Spaces:
Running
on
Zero
Running
on
Zero
File size: 10,115 Bytes
ff49a48 |
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 |
from dataclasses import dataclass, field
from typing import Any, Dict, Optional
import torch
import torch.nn.functional as F
from torch import nn
from ...utils import BaseModule
from .basic_transformer_block import BasicTransformerBlock
class Transformer1D(BaseModule):
"""
A 1D Transformer model for sequence data.
Parameters:
num_attention_heads (`int`, *optional*, defaults to 16): The number of heads to use for multi-head attention.
attention_head_dim (`int`, *optional*, defaults to 88): The number of channels in each head.
in_channels (`int`, *optional*):
The number of channels in the input and output (specify if the input is **continuous**).
num_layers (`int`, *optional*, defaults to 1): The number of layers of Transformer blocks to use.
dropout (`float`, *optional*, defaults to 0.0): The dropout probability to use.
cross_attention_dim (`int`, *optional*): The number of `encoder_hidden_states` dimensions to use.
activation_fn (`str`, *optional*, defaults to `"geglu"`): Activation function to use in feed-forward.
num_embeds_ada_norm ( `int`, *optional*):
The number of diffusion steps used during training. Pass if at least one of the norm_layers is
`AdaLayerNorm`. This is fixed during training since it is used to learn a number of embeddings that are
added to the hidden states.
During inference, you can denoise for up to but not more steps than `num_embeds_ada_norm`.
attention_bias (`bool`, *optional*):
Configure if the `TransformerBlocks` attention should contain a bias parameter.
"""
@dataclass
class Config(BaseModule.Config):
num_attention_heads: int = 16
attention_head_dim: int = 88
in_channels: Optional[int] = None
out_channels: Optional[int] = None
num_layers: int = 1
dropout: float = 0.0
norm_num_groups: int = 32
cross_attention_dim: Optional[int] = None
attention_bias: bool = False
activation_fn: str = "geglu"
only_cross_attention: bool = False
double_self_attention: bool = False
upcast_attention: bool = False
norm_type: str = "layer_norm"
norm_elementwise_affine: bool = True
gradient_checkpointing: bool = False
cfg: Config
def configure(self) -> None:
self.num_attention_heads = self.cfg.num_attention_heads
self.attention_head_dim = self.cfg.attention_head_dim
inner_dim = self.num_attention_heads * self.attention_head_dim
linear_cls = nn.Linear
# 2. Define input layers
self.in_channels = self.cfg.in_channels
self.norm = torch.nn.GroupNorm(
num_groups=self.cfg.norm_num_groups,
num_channels=self.cfg.in_channels,
eps=1e-6,
affine=True,
)
self.proj_in = linear_cls(self.cfg.in_channels, inner_dim)
# 3. Define transformers blocks
self.transformer_blocks = nn.ModuleList(
[
BasicTransformerBlock(
inner_dim,
self.num_attention_heads,
self.attention_head_dim,
dropout=self.cfg.dropout,
cross_attention_dim=self.cfg.cross_attention_dim,
activation_fn=self.cfg.activation_fn,
attention_bias=self.cfg.attention_bias,
only_cross_attention=self.cfg.only_cross_attention,
double_self_attention=self.cfg.double_self_attention,
upcast_attention=self.cfg.upcast_attention,
norm_type=self.cfg.norm_type,
norm_elementwise_affine=self.cfg.norm_elementwise_affine,
)
for d in range(self.cfg.num_layers)
]
)
# 4. Define output layers
self.out_channels = (
self.cfg.in_channels
if self.cfg.out_channels is None
else self.cfg.out_channels
)
self.proj_out = linear_cls(inner_dim, self.cfg.in_channels)
self.gradient_checkpointing = self.cfg.gradient_checkpointing
def forward(
self,
hidden_states: torch.Tensor,
encoder_hidden_states: Optional[torch.Tensor] = None,
attention_mask: Optional[torch.Tensor] = None,
encoder_attention_mask: Optional[torch.Tensor] = None,
):
"""
The [`Transformer1DModel`] forward method.
Args:
hidden_states (`torch.LongTensor` of shape `(batch size, num latent pixels)` if discrete, `torch.FloatTensor` of shape `(batch size, channel, height, width)` if continuous):
Input `hidden_states`.
encoder_hidden_states ( `torch.FloatTensor` of shape `(batch size, sequence len, embed dims)`, *optional*):
Conditional embeddings for cross attention layer. If not given, cross-attention defaults to
self-attention.
timestep ( `torch.LongTensor`, *optional*):
Used to indicate denoising step. Optional timestep to be applied as an embedding in `AdaLayerNorm`.
class_labels ( `torch.LongTensor` of shape `(batch size, num classes)`, *optional*):
Used to indicate class labels conditioning. Optional class labels to be applied as an embedding in
`AdaLayerZeroNorm`.
cross_attention_kwargs ( `Dict[str, Any]`, *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).
attention_mask ( `torch.Tensor`, *optional*):
An attention mask of shape `(batch, key_tokens)` is applied to `encoder_hidden_states`. If `1` the mask
is kept, otherwise if `0` it is discarded. Mask will be converted into a bias, which adds large
negative values to the attention scores corresponding to "discard" tokens.
encoder_attention_mask ( `torch.Tensor`, *optional*):
Cross-attention mask applied to `encoder_hidden_states`. Two formats supported:
* Mask `(batch, sequence_length)` True = keep, False = discard.
* Bias `(batch, 1, sequence_length)` 0 = keep, -10000 = discard.
If `ndim == 2`: will be interpreted as a mask, then converted into a bias consistent with the format
above. This bias will be added to the cross-attention scores.
return_dict (`bool`, *optional*, defaults to `True`):
Whether or not to return a [`~models.unet_2d_condition.UNet2DConditionOutput`] 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.
"""
# ensure attention_mask is a bias, and give it a singleton query_tokens dimension.
# we may have done this conversion already, e.g. if we came here via UNet2DConditionModel#forward.
# we can tell by counting dims; if ndim == 2: it's a mask rather than a bias.
# expects mask of shape:
# [batch, key_tokens]
# adds singleton query_tokens dimension:
# [batch, 1, key_tokens]
# this helps to broadcast it as a bias over attention scores, which will be in one of the following shapes:
# [batch, heads, query_tokens, key_tokens] (e.g. torch sdp attn)
# [batch * heads, query_tokens, key_tokens] (e.g. xformers or classic attn)
if attention_mask is not None and attention_mask.ndim == 2:
# assume that mask is expressed as:
# (1 = keep, 0 = discard)
# convert mask into a bias that can be added to attention scores:
# (keep = +0, discard = -10000.0)
attention_mask = (1 - attention_mask.to(hidden_states.dtype)) * -10000.0
attention_mask = attention_mask.unsqueeze(1)
# convert encoder_attention_mask to a bias the same way we do for attention_mask
if encoder_attention_mask is not None and encoder_attention_mask.ndim == 2:
encoder_attention_mask = (
1 - encoder_attention_mask.to(hidden_states.dtype)
) * -10000.0
encoder_attention_mask = encoder_attention_mask.unsqueeze(1)
# 1. Input
batch, _, seq_len = hidden_states.shape
residual = hidden_states
hidden_states = self.norm(hidden_states)
inner_dim = hidden_states.shape[1]
hidden_states = hidden_states.permute(0, 2, 1).reshape(
batch, seq_len, inner_dim
)
hidden_states = self.proj_in(hidden_states)
# 2. Blocks
for block in self.transformer_blocks:
if self.training and self.gradient_checkpointing:
hidden_states = torch.utils.checkpoint.checkpoint(
block,
hidden_states,
attention_mask,
encoder_hidden_states,
encoder_attention_mask,
use_reentrant=False,
)
else:
hidden_states = block(
hidden_states,
attention_mask=attention_mask,
encoder_hidden_states=encoder_hidden_states,
encoder_attention_mask=encoder_attention_mask,
)
# 3. Output
hidden_states = self.proj_out(hidden_states)
hidden_states = (
hidden_states.reshape(batch, seq_len, inner_dim)
.permute(0, 2, 1)
.contiguous()
)
output = hidden_states + residual
return output
|