VictorMylle/Kohya-ss-sd-scripts
1
0
Fork 1
mirror of https://github.com/kohya-ss/sd-scripts.git synced 2026-04-08 22:35:09 +00:00
Code Issues Packages Projects Releases Wiki Activity

fix eps value, enable xformers, etc.

Browse Source
This commit is contained in:
Kohya S
2023-06-03 21:29:27 +09:00
parent 5db792b10b
commit c0a7df9ee1
3 changed files with 171 additions and 78 deletions
Download Patch File Download Diff File Expand all files Collapse all files

63
gen_img_diffusers.py
View File

@@ -317,7 +317,7 @@ def replace_unet_modules(unet: diffusers.models.unet_2d_condition.UNet2DConditio
if mem_eff_attn: if mem_eff_attn:
replace_unet_cross_attn_to_memory_efficient() replace_unet_cross_attn_to_memory_efficient()
elif xformers: elif xformers:
replace_unet_cross_attn_to_xformers() replace_unet_cross_attn_to_xformers(unet)
def replace_unet_cross_attn_to_memory_efficient(): def replace_unet_cross_attn_to_memory_efficient():
@@ -357,50 +357,55 @@ def replace_unet_cross_attn_to_memory_efficient():
out = self.to_out[1](out) out = self.to_out[1](out)
return out return out
diffusers.models.attention.CrossAttention.forward = forward_flash_attn # diffusers.models.attention.CrossAttention.forward = forward_flash_attn
# TODO U-Net側に移す
from library.original_unet import CrossAttention
CrossAttention.forward = forward_flash_attn
def replace_unet_cross_attn_to_xformers(): def replace_unet_cross_attn_to_xformers(unet:UNet2DConditionModel):
print("CrossAttention.forward has been replaced to enable xformers and NAI style Hypernetwork") print("CrossAttention.forward has been replaced to enable xformers and NAI style Hypernetwork")
try: try:
import xformers.ops import xformers.ops
except ImportError: except ImportError:
raise ImportError("No xformers / xformersがインストールされていないようです") raise ImportError("No xformers / xformersがインストールされていないようです")
unet.set_use_memory_efficient_attention_xformers(True)
def forward_xformers(self, x, context=None, mask=None): # def forward_xformers(self, x, context=None, mask=None):
h = self.heads # h = self.heads
q_in = self.to_q(x) # q_in = self.to_q(x)
context = default(context, x) # context = default(context, x)
context = context.to(x.dtype) # context = context.to(x.dtype)
if hasattr(self, "hypernetwork") and self.hypernetwork is not None: # if hasattr(self, "hypernetwork") and self.hypernetwork is not None:
context_k, context_v = self.hypernetwork.forward(x, context) # context_k, context_v = self.hypernetwork.forward(x, context)
context_k = context_k.to(x.dtype) # context_k = context_k.to(x.dtype)
context_v = context_v.to(x.dtype) # context_v = context_v.to(x.dtype)
else: # else:
context_k = context # context_k = context
context_v = context # context_v = context
k_in = self.to_k(context_k) # k_in = self.to_k(context_k)
v_in = self.to_v(context_v) # v_in = self.to_v(context_v)
q, k, v = map(lambda t: rearrange(t, "b n (h d) -> b n h d", h=h), (q_in, k_in, v_in)) # q, k, v = map(lambda t: rearrange(t, "b n (h d) -> b n h d", h=h), (q_in, k_in, v_in))
del q_in, k_in, v_in # del q_in, k_in, v_in
q = q.contiguous() # q = q.contiguous()
k = k.contiguous() # k = k.contiguous()
v = v.contiguous() # v = v.contiguous()
out = xformers.ops.memory_efficient_attention(q, k, v, attn_bias=None) # 最適なのを選んでくれる # out = xformers.ops.memory_efficient_attention(q, k, v, attn_bias=None) # 最適なのを選んでくれる
out = rearrange(out, "b n h d -> b n (h d)", h=h) # out = rearrange(out, "b n h d -> b n (h d)", h=h)
# diffusers 0.7.0~ # # diffusers 0.7.0~
out = self.to_out[0](out) # out = self.to_out[0](out)
out = self.to_out[1](out) # out = self.to_out[1](out)
return out # return out
diffusers.models.attention.CrossAttention.forward = forward_xformers # diffusers.models.attention.CrossAttention.forward = forward_xformers
def replace_vae_modules(vae: diffusers.models.AutoencoderKL, mem_eff_attn, xformers): def replace_vae_modules(vae: diffusers.models.AutoencoderKL, mem_eff_attn, xformers):

120
library/original_unet.py
View File

@@ -1,10 +1,10 @@
# Diffusers 0.10.2からStable Diffusionに必要な部分だけを持ってくる # Diffusers 0.10.2からStable Diffusionに必要な部分だけを持ってくる
# 条件分岐等で不要な部分は削除している
# コードの多くはDiffusersからコピーしている # コードの多くはDiffusersからコピーしている
# コードが冗長になる部分はコメント等を適宜削除する
# 制約として、モデルのstate_dictがDiffusers 0.10.2のものと同じ形式である必要がある # 制約として、モデルのstate_dictがDiffusers 0.10.2のものと同じ形式である必要がある
# Copy from Diffusers 0.10.2 for Stable Diffusion. Most of the code is copied from Diffusers. # Copy from Diffusers 0.10.2 for Stable Diffusion. Most of the code is copied from Diffusers.
# Remove redundant code by deleting comments, etc. as appropriate # Unnecessary parts are deleted by condition branching.
# As a constraint, the state_dict of the model must be in the same format as that of Diffusers 0.10.2 # As a constraint, the state_dict of the model must be in the same format as that of Diffusers 0.10.2
""" """
@@ -111,6 +111,7 @@ from typing import Dict, Optional, Tuple, Union
import torch import torch
from torch import nn from torch import nn
from torch.nn import functional as F from torch.nn import functional as F
from einops import rearrange
BLOCK_OUT_CHANNELS: Tuple[int] = (320, 640, 1280, 1280) BLOCK_OUT_CHANNELS: Tuple[int] = (320, 640, 1280, 1280)
TIMESTEP_INPUT_DIM = BLOCK_OUT_CHANNELS[0] TIMESTEP_INPUT_DIM = BLOCK_OUT_CHANNELS[0]
@@ -121,8 +122,8 @@ LAYERS_PER_BLOCK: int = 2
LAYERS_PER_BLOCK_UP: int = LAYERS_PER_BLOCK + 1 LAYERS_PER_BLOCK_UP: int = LAYERS_PER_BLOCK + 1
TIME_EMBED_FLIP_SIN_TO_COS: bool = True TIME_EMBED_FLIP_SIN_TO_COS: bool = True
TIME_EMBED_FREQ_SHIFT: int = 0 TIME_EMBED_FREQ_SHIFT: int = 0
RESNET_GROUPS: int = 32 NORM_GROUPS: int = 32
RESNET_EPS: float = 1e-6 NORM_EPS: float = 1e-5
TRANSFORMER_NORM_NUM_GROUPS = 32 TRANSFORMER_NORM_NUM_GROUPS = 32
DOWN_BLOCK_TYPES = ["CrossAttnDownBlock2D", "CrossAttnDownBlock2D", "CrossAttnDownBlock2D", "DownBlock2D"] DOWN_BLOCK_TYPES = ["CrossAttnDownBlock2D", "CrossAttnDownBlock2D", "CrossAttnDownBlock2D", "DownBlock2D"]
@@ -233,13 +234,13 @@ class ResnetBlock2D(nn.Module):
self.in_channels = in_channels self.in_channels = in_channels
self.out_channels = out_channels self.out_channels = out_channels
self.norm1 = torch.nn.GroupNorm(num_groups=RESNET_GROUPS, num_channels=in_channels, eps=RESNET_EPS, affine=True) self.norm1 = torch.nn.GroupNorm(num_groups=NORM_GROUPS, num_channels=in_channels, eps=NORM_EPS, affine=True)
self.conv1 = torch.nn.Conv2d(in_channels, out_channels, kernel_size=3, stride=1, padding=1) self.conv1 = torch.nn.Conv2d(in_channels, out_channels, kernel_size=3, stride=1, padding=1)
self.time_emb_proj = torch.nn.Linear(TIME_EMBED_DIM, out_channels) self.time_emb_proj = torch.nn.Linear(TIME_EMBED_DIM, out_channels)
self.norm2 = torch.nn.GroupNorm(num_groups=RESNET_GROUPS, num_channels=out_channels, eps=RESNET_EPS, affine=True) self.norm2 = torch.nn.GroupNorm(num_groups=NORM_GROUPS, num_channels=out_channels, eps=NORM_EPS, affine=True)
self.conv2 = torch.nn.Conv2d(out_channels, out_channels, kernel_size=3, stride=1, padding=1) self.conv2 = torch.nn.Conv2d(out_channels, out_channels, kernel_size=3, stride=1, padding=1)
# if non_linearity == "swish": # if non_linearity == "swish":
@@ -304,6 +305,9 @@ class DownBlock2D(nn.Module):
self.gradient_checkpointing = False self.gradient_checkpointing = False
def set_use_memory_efficient_attention_xformers(self, value):
pass
def forward(self, hidden_states, temb=None): def forward(self, hidden_states, temb=None):
output_states = () output_states = ()
@@ -372,6 +376,11 @@ class CrossAttention(nn.Module):
self.to_out.append(nn.Linear(inner_dim, query_dim)) self.to_out.append(nn.Linear(inner_dim, query_dim))
# no dropout here # no dropout here
self.use_memory_efficient_attention_xformers = False
def set_use_memory_efficient_attention_xformers(self, value):
self.use_memory_efficient_attention_xformers = value
def reshape_heads_to_batch_dim(self, tensor): def reshape_heads_to_batch_dim(self, tensor):
batch_size, seq_len, dim = tensor.shape batch_size, seq_len, dim = tensor.shape
head_size = self.heads head_size = self.heads
@@ -387,6 +396,9 @@ class CrossAttention(nn.Module):
return tensor return tensor
def forward(self, hidden_states, context=None, mask=None): def forward(self, hidden_states, context=None, mask=None):
if self.use_memory_efficient_attention_xformers:
return self.forward_memory_efficient_xformers(hidden_states, context, mask)
query = self.to_q(hidden_states) query = self.to_q(hidden_states)
context = context if context is not None else hidden_states context = context if context is not None else hidden_states
key = self.to_k(context) key = self.to_k(context)
@@ -427,6 +439,30 @@ class CrossAttention(nn.Module):
hidden_states = self.reshape_batch_dim_to_heads(hidden_states) hidden_states = self.reshape_batch_dim_to_heads(hidden_states)
return hidden_states return hidden_states
# TODO support Hypernetworks
def forward_memory_efficient_xformers(self, x, context=None, mask=None):
import xformers.ops
h = self.heads
q_in = self.to_q(x)
context = context if context is not None else x
context = context.to(x.dtype)
k_in = self.to_k(context)
v_in = self.to_v(context)
q, k, v = map(lambda t: rearrange(t, "b n (h d) -> b n h d", h=h), (q_in, k_in, v_in))
del q_in, k_in, v_in
q = q.contiguous()
k = k.contiguous()
v = v.contiguous()
out = xformers.ops.memory_efficient_attention(q, k, v, attn_bias=None) # 最適なのを選んでくれる
out = rearrange(out, "b n h d -> b n (h d)", h=h)
out = self.to_out[0](out)
return out
# feedforward # feedforward
class GEGLU(nn.Module): class GEGLU(nn.Module):
@@ -506,8 +542,9 @@ class BasicTransformerBlock(nn.Module):
# 3. Feed-forward # 3. Feed-forward
self.norm3 = nn.LayerNorm(dim) self.norm3 = nn.LayerNorm(dim)
def set_use_memory_efficient_attention_xformers(self, use_memory_efficient_attention_xformers: bool): def set_use_memory_efficient_attention_xformers(self, value: bool):
raise NotImplementedError("Memory efficient attention is not implemented for this model.") self.attn1.set_use_memory_efficient_attention_xformers(value)
self.attn2.set_use_memory_efficient_attention_xformers(value)
def forward(self, hidden_states, context=None, timestep=None): def forward(self, hidden_states, context=None, timestep=None):
# 1. Self-Attention # 1. Self-Attention
@@ -566,6 +603,10 @@ class Transformer2DModel(nn.Module):
else: else:
self.proj_out = nn.Conv2d(inner_dim, in_channels, kernel_size=1, stride=1, padding=0) self.proj_out = nn.Conv2d(inner_dim, in_channels, kernel_size=1, stride=1, padding=0)
def set_use_memory_efficient_attention_xformers(self, value):
for transformer in self.transformer_blocks:
transformer.set_use_memory_efficient_attention_xformers(value)
def forward(self, hidden_states, encoder_hidden_states=None, timestep=None, return_dict: bool = True): def forward(self, hidden_states, encoder_hidden_states=None, timestep=None, return_dict: bool = True):
# 1. Input # 1. Input
batch, _, height, weight = hidden_states.shape batch, _, height, weight = hidden_states.shape
@@ -643,6 +684,10 @@ class CrossAttnDownBlock2D(nn.Module):
self.gradient_checkpointing = False self.gradient_checkpointing = False
def set_use_memory_efficient_attention_xformers(self, value):
for attn in self.attentions:
attn.set_use_memory_efficient_attention_xformers(value)
def forward(self, hidden_states, temb=None, encoder_hidden_states=None): def forward(self, hidden_states, temb=None, encoder_hidden_states=None):
output_states = () output_states = ()
@@ -714,11 +759,37 @@ class UNetMidBlock2DCrossAttn(nn.Module):
self.attentions = nn.ModuleList(attentions) self.attentions = nn.ModuleList(attentions)
self.resnets = nn.ModuleList(resnets) self.resnets = nn.ModuleList(resnets)
self.gradient_checkpointing = False
def set_use_memory_efficient_attention_xformers(self, value):
for attn in self.attentions:
attn.set_use_memory_efficient_attention_xformers(value)
def forward(self, hidden_states, temb=None, encoder_hidden_states=None): def forward(self, hidden_states, temb=None, encoder_hidden_states=None):
hidden_states = self.resnets[0](hidden_states, temb) for i, resnet in enumerate(self.resnets):
for attn, resnet in zip(self.attentions, self.resnets[1:]): attn = None if i == 0 else self.attentions[i - 1]
hidden_states = attn(hidden_states, encoder_hidden_states).sample
hidden_states = resnet(hidden_states, temb) 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
if attn is not None:
hidden_states = torch.utils.checkpoint.checkpoint(
create_custom_forward(attn, return_dict=False), hidden_states, encoder_hidden_states
)[0]
hidden_states = torch.utils.checkpoint.checkpoint(create_custom_forward(resnet), hidden_states, temb)
else:
if attn is not None:
hidden_states = attn(hidden_states, encoder_hidden_states).sample
hidden_states = resnet(hidden_states, temb)
return hidden_states return hidden_states
@@ -792,6 +863,9 @@ class UpBlock2D(nn.Module):
self.gradient_checkpointing = False self.gradient_checkpointing = False
def set_use_memory_efficient_attention_xformers(self, value):
pass
def forward(self, hidden_states, res_hidden_states_tuple, temb=None, upsample_size=None): def forward(self, hidden_states, res_hidden_states_tuple, temb=None, upsample_size=None):
for resnet in self.resnets: for resnet in self.resnets:
# pop res hidden states # pop res hidden states
@@ -868,6 +942,10 @@ class CrossAttnUpBlock2D(nn.Module):
self.gradient_checkpointing = False self.gradient_checkpointing = False
def set_use_memory_efficient_attention_xformers(self, value):
for attn in self.attentions:
attn.set_use_memory_efficient_attention_xformers(value)
def forward( def forward(
self, self,
hidden_states, hidden_states,
@@ -991,6 +1069,8 @@ class UNet2DConditionModel(nn.Module):
self.sample_size = sample_size self.sample_size = sample_size
# state_dictの書式が変わるのでmoduleの持ち方は変えられない
# input # input
self.conv_in = nn.Conv2d(IN_CHANNELS, BLOCK_OUT_CHANNELS[0], kernel_size=3, padding=(1, 1)) self.conv_in = nn.Conv2d(IN_CHANNELS, BLOCK_OUT_CHANNELS[0], kernel_size=3, padding=(1, 1))
@@ -1069,7 +1149,7 @@ class UNet2DConditionModel(nn.Module):
prev_output_channel = output_channel prev_output_channel = output_channel
# out # out
self.conv_norm_out = nn.GroupNorm(num_channels=BLOCK_OUT_CHANNELS[0], num_groups=RESNET_GROUPS, eps=RESNET_EPS) self.conv_norm_out = nn.GroupNorm(num_channels=BLOCK_OUT_CHANNELS[0], num_groups=NORM_GROUPS, eps=NORM_EPS)
self.conv_act = nn.SiLU() self.conv_act = nn.SiLU()
self.conv_out = nn.Conv2d(BLOCK_OUT_CHANNELS[0], OUT_CHANNELS, kernel_size=3, padding=1) self.conv_out = nn.Conv2d(BLOCK_OUT_CHANNELS[0], OUT_CHANNELS, kernel_size=3, padding=1)
@@ -1088,16 +1168,20 @@ class UNet2DConditionModel(nn.Module):
return any(hasattr(m, "gradient_checkpointing") and m.gradient_checkpointing for m in self.modules()) return any(hasattr(m, "gradient_checkpointing") and m.gradient_checkpointing for m in self.modules())
def enable_gradient_checkpointing(self): def enable_gradient_checkpointing(self):
self._set_gradient_checkpointing(self, value=True) self.set_gradient_checkpointing(value=True)
def disable_gradient_checkpointing(self): def disable_gradient_checkpointing(self):
self._set_gradient_checkpointing(self, value=False) self.set_gradient_checkpointing(value=False)
def set_use_memory_efficient_attention_xformers(self, valid: bool) -> None: def set_use_memory_efficient_attention_xformers(self, valid: bool) -> None:
raise NotImplementedError("Memory efficient attention is not supported for this model.") modules = self.down_blocks + [self.mid_block] + self.up_blocks
for module in modules:
module.set_use_memory_efficient_attention_xformers(valid)
def _set_gradient_checkpointing(self, module, value=False): def set_gradient_checkpointing(self, value=False):
if isinstance(module, (CrossAttnDownBlock2D, DownBlock2D, CrossAttnUpBlock2D, UpBlock2D)): modules = self.down_blocks + [self.mid_block] + self.up_blocks
for module in modules:
print(module.__class__.__name__, module.gradient_checkpointing, "->", value)
module.gradient_checkpointing = value module.gradient_checkpointing = value
# endregion # endregion

66
library/train_util.py
View File

@@ -1792,7 +1792,7 @@ def replace_unet_modules(unet: diffusers.models.unet_2d_condition.UNet2DConditio
if mem_eff_attn: if mem_eff_attn:
replace_unet_cross_attn_to_memory_efficient() replace_unet_cross_attn_to_memory_efficient()
elif xformers: elif xformers:
replace_unet_cross_attn_to_xformers() replace_unet_cross_attn_to_xformers(unet)
def replace_unet_cross_attn_to_memory_efficient(): def replace_unet_cross_attn_to_memory_efficient():
@@ -1827,55 +1827,59 @@ def replace_unet_cross_attn_to_memory_efficient():
out = rearrange(out, "b h n d -> b n (h d)") out = rearrange(out, "b h n d -> b n (h d)")
# diffusers 0.7.0~ わざわざ変えるなよ (;´Д｀)
out = self.to_out[0](out) out = self.to_out[0](out)
out = self.to_out[1](out) # out = self.to_out[1](out)
return out return out
diffusers.models.attention.CrossAttention.forward = forward_flash_attn # diffusers.models.attention.CrossAttention.forward = forward_flash_attn
from library.original_unet import CrossAttention
CrossAttention.forward = forward_flash_attn
def replace_unet_cross_attn_to_xformers(): def replace_unet_cross_attn_to_xformers(unet):
print("CrossAttention.forward has been replaced to enable xformers.") print("CrossAttention.forward has been replaced to enable xformers.")
try: try:
import xformers.ops import xformers.ops
except ImportError: except ImportError:
raise ImportError("No xformers / xformersがインストールされていないようです") raise ImportError("No xformers / xformersがインストールされていないようです")
def forward_xformers(self, x, context=None, mask=None): unet.set_use_memory_efficient_attention_xformers(True)
h = self.heads
q_in = self.to_q(x)
context = default(context, x) # def forward_xformers(self, x, context=None, mask=None):
context = context.to(x.dtype) # h = self.heads
# q_in = self.to_q(x)
if hasattr(self, "hypernetwork") and self.hypernetwork is not None: # context = default(context, x)
context_k, context_v = self.hypernetwork.forward(x, context) # context = context.to(x.dtype)
context_k = context_k.to(x.dtype)
context_v = context_v.to(x.dtype)
else:
context_k = context
context_v = context
k_in = self.to_k(context_k) # if hasattr(self, "hypernetwork") and self.hypernetwork is not None:
v_in = self.to_v(context_v) # context_k, context_v = self.hypernetwork.forward(x, context)
# context_k = context_k.to(x.dtype)
# context_v = context_v.to(x.dtype)
# else:
# context_k = context
# context_v = context
q, k, v = map(lambda t: rearrange(t, "b n (h d) -> b n h d", h=h), (q_in, k_in, v_in)) # k_in = self.to_k(context_k)
del q_in, k_in, v_in # v_in = self.to_v(context_v)
q = q.contiguous() # q, k, v = map(lambda t: rearrange(t, "b n (h d) -> b n h d", h=h), (q_in, k_in, v_in))
k = k.contiguous() # del q_in, k_in, v_in
v = v.contiguous()
out = xformers.ops.memory_efficient_attention(q, k, v, attn_bias=None) # 最適なのを選んでくれる
out = rearrange(out, "b n h d -> b n (h d)", h=h) # q = q.contiguous()
# k = k.contiguous()
# v = v.contiguous()
# out = xformers.ops.memory_efficient_attention(q, k, v, attn_bias=None) # 最適なのを選んでくれる
# diffusers 0.7.0~ # out = rearrange(out, "b n h d -> b n (h d)", h=h)
out = self.to_out[0](out)
out = self.to_out[1](out)
return out
diffusers.models.attention.CrossAttention.forward = forward_xformers # # diffusers 0.7.0~
# out = self.to_out[0](out)
# out = self.to_out[1](out)
# return out
# diffusers.models.attention.CrossAttention.forward = forward_xformers
""" """