86 lines
4.1 KiB
Python
86 lines
4.1 KiB
Python
#
|
|
# Composable-Diffusion with Lora
|
|
#
|
|
import torch
|
|
import gradio as gr
|
|
|
|
import composable_lora
|
|
import modules.scripts as scripts
|
|
from modules import script_callbacks
|
|
from modules.processing import StableDiffusionProcessing
|
|
|
|
def unload():
|
|
torch.nn.Linear.forward = torch.nn.Linear_forward_before_lora
|
|
torch.nn.Conv2d.forward = torch.nn.Conv2d_forward_before_lora
|
|
|
|
if not hasattr(torch.nn, 'Linear_forward_before_lora'):
|
|
torch.nn.Linear_forward_before_lora = torch.nn.Linear.forward
|
|
|
|
if not hasattr(torch.nn, 'Conv2d_forward_before_lora'):
|
|
torch.nn.Conv2d_forward_before_lora = torch.nn.Conv2d.forward
|
|
|
|
torch.nn.Linear.forward = composable_lora.lora_Linear_forward
|
|
torch.nn.Conv2d.forward = composable_lora.lora_Conv2d_forward
|
|
|
|
script_callbacks.on_script_unloaded(unload)
|
|
|
|
class ComposableLoraScript(scripts.Script):
|
|
def title(self):
|
|
return "Composable Lora"
|
|
|
|
def show(self, is_img2img):
|
|
return scripts.AlwaysVisible
|
|
|
|
def ui(self, is_img2img):
|
|
with gr.Group():
|
|
with gr.Accordion("Composable Lora", open=False):
|
|
enabled = gr.Checkbox(value=False, label="Enabled")
|
|
opt_composable_with_step = gr.Checkbox(value=False, label="Composable LoRA with step")
|
|
opt_uc_text_model_encoder = gr.Checkbox(value=False, label="Use Lora in uc text model encoder")
|
|
opt_uc_diffusion_model = gr.Checkbox(value=False, label="Use Lora in uc diffusion model")
|
|
opt_plot_lora_weight = gr.Checkbox(value=False, label="Plot the LoRA weight in all steps")
|
|
opt_single_no_uc = gr.Checkbox(value=False, label="Don't use LoRA in uc if there're no subprompts")
|
|
unload_ext = gr.Checkbox(value=False, label="Unload (If you got a corrupted image, try uncheck [Enabled] and checking this option and generate an image without LoRA, and then turn off this option.)")
|
|
def enabled_changed(opt_enabled: bool, opt_unload_ext: bool):
|
|
if opt_enabled:
|
|
unload_ext.interactive=False
|
|
return False
|
|
else:
|
|
unload_ext.interactive=True
|
|
return opt_unload_ext
|
|
enabled.change(enabled_changed, inputs=[enabled, unload_ext], outputs=[unload_ext])
|
|
return [enabled, opt_composable_with_step, opt_uc_text_model_encoder, opt_uc_diffusion_model, opt_plot_lora_weight, opt_single_no_uc, unload_ext]
|
|
|
|
def process(self, p: StableDiffusionProcessing, enabled: bool, opt_composable_with_step: bool, opt_uc_text_model_encoder: bool, opt_uc_diffusion_model: bool, opt_plot_lora_weight: bool, opt_single_no_uc: bool, unload_ext : bool):
|
|
composable_lora.enabled = enabled
|
|
composable_lora.opt_uc_text_model_encoder = opt_uc_text_model_encoder
|
|
composable_lora.opt_uc_diffusion_model = opt_uc_diffusion_model
|
|
composable_lora.opt_composable_with_step = opt_composable_with_step
|
|
composable_lora.opt_plot_lora_weight = opt_plot_lora_weight
|
|
composable_lora.opt_single_no_uc = opt_single_no_uc
|
|
|
|
composable_lora.num_batches = p.batch_size
|
|
composable_lora.num_steps = p.steps
|
|
|
|
composable_lora.backup_lora_Linear_forward = torch.nn.Linear.forward
|
|
composable_lora.backup_lora_Conv2d_forward = torch.nn.Conv2d.forward
|
|
if (composable_lora.should_reload() or (torch.nn.Linear.forward != composable_lora.lora_Linear_forward)):
|
|
if enabled or not unload_ext:
|
|
torch.nn.Linear.forward = composable_lora.lora_Linear_forward
|
|
torch.nn.Conv2d.forward = composable_lora.lora_Conv2d_forward
|
|
|
|
composable_lora.reset_step_counters()
|
|
|
|
prompt = p.all_prompts[0]
|
|
composable_lora.load_prompt_loras(prompt)
|
|
|
|
def process_batch(self, p: StableDiffusionProcessing, *args, **kwargs):
|
|
composable_lora.reset_counters()
|
|
|
|
def postprocess(self, p, processed, *args):
|
|
torch.nn.Linear.forward = composable_lora.backup_lora_Linear_forward
|
|
torch.nn.Conv2d.forward = composable_lora.backup_lora_Conv2d_forward
|
|
if composable_lora.enabled:
|
|
if composable_lora.opt_plot_lora_weight:
|
|
processed.images.extend([composable_lora.plot_lora()])
|