Spaces:
Configuration error
Configuration error
File size: 3,488 Bytes
87e21d1 |
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 83 |
# Copyright 2024 MIT Han Lab
#
# Licensed under the Apache License, Version 2.0 (the "License");
# you may not use this file except in compliance with the License.
# You may obtain a copy of the License at
#
# http://www.apache.org/licenses/LICENSE-2.0
#
# Unless required by applicable law or agreed to in writing, software
# distributed under the License is distributed on an "AS IS" BASIS,
# WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied.
# See the License for the specific language governing permissions and
# limitations under the License.
#
# SPDX-License-Identifier: Apache-2.0
from typing import Callable, Optional
import diffusers
import torch
from huggingface_hub import PyTorchModelHubMixin
from torch import nn
from ..efficientvit.models.efficientvit.dc_ae import DCAE, DCAEConfig, dc_ae_f32c32, dc_ae_f64c128, dc_ae_f128c512
__all__ = ["create_dc_ae_model_cfg", "DCAE_HF", "AutoencoderKL"]
REGISTERED_DCAE_MODEL: dict[str, tuple[Callable, Optional[str]]] = {
"dc-ae-f32c32-in-1.0": (dc_ae_f32c32, None),
"dc-ae-f64c128-in-1.0": (dc_ae_f64c128, None),
"dc-ae-f128c512-in-1.0": (dc_ae_f128c512, None),
#################################################################################################
"dc-ae-f32c32-mix-1.0": (dc_ae_f32c32, None),
"dc-ae-f64c128-mix-1.0": (dc_ae_f64c128, None),
"dc-ae-f128c512-mix-1.0": (dc_ae_f128c512, None),
#################################################################################################
"dc-ae-f32c32-sana-1.0": (dc_ae_f32c32, None),
}
def create_dc_ae_model_cfg(name: str, pretrained_path: Optional[str] = None) -> DCAEConfig:
assert name in REGISTERED_DCAE_MODEL, f"{name} is not supported"
dc_ae_cls, default_pt_path = REGISTERED_DCAE_MODEL[name]
pretrained_path = default_pt_path if pretrained_path is None else pretrained_path
model_cfg = dc_ae_cls(name, pretrained_path)
return model_cfg
class DCAE_HF(DCAE, PyTorchModelHubMixin):
def __init__(self, model_name: str):
cfg = create_dc_ae_model_cfg(model_name)
DCAE.__init__(self, cfg)
class AutoencoderKL(nn.Module):
def __init__(self, model_name: str):
super().__init__()
self.model_name = model_name
if self.model_name in ["stabilityai/sd-vae-ft-ema"]:
self.model = diffusers.models.AutoencoderKL.from_pretrained(self.model_name)
self.spatial_compression_ratio = 8
elif self.model_name == "flux-vae":
from diffusers import FluxPipeline
pipe = FluxPipeline.from_pretrained("black-forest-labs/FLUX.1-schnell", torch_dtype=torch.bfloat16)
self.model = diffusers.models.AutoencoderKL.from_pretrained(pipe.vae.config._name_or_path)
self.spatial_compression_ratio = 8
else:
raise ValueError(f"{self.model_name} is not supported for AutoencoderKL")
def encode(self, x: torch.Tensor) -> torch.Tensor:
if self.model_name in ["stabilityai/sd-vae-ft-ema", "flux-vae"]:
return self.model.encode(x).latent_dist.sample()
else:
raise ValueError(f"{self.model_name} is not supported for AutoencoderKL")
def decode(self, latent: torch.Tensor) -> torch.Tensor:
if self.model_name in ["stabilityai/sd-vae-ft-ema", "flux-vae"]:
return self.model.decode(latent).sample
else:
raise ValueError(f"{self.model_name} is not supported for AutoencoderKL")
|