1580158
/
IP-Adapter
mirror of https://github.com/tencent-ailab/IP-Adapter
1
0
Fork 0
Code Issues Packages Projects Releases Wiki Activity

style: Format code indent, import order and remove trailing whitespace

pull/134/head
jetthu 2023-11-08 07:13:26 +08:00
parent 757123e0c5
commit bd7d2c1998
5 changed files with 161 additions and 122 deletions
Show Stats Download Patch File Download Diff File Expand all files Collapse all files

9
ip_adapter/__init__.py
View File

@ -1 +1,8 @@
from .ip_adapter import IPAdapter, IPAdapterXL, IPAdapterPlus, IPAdapterPlusXL
from .ip_adapter import IPAdapter, IPAdapterPlus, IPAdapterPlusXL, IPAdapterXL
__all__ = [
"IPAdapter",
"IPAdapterPlus",
"IPAdapterPlusXL",
"IPAdapterXL",
]

57
ip_adapter/attention_processor.py
View File

@ -8,6 +8,7 @@ class AttnProcessor(nn.Module):
r"""
Default processor for performing attention-related computations.
"""
def __init__(
self,
hidden_size=None,
@ -74,8 +75,8 @@ class AttnProcessor(nn.Module):
hidden_states = hidden_states / attn.rescale_output_factor
return hidden_states
class IPAttnProcessor(nn.Module):
r"""
Attention processor for IP-Adapater.
@ -135,7 +136,10 @@ class IPAttnProcessor(nn.Module):
else:
# get encoder_hidden_states, ip_hidden_states
end_pos = encoder_hidden_states.shape[1] - self.num_tokens
encoder_hidden_states, ip_hidden_states = encoder_hidden_states[:, :end_pos, :], encoder_hidden_states[:, end_pos:, :]
encoder_hidden_states, ip_hidden_states = (
encoder_hidden_states[:, :end_pos, :],
encoder_hidden_states[:, end_pos:, :],
)
if attn.norm_cross:
encoder_hidden_states = attn.norm_encoder_hidden_states(encoder_hidden_states)
@ -149,18 +153,18 @@ class IPAttnProcessor(nn.Module):
attention_probs = attn.get_attention_scores(query, key, attention_mask)
hidden_states = torch.bmm(attention_probs, value)
hidden_states = attn.batch_to_head_dim(hidden_states)
# for ip-adapter
ip_key = self.to_k_ip(ip_hidden_states)
ip_value = self.to_v_ip(ip_hidden_states)
ip_key = attn.head_to_batch_dim(ip_key)
ip_value = attn.head_to_batch_dim(ip_value)
ip_attention_probs = attn.get_attention_scores(query, ip_key, None)
ip_hidden_states = torch.bmm(ip_attention_probs, ip_value)
ip_hidden_states = attn.batch_to_head_dim(ip_hidden_states)
hidden_states = hidden_states + self.scale * ip_hidden_states
# linear proj
@ -177,12 +181,13 @@ class IPAttnProcessor(nn.Module):
hidden_states = hidden_states / attn.rescale_output_factor
return hidden_states
class AttnProcessor2_0(torch.nn.Module):
r"""
Processor for implementing scaled dot-product attention (enabled by default if you're using PyTorch 2.0).
"""
def __init__(
self,
hidden_size=None,
@ -265,8 +270,8 @@ class AttnProcessor2_0(torch.nn.Module):
hidden_states = hidden_states / attn.rescale_output_factor
return hidden_states
class IPAttnProcessor2_0(torch.nn.Module):
r"""
Attention processor for IP-Adapater for PyTorch 2.0.
@ -334,7 +339,10 @@ class IPAttnProcessor2_0(torch.nn.Module):
else:
# get encoder_hidden_states, ip_hidden_states
end_pos = encoder_hidden_states.shape[1] - self.num_tokens
encoder_hidden_states, ip_hidden_states = encoder_hidden_states[:, :end_pos, :], encoder_hidden_states[:, end_pos:, :]
encoder_hidden_states, ip_hidden_states = (
encoder_hidden_states[:, :end_pos, :],
encoder_hidden_states[:, end_pos:, :],
)
if attn.norm_cross:
encoder_hidden_states = attn.norm_encoder_hidden_states(encoder_hidden_states)
@ -357,11 +365,11 @@ class IPAttnProcessor2_0(torch.nn.Module):
hidden_states = hidden_states.transpose(1, 2).reshape(batch_size, -1, attn.heads * head_dim)
hidden_states = hidden_states.to(query.dtype)
# for ip-adapter
ip_key = self.to_k_ip(ip_hidden_states)
ip_value = self.to_v_ip(ip_hidden_states)
ip_key = ip_key.view(batch_size, -1, attn.heads, head_dim).transpose(1, 2)
ip_value = ip_value.view(batch_size, -1, attn.heads, head_dim).transpose(1, 2)
@ -370,10 +378,10 @@ class IPAttnProcessor2_0(torch.nn.Module):
ip_hidden_states = F.scaled_dot_product_attention(
query, ip_key, ip_value, attn_mask=None, dropout_p=0.0, is_causal=False
)
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)
hidden_states = hidden_states + self.scale * ip_hidden_states
# linear proj
@ -401,14 +409,7 @@ class CNAttnProcessor:
def __init__(self, num_tokens=4):
self.num_tokens = num_tokens
def __call__(
self,
attn,
hidden_states,
encoder_hidden_states=None,
attention_mask=None,
temb=None
):
def __call__(self, attn, hidden_states, encoder_hidden_states=None, attention_mask=None, temb=None):
residual = hidden_states
if attn.spatial_norm is not None:
@ -434,7 +435,7 @@ class CNAttnProcessor:
encoder_hidden_states = hidden_states
else:
end_pos = encoder_hidden_states.shape[1] - self.num_tokens
encoder_hidden_states = encoder_hidden_states[:, :end_pos] # only use text
encoder_hidden_states = encoder_hidden_states[:, :end_pos] # only use text
if attn.norm_cross:
encoder_hidden_states = attn.norm_encoder_hidden_states(encoder_hidden_states)
@ -470,7 +471,7 @@ class CNAttnProcessor2_0:
Processor for implementing scaled dot-product attention (enabled by default if you're using PyTorch 2.0).
"""
def __init__(self, num_tokens=4):
def __init__(self, num_tokens=4):
if not hasattr(F, "scaled_dot_product_attention"):
raise ImportError("AttnProcessor2_0 requires PyTorch 2.0, to use it, please upgrade PyTorch to 2.0.")
self.num_tokens = num_tokens
@ -513,7 +514,7 @@ class CNAttnProcessor2_0:
encoder_hidden_states = hidden_states
else:
end_pos = encoder_hidden_states.shape[1] - self.num_tokens
encoder_hidden_states = encoder_hidden_states[:, :end_pos] # only use text
encoder_hidden_states = encoder_hidden_states[:, :end_pos] # only use text
if attn.norm_cross:
encoder_hidden_states = attn.norm_encoder_hidden_states(encoder_hidden_states)
@ -550,4 +551,4 @@ class CNAttnProcessor2_0:
hidden_states = hidden_states / attn.rescale_output_factor
return hidden_states
return hidden_states

17
ip_adapter/custom_pipelines.py
View File

@ -1,13 +1,12 @@
from typing import Any, Callable, Dict, List, Optional, Tuple, Union
import torch
from diffusers import StableDiffusionXLPipeline
from diffusers.pipelines.stable_diffusion_xl import StableDiffusionXLPipelineOutput
from diffusers.pipelines.stable_diffusion_xl.pipeline_stable_diffusion_xl import rescale_noise_cfg
from .utils import is_torch2_available
if is_torch2_available():
from .attention_processor import IPAttnProcessor2_0 as IPAttnProcessor
else:
@ -15,16 +14,15 @@ else:
class StableDiffusionXLCustomPipeline(StableDiffusionXLPipeline):
def set_scale(self, scale):
for attn_processor in self.unet.attn_processors.values():
if isinstance(attn_processor, IPAttnProcessor):
attn_processor.scale = scale
@torch.no_grad()
def __call__(
def __call__( # noqa: C901
self,
prompt: Union[str, List[str]] = None,
prompt: Optional[Union[str, List[str]]] = None,
prompt_2: Optional[Union[str, List[str]]] = None,
height: Optional[int] = None,
width: Optional[int] = None,
@ -316,7 +314,7 @@ class StableDiffusionXLCustomPipeline(StableDiffusionXLPipeline):
)
num_inference_steps = len(list(filter(lambda ts: ts >= discrete_timestep_cutoff, timesteps)))
timesteps = timesteps[:num_inference_steps]
# get init conditioning scale
for attn_processor in self.unet.attn_processors.values():
if isinstance(attn_processor, IPAttnProcessor):
@ -325,9 +323,8 @@ class StableDiffusionXLCustomPipeline(StableDiffusionXLPipeline):
with self.progress_bar(total=num_inference_steps) as progress_bar:
for i, t in enumerate(timesteps):
if (i / len(timesteps) < control_guidance_start) or ((i + 1) / len(timesteps) > control_guidance_end):
self.set_scale(0.)
self.set_scale(0.0)
else:
self.set_scale(conditioning_scale)
@ -381,7 +378,7 @@ class StableDiffusionXLCustomPipeline(StableDiffusionXLPipeline):
else:
image = latents
if not output_type == "latent":
if output_type != "latent":
# apply watermark if available
if self.watermark is not None:
image = self.watermark.apply_watermark(image)
@ -394,4 +391,4 @@ class StableDiffusionXLCustomPipeline(StableDiffusionXLPipeline):
if not return_dict:
return (image,)
return StableDiffusionXLPipelineOutput(images=image)
return StableDiffusionXLPipelineOutput(images=image)

170
ip_adapter/ip_adapter.py
View File

@ -4,55 +4,67 @@ from typing import List
import torch
from diffusers import StableDiffusionPipeline
from diffusers.pipelines.controlnet import MultiControlNetModel
from transformers import CLIPVisionModelWithProjection, CLIPImageProcessor
from PIL import Image
from safetensors import safe_open
from transformers import CLIPImageProcessor, CLIPVisionModelWithProjection
from .utils import is_torch2_available
if is_torch2_available():
from .attention_processor import IPAttnProcessor2_0 as IPAttnProcessor, AttnProcessor2_0 as AttnProcessor, CNAttnProcessor2_0 as CNAttnProcessor
from .attention_processor import (
AttnProcessor2_0 as AttnProcessor,
)
from .attention_processor import (
CNAttnProcessor2_0 as CNAttnProcessor,
)
from .attention_processor import (
IPAttnProcessor2_0 as IPAttnProcessor,
)
else:
from .attention_processor import IPAttnProcessor, AttnProcessor, CNAttnProcessor
from .attention_processor import AttnProcessor, CNAttnProcessor, IPAttnProcessor
from .resampler import Resampler
class ImageProjModel(torch.nn.Module):
"""Projection Model"""
def __init__(self, cross_attention_dim=1024, clip_embeddings_dim=1024, clip_extra_context_tokens=4):
super().__init__()
self.cross_attention_dim = cross_attention_dim
self.clip_extra_context_tokens = clip_extra_context_tokens
self.proj = torch.nn.Linear(clip_embeddings_dim, self.clip_extra_context_tokens * cross_attention_dim)
self.norm = torch.nn.LayerNorm(cross_attention_dim)
def forward(self, image_embeds):
embeds = image_embeds
clip_extra_context_tokens = self.proj(embeds).reshape(-1, self.clip_extra_context_tokens, self.cross_attention_dim)
clip_extra_context_tokens = self.proj(embeds).reshape(
-1, self.clip_extra_context_tokens, self.cross_attention_dim
)
clip_extra_context_tokens = self.norm(clip_extra_context_tokens)
return clip_extra_context_tokens
class IPAdapter:
def __init__(self, sd_pipe, image_encoder_path, ip_ckpt, device, num_tokens=4):
self.device = device
self.image_encoder_path = image_encoder_path
self.ip_ckpt = ip_ckpt
self.num_tokens = num_tokens
self.pipe = sd_pipe.to(self.device)
self.set_ip_adapter()
# load image encoder
self.image_encoder = CLIPVisionModelWithProjection.from_pretrained(self.image_encoder_path).to(self.device, dtype=torch.float16)
self.image_encoder = CLIPVisionModelWithProjection.from_pretrained(self.image_encoder_path).to(
self.device, dtype=torch.float16
)
self.clip_image_processor = CLIPImageProcessor()
# image proj model
self.image_proj_model = self.init_proj()
self.load_ip_adapter()
def init_proj(self):
image_proj_model = ImageProjModel(
cross_attention_dim=self.pipe.unet.config.cross_attention_dim,
@ -60,7 +72,7 @@ class IPAdapter:
clip_extra_context_tokens=self.num_tokens,
).to(self.device, dtype=torch.float16)
return image_proj_model
def set_ip_adapter(self):
unet = self.pipe.unet
attn_procs = {}
@ -77,8 +89,12 @@ class IPAdapter:
if cross_attention_dim is None:
attn_procs[name] = AttnProcessor()
else:
attn_procs[name] = IPAttnProcessor(hidden_size=hidden_size, cross_attention_dim=cross_attention_dim,
scale=1.0,num_tokens= self.num_tokens).to(self.device, dtype=torch.float16)
attn_procs[name] = IPAttnProcessor(
hidden_size=hidden_size,
cross_attention_dim=cross_attention_dim,
scale=1.0,
num_tokens=self.num_tokens,
).to(self.device, dtype=torch.float16)
unet.set_attn_processor(attn_procs)
if hasattr(self.pipe, "controlnet"):
if isinstance(self.pipe.controlnet, MultiControlNetModel):
@ -86,7 +102,7 @@ class IPAdapter:
controlnet.set_attn_processor(CNAttnProcessor(num_tokens=self.num_tokens))
else:
self.pipe.controlnet.set_attn_processor(CNAttnProcessor(num_tokens=self.num_tokens))
def load_ip_adapter(self):
if os.path.splitext(self.ip_ckpt)[-1] == ".safetensors":
state_dict = {"image_proj": {}, "ip_adapter": {}}
@ -101,7 +117,7 @@ class IPAdapter:
self.image_proj_model.load_state_dict(state_dict["image_proj"])
ip_layers = torch.nn.ModuleList(self.pipe.unet.attn_processors.values())
ip_layers.load_state_dict(state_dict["ip_adapter"])
@torch.inference_mode()
def get_image_embeds(self, pil_image=None, clip_image_embeds=None):
if pil_image is not None:
@ -114,12 +130,12 @@ class IPAdapter:
image_prompt_embeds = self.image_proj_model(clip_image_embeds)
uncond_image_prompt_embeds = self.image_proj_model(torch.zeros_like(clip_image_embeds))
return image_prompt_embeds, uncond_image_prompt_embeds
def set_scale(self, scale):
for attn_processor in self.pipe.unet.attn_processors.values():
if isinstance(attn_processor, IPAttnProcessor):
attn_processor.scale = scale
def generate(
self,
pil_image=None,
@ -136,24 +152,23 @@ class IPAdapter:
self.set_scale(scale)
if pil_image is not None:
if isinstance(pil_image, Image.Image):
num_prompts = 1
else:
num_prompts = len(pil_image)
num_prompts = 1 if isinstance(pil_image, Image.Image) else len(pil_image)
else:
num_prompts = clip_image_embeds.size(0)
if prompt is None:
prompt = "best quality, high quality"
if negative_prompt is None:
negative_prompt = "monochrome, lowres, bad anatomy, worst quality, low quality"
if not isinstance(prompt, List):
prompt = [prompt] * num_prompts
if not isinstance(negative_prompt, List):
negative_prompt = [negative_prompt] * num_prompts
image_prompt_embeds, uncond_image_prompt_embeds = self.get_image_embeds(pil_image=pil_image, clip_image_embeds=clip_image_embeds)
image_prompt_embeds, uncond_image_prompt_embeds = self.get_image_embeds(
pil_image=pil_image, clip_image_embeds=clip_image_embeds
)
bs_embed, seq_len, _ = image_prompt_embeds.shape
image_prompt_embeds = image_prompt_embeds.repeat(1, num_samples, 1)
image_prompt_embeds = image_prompt_embeds.view(bs_embed * num_samples, seq_len, -1)
@ -162,11 +177,16 @@ class IPAdapter:
with torch.inference_mode():
prompt_embeds = self.pipe._encode_prompt(
prompt, device=self.device, num_images_per_prompt=num_samples, do_classifier_free_guidance=True, negative_prompt=negative_prompt)
prompt,
device=self.device,
num_images_per_prompt=num_samples,
do_classifier_free_guidance=True,
negative_prompt=negative_prompt,
)
negative_prompt_embeds_, prompt_embeds_ = prompt_embeds.chunk(2)
prompt_embeds = torch.cat([prompt_embeds_, image_prompt_embeds], dim=1)
negative_prompt_embeds = torch.cat([negative_prompt_embeds_, uncond_image_prompt_embeds], dim=1)
generator = torch.Generator(self.device).manual_seed(seed) if seed is not None else None
images = self.pipe(
prompt_embeds=prompt_embeds,
@ -176,13 +196,13 @@ class IPAdapter:
generator=generator,
**kwargs,
).images
return images
class IPAdapterXL(IPAdapter):
"""SDXL"""
def generate(
self,
pil_image,
@ -195,22 +215,19 @@ class IPAdapterXL(IPAdapter):
**kwargs,
):
self.set_scale(scale)
if isinstance(pil_image, Image.Image):
num_prompts = 1
else:
num_prompts = len(pil_image)
num_prompts = 1 if isinstance(pil_image, Image.Image) else len(pil_image)
if prompt is None:
prompt = "best quality, high quality"
if negative_prompt is None:
negative_prompt = "monochrome, lowres, bad anatomy, worst quality, low quality"
if not isinstance(prompt, List):
prompt = [prompt] * num_prompts
if not isinstance(negative_prompt, List):
negative_prompt = [negative_prompt] * num_prompts
image_prompt_embeds, uncond_image_prompt_embeds = self.get_image_embeds(pil_image)
bs_embed, seq_len, _ = image_prompt_embeds.shape
image_prompt_embeds = image_prompt_embeds.repeat(1, num_samples, 1)
@ -219,11 +236,20 @@ class IPAdapterXL(IPAdapter):
uncond_image_prompt_embeds = uncond_image_prompt_embeds.view(bs_embed * num_samples, seq_len, -1)
with torch.inference_mode():
prompt_embeds, negative_prompt_embeds, pooled_prompt_embeds, negative_pooled_prompt_embeds = self.pipe.encode_prompt(
prompt, num_images_per_prompt=num_samples, do_classifier_free_guidance=True, negative_prompt=negative_prompt)
(
prompt_embeds,
negative_prompt_embeds,
pooled_prompt_embeds,
negative_pooled_prompt_embeds,
) = self.pipe.encode_prompt(
prompt,
num_images_per_prompt=num_samples,
do_classifier_free_guidance=True,
negative_prompt=negative_prompt,
)
prompt_embeds = torch.cat([prompt_embeds, image_prompt_embeds], dim=1)
negative_prompt_embeds = torch.cat([negative_prompt_embeds, uncond_image_prompt_embeds], dim=1)
generator = torch.Generator(self.device).manual_seed(seed) if seed is not None else None
images = self.pipe(
prompt_embeds=prompt_embeds,
@ -234,10 +260,10 @@ class IPAdapterXL(IPAdapter):
generator=generator,
**kwargs,
).images
return images
class IPAdapterPlus(IPAdapter):
"""IP-Adapter with fine-grained features"""
@ -250,10 +276,10 @@ class IPAdapterPlus(IPAdapter):
num_queries=self.num_tokens,
embedding_dim=self.image_encoder.config.hidden_size,
output_dim=self.pipe.unet.config.cross_attention_dim,
ff_mult=4
ff_mult=4,
).to(self.device, dtype=torch.float16)
return image_proj_model
@torch.inference_mode()
def get_image_embeds(self, pil_image=None, clip_image_embeds=None):
if isinstance(pil_image, Image.Image):
@ -262,7 +288,9 @@ class IPAdapterPlus(IPAdapter):
clip_image = clip_image.to(self.device, dtype=torch.float16)
clip_image_embeds = self.image_encoder(clip_image, output_hidden_states=True).hidden_states[-2]
image_prompt_embeds = self.image_proj_model(clip_image_embeds)
uncond_clip_image_embeds = self.image_encoder(torch.zeros_like(clip_image), output_hidden_states=True).hidden_states[-2]
uncond_clip_image_embeds = self.image_encoder(
torch.zeros_like(clip_image), output_hidden_states=True
).hidden_states[-2]
uncond_image_prompt_embeds = self.image_proj_model(uncond_clip_image_embeds)
return image_prompt_embeds, uncond_image_prompt_embeds
@ -279,10 +307,10 @@ class IPAdapterPlusXL(IPAdapter):
num_queries=self.num_tokens,
embedding_dim=self.image_encoder.config.hidden_size,
output_dim=self.pipe.unet.config.cross_attention_dim,
ff_mult=4
ff_mult=4,
).to(self.device, dtype=torch.float16)
return image_proj_model
@torch.inference_mode()
def get_image_embeds(self, pil_image):
if isinstance(pil_image, Image.Image):
@ -291,10 +319,12 @@ class IPAdapterPlusXL(IPAdapter):
clip_image = clip_image.to(self.device, dtype=torch.float16)
clip_image_embeds = self.image_encoder(clip_image, output_hidden_states=True).hidden_states[-2]
image_prompt_embeds = self.image_proj_model(clip_image_embeds)
uncond_clip_image_embeds = self.image_encoder(torch.zeros_like(clip_image), output_hidden_states=True).hidden_states[-2]
uncond_clip_image_embeds = self.image_encoder(
torch.zeros_like(clip_image), output_hidden_states=True
).hidden_states[-2]
uncond_image_prompt_embeds = self.image_proj_model(uncond_clip_image_embeds)
return image_prompt_embeds, uncond_image_prompt_embeds
def generate(
self,
pil_image,
@ -307,22 +337,19 @@ class IPAdapterPlusXL(IPAdapter):
**kwargs,
):
self.set_scale(scale)
if isinstance(pil_image, Image.Image):
num_prompts = 1
else:
num_prompts = len(pil_image)
num_prompts = 1 if isinstance(pil_image, Image.Image) else len(pil_image)
if prompt is None:
prompt = "best quality, high quality"
if negative_prompt is None:
negative_prompt = "monochrome, lowres, bad anatomy, worst quality, low quality"
if not isinstance(prompt, List):
prompt = [prompt] * num_prompts
if not isinstance(negative_prompt, List):
negative_prompt = [negative_prompt] * num_prompts
image_prompt_embeds, uncond_image_prompt_embeds = self.get_image_embeds(pil_image)
bs_embed, seq_len, _ = image_prompt_embeds.shape
image_prompt_embeds = image_prompt_embeds.repeat(1, num_samples, 1)
@ -331,11 +358,20 @@ class IPAdapterPlusXL(IPAdapter):
uncond_image_prompt_embeds = uncond_image_prompt_embeds.view(bs_embed * num_samples, seq_len, -1)
with torch.inference_mode():
prompt_embeds, negative_prompt_embeds, pooled_prompt_embeds, negative_pooled_prompt_embeds = self.pipe.encode_prompt(
prompt, num_images_per_prompt=num_samples, do_classifier_free_guidance=True, negative_prompt=negative_prompt)
(
prompt_embeds,
negative_prompt_embeds,
pooled_prompt_embeds,
negative_pooled_prompt_embeds,
) = self.pipe.encode_prompt(
prompt,
num_images_per_prompt=num_samples,
do_classifier_free_guidance=True,
negative_prompt=negative_prompt,
)
prompt_embeds = torch.cat([prompt_embeds, image_prompt_embeds], dim=1)
negative_prompt_embeds = torch.cat([negative_prompt_embeds, uncond_image_prompt_embeds], dim=1)
generator = torch.Generator(self.device).manual_seed(seed) if seed is not None else None
images = self.pipe(
prompt_embeds=prompt_embeds,
@ -346,5 +382,5 @@ class IPAdapterPlusXL(IPAdapter):
generator=generator,
**kwargs,
).images
return images

30
ip_adapter/resampler.py
View File

@ -14,11 +14,11 @@ def FeedForward(dim, mult=4):
nn.GELU(),
nn.Linear(inner_dim, dim, bias=False),
)
def reshape_tensor(x, heads):
bs, length, width = x.shape
#(bs, length, width) --> (bs, length, n_heads, dim_per_head)
# (bs, length, width) --> (bs, length, n_heads, dim_per_head)
x = x.view(bs, length, heads, -1)
# (bs, length, n_heads, dim_per_head) --> (bs, n_heads, length, dim_per_head)
x = x.transpose(1, 2)
@ -42,7 +42,6 @@ class PerceiverAttention(nn.Module):
self.to_kv = nn.Linear(dim, inner_dim * 2, bias=False)
self.to_out = nn.Linear(inner_dim, dim, bias=False)
def forward(self, x, latents):
"""
Args:
@ -53,23 +52,23 @@ class PerceiverAttention(nn.Module):
"""
x = self.norm1(x)
latents = self.norm2(latents)
b, l, _ = latents.shape
q = self.to_q(latents)
kv_input = torch.cat((x, latents), dim=-2)
k, v = self.to_kv(kv_input).chunk(2, dim=-1)
q = reshape_tensor(q, self.heads)
k = reshape_tensor(k, self.heads)
v = reshape_tensor(v, self.heads)
# attention
scale = 1 / math.sqrt(math.sqrt(self.dim_head))
weight = (q * scale) @ (k * scale).transpose(-2, -1) # More stable with f16 than dividing afterwards
weight = (q * scale) @ (k * scale).transpose(-2, -1) # More stable with f16 than dividing afterwards
weight = torch.softmax(weight.float(), dim=-1).type(weight.dtype)
out = weight @ v
out = out.permute(0, 2, 1, 3).reshape(b, l, -1)
return self.to_out(out)
@ -88,14 +87,14 @@ class Resampler(nn.Module):
ff_mult=4,
):
super().__init__()
self.latents = nn.Parameter(torch.randn(1, num_queries, dim) / dim**0.5)
self.proj_in = nn.Linear(embedding_dim, dim)
self.proj_out = nn.Linear(dim, output_dim)
self.norm_out = nn.LayerNorm(output_dim)
self.layers = nn.ModuleList([])
for _ in range(depth):
self.layers.append(
@ -108,14 +107,13 @@ class Resampler(nn.Module):
)
def forward(self, x):
latents = self.latents.repeat(x.size(0), 1, 1)
x = self.proj_in(x)
for attn, ff in self.layers:
latents = attn(x, latents) + latents
latents = ff(latents) + latents
latents = self.proj_out(latents)
return self.norm_out(latents)
return self.norm_out(latents)