Instructions to use BiliSakura/DiT-diffusers with libraries, inference providers, notebooks, and local apps. Follow these links to get started.
- Libraries
- Diffusers
How to use BiliSakura/DiT-diffusers with Diffusers:
pip install -U diffusers transformers accelerate
import torch from diffusers import DiffusionPipeline # switch to "mps" for apple devices pipe = DiffusionPipeline.from_pretrained("BiliSakura/DiT-diffusers", dtype=torch.bfloat16, device_map="cuda") prompt = "Astronaut in a jungle, cold color palette, muted colors, detailed, 8k" image = pipe(prompt).images[0] - Notebooks
- Google Colab
- Kaggle
- Local Apps Settings
- Draw Things
- DiffusionBee
Fix generator determinism: forward generator through scheduler steps and seeded noise
Browse files- DiT-XL-2-256/pipeline.py +5 -61
- DiT-XL-2-512/pipeline.py +5 -61
DiT-XL-2-256/pipeline.py
CHANGED
|
@@ -1,56 +1,16 @@
|
|
| 1 |
-
|
| 2 |
-
|
| 3 |
-
|
| 4 |
-
# you may not use this file except in compliance with the License.
|
| 5 |
-
# You may obtain a copy of the License at
|
| 6 |
-
#
|
| 7 |
-
# http://www.apache.org/licenses/LICENSE-2.0
|
| 8 |
-
#
|
| 9 |
-
# Unless required by applicable law or agreed to in writing, software
|
| 10 |
-
# distributed under the License is distributed on an "AS IS" BASIS,
|
| 11 |
-
# WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied.
|
| 12 |
-
# See the License for the specific language governing permissions and
|
| 13 |
-
# limitations under the License.
|
| 14 |
|
| 15 |
import inspect
|
| 16 |
from typing import Any, Dict, List, Optional, Tuple, Union
|
| 17 |
|
| 18 |
import torch
|
| 19 |
-
|
| 20 |
from diffusers.image_processor import VaeImageProcessor
|
| 21 |
from diffusers.pipelines.pipeline_utils import DiffusionPipeline, ImagePipelineOutput
|
| 22 |
from diffusers.schedulers import KarrasDiffusionSchedulers
|
| 23 |
-
from diffusers.utils import replace_example_docstring
|
| 24 |
from diffusers.utils.torch_utils import randn_tensor
|
| 25 |
|
| 26 |
-
EXAMPLE_DOC_STRING = """
|
| 27 |
-
Examples:
|
| 28 |
-
```py
|
| 29 |
-
>>> from pathlib import Path
|
| 30 |
-
>>> import torch
|
| 31 |
-
>>> from diffusers import DiffusionPipeline
|
| 32 |
-
|
| 33 |
-
>>> model_dir = Path("path/to/BiliSakura/DiT-diffusers/DiT-XL-2-256")
|
| 34 |
-
>>> pipe = DiffusionPipeline.from_pretrained(
|
| 35 |
-
... str(model_dir),
|
| 36 |
-
... local_files_only=True,
|
| 37 |
-
... custom_pipeline=str(model_dir / "pipeline.py"),
|
| 38 |
-
... trust_remote_code=True,
|
| 39 |
-
... torch_dtype=torch.float16,
|
| 40 |
-
... )
|
| 41 |
-
>>> pipe = pipe.to("cuda")
|
| 42 |
-
|
| 43 |
-
>>> class_id = pipe.get_label_ids("golden retriever")[0]
|
| 44 |
-
>>> image = pipe(
|
| 45 |
-
... class_labels=class_id,
|
| 46 |
-
... height=256,
|
| 47 |
-
... width=256,
|
| 48 |
-
... num_inference_steps=250,
|
| 49 |
-
... guidance_scale=4.0,
|
| 50 |
-
... ).images[0]
|
| 51 |
-
```
|
| 52 |
-
"""
|
| 53 |
-
|
| 54 |
|
| 55 |
class DiTPipeline(DiffusionPipeline):
|
| 56 |
r"""Class-conditional Diffusion Transformer pipeline with custom pipeline loading support."""
|
|
@@ -104,7 +64,6 @@ class DiTPipeline(DiffusionPipeline):
|
|
| 104 |
return self._id2label
|
| 105 |
|
| 106 |
def get_label_ids(self, label: Union[str, List[str]]) -> List[int]:
|
| 107 |
-
r"""Map English ImageNet labels to class ids."""
|
| 108 |
labels = [label] if isinstance(label, str) else label
|
| 109 |
if not self.labels:
|
| 110 |
raise ValueError("No id2label mapping is available in this checkpoint.")
|
|
@@ -133,8 +92,8 @@ class DiTPipeline(DiffusionPipeline):
|
|
| 133 |
@staticmethod
|
| 134 |
def prepare_extra_step_kwargs(
|
| 135 |
scheduler: KarrasDiffusionSchedulers,
|
| 136 |
-
generator: Optional[Union[torch.Generator, List[torch.Generator]]],
|
| 137 |
-
eta: float,
|
| 138 |
) -> Dict[str, Any]:
|
| 139 |
kwargs: Dict[str, Any] = {}
|
| 140 |
step_params = set(inspect.signature(scheduler.step).parameters.keys())
|
|
@@ -153,7 +112,6 @@ class DiTPipeline(DiffusionPipeline):
|
|
| 153 |
return timestep.expand(batch_size)
|
| 154 |
|
| 155 |
@torch.no_grad()
|
| 156 |
-
@replace_example_docstring(EXAMPLE_DOC_STRING)
|
| 157 |
def __call__(
|
| 158 |
self,
|
| 159 |
class_labels: Union[int, str, List[Union[int, str]], torch.Tensor] = 207,
|
|
@@ -167,13 +125,6 @@ class DiTPipeline(DiffusionPipeline):
|
|
| 167 |
output_type: str = "pil",
|
| 168 |
return_dict: bool = True,
|
| 169 |
) -> Union[ImagePipelineOutput, Tuple]:
|
| 170 |
-
r"""
|
| 171 |
-
Generate class-conditional samples from a DiT checkpoint.
|
| 172 |
-
|
| 173 |
-
Examples:
|
| 174 |
-
<!-- this section is replaced by replace_example_docstring -->
|
| 175 |
-
"""
|
| 176 |
-
# Stage 1: check inputs
|
| 177 |
class_labels_list = self._normalize_class_labels(class_labels)
|
| 178 |
batch_size = len(class_labels_list)
|
| 179 |
native_size = int(getattr(self.transformer.config, "sample_size", 32)) * self.vae_scale_factor
|
|
@@ -187,7 +138,6 @@ class DiTPipeline(DiffusionPipeline):
|
|
| 187 |
if output_type not in {"pil", "np", "pt", "latent"}:
|
| 188 |
raise ValueError(f"Unsupported `output_type`: {output_type}")
|
| 189 |
|
| 190 |
-
# Stage 2: define call parameters
|
| 191 |
device = self._execution_device
|
| 192 |
do_classifier_free_guidance = float(guidance_scale) > 1.0
|
| 193 |
latent_h = height // self.vae_scale_factor
|
|
@@ -195,17 +145,14 @@ class DiTPipeline(DiffusionPipeline):
|
|
| 195 |
latent_channels = int(getattr(self.transformer.config, "in_channels", 4))
|
| 196 |
extra_step_kwargs = self.prepare_extra_step_kwargs(self.scheduler, generator=generator, eta=float(eta))
|
| 197 |
|
| 198 |
-
# Stage 3: encode class conditioning
|
| 199 |
class_labels_tensor = torch.tensor(class_labels_list, device=device, dtype=torch.long)
|
| 200 |
if do_classifier_free_guidance:
|
| 201 |
null_class = int(self.config.null_class_id)
|
| 202 |
uncond = torch.full((batch_size,), null_class, device=device, dtype=torch.long)
|
| 203 |
class_labels_tensor = torch.cat([class_labels_tensor, uncond], dim=0)
|
| 204 |
|
| 205 |
-
# Stage 4: prepare timesteps
|
| 206 |
self.scheduler.set_timesteps(num_inference_steps, device=device)
|
| 207 |
|
| 208 |
-
# Stage 5: prepare latent variables
|
| 209 |
if latents is None:
|
| 210 |
latents = randn_tensor(
|
| 211 |
(batch_size, latent_channels, latent_h, latent_w),
|
|
@@ -219,9 +166,6 @@ class DiTPipeline(DiffusionPipeline):
|
|
| 219 |
if tuple(latents.shape) != expected:
|
| 220 |
raise ValueError(f"Invalid `latents` shape: {tuple(latents.shape)}. Expected {expected}.")
|
| 221 |
|
| 222 |
-
# Stage 6: prepare extra step kwargs (already done above for clarity)
|
| 223 |
-
|
| 224 |
-
# Stage 7: run denoising loop
|
| 225 |
for timestep in self.progress_bar(self.scheduler.timesteps):
|
| 226 |
latent_model_input = latents
|
| 227 |
if do_classifier_free_guidance:
|
|
|
|
| 1 |
+
"""Hub custom pipeline: DiTPipeline."""
|
| 2 |
+
|
| 3 |
+
from __future__ import annotations
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| 4 |
|
| 5 |
import inspect
|
| 6 |
from typing import Any, Dict, List, Optional, Tuple, Union
|
| 7 |
|
| 8 |
import torch
|
|
|
|
| 9 |
from diffusers.image_processor import VaeImageProcessor
|
| 10 |
from diffusers.pipelines.pipeline_utils import DiffusionPipeline, ImagePipelineOutput
|
| 11 |
from diffusers.schedulers import KarrasDiffusionSchedulers
|
|
|
|
| 12 |
from diffusers.utils.torch_utils import randn_tensor
|
| 13 |
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| 14 |
|
| 15 |
class DiTPipeline(DiffusionPipeline):
|
| 16 |
r"""Class-conditional Diffusion Transformer pipeline with custom pipeline loading support."""
|
|
|
|
| 64 |
return self._id2label
|
| 65 |
|
| 66 |
def get_label_ids(self, label: Union[str, List[str]]) -> List[int]:
|
|
|
|
| 67 |
labels = [label] if isinstance(label, str) else label
|
| 68 |
if not self.labels:
|
| 69 |
raise ValueError("No id2label mapping is available in this checkpoint.")
|
|
|
|
| 92 |
@staticmethod
|
| 93 |
def prepare_extra_step_kwargs(
|
| 94 |
scheduler: KarrasDiffusionSchedulers,
|
| 95 |
+
generator: Optional[Union[torch.Generator, List[torch.Generator]]] = None,
|
| 96 |
+
eta: float = 0.0,
|
| 97 |
) -> Dict[str, Any]:
|
| 98 |
kwargs: Dict[str, Any] = {}
|
| 99 |
step_params = set(inspect.signature(scheduler.step).parameters.keys())
|
|
|
|
| 112 |
return timestep.expand(batch_size)
|
| 113 |
|
| 114 |
@torch.no_grad()
|
|
|
|
| 115 |
def __call__(
|
| 116 |
self,
|
| 117 |
class_labels: Union[int, str, List[Union[int, str]], torch.Tensor] = 207,
|
|
|
|
| 125 |
output_type: str = "pil",
|
| 126 |
return_dict: bool = True,
|
| 127 |
) -> Union[ImagePipelineOutput, Tuple]:
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| 128 |
class_labels_list = self._normalize_class_labels(class_labels)
|
| 129 |
batch_size = len(class_labels_list)
|
| 130 |
native_size = int(getattr(self.transformer.config, "sample_size", 32)) * self.vae_scale_factor
|
|
|
|
| 138 |
if output_type not in {"pil", "np", "pt", "latent"}:
|
| 139 |
raise ValueError(f"Unsupported `output_type`: {output_type}")
|
| 140 |
|
|
|
|
| 141 |
device = self._execution_device
|
| 142 |
do_classifier_free_guidance = float(guidance_scale) > 1.0
|
| 143 |
latent_h = height // self.vae_scale_factor
|
|
|
|
| 145 |
latent_channels = int(getattr(self.transformer.config, "in_channels", 4))
|
| 146 |
extra_step_kwargs = self.prepare_extra_step_kwargs(self.scheduler, generator=generator, eta=float(eta))
|
| 147 |
|
|
|
|
| 148 |
class_labels_tensor = torch.tensor(class_labels_list, device=device, dtype=torch.long)
|
| 149 |
if do_classifier_free_guidance:
|
| 150 |
null_class = int(self.config.null_class_id)
|
| 151 |
uncond = torch.full((batch_size,), null_class, device=device, dtype=torch.long)
|
| 152 |
class_labels_tensor = torch.cat([class_labels_tensor, uncond], dim=0)
|
| 153 |
|
|
|
|
| 154 |
self.scheduler.set_timesteps(num_inference_steps, device=device)
|
| 155 |
|
|
|
|
| 156 |
if latents is None:
|
| 157 |
latents = randn_tensor(
|
| 158 |
(batch_size, latent_channels, latent_h, latent_w),
|
|
|
|
| 166 |
if tuple(latents.shape) != expected:
|
| 167 |
raise ValueError(f"Invalid `latents` shape: {tuple(latents.shape)}. Expected {expected}.")
|
| 168 |
|
|
|
|
|
|
|
|
|
|
| 169 |
for timestep in self.progress_bar(self.scheduler.timesteps):
|
| 170 |
latent_model_input = latents
|
| 171 |
if do_classifier_free_guidance:
|
DiT-XL-2-512/pipeline.py
CHANGED
|
@@ -1,56 +1,16 @@
|
|
| 1 |
-
|
| 2 |
-
|
| 3 |
-
|
| 4 |
-
# you may not use this file except in compliance with the License.
|
| 5 |
-
# You may obtain a copy of the License at
|
| 6 |
-
#
|
| 7 |
-
# http://www.apache.org/licenses/LICENSE-2.0
|
| 8 |
-
#
|
| 9 |
-
# Unless required by applicable law or agreed to in writing, software
|
| 10 |
-
# distributed under the License is distributed on an "AS IS" BASIS,
|
| 11 |
-
# WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied.
|
| 12 |
-
# See the License for the specific language governing permissions and
|
| 13 |
-
# limitations under the License.
|
| 14 |
|
| 15 |
import inspect
|
| 16 |
from typing import Any, Dict, List, Optional, Tuple, Union
|
| 17 |
|
| 18 |
import torch
|
| 19 |
-
|
| 20 |
from diffusers.image_processor import VaeImageProcessor
|
| 21 |
from diffusers.pipelines.pipeline_utils import DiffusionPipeline, ImagePipelineOutput
|
| 22 |
from diffusers.schedulers import KarrasDiffusionSchedulers
|
| 23 |
-
from diffusers.utils import replace_example_docstring
|
| 24 |
from diffusers.utils.torch_utils import randn_tensor
|
| 25 |
|
| 26 |
-
EXAMPLE_DOC_STRING = """
|
| 27 |
-
Examples:
|
| 28 |
-
```py
|
| 29 |
-
>>> from pathlib import Path
|
| 30 |
-
>>> import torch
|
| 31 |
-
>>> from diffusers import DiffusionPipeline
|
| 32 |
-
|
| 33 |
-
>>> model_dir = Path("path/to/BiliSakura/DiT-diffusers/DiT-XL-2-512")
|
| 34 |
-
>>> pipe = DiffusionPipeline.from_pretrained(
|
| 35 |
-
... str(model_dir),
|
| 36 |
-
... local_files_only=True,
|
| 37 |
-
... custom_pipeline=str(model_dir / "pipeline.py"),
|
| 38 |
-
... trust_remote_code=True,
|
| 39 |
-
... torch_dtype=torch.float16,
|
| 40 |
-
... )
|
| 41 |
-
>>> pipe = pipe.to("cuda")
|
| 42 |
-
|
| 43 |
-
>>> class_id = pipe.get_label_ids("golden retriever")[0]
|
| 44 |
-
>>> image = pipe(
|
| 45 |
-
... class_labels=class_id,
|
| 46 |
-
... height=512,
|
| 47 |
-
... width=512,
|
| 48 |
-
... num_inference_steps=250,
|
| 49 |
-
... guidance_scale=4.0,
|
| 50 |
-
... ).images[0]
|
| 51 |
-
```
|
| 52 |
-
"""
|
| 53 |
-
|
| 54 |
|
| 55 |
class DiTPipeline(DiffusionPipeline):
|
| 56 |
r"""Class-conditional Diffusion Transformer pipeline with custom pipeline loading support."""
|
|
@@ -104,7 +64,6 @@ class DiTPipeline(DiffusionPipeline):
|
|
| 104 |
return self._id2label
|
| 105 |
|
| 106 |
def get_label_ids(self, label: Union[str, List[str]]) -> List[int]:
|
| 107 |
-
r"""Map English ImageNet labels to class ids."""
|
| 108 |
labels = [label] if isinstance(label, str) else label
|
| 109 |
if not self.labels:
|
| 110 |
raise ValueError("No id2label mapping is available in this checkpoint.")
|
|
@@ -133,8 +92,8 @@ class DiTPipeline(DiffusionPipeline):
|
|
| 133 |
@staticmethod
|
| 134 |
def prepare_extra_step_kwargs(
|
| 135 |
scheduler: KarrasDiffusionSchedulers,
|
| 136 |
-
generator: Optional[Union[torch.Generator, List[torch.Generator]]],
|
| 137 |
-
eta: float,
|
| 138 |
) -> Dict[str, Any]:
|
| 139 |
kwargs: Dict[str, Any] = {}
|
| 140 |
step_params = set(inspect.signature(scheduler.step).parameters.keys())
|
|
@@ -153,7 +112,6 @@ class DiTPipeline(DiffusionPipeline):
|
|
| 153 |
return timestep.expand(batch_size)
|
| 154 |
|
| 155 |
@torch.no_grad()
|
| 156 |
-
@replace_example_docstring(EXAMPLE_DOC_STRING)
|
| 157 |
def __call__(
|
| 158 |
self,
|
| 159 |
class_labels: Union[int, str, List[Union[int, str]], torch.Tensor] = 207,
|
|
@@ -167,13 +125,6 @@ class DiTPipeline(DiffusionPipeline):
|
|
| 167 |
output_type: str = "pil",
|
| 168 |
return_dict: bool = True,
|
| 169 |
) -> Union[ImagePipelineOutput, Tuple]:
|
| 170 |
-
r"""
|
| 171 |
-
Generate class-conditional samples from a DiT checkpoint.
|
| 172 |
-
|
| 173 |
-
Examples:
|
| 174 |
-
<!-- this section is replaced by replace_example_docstring -->
|
| 175 |
-
"""
|
| 176 |
-
# Stage 1: check inputs
|
| 177 |
class_labels_list = self._normalize_class_labels(class_labels)
|
| 178 |
batch_size = len(class_labels_list)
|
| 179 |
native_size = int(getattr(self.transformer.config, "sample_size", 32)) * self.vae_scale_factor
|
|
@@ -187,7 +138,6 @@ class DiTPipeline(DiffusionPipeline):
|
|
| 187 |
if output_type not in {"pil", "np", "pt", "latent"}:
|
| 188 |
raise ValueError(f"Unsupported `output_type`: {output_type}")
|
| 189 |
|
| 190 |
-
# Stage 2: define call parameters
|
| 191 |
device = self._execution_device
|
| 192 |
do_classifier_free_guidance = float(guidance_scale) > 1.0
|
| 193 |
latent_h = height // self.vae_scale_factor
|
|
@@ -195,17 +145,14 @@ class DiTPipeline(DiffusionPipeline):
|
|
| 195 |
latent_channels = int(getattr(self.transformer.config, "in_channels", 4))
|
| 196 |
extra_step_kwargs = self.prepare_extra_step_kwargs(self.scheduler, generator=generator, eta=float(eta))
|
| 197 |
|
| 198 |
-
# Stage 3: encode class conditioning
|
| 199 |
class_labels_tensor = torch.tensor(class_labels_list, device=device, dtype=torch.long)
|
| 200 |
if do_classifier_free_guidance:
|
| 201 |
null_class = int(self.config.null_class_id)
|
| 202 |
uncond = torch.full((batch_size,), null_class, device=device, dtype=torch.long)
|
| 203 |
class_labels_tensor = torch.cat([class_labels_tensor, uncond], dim=0)
|
| 204 |
|
| 205 |
-
# Stage 4: prepare timesteps
|
| 206 |
self.scheduler.set_timesteps(num_inference_steps, device=device)
|
| 207 |
|
| 208 |
-
# Stage 5: prepare latent variables
|
| 209 |
if latents is None:
|
| 210 |
latents = randn_tensor(
|
| 211 |
(batch_size, latent_channels, latent_h, latent_w),
|
|
@@ -219,9 +166,6 @@ class DiTPipeline(DiffusionPipeline):
|
|
| 219 |
if tuple(latents.shape) != expected:
|
| 220 |
raise ValueError(f"Invalid `latents` shape: {tuple(latents.shape)}. Expected {expected}.")
|
| 221 |
|
| 222 |
-
# Stage 6: prepare extra step kwargs (already done above for clarity)
|
| 223 |
-
|
| 224 |
-
# Stage 7: run denoising loop
|
| 225 |
for timestep in self.progress_bar(self.scheduler.timesteps):
|
| 226 |
latent_model_input = latents
|
| 227 |
if do_classifier_free_guidance:
|
|
|
|
| 1 |
+
"""Hub custom pipeline: DiTPipeline."""
|
| 2 |
+
|
| 3 |
+
from __future__ import annotations
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| 4 |
|
| 5 |
import inspect
|
| 6 |
from typing import Any, Dict, List, Optional, Tuple, Union
|
| 7 |
|
| 8 |
import torch
|
|
|
|
| 9 |
from diffusers.image_processor import VaeImageProcessor
|
| 10 |
from diffusers.pipelines.pipeline_utils import DiffusionPipeline, ImagePipelineOutput
|
| 11 |
from diffusers.schedulers import KarrasDiffusionSchedulers
|
|
|
|
| 12 |
from diffusers.utils.torch_utils import randn_tensor
|
| 13 |
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| 14 |
|
| 15 |
class DiTPipeline(DiffusionPipeline):
|
| 16 |
r"""Class-conditional Diffusion Transformer pipeline with custom pipeline loading support."""
|
|
|
|
| 64 |
return self._id2label
|
| 65 |
|
| 66 |
def get_label_ids(self, label: Union[str, List[str]]) -> List[int]:
|
|
|
|
| 67 |
labels = [label] if isinstance(label, str) else label
|
| 68 |
if not self.labels:
|
| 69 |
raise ValueError("No id2label mapping is available in this checkpoint.")
|
|
|
|
| 92 |
@staticmethod
|
| 93 |
def prepare_extra_step_kwargs(
|
| 94 |
scheduler: KarrasDiffusionSchedulers,
|
| 95 |
+
generator: Optional[Union[torch.Generator, List[torch.Generator]]] = None,
|
| 96 |
+
eta: float = 0.0,
|
| 97 |
) -> Dict[str, Any]:
|
| 98 |
kwargs: Dict[str, Any] = {}
|
| 99 |
step_params = set(inspect.signature(scheduler.step).parameters.keys())
|
|
|
|
| 112 |
return timestep.expand(batch_size)
|
| 113 |
|
| 114 |
@torch.no_grad()
|
|
|
|
| 115 |
def __call__(
|
| 116 |
self,
|
| 117 |
class_labels: Union[int, str, List[Union[int, str]], torch.Tensor] = 207,
|
|
|
|
| 125 |
output_type: str = "pil",
|
| 126 |
return_dict: bool = True,
|
| 127 |
) -> Union[ImagePipelineOutput, Tuple]:
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| 128 |
class_labels_list = self._normalize_class_labels(class_labels)
|
| 129 |
batch_size = len(class_labels_list)
|
| 130 |
native_size = int(getattr(self.transformer.config, "sample_size", 32)) * self.vae_scale_factor
|
|
|
|
| 138 |
if output_type not in {"pil", "np", "pt", "latent"}:
|
| 139 |
raise ValueError(f"Unsupported `output_type`: {output_type}")
|
| 140 |
|
|
|
|
| 141 |
device = self._execution_device
|
| 142 |
do_classifier_free_guidance = float(guidance_scale) > 1.0
|
| 143 |
latent_h = height // self.vae_scale_factor
|
|
|
|
| 145 |
latent_channels = int(getattr(self.transformer.config, "in_channels", 4))
|
| 146 |
extra_step_kwargs = self.prepare_extra_step_kwargs(self.scheduler, generator=generator, eta=float(eta))
|
| 147 |
|
|
|
|
| 148 |
class_labels_tensor = torch.tensor(class_labels_list, device=device, dtype=torch.long)
|
| 149 |
if do_classifier_free_guidance:
|
| 150 |
null_class = int(self.config.null_class_id)
|
| 151 |
uncond = torch.full((batch_size,), null_class, device=device, dtype=torch.long)
|
| 152 |
class_labels_tensor = torch.cat([class_labels_tensor, uncond], dim=0)
|
| 153 |
|
|
|
|
| 154 |
self.scheduler.set_timesteps(num_inference_steps, device=device)
|
| 155 |
|
|
|
|
| 156 |
if latents is None:
|
| 157 |
latents = randn_tensor(
|
| 158 |
(batch_size, latent_channels, latent_h, latent_w),
|
|
|
|
| 166 |
if tuple(latents.shape) != expected:
|
| 167 |
raise ValueError(f"Invalid `latents` shape: {tuple(latents.shape)}. Expected {expected}.")
|
| 168 |
|
|
|
|
|
|
|
|
|
|
| 169 |
for timestep in self.progress_bar(self.scheduler.timesteps):
|
| 170 |
latent_model_input = latents
|
| 171 |
if do_classifier_free_guidance:
|