
|
||||
"lr_scheduler_num_cycles": "",
|
||||
"lr_scheduler_power": "",
|
||||
"lr_warmup": 0,
|
||||
"max_bucket_reso": 2048,
|
||||
"max_data_loader_n_workers": "0",
|
||||
"max_resolution": "1024,1024",
|
||||
"max_timestep": 1000,
|
||||
"max_token_length": "75",
|
||||
"max_train_epochs": "",
|
||||
"max_train_steps": "320",
|
||||
"mem_eff_attn": false,
|
||||
"mid_lr_weight": "",
|
||||
"min_bucket_reso": 256,
|
||||
"min_snr_gamma": 0,
|
||||
"min_timestep": 0,
|
||||
"mixed_precision": "bf16",
|
||||
"module_dropout": 0,
|
||||
"multires_noise_discount": 0.2,
|
||||
"multires_noise_iterations": 8,
|
||||
"network_alpha": 32,
|
||||
"network_dim": 32,
|
||||
"network_dropout": 0,
|
||||
"no_token_padding": false,
|
||||
"noise_offset": 0.0357,
|
||||
"noise_offset_type": "Original",
|
||||
"num_cpu_threads_per_process": 2,
|
||||
"optimizer": "Prodigy",
|
||||
"optimizer_args": "decouple=True weight_decay=0.5 betas=0.9,0.99 use_bias_correction=False",
|
||||
"persistent_data_loader_workers": false,
|
||||
"prior_loss_weight": 1.0,
|
||||
"random_crop": false,
|
||||
"rank_dropout": 0,
|
||||
"save_every_n_epochs": 10,
|
||||
"save_every_n_steps": 0,
|
||||
"save_last_n_steps": 0,
|
||||
"save_last_n_steps_state": 0,
|
||||
"save_precision": "bf16",
|
||||
"scale_v_pred_loss_like_noise_pred": false,
|
||||
"scale_weight_norms": 1,
|
||||
"sdxl": true,
|
||||
"sdxl_cache_text_encoder_outputs": false,
|
||||
"sdxl_no_half_vae": true,
|
||||
"seed": "12345",
|
||||
"shuffle_caption": false,
|
||||
"stop_text_encoder_training": 0,
|
||||
"text_encoder_lr": 1.0,
|
||||
"train_batch_size": 4,
|
||||
"train_on_input": false,
|
||||
"training_comment": "",
|
||||
"unet_lr": 1.0,
|
||||
"unit": 1,
|
||||
"up_lr_weight": "",
|
||||
"use_cp": true,
|
||||
"use_wandb": false,
|
||||
"v2": false,
|
||||
"v_parameterization": false,
|
||||
"vae_batch_size": 0,
|
||||
"weighted_captions": false,
|
||||
"xformers": true
|
||||
}
|
||||
|
|
@ -6,7 +6,7 @@
|
|||
"block_dims": "",
|
||||
"block_lr_zero_threshold": "",
|
||||
"bucket_no_upscale": true,
|
||||
"bucket_reso_steps": 32,
|
||||
"bucket_reso_steps": 64,
|
||||
"cache_latents": true,
|
||||
"cache_latents_to_disk": true,
|
||||
"caption_dropout_every_n_epochs": 0.0,
|
||||
|
|
|
|||
Loading…
Reference in New Issue