Foreign-Mirrors/Fooocus
1
0
Fork 0
mirror of https://github.com/lllyasviel/Fooocus.git synced 2026-08-16 13:13:16 +02:00
Code Issues Packages Projects Releases Wiki Activity

2.1.782

Browse Source 
2.1.782
This commit is contained in:
lllyasviel
2023-11-11 01:43:01 -08:00
parent a9bb1079cf
commit 4fe08161a5
48 changed files with 1041 additions and 2639 deletions
Download Patch File Download Diff File Expand all files Collapse all files
backend/headless/fcbh/model_base.py
+31 -28
View File
@@ -1,11 +1,9 @@
import torch
from fcbh.ldm.modules.diffusionmodules.openaimodel import UNetModel
from fcbh.ldm.modules.encoders.noise_aug_modules import CLIPEmbeddingNoiseAugmentation
from fcbh.ldm.modules.diffusionmodules.util import make_beta_schedule
from fcbh.ldm.modules.diffusionmodules.openaimodel import Timestep
import fcbh.model_management
import fcbh.conds
import numpy as np
from enum import Enum
from . import utils
@@ -13,6 +11,23 @@ class ModelType(Enum):
EPS = 1
V_PREDICTION = 2
from fcbh.model_sampling import EPS, V_PREDICTION, ModelSamplingDiscrete
def model_sampling(model_config, model_type):
if model_type == ModelType.EPS:
c = EPS
elif model_type == ModelType.V_PREDICTION:
c = V_PREDICTION
s = ModelSamplingDiscrete
class ModelSampling(s, c):
pass
return ModelSampling(model_config)
class BaseModel(torch.nn.Module):
def __init__(self, model_config, model_type=ModelType.EPS, device=None):
super().__init__()
@@ -20,10 +35,12 @@ class BaseModel(torch.nn.Module):
unet_config = model_config.unet_config
self.latent_format = model_config.latent_format
self.model_config = model_config
self.register_schedule(given_betas=None, beta_schedule=model_config.beta_schedule, timesteps=1000, linear_start=0.00085, linear_end=0.012, cosine_s=8e-3)
if not unet_config.get("disable_unet_model_creation", False):
self.diffusion_model = UNetModel(**unet_config, device=device)
self.model_type = model_type
self.model_sampling = model_sampling(model_config, model_type)
self.adm_channels = unet_config.get("adm_in_channels", None)
if self.adm_channels is None:
self.adm_channels = 0
@@ -31,39 +48,25 @@ class BaseModel(torch.nn.Module):
print("model_type", model_type.name)
print("adm", self.adm_channels)
def register_schedule(self, given_betas=None, beta_schedule="linear", timesteps=1000,
linear_start=1e-4, linear_end=2e-2, cosine_s=8e-3):
if given_betas is not None:
betas = given_betas
else:
betas = make_beta_schedule(beta_schedule, timesteps, linear_start=linear_start, linear_end=linear_end, cosine_s=cosine_s)
alphas = 1. - betas
alphas_cumprod = np.cumprod(alphas, axis=0)
alphas_cumprod_prev = np.append(1., alphas_cumprod[:-1])
timesteps, = betas.shape
self.num_timesteps = int(timesteps)
self.linear_start = linear_start
self.linear_end = linear_end
self.register_buffer('betas', torch.tensor(betas, dtype=torch.float32))
self.register_buffer('alphas_cumprod', torch.tensor(alphas_cumprod, dtype=torch.float32))
self.register_buffer('alphas_cumprod_prev', torch.tensor(alphas_cumprod_prev, dtype=torch.float32))
def apply_model(self, x, t, c_concat=None, c_crossattn=None, control=None, transformer_options={}, **kwargs):
sigma = t
xc = self.model_sampling.calculate_input(sigma, x)
if c_concat is not None:
xc = torch.cat([x] + [c_concat], dim=1)
else:
xc = x
xc = torch.cat([xc] + [c_concat], dim=1)
context = c_crossattn
dtype = self.get_dtype()
xc = xc.to(dtype)
t = t.to(dtype)
t = self.model_sampling.timestep(t).float()
context = context.to(dtype)
extra_conds = {}
for o in kwargs:
extra_conds[o] = kwargs[o].to(dtype)
return self.diffusion_model(xc, t, context=context, control=control, transformer_options=transformer_options, **extra_conds).float()
extra = kwargs[o]
if hasattr(extra, "to"):
extra = extra.to(dtype)
extra_conds[o] = extra
model_output = self.diffusion_model(xc, t, context=context, control=control, transformer_options=transformer_options, **extra_conds).float()
return self.model_sampling.calculate_denoised(sigma, model_output, x)
def get_dtype(self):
return self.diffusion_model.dtype