Spaces:
Runtime error
Runtime error
File size: 3,042 Bytes
ad93086 |
1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 18 19 20 21 22 23 24 25 26 27 28 29 30 31 32 33 34 35 36 37 38 39 40 41 42 43 44 45 46 47 48 49 50 51 52 53 54 55 56 57 58 59 60 61 62 63 64 65 66 67 68 69 70 71 72 73 74 75 76 77 78 79 80 81 82 |
import torch
from huggingface_guess import model_list
from backend.diffusion_engine.base import ForgeDiffusionEngine, ForgeObjects
from backend.patcher.clip import CLIP
from backend.patcher.vae import VAE
from backend.patcher.unet import UnetPatcher
from backend.text_processing.classic_engine import ClassicTextProcessingEngine
from backend.args import dynamic_args
from backend import memory_management
class StableDiffusion2(ForgeDiffusionEngine):
matched_guesses = [model_list.SD20]
def __init__(self, estimated_config, huggingface_components):
super().__init__(estimated_config, huggingface_components)
clip = CLIP(
model_dict={
'clip_h': huggingface_components['text_encoder']
},
tokenizer_dict={
'clip_h': huggingface_components['tokenizer']
}
)
vae = VAE(model=huggingface_components['vae'])
unet = UnetPatcher.from_model(
model=huggingface_components['unet'],
diffusers_scheduler=huggingface_components['scheduler'],
config=estimated_config
)
self.text_processing_engine = ClassicTextProcessingEngine(
text_encoder=clip.cond_stage_model.clip_h,
tokenizer=clip.tokenizer.clip_h,
embedding_dir=dynamic_args['embedding_dir'],
embedding_key='clip_h',
embedding_expected_shape=1024,
emphasis_name=dynamic_args['emphasis_name'],
text_projection=False,
minimal_clip_skip=1,
clip_skip=1,
return_pooled=False,
final_layer_norm=True,
)
self.forge_objects = ForgeObjects(unet=unet, clip=clip, vae=vae, clipvision=None)
self.forge_objects_original = self.forge_objects.shallow_copy()
self.forge_objects_after_applying_lora = self.forge_objects.shallow_copy()
# WebUI Legacy
self.is_sd2 = True
def set_clip_skip(self, clip_skip):
self.text_processing_engine.clip_skip = clip_skip
@torch.inference_mode()
def get_learned_conditioning(self, prompt: list[str]):
memory_management.load_model_gpu(self.forge_objects.clip.patcher)
cond = self.text_processing_engine(prompt)
return cond
@torch.inference_mode()
def get_prompt_lengths_on_ui(self, prompt):
_, token_count = self.text_processing_engine.process_texts([prompt])
return token_count, self.text_processing_engine.get_target_prompt_token_count(token_count)
@torch.inference_mode()
def encode_first_stage(self, x):
sample = self.forge_objects.vae.encode(x.movedim(1, -1) * 0.5 + 0.5)
sample = self.forge_objects.vae.first_stage_model.process_in(sample)
return sample.to(x)
@torch.inference_mode()
def decode_first_stage(self, x):
sample = self.forge_objects.vae.first_stage_model.process_out(x)
sample = self.forge_objects.vae.decode(sample).movedim(-1, 1) * 2.0 - 1.0
return sample.to(x)
|