{{ message }}
Repository navigation
Expand file tree
/
Copy pathcontrolnet_flux.py
More file actions
386 lines (322 loc) · 16.4 KB
/
Copy pathcontrolnet_flux.py
File metadata and controls
386 lines (322 loc) · 16.4 KB
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
84
85
86
87
88
89
90
91
92
93
94
95
96
97
98
99
100
101
102
103
104
105
106
107
108
109
110
111
112
113
114
115
116
117
118
119
120
121
122
123
124
125
126
127
128
129
130
131
132
133
134
135
136
137
138
139
140
141
142
143
144
145
146
147
148
149
150
151
152
153
154
155
156
157
158
159
160
161
162
163
164
165
166
167
168
169
170
171
172
173
174
175
176
177
178
179
180
181
182
183
184
185
186
187
188
189
190
191
192
193
194
195
196
197
198
199
200
201
202
203
204
205
206
207
208
209
210
211
212
213
214
215
216
217
218
219
220
221
222
223
224
225
226
227
228
229
230
231
232
233
234
235
236
237
238
239
240
241
242
243
244
245
246
247
248
249
250
251
252
253
254
255
256
257
258
259
260
261
262
263
264
265
266
267
268
269
270
271
272
273
274
275
276
277
278
279
280
281
282
283
284
285
286
287
288
289
290
291
292
293
294
295
296
297
298
299
300
301
302
303
304
305
306
307
308
309
310
311
312
313
314
315
316
317
318
319
320
321
322
323
324
325
326
327
328
329
330
331
332
333
334
335
336
337
338
339
340
341
342
343
344
345
346
347
348
349
350
351
352
353
354
355
356
357
358
359
360
361
362
363
364
365
366
367
368
369
370
371
372
373
374
375
376
377
378
379
380
381
382
383
384
385
386
# Copyright 2024 Black Forest Labs, The HuggingFace Team and The InstantX Team. All rights reserved.
#
# 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.
from dataclasses import dataclass
from typing import Any, Dict, List, Optional, Tuple, Union
import torch
import torch.nn as nn
from ..configuration_utils import ConfigMixin, register_to_config
from ..loaders import PeftAdapterMixin
from ..models.attention_processor import AttentionProcessor
from ..models.modeling_utils import ModelMixin
from ..utils import USE_PEFT_BACKEND, is_torch_version, logging, scale_lora_layers, unscale_lora_layers
from .controlnet import BaseOutput, zero_module
from .embeddings import CombinedTimestepGuidanceTextProjEmbeddings, CombinedTimestepTextProjEmbeddings, FluxPosEmbed
from .modeling_outputs import Transformer2DModelOutput
from .transformers.transformer_flux import FluxSingleTransformerBlock, FluxTransformerBlock
logger = logging.get_logger(__name__) # pylint: disable=invalid-name
@dataclass
class FluxControlNetOutput(BaseOutput):
controlnet_block_samples: Tuple[torch.Tensor]
controlnet_single_block_samples: Tuple[torch.Tensor]
class FluxControlNetModel(ModelMixin, ConfigMixin, PeftAdapterMixin):
_supports_gradient_checkpointing = True
@register_to_config
def __init__(
self,
patch_size: int = 1,
in_channels: int = 64,
num_layers: int = 19,
num_single_layers: int = 38,
attention_head_dim: int = 128,
num_attention_heads: int = 24,
joint_attention_dim: int = 4096,
pooled_projection_dim: int = 768,
guidance_embeds: bool = False,
axes_dims_rope: List[int] = [16, 56, 56],
):
super().__init__()
self.out_channels = in_channels
self.inner_dim = num_attention_heads * attention_head_dim
self.pos_embed = FluxPosEmbed(theta=10000, axes_dim=axes_dims_rope)
text_time_guidance_cls = (
CombinedTimestepGuidanceTextProjEmbeddings if guidance_embeds else CombinedTimestepTextProjEmbeddings
)
self.time_text_embed = text_time_guidance_cls(
embedding_dim=self.inner_dim, pooled_projection_dim=pooled_projection_dim
)
self.context_embedder = nn.Linear(joint_attention_dim, self.inner_dim)
self.x_embedder = torch.nn.Linear(in_channels, self.inner_dim)
self.transformer_blocks = nn.ModuleList(
[
FluxTransformerBlock(
dim=self.inner_dim,
num_attention_heads=num_attention_heads,
attention_head_dim=attention_head_dim,
)
for i in range(num_layers)
]
)
self.single_transformer_blocks = nn.ModuleList(
[
FluxSingleTransformerBlock(
dim=self.inner_dim,
num_attention_heads=num_attention_heads,
attention_head_dim=attention_head_dim,
)
for i in range(num_single_layers)
]
)
# controlnet_blocks
self.controlnet_blocks = nn.ModuleList([])
for _ in range(len(self.transformer_blocks)):
self.controlnet_blocks.append(zero_module(nn.Linear(self.inner_dim, self.inner_dim)))
self.controlnet_single_blocks = nn.ModuleList([])
for _ in range(len(self.single_transformer_blocks)):
self.controlnet_single_blocks.append(zero_module(nn.Linear(self.inner_dim, self.inner_dim)))
self.controlnet_x_embedder = zero_module(torch.nn.Linear(in_channels, self.inner_dim))
self.gradient_checkpointing = False
@property
# Copied from diffusers.models.unets.unet_2d_condition.UNet2DConditionModel.attn_processors
def attn_processors(self):
r"""
Returns:
`dict` of attention processors: A dictionary containing all attention processors used in the model with
indexed by its weight name.
"""
# set recursively
processors = {}
def fn_recursive_add_processors(name: str, module: torch.nn.Module, processors: Dict[str, AttentionProcessor]):
if hasattr(module, "get_processor"):
processors[f"{name}.processor"] = module.get_processor()
for sub_name, child in module.named_children():
fn_recursive_add_processors(f"{name}.{sub_name}", child, processors)
return processors
for name, module in self.named_children():
fn_recursive_add_processors(name, module, processors)
return processors
# Copied from diffusers.models.unets.unet_2d_condition.UNet2DConditionModel.set_attn_processor
def set_attn_processor(self, processor):
r"""
Sets the attention processor to use to compute attention.
Parameters:
processor (`dict` of `AttentionProcessor` or only `AttentionProcessor`):
The instantiated processor class or a dictionary of processor classes that will be set as the processor
for **all** `Attention` layers.
If `processor` is a dict, the key needs to define the path to the corresponding cross attention
processor. This is strongly recommended when setting trainable attention processors.
"""
count = len(self.attn_processors.keys())
if isinstance(processor, dict) and len(processor) != count:
raise ValueError(
f"A dict of processors was passed, but the number of processors {len(processor)} does not match the"
f" number of attention layers: {count}. Please make sure to pass {count} processor classes."
)
def fn_recursive_attn_processor(name: str, module: torch.nn.Module, processor):
if hasattr(module, "set_processor"):
if not isinstance(processor, dict):
module.set_processor(processor)
else:
module.set_processor(processor.pop(f"{name}.processor"))
for sub_name, child in module.named_children():
fn_recursive_attn_processor(f"{name}.{sub_name}", child, processor)
for name, module in self.named_children():
fn_recursive_attn_processor(name, module, processor)
def _set_gradient_checkpointing(self, module, value=False):
if hasattr(module, "gradient_checkpointing"):
module.gradient_checkpointing = value
@classmethod
def from_transformer(
cls,
transformer,
num_layers=4,
num_single_layers=10,
attention_head_dim: int = 128,
num_attention_heads: int = 24,
load_weights_from_transformer=True,
):
config = transformer.config
config["num_layers"] = num_layers
config["num_single_layers"] = num_single_layers
config["attention_head_dim"] = attention_head_dim
config["num_attention_heads"] = num_attention_heads
controlnet = cls(**config)
if load_weights_from_transformer:
controlnet.pos_embed.load_state_dict(transformer.pos_embed.state_dict())
controlnet.time_text_embed.load_state_dict(transformer.time_text_embed.state_dict())
controlnet.context_embedder.load_state_dict(transformer.context_embedder.state_dict())
controlnet.x_embedder.load_state_dict(transformer.x_embedder.state_dict())
controlnet.transformer_blocks.load_state_dict(transformer.transformer_blocks.state_dict(), strict=False)
controlnet.single_transformer_blocks.load_state_dict(
transformer.single_transformer_blocks.state_dict(), strict=False
)
controlnet.controlnet_x_embedder = zero_module(controlnet.controlnet_x_embedder)
return controlnet
def forward(
self,
hidden_states: torch.Tensor,
controlnet_cond: torch.Tensor,
conditioning_scale: float = 1.0,
encoder_hidden_states: torch.Tensor = None,
pooled_projections: torch.Tensor = None,
timestep: torch.LongTensor = None,
img_ids: torch.Tensor = None,
txt_ids: torch.Tensor = None,
guidance: torch.Tensor = None,
joint_attention_kwargs: Optional[Dict[str, Any]] = None,
return_dict: bool = True,
) -> Union[torch.FloatTensor, Transformer2DModelOutput]:
"""
The [`FluxTransformer2DModel`] forward method.
Args:
hidden_states (`torch.FloatTensor` of shape `(batch size, channel, height, width)`):
Input `hidden_states`.
encoder_hidden_states (`torch.FloatTensor` of shape `(batch size, sequence_len, embed_dims)`):
Conditional embeddings (embeddings computed from the input conditions such as prompts) to use.
pooled_projections (`torch.FloatTensor` of shape `(batch_size, projection_dim)`): Embeddings projected
from the embeddings of input conditions.
timestep ( `torch.LongTensor`):
Used to indicate denoising step.
block_controlnet_hidden_states: (`list` of `torch.Tensor`):
A list of tensors that if specified are added to the residuals of transformer blocks.
joint_attention_kwargs (`dict`, *optional*):
A kwargs dictionary that if specified is passed along to the `AttentionProcessor` as defined under
`self.processor` in
[diffusers.models.attention_processor](https://github.com/huggingface/diffusers/blob/main/src/diffusers/models/attention_processor.py).
return_dict (`bool`, *optional*, defaults to `True`):
Whether or not to return a [`~models.transformer_2d.Transformer2DModelOutput`] instead of a plain
tuple.
Returns:
If `return_dict` is True, an [`~models.transformer_2d.Transformer2DModelOutput`] is returned, otherwise a
`tuple` where the first element is the sample tensor.
"""
if joint_attention_kwargs is not None:
joint_attention_kwargs = joint_attention_kwargs.copy()
lora_scale = joint_attention_kwargs.pop("scale", 1.0)
else:
lora_scale = 1.0
if USE_PEFT_BACKEND:
# weight the lora layers by setting `lora_scale` for each PEFT layer
scale_lora_layers(self, lora_scale)
else:
if joint_attention_kwargs is not None and joint_attention_kwargs.get("scale", None) is not None:
logger.warning(
"Passing `scale` via `joint_attention_kwargs` when not using the PEFT backend is ineffective."
)
hidden_states = self.x_embedder(hidden_states)
# add
hidden_states = hidden_states + self.controlnet_x_embedder(controlnet_cond)
timestep = timestep.to(hidden_states.dtype) * 1000
if guidance is not None:
guidance = guidance.to(hidden_states.dtype) * 1000
else:
guidance = None
temb = (
self.time_text_embed(timestep, pooled_projections)
if guidance is None
else self.time_text_embed(timestep, guidance, pooled_projections)
)
encoder_hidden_states = self.context_embedder(encoder_hidden_states)
if txt_ids.ndim == 3:
logger.warning(
"Passing `txt_ids` 3d torch.Tensor is deprecated."
"Please remove the batch dimension and pass it as a 2d torch Tensor"
)
txt_ids = txt_ids[0]
if img_ids.ndim == 3:
logger.warning(
"Passing `img_ids` 3d torch.Tensor is deprecated."
"Please remove the batch dimension and pass it as a 2d torch Tensor"
)
img_ids = img_ids[0]
ids = torch.cat((txt_ids, img_ids), dim=0)
image_rotary_emb = self.pos_embed(ids)
block_samples = ()
for index_block, block in enumerate(self.transformer_blocks):
if self.training and self.gradient_checkpointing:
def create_custom_forward(module, return_dict=None):
def custom_forward(*inputs):
if return_dict is not None:
return module(*inputs, return_dict=return_dict)
else:
return module(*inputs)
return custom_forward
ckpt_kwargs: Dict[str, Any] = {"use_reentrant": False} if is_torch_version(">=", "1.11.0") else {}
encoder_hidden_states, hidden_states = torch.utils.checkpoint.checkpoint(
create_custom_forward(block),
hidden_states,
encoder_hidden_states,
temb,
image_rotary_emb,
**ckpt_kwargs,
)
else:
encoder_hidden_states, hidden_states = block(
hidden_states=hidden_states,
encoder_hidden_states=encoder_hidden_states,
temb=temb,
image_rotary_emb=image_rotary_emb,
)
block_samples = block_samples + (hidden_states,)
hidden_states = torch.cat([encoder_hidden_states, hidden_states], dim=1)
single_block_samples = ()
for index_block, block in enumerate(self.single_transformer_blocks):
if self.training and self.gradient_checkpointing:
def create_custom_forward(module, return_dict=None):
def custom_forward(*inputs):
if return_dict is not None:
return module(*inputs, return_dict=return_dict)
else:
return module(*inputs)
return custom_forward
ckpt_kwargs: Dict[str, Any] = {"use_reentrant": False} if is_torch_version(">=", "1.11.0") else {}
hidden_states = torch.utils.checkpoint.checkpoint(
create_custom_forward(block),
hidden_states,
temb,
image_rotary_emb,
**ckpt_kwargs,
)
else:
hidden_states = block(
hidden_states=hidden_states,
temb=temb,
image_rotary_emb=image_rotary_emb,
)
single_block_samples = single_block_samples + (hidden_states[:, encoder_hidden_states.shape[1] :],)
# controlnet block
controlnet_block_samples = ()
for block_sample, controlnet_block in zip(block_samples, self.controlnet_blocks):
block_sample = controlnet_block(block_sample)
controlnet_block_samples = controlnet_block_samples + (block_sample,)
controlnet_single_block_samples = ()
for single_block_sample, controlnet_block in zip(single_block_samples, self.controlnet_single_blocks):
single_block_sample = controlnet_block(single_block_sample)
controlnet_single_block_samples = controlnet_single_block_samples + (single_block_sample,)
# scaling
controlnet_block_samples = [sample * conditioning_scale for sample in controlnet_block_samples]
controlnet_single_block_samples = [sample * conditioning_scale for sample in controlnet_single_block_samples]
#
controlnet_block_samples = None if len(controlnet_block_samples) == 0 else controlnet_block_samples
controlnet_single_block_samples = (
None if len(controlnet_single_block_samples) == 0 else controlnet_single_block_samples
)
if USE_PEFT_BACKEND:
# remove `lora_scale` from each PEFT layer
unscale_lora_layers(self, lora_scale)
if not return_dict:
return (controlnet_block_samples, controlnet_single_block_samples)
return FluxControlNetOutput(
controlnet_block_samples=controlnet_block_samples,
controlnet_single_block_samples=controlnet_single_block_samples,
)
You can’t perform that action at this time.