evan/ai-toolkit
1
0 Fork 0
Code Issues Pull Requests Actions Packages Projects Releases Wiki Activity

Added specialized scaler training to ip adapters

Browse Source
This commit is contained in:
Jaret Burkett committed 2024-04-05 08:17:09 -06:00
1 parent 427847ac4c
commit 7284aab7c0
7 files changed
+182 -29
Download Patch File Download Diff File Expand all files Collapse all files

No files matched your search

extensions_built_in/sd_trainer/SDTrainer.py
+4 -2
View File
@@ -1243,8 +1243,10 @@ class SDTrainer(BaseSDTrainProcess):
has_been_preprocessed=True, has_been_preprocessed=True,
quad_count=quad_count quad_count=quad_count
) )
# else: else:
# raise ValueError("Adapter images now must be loaded with dataloader or be a reg image") print("No Clip Image")
print([file_item.path for file_item in batch.file_items])
raise ValueError("Could not find clip image")
if not self.adapter_config.train_image_encoder: if not self.adapter_config.train_image_encoder:
# we are not training the image encoder, so we need to detach the embeds # we are not training the image encoder, so we need to detach the embeds
jobs/process/BaseSDTrainProcess.py
+12 -5
View File
@@ -1293,11 +1293,18 @@ class BaseSDTrainProcess(BaseTrainProcess):
if self.adapter_config is not None: if self.adapter_config is not None:
self.setup_adapter() self.setup_adapter()
if self.adapter_config.train: if self.adapter_config.train:
# set trainable params
params.append({ if isinstance(self.adapter, IPAdapter):
'params': self.adapter.parameters(), # we have custom LR groups for IPAdapter
'lr': self.train_config.adapter_lr adapter_param_groups = self.adapter.get_parameter_groups(self.train_config.adapter_lr)
}) for group in adapter_param_groups:
params.append(group)
else:
# set trainable params
params.append({
'params': self.adapter.parameters(),
'lr': self.train_config.adapter_lr
})
if self.train_config.gradient_checkpointing: if self.train_config.gradient_checkpointing:
self.adapter.enable_gradient_checkpointing() self.adapter.enable_gradient_checkpointing()
toolkit/config.py
+1 -3
View File
@@ -43,9 +43,7 @@ def preprocess_config(config: OrderedDict, name: str = None):
if "name" not in config["config"] and name is None: if "name" not in config["config"] and name is None:
raise ValueError("config file must have a config.name key") raise ValueError("config file must have a config.name key")
# we need to replace tags. For now just [name] # we need to replace tags. For now just [name]
if name is not None: if name is None:
config["config"]["name"] = name
else:
name = config["config"]["name"] name = config["config"]["name"]
config_string = json.dumps(config) config_string = json.dumps(config)
config_string = config_string.replace("[name]", name) config_string = config_string.replace("[name]", name)
toolkit/config_modules.py
+6
View File
@@ -181,6 +181,12 @@ class AdapterConfig:
self.text_encoder_path: str = kwargs.get('text_encoder_path', None) self.text_encoder_path: str = kwargs.get('text_encoder_path', None)
self.text_encoder_arch: str = kwargs.get('text_encoder_arch', 'clip') # clip t5 self.text_encoder_arch: str = kwargs.get('text_encoder_arch', 'clip') # clip t5
self.train_scaler: bool = kwargs.get('train_scaler', False)
self.scaler_lr: Optional[float] = kwargs.get('scaler_lr', None)
# trains with a scaler to easy channel bias but merges it in on save
self.merge_scaler: bool = kwargs.get('merge_scaler', False)
class EmbeddingConfig: class EmbeddingConfig:
def __init__(self, **kwargs): def __init__(self, **kwargs):
toolkit/ip_adapter.py
+105 -6
View File
@@ -80,9 +80,14 @@ class MLPProjModelClipFace(torch.nn.Module):
class CustomIPAttentionProcessor(IPAttnProcessor2_0): class CustomIPAttentionProcessor(IPAttnProcessor2_0):
def __init__(self, hidden_size, cross_attention_dim, scale=1.0, num_tokens=4, adapter=None): def __init__(self, hidden_size, cross_attention_dim, scale=1.0, num_tokens=4, adapter=None, train_scaler=False):
super().__init__(hidden_size, cross_attention_dim, scale=scale, num_tokens=num_tokens) super().__init__(hidden_size, cross_attention_dim, scale=scale, num_tokens=num_tokens)
self.adapter_ref: weakref.ref = weakref.ref(adapter) self.adapter_ref: weakref.ref = weakref.ref(adapter)
self.train_scaler = train_scaler
if train_scaler:
# self.ip_scaler = torch.nn.Parameter(torch.ones([num_tokens], dtype=torch.float32) * 0.9999)
self.ip_scaler = torch.nn.Parameter(torch.ones([1], dtype=torch.float32) * 0.9999)
self.ip_scaler.requires_grad_(True)
def __call__( def __call__(
self, self,
@@ -169,6 +174,13 @@ class CustomIPAttentionProcessor(IPAttnProcessor2_0):
# will be none if disabled # will be none if disabled
if ip_hidden_states is not None: if ip_hidden_states is not None:
# apply scaler
if self.train_scaler:
weight = self.ip_scaler
# reshape to (1, self.num_tokens, 1)
weight = weight.view(1, -1, 1)
ip_hidden_states = ip_hidden_states * weight
# for ip-adapter # for ip-adapter
ip_key = self.to_k_ip(ip_hidden_states) ip_key = self.to_k_ip(ip_hidden_states)
ip_value = self.to_v_ip(ip_hidden_states) ip_value = self.to_v_ip(ip_hidden_states)
@@ -185,7 +197,8 @@ class CustomIPAttentionProcessor(IPAttnProcessor2_0):
ip_hidden_states = ip_hidden_states.transpose(1, 2).reshape(batch_size, -1, attn.heads * head_dim) ip_hidden_states = ip_hidden_states.transpose(1, 2).reshape(batch_size, -1, attn.heads * head_dim)
ip_hidden_states = ip_hidden_states.to(query.dtype) ip_hidden_states = ip_hidden_states.to(query.dtype)
hidden_states = hidden_states + self.scale * ip_hidden_states scale = self.scale
hidden_states = hidden_states + scale * ip_hidden_states
# linear proj # linear proj
hidden_states = attn.to_out[0](hidden_states) hidden_states = attn.to_out[0](hidden_states)
@@ -202,6 +215,21 @@ class CustomIPAttentionProcessor(IPAttnProcessor2_0):
return hidden_states return hidden_states
# this ensures that the ip_scaler is not changed when we load the model
# def _apply(self, fn):
# if hasattr(self, "ip_scaler"):
# # Overriding the _apply method to prevent the special_parameter from changing dtype
# self.ip_scaler = fn(self.ip_scaler)
# # Temporarily set the special_parameter to None to exclude it from default _apply processing
# ip_scaler = self.ip_scaler
# self.ip_scaler = None
# super(CustomIPAttentionProcessor, self)._apply(fn)
# # Restore the special_parameter after the default _apply processing
# self.ip_scaler = ip_scaler
# return self
# else:
# return super(CustomIPAttentionProcessor, self)._apply(fn)
# loosely based on # ref https://github.com/tencent-ailab/IP-Adapter/blob/main/tutorial_train.py # loosely based on # ref https://github.com/tencent-ailab/IP-Adapter/blob/main/tutorial_train.py
class IPAdapter(torch.nn.Module): class IPAdapter(torch.nn.Module):
@@ -485,7 +513,8 @@ class IPAdapter(torch.nn.Module):
cross_attention_dim=cross_attention_dim, cross_attention_dim=cross_attention_dim,
scale=1.0, scale=1.0,
num_tokens=self.config.num_tokens, num_tokens=self.config.num_tokens,
adapter=self adapter=self,
train_scaler=self.config.train_scaler or self.config.merge_scaler
) )
if self.sd_ref().is_pixart: if self.sd_ref().is_pixart:
# pixart is much more sensitive # pixart is much more sensitive
@@ -494,7 +523,7 @@ class IPAdapter(torch.nn.Module):
"to_v_ip.weight": weights["to_v_ip.weight"] * 0.01, "to_v_ip.weight": weights["to_v_ip.weight"] * 0.01,
} }
attn_procs[name].load_state_dict(weights) attn_procs[name].load_state_dict(weights, strict=False)
attn_processor_names.append(name) attn_processor_names.append(name)
print(f"Attn Processors") print(f"Attn Processors")
print(attn_processor_names) print(attn_processor_names)
@@ -568,9 +597,34 @@ class IPAdapter(torch.nn.Module):
state_dict = OrderedDict() state_dict = OrderedDict()
if self.config.train_only_image_encoder: if self.config.train_only_image_encoder:
return self.image_encoder.state_dict() return self.image_encoder.state_dict()
if self.config.train_scaler:
state_dict["ip_scale"] = self.adapter_modules.state_dict()
# remove items that are not scalers
for key in list(state_dict["ip_scale"].keys()):
if not key.endswith("ip_scaler"):
del state_dict["ip_scale"][key]
return state_dict
state_dict["image_proj"] = self.image_proj_model.state_dict() state_dict["image_proj"] = self.image_proj_model.state_dict()
state_dict["ip_adapter"] = self.adapter_modules.state_dict() state_dict["ip_adapter"] = self.adapter_modules.state_dict()
# handle merge scaler training
if self.config.merge_scaler:
for key in list(state_dict["ip_adapter"].keys()):
if key.endswith("ip_scaler"):
# merge in the scaler so we dont have to save it and it will be compatible with other ip adapters
scale = state_dict["ip_adapter"][key].clone()
key_start = key.split(".")[-2]
# reshape to (1, 1)
scale = scale.view(1, 1)
del state_dict["ip_adapter"][key]
# find the to_k_ip and to_v_ip keys
for key2 in list(state_dict["ip_adapter"].keys()):
if key2.endswith(f"{key_start}.to_k_ip.weight"):
state_dict["ip_adapter"][key2] = state_dict["ip_adapter"][key2].clone() * scale
if key2.endswith(f"{key_start}.to_v_ip.weight"):
state_dict["ip_adapter"][key2] = state_dict["ip_adapter"][key2].clone() * scale
if self.config.train_image_encoder: if self.config.train_image_encoder:
state_dict["image_encoder"] = self.image_encoder.state_dict() state_dict["image_encoder"] = self.image_encoder.state_dict()
if self.preprocessor is not None: if self.preprocessor is not None:
@@ -866,18 +920,61 @@ class IPAdapter(torch.nn.Module):
self.image_proj_model.train(mode) self.image_proj_model.train(mode)
return super().train(mode) return super().train(mode)
def parameters(self, recurse: bool = True) -> Iterator[Parameter]: def get_parameter_groups(self, adapter_lr):
param_groups = []
# when training just scaler, we do not train anything else
if not self.config.train_scaler:
param_groups.append({
"params": self.get_non_scaler_parameters(),
"lr": adapter_lr,
})
if self.config.train_scaler or self.config.merge_scaler:
scaler_lr = adapter_lr if self.config.scaler_lr is None else self.config.scaler_lr
param_groups.append({
"params": self.get_scaler_parameters(),
"lr": scaler_lr,
})
return param_groups
def get_scaler_parameters(self):
# only get the scalera from the adapter modules
for attn_processor in self.adapter_modules:
# only get the scaler
# check if it has ip_scaler attribute
if hasattr(attn_processor, "ip_scaler"):
scaler_param = attn_processor.ip_scaler
yield scaler_param
def get_non_scaler_parameters(self, recurse: bool = True) -> Iterator[Parameter]:
if self.config.train_only_image_encoder: if self.config.train_only_image_encoder:
yield from self.image_encoder.parameters(recurse) yield from self.image_encoder.parameters(recurse)
return return
if self.config.train_scaler:
# no params
return
for attn_processor in self.adapter_modules: for attn_processor in self.adapter_modules:
yield from attn_processor.parameters(recurse) if self.config.train_scaler or self.config.merge_scaler:
# todo remove scaler
if hasattr(attn_processor, "to_k_ip"):
# yield the linear layer
yield from attn_processor.to_k_ip.parameters(recurse)
if hasattr(attn_processor, "to_v_ip"):
# yield the linear layer
yield from attn_processor.to_v_ip.parameters(recurse)
else:
yield from attn_processor.parameters(recurse)
yield from self.image_proj_model.parameters(recurse) yield from self.image_proj_model.parameters(recurse)
if self.config.train_image_encoder: if self.config.train_image_encoder:
yield from self.image_encoder.parameters(recurse) yield from self.image_encoder.parameters(recurse)
if self.preprocessor is not None: if self.preprocessor is not None:
yield from self.preprocessor.parameters(recurse) yield from self.preprocessor.parameters(recurse)
def parameters(self, recurse: bool = True) -> Iterator[Parameter]:
yield from self.get_non_scaler_parameters(recurse)
if self.config.train_scaler or self.config.merge_scaler:
yield from self.get_scaler_parameters()
def merge_in_weights(self, state_dict: Mapping[str, Any]): def merge_in_weights(self, state_dict: Mapping[str, Any]):
# merge in img_proj weights # merge in img_proj weights
current_img_proj_state_dict = self.image_proj_model.state_dict() current_img_proj_state_dict = self.image_proj_model.state_dict()
@@ -975,6 +1072,8 @@ class IPAdapter(torch.nn.Module):
def load_state_dict(self, state_dict: Mapping[str, Any], strict: bool = True): def load_state_dict(self, state_dict: Mapping[str, Any], strict: bool = True):
strict = False strict = False
if self.config.train_scaler and 'ip_scale' in state_dict:
self.adapter_modules.load_state_dict(state_dict["ip_scale"], strict=False)
if 'ip_adapter' in state_dict: if 'ip_adapter' in state_dict:
try: try:
self.image_proj_model.load_state_dict(state_dict["image_proj"], strict=strict) self.image_proj_model.load_state_dict(state_dict["image_proj"], strict=strict)
toolkit/models/ilora.py
+29 -10
View File
@@ -5,6 +5,10 @@ import torch.nn as nn
from typing import TYPE_CHECKING from typing import TYPE_CHECKING
from toolkit.models.clip_fusion import ZipperBlock from toolkit.models.clip_fusion import ZipperBlock
from toolkit.models.zipper_resampler import ZipperModule, ZipperResampler from toolkit.models.zipper_resampler import ZipperModule, ZipperResampler
import sys
from toolkit.paths import REPOS_ROOT
sys.path.append(REPOS_ROOT)
from ipadapter.ip_adapter.resampler import Resampler
if TYPE_CHECKING: if TYPE_CHECKING:
from toolkit.lora_special import LoRAModule from toolkit.lora_special import LoRAModule
@@ -50,7 +54,7 @@ class InstantLoRAMidModule(torch.nn.Module):
raise e raise e
# apply tanh to limit values to -1 to 1 # apply tanh to limit values to -1 to 1
# scaler = torch.tanh(scaler) # scaler = torch.tanh(scaler)
return x * scaler return x * (scaler + 1.0)
class InstantLoRAModule(torch.nn.Module): class InstantLoRAModule(torch.nn.Module):
@@ -78,15 +82,30 @@ class InstantLoRAModule(torch.nn.Module):
lora_modules = self.sd_ref().network.get_all_modules() lora_modules = self.sd_ref().network.get_all_modules()
# resample the output so each module gets one token with a size of its dim so we can multiply by that # resample the output so each module gets one token with a size of its dim so we can multiply by that
self.resampler = ZipperResampler( # self.resampler = ZipperResampler(
in_size=self.vision_hidden_size, # in_size=self.vision_hidden_size,
in_tokens=self.vision_tokens, # in_tokens=self.vision_tokens,
out_size=self.dim, # out_size=self.dim,
out_tokens=len(lora_modules), # out_tokens=len(lora_modules),
hidden_size=self.vision_hidden_size, # hidden_size=self.vision_hidden_size,
hidden_tokens=self.vision_tokens, # hidden_tokens=self.vision_tokens,
num_blocks=1, # num_blocks=1,
) # )
# heads = 20
heads = 12
dim = 1280
output_dim = self.dim
self.resampler = Resampler(
dim=dim,
depth=4,
dim_head=64,
heads=heads,
num_queries=len(lora_modules),
embedding_dim=self.vision_hidden_size,
max_seq_len=self.vision_tokens,
output_dim=output_dim,
ff_mult=4
)
for idx, lora_module in enumerate(lora_modules): for idx, lora_module in enumerate(lora_modules):
# add a new mid module that will take the original forward and add a vector to it # add a new mid module that will take the original forward and add a vector to it
toolkit/stable_diffusion_model.py
+25 -3
View File
@@ -39,7 +39,7 @@ from toolkit.pipelines import CustomStableDiffusionXLPipeline, CustomStableDiffu
StableDiffusionKDiffusionXLPipeline, StableDiffusionXLRefinerPipeline StableDiffusionKDiffusionXLPipeline, StableDiffusionXLRefinerPipeline
from diffusers import StableDiffusionPipeline, StableDiffusionXLPipeline, T2IAdapter, DDPMScheduler, \ from diffusers import StableDiffusionPipeline, StableDiffusionXLPipeline, T2IAdapter, DDPMScheduler, \
StableDiffusionXLAdapterPipeline, StableDiffusionAdapterPipeline, DiffusionPipeline, \ StableDiffusionXLAdapterPipeline, StableDiffusionAdapterPipeline, DiffusionPipeline, \
StableDiffusionXLImg2ImgPipeline, LCMScheduler, Transformer2DModel StableDiffusionXLImg2ImgPipeline, LCMScheduler, Transformer2DModel, AutoencoderTiny
import diffusers import diffusers
from diffusers import \ from diffusers import \
AutoencoderKL, \ AutoencoderKL, \
@@ -872,8 +872,10 @@ class StableDiffusion:
is_input_scaled=False, is_input_scaled=False,
detach_unconditional=False, detach_unconditional=False,
rescale_cfg=None, rescale_cfg=None,
return_conditional_pred=False,
**kwargs, **kwargs,
): ):
conditional_pred = None
# get the embeddings # get the embeddings
if text_embeddings is None and conditional_embeddings is None: if text_embeddings is None and conditional_embeddings is None:
raise ValueError("Either text_embeddings or conditional_embeddings must be specified") raise ValueError("Either text_embeddings or conditional_embeddings must be specified")
@@ -1024,9 +1026,12 @@ class StableDiffusion:
**kwargs, **kwargs,
).sample ).sample
conditional_pred = noise_pred
if do_classifier_free_guidance: if do_classifier_free_guidance:
# perform guidance # perform guidance
noise_pred_uncond, noise_pred_text = noise_pred.chunk(2) noise_pred_uncond, noise_pred_text = noise_pred.chunk(2)
conditional_pred = noise_pred_text
noise_pred = noise_pred_uncond + guidance_scale * ( noise_pred = noise_pred_uncond + guidance_scale * (
noise_pred_text - noise_pred_uncond noise_pred_text - noise_pred_uncond
) )
@@ -1112,9 +1117,12 @@ class StableDiffusion:
**kwargs, **kwargs,
).sample ).sample
conditional_pred = noise_pred
if do_classifier_free_guidance: if do_classifier_free_guidance:
# perform guidance # perform guidance
noise_pred_uncond, noise_pred_text = noise_pred.chunk(2, dim=0) noise_pred_uncond, noise_pred_text = noise_pred.chunk(2, dim=0)
conditional_pred = noise_pred_text
if detach_unconditional: if detach_unconditional:
noise_pred_uncond = noise_pred_uncond.detach() noise_pred_uncond = noise_pred_uncond.detach()
noise_pred = noise_pred_uncond + guidance_scale * ( noise_pred = noise_pred_uncond + guidance_scale * (
@@ -1141,6 +1149,8 @@ class StableDiffusion:
# Based on 3.4. in https://arxiv.org/pdf/2305.08891.pdf # Based on 3.4. in https://arxiv.org/pdf/2305.08891.pdf
noise_pred = rescale_noise_cfg(noise_pred, noise_pred_text, guidance_rescale=guidance_rescale) noise_pred = rescale_noise_cfg(noise_pred, noise_pred_text, guidance_rescale=guidance_rescale)
if return_conditional_pred:
return noise_pred, conditional_pred
return noise_pred return noise_pred
def step_scheduler(self, model_input, latent_input, timestep_tensor, noise_scheduler=None): def step_scheduler(self, model_input, latent_input, timestep_tensor, noise_scheduler=None):
@@ -1187,23 +1197,30 @@ class StableDiffusion:
bleed_ratio: float = 0.5, bleed_ratio: float = 0.5,
bleed_latents: torch.FloatTensor = None, bleed_latents: torch.FloatTensor = None,
is_input_scaled=False, is_input_scaled=False,
return_first_prediction=False,
**kwargs, **kwargs,
): ):
timesteps_to_run = self.noise_scheduler.timesteps[start_timesteps:total_timesteps] timesteps_to_run = self.noise_scheduler.timesteps[start_timesteps:total_timesteps]
first_prediction = None
for timestep in tqdm(timesteps_to_run, leave=False): for timestep in tqdm(timesteps_to_run, leave=False):
timestep = timestep.unsqueeze_(0) timestep = timestep.unsqueeze_(0)
noise_pred = self.predict_noise( noise_pred, conditional_pred = self.predict_noise(
latents, latents,
text_embeddings, text_embeddings,
timestep, timestep,
guidance_scale=guidance_scale, guidance_scale=guidance_scale,
add_time_ids=add_time_ids, add_time_ids=add_time_ids,
is_input_scaled=is_input_scaled, is_input_scaled=is_input_scaled,
return_conditional_pred=True,
**kwargs, **kwargs,
) )
# some schedulers need to run separately, so do that. (euler for example) # some schedulers need to run separately, so do that. (euler for example)
if return_first_prediction and first_prediction is None:
first_prediction = conditional_pred
latents = self.step_scheduler(noise_pred, latents, timestep) latents = self.step_scheduler(noise_pred, latents, timestep)
# if not last step, and bleeding, bleed in some latents # if not last step, and bleeding, bleed in some latents
@@ -1214,6 +1231,8 @@ class StableDiffusion:
is_input_scaled = False is_input_scaled = False
# return latents_steps # return latents_steps
if return_first_prediction:
return latents, first_prediction
return latents return latents
def encode_prompt( def encode_prompt(
@@ -1311,7 +1330,10 @@ class StableDiffusion:
image_list[i] = Resize((image.shape[1] // 8 * 8, image.shape[2] // 8 * 8))(image) image_list[i] = Resize((image.shape[1] // 8 * 8, image.shape[2] // 8 * 8))(image)
images = torch.stack(image_list) images = torch.stack(image_list)
latents = self.vae.encode(images).latent_dist.sample() if isinstance(self.vae, AutoencoderTiny):
latents = self.vae.encode(images, return_dict=False)[0]
else:
latents = self.vae.encode(images).latent_dist.sample()
# latents = self.vae.encode(images, return_dict=False)[0] # latents = self.vae.encode(images, return_dict=False)[0]
latents = latents * self.vae.config['scaling_factor'] latents = latents * self.vae.config['scaling_factor']
latents = latents.to(device, dtype=dtype) latents = latents.to(device, dtype=dtype)