File size: 7,376 Bytes
4258b21 |
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 |
from dataclasses import dataclass
from typing import Any, Dict, List, Optional, Tuple, Union
import torch
import torch.nn as nn
import torch.utils.checkpoint
from collections import OrderedDict
from .outputs import BaseOutput
from model.diffusers.models.unet_2d_condition import UNet2DConditionModel
all_feature_dic={}
def clear_feature_dic():
global all_feature_dic
all_feature_dic={}
all_feature_dic["low"]=[]
all_feature_dic["mid"]=[]
all_feature_dic["high"]=[]
all_feature_dic["highest"]=[]
def get_feature_dic():
global all_feature_dic
return all_feature_dic
class UNet2DConditionOutput(BaseOutput):
"""
Args:
sample (`torch.FloatTensor` of shape `(batch_size, num_channels, height, width)`):
Hidden states conditioned on `encoder_hidden_states` input. Output of last layer of model.
"""
sample: torch.FloatTensor
class UNet2D(UNet2DConditionModel):
def forward(
self,
sample: torch.FloatTensor,
timestep: Union[torch.Tensor, float, int],
encoder_hidden_states: torch.Tensor,
return_dict: bool = True,
) -> Union[UNet2DConditionOutput, Tuple]:
"""r
Args:
sample (`torch.FloatTensor`): (batch, channel, height, width) noisy inputs tensor
timestep (`torch.FloatTensor` or `float` or `int): (batch) timesteps
encoder_hidden_states (`torch.FloatTensor`): (batch, channel, height, width) encoder hidden states
return_dict (`bool`, *optional*, defaults to `True`):
Whether or not to return a [`models.unet_2d_condition.UNet2DConditionOutput`] instead of a plain tuple.
Returns:
[`~models.unet_2d_condition.UNet2DConditionOutput`] or `tuple`:
[`~models.unet_2d_condition.UNet2DConditionOutput`] if `return_dict` is True, otherwise a `tuple`. When
returning a tuple, the first element is the sample tensor.
"""
# 0. center input if necessary
if self.config.center_input_sample:
sample = 2 * sample - 1.0
# 1. time
timesteps = timestep
if not torch.is_tensor(timesteps):
timesteps = torch.tensor([timesteps], dtype=torch.long, device=sample.device)
elif torch.is_tensor(timesteps) and len(timesteps.shape) == 0:
timesteps = timesteps.to(dtype=torch.float32)
timesteps = timesteps[None].to(device=sample.device)
# if timesteps[0]==0:
# flag_time=True
# else:
# flag_time=False
if timesteps[0]==0 or timesteps[0]==1:
flag_time=True
else:
flag_time=False
# broadcast to batch dimension in a way that's compatible with ONNX/Core ML
timesteps = timesteps.expand(sample.shape[0])
t_emb = self.time_proj(timesteps)
emb = self.time_embedding(t_emb)
# 2. pre-process
sample = self.conv_in(sample)
if flag_time:
reshape_h=sample.reshape(int(sample.size()[0]/2),int(sample.size()[1]*2),sample.size()[2],sample.size()[3])
if reshape_h.size()[2]==8:
all_feature_dic["low"].append(reshape_h)
elif reshape_h.size()[2]==16:
all_feature_dic["mid"].append(reshape_h)
elif reshape_h.size()[2]==32:
all_feature_dic["high"].append(reshape_h)
elif reshape_h.size()[2]==64:
all_feature_dic["highest"].append(reshape_h)
# 3. down
down_block_res_samples = (sample,)
for downsample_block in self.down_blocks:
if hasattr(downsample_block, "attentions") and downsample_block.attentions is not None:
sample, res_samples = downsample_block(
hidden_states=sample, temb=emb, encoder_hidden_states=encoder_hidden_states
)
else:
sample, res_samples = downsample_block(hidden_states=sample, temb=emb)
if flag_time:
for h in res_samples:
reshape_h=h.reshape(int(h.size()[0]/2),int(h.size()[1]*2),h.size()[2],h.size()[3])
# print(reshape_h.size()[2])
if reshape_h.size()[2]==8:
all_feature_dic["low"].append(reshape_h)
elif reshape_h.size()[2]==16:
all_feature_dic["mid"].append(reshape_h)
elif reshape_h.size()[2]==32:
all_feature_dic["high"].append(reshape_h)
elif reshape_h.size()[2]==64:
all_feature_dic["highest"].append(reshape_h)
down_block_res_samples += res_samples
# 4. mid
sample = self.mid_block(sample, emb, encoder_hidden_states=encoder_hidden_states)
if flag_time:
reshape_h=sample.reshape(int(sample.size()[0]/2),int(sample.size()[1]*2),sample.size()[2],sample.size()[3])
if reshape_h.size()[2]==8:
all_feature_dic["low"].append(reshape_h)
elif reshape_h.size()[2]==16:
all_feature_dic["mid"].append(reshape_h)
elif reshape_h.size()[2]==32:
all_feature_dic["high"].append(reshape_h)
elif reshape_h.size()[2]==64:
all_feature_dic["highest"].append(reshape_h)
# 5. up
for upsample_block in self.up_blocks:
res_samples = down_block_res_samples[-len(upsample_block.resnets) :]
down_block_res_samples = down_block_res_samples[: -len(upsample_block.resnets)]
if hasattr(upsample_block, "attentions") and upsample_block.attentions is not None:
sample, up_samples = upsample_block(
hidden_states=sample,
temb=emb,
res_hidden_states_tuple=res_samples,
encoder_hidden_states=encoder_hidden_states,
)
else:
sample, up_samples = upsample_block(hidden_states=sample, temb=emb, res_hidden_states_tuple=res_samples)
if flag_time:
for h in up_samples:
reshape_h=h.reshape(int(h.size()[0]/2),int(h.size()[1]*2),h.size()[2],h.size()[3])
# reshape_h=sample.reshape(int(sample.size()[0]/2),int(sample.size()[1]*2),sample.size()[2],sample.size()[3])
if reshape_h.size()[2]==8:
all_feature_dic["low"].append(reshape_h)
elif reshape_h.size()[2]==16:
all_feature_dic["mid"].append(reshape_h)
elif reshape_h.size()[2]==32:
all_feature_dic["high"].append(reshape_h)
elif reshape_h.size()[2]==64:
all_feature_dic["highest"].append(reshape_h)
# 6. post-process
# make sure hidden states is in float32
# when running in half-precision
sample = self.conv_norm_out(sample.float()).type(sample.dtype)
sample = self.conv_act(sample)
sample = self.conv_out(sample)
if not return_dict:
return (sample,)
return UNet2DConditionOutput(sample=sample) |