Delete ip_adapter

ip_adapter/__init__.py

@@ -1 +0,0 @@
-from .ip_adapter import IPAdapter, IPAdapterXL, IPAdapterPlus

ip_adapter/attention_processor.py

@@ -1,553 +0,0 @@
-# modified from https://github.com/huggingface/diffusers/blob/main/src/diffusers/models/attention_processor.py
-import torch
-import torch.nn as nn
-import torch.nn.functional as F
-
-
-class AttnProcessor(nn.Module):
-    r"""
-    Default processor for performing attention-related computations.
-    """
-    def __init__(
-        self,
-        hidden_size=None,
-        cross_attention_dim=None,
-    ):
-        super().__init__()
-
-    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:
-            hidden_states = attn.spatial_norm(hidden_states, temb)
-
-        input_ndim = hidden_states.ndim
-
-        if input_ndim == 4:
-            batch_size, channel, height, width = hidden_states.shape
-            hidden_states = hidden_states.view(batch_size, channel, height * width).transpose(1, 2)
-
-        batch_size, sequence_length, _ = (
-            hidden_states.shape if encoder_hidden_states is None else encoder_hidden_states.shape
-        )
-        attention_mask = attn.prepare_attention_mask(attention_mask, sequence_length, batch_size)
-
-        if attn.group_norm is not None:
-            hidden_states = attn.group_norm(hidden_states.transpose(1, 2)).transpose(1, 2)
-
-        query = attn.to_q(hidden_states)
-
-        if encoder_hidden_states is None:
-            encoder_hidden_states = hidden_states
-        elif attn.norm_cross:
-            encoder_hidden_states = attn.norm_encoder_hidden_states(encoder_hidden_states)
-
-        key = attn.to_k(encoder_hidden_states)
-        value = attn.to_v(encoder_hidden_states)
-
-        query = attn.head_to_batch_dim(query)
-        key = attn.head_to_batch_dim(key)
-        value = attn.head_to_batch_dim(value)
-
-        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)
-
-        # linear proj
-        hidden_states = attn.to_out[0](hidden_states)
-        # dropout
-        hidden_states = attn.to_out[1](hidden_states)
-
-        if input_ndim == 4:
-            hidden_states = hidden_states.transpose(-1, -2).reshape(batch_size, channel, height, width)
-
-        if attn.residual_connection:
-            hidden_states = hidden_states + residual
-
-        hidden_states = hidden_states / attn.rescale_output_factor
-
-        return hidden_states
-    
-    
-class IPAttnProcessor(nn.Module):
-    r"""
-    Attention processor for IP-Adapater.
-    Args:
-        hidden_size (`int`):
-            The hidden size of the attention layer.
-        cross_attention_dim (`int`):
-            The number of channels in the `encoder_hidden_states`.
-        scale (`float`, defaults to 1.0):
-            the weight scale of image prompt.
-        num_tokens (`int`, defaults to 4 when do ip_adapter_plus it should be 16):
-            The context length of the image features.
-    """
-
-    def __init__(self, hidden_size, cross_attention_dim=None, scale=1.0, num_tokens=4):
-        super().__init__()
-
-        self.hidden_size = hidden_size
-        self.cross_attention_dim = cross_attention_dim
-        self.scale = scale
-        self.num_tokens = num_tokens
-
-        self.to_k_ip = nn.Linear(cross_attention_dim or hidden_size, hidden_size, bias=False)
-        self.to_v_ip = nn.Linear(cross_attention_dim or hidden_size, hidden_size, bias=False)
-
-    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:
-            hidden_states = attn.spatial_norm(hidden_states, temb)
-
-        input_ndim = hidden_states.ndim
-
-        if input_ndim == 4:
-            batch_size, channel, height, width = hidden_states.shape
-            hidden_states = hidden_states.view(batch_size, channel, height * width).transpose(1, 2)
-
-        batch_size, sequence_length, _ = (
-            hidden_states.shape if encoder_hidden_states is None else encoder_hidden_states.shape
-        )
-        attention_mask = attn.prepare_attention_mask(attention_mask, sequence_length, batch_size)
-
-        if attn.group_norm is not None:
-            hidden_states = attn.group_norm(hidden_states.transpose(1, 2)).transpose(1, 2)
-
-        query = attn.to_q(hidden_states)
-
-        if encoder_hidden_states is None:
-            encoder_hidden_states = hidden_states
-        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:, :]
-            if attn.norm_cross:
-                encoder_hidden_states = attn.norm_encoder_hidden_states(encoder_hidden_states)
-
-        key = attn.to_k(encoder_hidden_states)
-        value = attn.to_v(encoder_hidden_states)
-
-        query = attn.head_to_batch_dim(query)
-        key = attn.head_to_batch_dim(key)
-        value = attn.head_to_batch_dim(value)
-
-        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
-        hidden_states = attn.to_out[0](hidden_states)
-        # dropout
-        hidden_states = attn.to_out[1](hidden_states)
-
-        if input_ndim == 4:
-            hidden_states = hidden_states.transpose(-1, -2).reshape(batch_size, channel, height, width)
-
-        if attn.residual_connection:
-            hidden_states = hidden_states + residual
-
-        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,
-        cross_attention_dim=None,
-    ):
-        super().__init__()
-        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.")
-
-    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:
-            hidden_states = attn.spatial_norm(hidden_states, temb)
-
-        input_ndim = hidden_states.ndim
-
-        if input_ndim == 4:
-            batch_size, channel, height, width = hidden_states.shape
-            hidden_states = hidden_states.view(batch_size, channel, height * width).transpose(1, 2)
-
-        batch_size, sequence_length, _ = (
-            hidden_states.shape if encoder_hidden_states is None else encoder_hidden_states.shape
-        )
-
-        if attention_mask is not None:
-            attention_mask = attn.prepare_attention_mask(attention_mask, sequence_length, batch_size)
-            # scaled_dot_product_attention expects attention_mask shape to be
-            # (batch, heads, source_length, target_length)
-            attention_mask = attention_mask.view(batch_size, attn.heads, -1, attention_mask.shape[-1])
-
-        if attn.group_norm is not None:
-            hidden_states = attn.group_norm(hidden_states.transpose(1, 2)).transpose(1, 2)
-
-        query = attn.to_q(hidden_states)
-
-        if encoder_hidden_states is None:
-            encoder_hidden_states = hidden_states
-        elif attn.norm_cross:
-            encoder_hidden_states = attn.norm_encoder_hidden_states(encoder_hidden_states)
-
-        key = attn.to_k(encoder_hidden_states)
-        value = attn.to_v(encoder_hidden_states)
-
-        inner_dim = key.shape[-1]
-        head_dim = inner_dim // attn.heads
-
-        query = query.view(batch_size, -1, attn.heads, head_dim).transpose(1, 2)
-
-        key = key.view(batch_size, -1, attn.heads, head_dim).transpose(1, 2)
-        value = value.view(batch_size, -1, attn.heads, head_dim).transpose(1, 2)
-
-        # the output of sdp = (batch, num_heads, seq_len, head_dim)
-        # TODO: add support for attn.scale when we move to Torch 2.1
-        hidden_states = F.scaled_dot_product_attention(
-            query, key, value, attn_mask=attention_mask, dropout_p=0.0, is_causal=False
-        )
-
-        hidden_states = hidden_states.transpose(1, 2).reshape(batch_size, -1, attn.heads * head_dim)
-        hidden_states = hidden_states.to(query.dtype)
-
-        # linear proj
-        hidden_states = attn.to_out[0](hidden_states)
-        # dropout
-        hidden_states = attn.to_out[1](hidden_states)
-
-        if input_ndim == 4:
-            hidden_states = hidden_states.transpose(-1, -2).reshape(batch_size, channel, height, width)
-
-        if attn.residual_connection:
-            hidden_states = hidden_states + residual
-
-        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.
-    Args:
-        hidden_size (`int`):
-            The hidden size of the attention layer.
-        cross_attention_dim (`int`):
-            The number of channels in the `encoder_hidden_states`.
-        scale (`float`, defaults to 1.0):
-            the weight scale of image prompt.
-        num_tokens (`int`, defaults to 4 when do ip_adapter_plus it should be 16):
-            The context length of the image features.
-    """
-
-    def __init__(self, hidden_size, cross_attention_dim=None, scale=1.0, num_tokens=4):
-        super().__init__()
-
-        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.hidden_size = hidden_size
-        self.cross_attention_dim = cross_attention_dim
-        self.scale = scale
-        self.num_tokens = num_tokens
-
-        self.to_k_ip = nn.Linear(cross_attention_dim or hidden_size, hidden_size, bias=False)
-        self.to_v_ip = nn.Linear(cross_attention_dim or hidden_size, hidden_size, bias=False)
-
-    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:
-            hidden_states = attn.spatial_norm(hidden_states, temb)
-
-        input_ndim = hidden_states.ndim
-
-        if input_ndim == 4:
-            batch_size, channel, height, width = hidden_states.shape
-            hidden_states = hidden_states.view(batch_size, channel, height * width).transpose(1, 2)
-
-        batch_size, sequence_length, _ = (
-            hidden_states.shape if encoder_hidden_states is None else encoder_hidden_states.shape
-        )
-
-        if attention_mask is not None:
-            attention_mask = attn.prepare_attention_mask(attention_mask, sequence_length, batch_size)
-            # scaled_dot_product_attention expects attention_mask shape to be
-            # (batch, heads, source_length, target_length)
-            attention_mask = attention_mask.view(batch_size, attn.heads, -1, attention_mask.shape[-1])
-
-        if attn.group_norm is not None:
-            hidden_states = attn.group_norm(hidden_states.transpose(1, 2)).transpose(1, 2)
-
-        query = attn.to_q(hidden_states)
-
-        if encoder_hidden_states is None:
-            encoder_hidden_states = hidden_states
-        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:, :]
-            if attn.norm_cross:
-                encoder_hidden_states = attn.norm_encoder_hidden_states(encoder_hidden_states)
-
-        key = attn.to_k(encoder_hidden_states)
-        value = attn.to_v(encoder_hidden_states)
-
-        inner_dim = key.shape[-1]
-        head_dim = inner_dim // attn.heads
-
-        query = query.view(batch_size, -1, attn.heads, head_dim).transpose(1, 2)
-
-        key = key.view(batch_size, -1, attn.heads, head_dim).transpose(1, 2)
-        value = value.view(batch_size, -1, attn.heads, head_dim).transpose(1, 2)
-
-        # the output of sdp = (batch, num_heads, seq_len, head_dim)
-        # TODO: add support for attn.scale when we move to Torch 2.1
-        hidden_states = F.scaled_dot_product_attention(
-            query, key, value, attn_mask=attention_mask, dropout_p=0.0, is_causal=False
-        )
-
-        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)
-
-        # the output of sdp = (batch, num_heads, seq_len, head_dim)
-        # TODO: add support for attn.scale when we move to Torch 2.1
-        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
-        hidden_states = attn.to_out[0](hidden_states)
-        # dropout
-        hidden_states = attn.to_out[1](hidden_states)
-
-        if input_ndim == 4:
-            hidden_states = hidden_states.transpose(-1, -2).reshape(batch_size, channel, height, width)
-
-        if attn.residual_connection:
-            hidden_states = hidden_states + residual
-
-        hidden_states = hidden_states / attn.rescale_output_factor
-
-        return hidden_states
-
-
-## for controlnet
-class CNAttnProcessor:
-    r"""
-    Default processor for performing attention-related computations.
-    """
-
-    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
-    ):
-        residual = hidden_states
-
-        if attn.spatial_norm is not None:
-            hidden_states = attn.spatial_norm(hidden_states, temb)
-
-        input_ndim = hidden_states.ndim
-
-        if input_ndim == 4:
-            batch_size, channel, height, width = hidden_states.shape
-            hidden_states = hidden_states.view(batch_size, channel, height * width).transpose(1, 2)
-
-        batch_size, sequence_length, _ = (
-            hidden_states.shape if encoder_hidden_states is None else encoder_hidden_states.shape
-        )
-        attention_mask = attn.prepare_attention_mask(attention_mask, sequence_length, batch_size)
-
-        if attn.group_norm is not None:
-            hidden_states = attn.group_norm(hidden_states.transpose(1, 2)).transpose(1, 2)
-
-        query = attn.to_q(hidden_states)
-
-        if encoder_hidden_states is None:
-            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
-            if attn.norm_cross:
-                encoder_hidden_states = attn.norm_encoder_hidden_states(encoder_hidden_states)
-
-        key = attn.to_k(encoder_hidden_states)
-        value = attn.to_v(encoder_hidden_states)
-
-        query = attn.head_to_batch_dim(query)
-        key = attn.head_to_batch_dim(key)
-        value = attn.head_to_batch_dim(value)
-
-        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)
-
-        # linear proj
-        hidden_states = attn.to_out[0](hidden_states)
-        # dropout
-        hidden_states = attn.to_out[1](hidden_states)
-
-        if input_ndim == 4:
-            hidden_states = hidden_states.transpose(-1, -2).reshape(batch_size, channel, height, width)
-
-        if attn.residual_connection:
-            hidden_states = hidden_states + residual
-
-        hidden_states = hidden_states / attn.rescale_output_factor
-
-        return hidden_states
-
-
-class CNAttnProcessor2_0:
-    r"""
-    Processor for implementing scaled dot-product attention (enabled by default if you're using PyTorch 2.0).
-    """
-
-    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
-
-    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:
-            hidden_states = attn.spatial_norm(hidden_states, temb)
-
-        input_ndim = hidden_states.ndim
-
-        if input_ndim == 4:
-            batch_size, channel, height, width = hidden_states.shape
-            hidden_states = hidden_states.view(batch_size, channel, height * width).transpose(1, 2)
-
-        batch_size, sequence_length, _ = (
-            hidden_states.shape if encoder_hidden_states is None else encoder_hidden_states.shape
-        )
-
-        if attention_mask is not None:
-            attention_mask = attn.prepare_attention_mask(attention_mask, sequence_length, batch_size)
-            # scaled_dot_product_attention expects attention_mask shape to be
-            # (batch, heads, source_length, target_length)
-            attention_mask = attention_mask.view(batch_size, attn.heads, -1, attention_mask.shape[-1])
-
-        if attn.group_norm is not None:
-            hidden_states = attn.group_norm(hidden_states.transpose(1, 2)).transpose(1, 2)
-
-        query = attn.to_q(hidden_states)
-
-        if encoder_hidden_states is None:
-            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
-            if attn.norm_cross:
-                encoder_hidden_states = attn.norm_encoder_hidden_states(encoder_hidden_states)
-
-        key = attn.to_k(encoder_hidden_states)
-        value = attn.to_v(encoder_hidden_states)
-
-        inner_dim = key.shape[-1]
-        head_dim = inner_dim // attn.heads
-
-        query = query.view(batch_size, -1, attn.heads, head_dim).transpose(1, 2)
-
-        key = key.view(batch_size, -1, attn.heads, head_dim).transpose(1, 2)
-        value = value.view(batch_size, -1, attn.heads, head_dim).transpose(1, 2)
-
-        # the output of sdp = (batch, num_heads, seq_len, head_dim)
-        # TODO: add support for attn.scale when we move to Torch 2.1
-        hidden_states = F.scaled_dot_product_attention(
-            query, key, value, attn_mask=attention_mask, dropout_p=0.0, is_causal=False
-        )
-
-        hidden_states = hidden_states.transpose(1, 2).reshape(batch_size, -1, attn.heads * head_dim)
-        hidden_states = hidden_states.to(query.dtype)
-
-        # linear proj
-        hidden_states = attn.to_out[0](hidden_states)
-        # dropout
-        hidden_states = attn.to_out[1](hidden_states)
-
-        if input_ndim == 4:
-            hidden_states = hidden_states.transpose(-1, -2).reshape(batch_size, channel, height, width)
-
-        if attn.residual_connection:
-            hidden_states = hidden_states + residual
-
-        hidden_states = hidden_states / attn.rescale_output_factor
-
-        return hidden_states

ip_adapter/ip_adapter.py

@@ -1,273 +0,0 @@
-import os
-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 .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
-else:
-    from .attention_processor import IPAttnProcessor, AttnProcessor, CNAttnProcessor
-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.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.bfloat16)
-        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,
-            clip_embeddings_dim=self.image_encoder.config.projection_dim,
-            clip_extra_context_tokens=self.num_tokens,
-        ).to(self.device, dtype=torch.bfloat16)
-        return image_proj_model
-        
-    def set_ip_adapter(self):
-        unet = self.pipe.unet
-        attn_procs = {}
-        for name in unet.attn_processors.keys():
-            cross_attention_dim = None if name.endswith("attn1.processor") else unet.config.cross_attention_dim
-            if name.startswith("mid_block"):
-                hidden_size = unet.config.block_out_channels[-1]
-            elif name.startswith("up_blocks"):
-                block_id = int(name[len("up_blocks.")])
-                hidden_size = list(reversed(unet.config.block_out_channels))[block_id]
-            elif name.startswith("down_blocks"):
-                block_id = int(name[len("down_blocks.")])
-                hidden_size = unet.config.block_out_channels[block_id]
-            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.bfloat16)
-        unet.set_attn_processor(attn_procs)
-        if hasattr(self.pipe, "controlnet"):
-            if isinstance(self.pipe.controlnet, MultiControlNetModel):
-                for controlnet in self.pipe.controlnet.nets:
-                    controlnet.set_attn_processor(CNAttnProcessor(num_tokens=self.num_tokens))
-            else:
-                self.pipe.controlnet.set_attn_processor(CNAttnProcessor(num_tokens=self.num_tokens))
-
-    def update_state_dict(self, state_dict):
-        image_proj_dict = {}
-        ip_adapter_dict = {}
-
-        for k in state_dict.keys():
-            if k.startswith("image_proj_model"):
-                image_proj_dict[k.replace("image_proj_model.", "")] = state_dict[k]
-            if k.startswith("adapter_modules"):
-                ip_adapter_dict[k.replace("adapter_modules.", "")] = state_dict[k]
-
-        dict = {'image_proj': image_proj_dict,
-                'ip_adapter' : ip_adapter_dict
-                } 
-        return dict
-    
-    def load_ip_adapter(self):
-        state_dict = torch.load(self.ip_ckpt, map_location="cpu")
-        if "image_proj_model.proj.weight" in state_dict.keys():
-            state_dict = self.update_state_dict(state_dict)
-        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):
-        if isinstance(pil_image, Image.Image):
-            pil_image = [pil_image]
-        clip_image = self.clip_image_processor(images=pil_image, return_tensors="pt").pixel_values
-        clip_image_embeds = self.image_encoder(clip_image.to(self.device, dtype=torch.bfloat16)).image_embeds
-        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,
-        prompt=None,
-        negative_prompt=None,
-        scale=1.0,
-        num_samples=4,
-        seed=-1,
-        guidance_scale=7.5,
-        num_inference_steps=30,
-        **kwargs,
-    ):
-        self.set_scale(scale)
-        
-        if isinstance(pil_image, List):
-            num_prompts = len(pil_image)
-        else:
-            num_prompts = 1
-        
-        # if isinstance(pil_image, Image.Image):
-        #     num_prompts = 1
-        # else:
-        #     num_prompts = len(pil_image)
-        #     print("num promp", num_prompts)
-        
-        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)
-        image_prompt_embeds = image_prompt_embeds.view(bs_embed * num_samples, seq_len, -1)
-        uncond_image_prompt_embeds = uncond_image_prompt_embeds.repeat(1, num_samples, 1)
-        uncond_image_prompt_embeds = uncond_image_prompt_embeds.view(bs_embed * num_samples, seq_len, -1)
-
-        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)
-            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,
-            negative_prompt_embeds=negative_prompt_embeds,
-            guidance_scale=guidance_scale,
-            num_inference_steps=num_inference_steps,
-            generator=generator,
-            **kwargs,
-        ).images
-        
-        return images
-    
-    
-class IPAdapterXL(IPAdapter):
-    """SDXL"""
-    
-    def generate(
-        self,
-        pil_image,
-        prompt=None,
-        negative_prompt=None,
-        scale=1.0,
-        num_samples=4,
-        seed=-1,
-        num_inference_steps=30,
-        **kwargs,
-    ):
-        self.set_scale(scale)
-        
-        if isinstance(pil_image, Image.Image):
-            num_prompts = 1
-        else:
-            num_prompts = 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)
-        image_prompt_embeds = image_prompt_embeds.view(bs_embed * num_samples, seq_len, -1)
-        uncond_image_prompt_embeds = uncond_image_prompt_embeds.repeat(1, num_samples, 1)
-        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 = 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,
-            negative_prompt_embeds=negative_prompt_embeds,
-            pooled_prompt_embeds=pooled_prompt_embeds,
-            negative_pooled_prompt_embeds=negative_pooled_prompt_embeds,
-            num_inference_steps=num_inference_steps,
-            generator=generator,
-            **kwargs,
-        ).images
-        
-        return images
-    
-    
-class IPAdapterPlus(IPAdapter):
-    """IP-Adapter with fine-grained features"""
-    
-    def init_proj(self):
-        image_proj_model = Resampler(
-            dim=self.pipe.unet.config.cross_attention_dim,
-            depth=4,
-            dim_head=64,
-            heads=12,
-            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
-        ).to(self.device, dtype=torch.bfloat16)
-        return image_proj_model
-    
-    @torch.inference_mode()
-    def get_image_embeds(self, pil_image):
-        if isinstance(pil_image, Image.Image):
-            pil_image = [pil_image]
-        clip_image = self.clip_image_processor(images=pil_image, return_tensors="pt").pixel_values
-        clip_image = clip_image.to(self.device, dtype=torch.bfloat16)
-        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_image_prompt_embeds = self.image_proj_model(uncond_clip_image_embeds)
-        return image_prompt_embeds, uncond_image_prompt_embeds

ip_adapter/resampler.py

@@ -1,121 +0,0 @@
-# modified from https://github.com/mlfoundations/open_flamingo/blob/main/open_flamingo/src/helpers.py
-import math
-
-import torch
-import torch.nn as nn
-
-
-# FFN
-def FeedForward(dim, mult=4):
-    inner_dim = int(dim * mult)
-    return nn.Sequential(
-        nn.LayerNorm(dim),
-        nn.Linear(dim, inner_dim, bias=False),
-        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)
-    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)
-    # (bs, n_heads, length, dim_per_head) --> (bs*n_heads, length, dim_per_head)
-    x = x.reshape(bs, heads, length, -1)
-    return x
-
-
-class PerceiverAttention(nn.Module):
-    def __init__(self, *, dim, dim_head=64, heads=8):
-        super().__init__()
-        self.scale = dim_head**-0.5
-        self.dim_head = dim_head
-        self.heads = heads
-        inner_dim = dim_head * heads
-
-        self.norm1 = nn.LayerNorm(dim)
-        self.norm2 = nn.LayerNorm(dim)
-
-        self.to_q = nn.Linear(dim, inner_dim, bias=False)
-        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:
-            x (torch.Tensor): image features
-                shape (b, n1, D)
-            latent (torch.Tensor): latent features
-                shape (b, n2, D)
-        """
-        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 = 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)
-
-
-class Resampler(nn.Module):
-    def __init__(
-        self,
-        dim=1024,
-        depth=8,
-        dim_head=64,
-        heads=16,
-        num_queries=8,
-        embedding_dim=768,
-        output_dim=1024,
-        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(
-                nn.ModuleList(
-                    [
-                        PerceiverAttention(dim=dim, dim_head=dim_head, heads=heads),
-                        FeedForward(dim=dim, mult=ff_mult),
-                    ]
-                )
-            )
-
-    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)

ip_adapter/utils.py

@@ -1,5 +0,0 @@
-import torch.nn.functional as F
-
-
-def is_torch2_available():
-    return hasattr(F, "scaled_dot_product_attention")