Merge pull request #38 from huggingface/one_attentino_module

Unify attention modules

src/diffusers/models/attention2d.py → src/diffusers/models/attention.py

 import math
 
 import torch
-import torch.nn.functional as F
 from torch import nn
 
 
 # unet_grad_tts.py
+# TODO(Patrick) - weird linear attention layer. Check with: https://github.com/huawei-noah/Speech-Backbones/issues/15
 class LinearAttention(torch.nn.Module):
     def __init__(self, dim, heads=4, dim_head=32):
         super(LinearAttention, self).__init__()
@@ -18,7 +18,6 @@ class LinearAttention(torch.nn.Module):
     def forward(self, x):
         b, c, h, w = x.shape
         qkv = self.to_qkv(x)
-        #        q, k, v = rearrange(qkv, "b (qkv heads c) h w -> qkv b heads c (h w)", heads=self.heads, qkv=3)
         q, k, v = (
             qkv.reshape(b, 3, self.heads, self.dim_head, h, w)
             .permute(1, 0, 2, 3, 4, 5)
@@ -27,12 +26,11 @@ class LinearAttention(torch.nn.Module):
         k = k.softmax(dim=-1)
         context = torch.einsum("bhdn,bhen->bhde", k, v)
         out = torch.einsum("bhde,bhdn->bhen", context, q)
-        #        out = rearrange(out, "b heads c (h w) -> b (heads c) h w", heads=self.heads, h=h, w=w)
         out = out.reshape(b, self.heads, self.dim_head, h, w).reshape(b, self.heads * self.dim_head, h, w)
         return self.to_out(out)
 
 
-# unet_glide.py & unet_ldm.py
+# the main attention block that is used for all models
 class AttentionBlock(nn.Module):
     """
     An attention block that allows spatial positions to attend to each other.
@@ -46,10 +44,13 @@ class AttentionBlock(nn.Module):
         channels,
         num_heads=1,
         num_head_channels=-1,
+        num_groups=32,
         use_checkpoint=False,
         encoder_channels=None,
         use_new_attention_order=False,  # TODO(Patrick) -> is never used, maybe delete?
         overwrite_qkv=False,
+        overwrite_linear=False,
+        rescale_output_factor=1.0,
     ):
         super().__init__()
         self.channels = channels
@@ -62,41 +63,67 @@ class AttentionBlock(nn.Module):
             self.num_heads = channels // num_head_channels
 
         self.use_checkpoint = use_checkpoint
-        self.norm = normalization(channels, swish=0.0)
-        self.qkv = conv_nd(1, channels, channels * 3, 1)
+        self.norm = nn.GroupNorm(num_channels=channels, num_groups=num_groups, eps=1e-5, affine=True)
+        self.qkv = nn.Conv1d(channels, channels * 3, 1)
         self.n_heads = self.num_heads
+        self.rescale_output_factor = rescale_output_factor
 
         if encoder_channels is not None:
-            self.encoder_kv = conv_nd(1, encoder_channels, channels * 2, 1)
+            self.encoder_kv = nn.Conv1d(encoder_channels, channels * 2, 1)
 
-        self.proj_out = zero_module(conv_nd(1, channels, channels, 1))
+        self.proj_out = zero_module(nn.Conv1d(channels, channels, 1))
 
         self.overwrite_qkv = overwrite_qkv
         if overwrite_qkv:
             in_channels = channels
+            self.norm = nn.GroupNorm(num_channels=channels, num_groups=num_groups, eps=1e-6)
             self.q = torch.nn.Conv2d(in_channels, in_channels, kernel_size=1, stride=1, padding=0)
             self.k = torch.nn.Conv2d(in_channels, in_channels, kernel_size=1, stride=1, padding=0)
             self.v = torch.nn.Conv2d(in_channels, in_channels, kernel_size=1, stride=1, padding=0)
             self.proj_out = torch.nn.Conv2d(in_channels, in_channels, kernel_size=1, stride=1, padding=0)
 
+        self.overwrite_linear = overwrite_linear
+        if self.overwrite_linear:
+            num_groups = min(channels // 4, 32)
+            self.norm = nn.GroupNorm(num_channels=channels, num_groups=num_groups, eps=1e-6)
+            self.NIN_0 = NIN(channels, channels)
+            self.NIN_1 = NIN(channels, channels)
+            self.NIN_2 = NIN(channels, channels)
+            self.NIN_3 = NIN(channels, channels)
+
+            self.GroupNorm_0 = nn.GroupNorm(num_groups=num_groups, num_channels=channels, eps=1e-6)
+
         self.is_overwritten = False
 
     def set_weights(self, module):
         if self.overwrite_qkv:
-            qkv_weight = torch.cat([module.q.weight.data, module.k.weight.data, module.v.weight.data], dim=0)[:, :, :, 0]
+            qkv_weight = torch.cat([module.q.weight.data, module.k.weight.data, module.v.weight.data], dim=0)[
+                :, :, :, 0
+            ]
             qkv_bias = torch.cat([module.q.bias.data, module.k.bias.data, module.v.bias.data], dim=0)
 
             self.qkv.weight.data = qkv_weight
             self.qkv.bias.data = qkv_bias
 
-            proj_out = zero_module(conv_nd(1, self.channels, self.channels, 1))
+            proj_out = zero_module(nn.Conv1d(self.channels, self.channels, 1))
             proj_out.weight.data = module.proj_out.weight.data[:, :, :, 0]
             proj_out.bias.data = module.proj_out.bias.data
 
             self.proj_out = proj_out
+        elif self.overwrite_linear:
+            self.qkv.weight.data = torch.concat(
+                [self.NIN_0.W.data.T, self.NIN_1.W.data.T, self.NIN_2.W.data.T], dim=0
+            )[:, :, None]
+            self.qkv.bias.data = torch.concat([self.NIN_0.b.data, self.NIN_1.b.data, self.NIN_2.b.data], dim=0)
+
+            self.proj_out.weight.data = self.NIN_3.W.data.T[:, :, None]
+            self.proj_out.bias.data = self.NIN_3.b.data
+
+            self.norm.weight.data = self.GroupNorm_0.weight.data
+            self.norm.bias.data = self.GroupNorm_0.bias.data
 
     def forward(self, x, encoder_out=None):
-        if self.overwrite_qkv and not self.is_overwritten:
+        if (self.overwrite_qkv or self.overwrite_linear) and not self.is_overwritten:
             self.set_weights(self)
             self.is_overwritten = True
 
@@ -124,69 +151,74 @@ class AttentionBlock(nn.Module):
         h = a.reshape(bs, -1, length)
 
         h = self.proj_out(h)
+        h = h.reshape(b, c, *spatial)
 
-        return x + h.reshape(b, c, *spatial)
-
-
-def conv_nd(dims, *args, **kwargs):
-    """
-    Create a 1D, 2D, or 3D convolution module.
-    """
-    if dims == 1:
-        return nn.Conv1d(*args, **kwargs)
-    elif dims == 2:
-        return nn.Conv2d(*args, **kwargs)
-    elif dims == 3:
-        return nn.Conv3d(*args, **kwargs)
-    raise ValueError(f"unsupported dimensions: {dims}")
-
-
-class GroupNorm32(nn.GroupNorm):
-    def __init__(self, num_groups, num_channels, swish, eps=1e-5, affine=True):
-        super().__init__(num_groups=num_groups, num_channels=num_channels, eps=eps, affine=affine)
-        self.swish = swish
-
-    def forward(self, x):
-        y = super().forward(x.float()).to(x.dtype)
-        if self.swish == 1.0:
-            y = F.silu(y)
-        elif self.swish:
-            y = y * F.sigmoid(y * float(self.swish))
-        return y
-
-
-def normalization(channels, swish=0.0, eps=1e-5):
-    """
-    Make a standard normalization layer, with an optional swish activation.
-
-    :param channels: number of input channels. :return: an nn.Module for normalization.
-    """
-    return GroupNorm32(num_channels=channels, num_groups=32, swish=swish, eps=eps, affine=True)
+        result = x + h
 
+        result = result / self.rescale_output_factor
 
-def zero_module(module):
-    """
-    Zero out the parameters of a module and return it.
-    """
-    for p in module.parameters():
-        p.detach().zero_()
-    return module
+        return result
 
 
 # unet_score_estimation.py
 # class AttnBlockpp(nn.Module):
 #    """Channel-wise self-attention block. Modified from DDPM."""
 #
-#    def __init__(self, channels, skip_rescale=False, init_scale=0.0):
+#    def __init__(
+#        self,
+#        channels,
+#        skip_rescale=False,
+#        init_scale=0.0,
+#        num_heads=1,
+#        num_head_channels=-1,
+#        use_checkpoint=False,
+#        encoder_channels=None,
+#        use_new_attention_order=False,  # TODO(Patrick) -> is never used, maybe delete?
+#        overwrite_qkv=False,
+#        overwrite_from_grad_tts=False,
+#    ):
 #        super().__init__()
-#        self.GroupNorm_0 = nn.GroupNorm(num_groups=min(channels // 4, 32), num_channels=channels, eps=1e-6)
+#        num_groups = min(channels // 4, 32)
+#        self.GroupNorm_0 = nn.GroupNorm(num_groups=num_groups, num_channels=channels, eps=1e-6)
 #        self.NIN_0 = NIN(channels, channels)
 #        self.NIN_1 = NIN(channels, channels)
 #        self.NIN_2 = NIN(channels, channels)
 #        self.NIN_3 = NIN(channels, channels, init_scale=init_scale)
 #        self.skip_rescale = skip_rescale
 #
+#        self.channels = channels
+#        if num_head_channels == -1:
+#            self.num_heads = num_heads
+#        else:
+#            assert (
+#                channels % num_head_channels == 0
+#            ), f"q,k,v channels {channels} is not divisible by num_head_channels {num_head_channels}"
+#            self.num_heads = channels // num_head_channels
+#
+#        self.use_checkpoint = use_checkpoint
+#        self.norm = nn.GroupNorm(num_channels=channels, num_groups=num_groups, eps=1e-6)
+#        self.qkv = nn.Conv1d(channels, channels * 3, 1)
+#        self.n_heads = self.num_heads
+#
+#        self.proj_out = zero_module(nn.Conv1d(channels, channels, 1))
+#
+#        self.is_weight_set = False
+#
+#    def set_weights(self):
+#        self.qkv.weight.data = torch.concat([self.NIN_0.W.data.T, self.NIN_1.W.data.T, self.NIN_2.W.data.T], dim=0)[:, :, None]
+#        self.qkv.bias.data = torch.concat([self.NIN_0.b.data, self.NIN_1.b.data, self.NIN_2.b.data], dim=0)
+#
+#        self.proj_out.weight.data = self.NIN_3.W.data.T[:, :, None]
+#        self.proj_out.bias.data = self.NIN_3.b.data
+#
+#        self.norm.weight.data = self.GroupNorm_0.weight.data
+#        self.norm.bias.data = self.GroupNorm_0.bias.data
+#
 #    def forward(self, x):
+#        if not self.is_weight_set:
+#            self.set_weights()
+#            self.is_weight_set = True
+#
 #        B, C, H, W = x.shape
 #        h = self.GroupNorm_0(x)
 #        q = self.NIN_0(h)
@@ -199,7 +231,59 @@ def zero_module(module):
 #        w = torch.reshape(w, (B, H, W, H, W))
 #        h = torch.einsum("bhwij,bcij->bchw", w, v)
 #        h = self.NIN_3(h)
+#
 #        if not self.skip_rescale:
-#            return x + h
+#            result = x + h
 #        else:
+#            result = (x + h) / np.sqrt(2.0)
+#
+#        result = self.forward_2(x)
+#
+#        return result
+#
+#    def forward_2(self, x, encoder_out=None):
+#        b, c, *spatial = x.shape
+#        hid_states = self.norm(x).view(b, c, -1)
+#
+#        qkv = self.qkv(hid_states)
+#        bs, width, length = qkv.shape
+#        assert width % (3 * self.n_heads) == 0
+#        ch = width // (3 * self.n_heads)
+#        q, k, v = qkv.reshape(bs * self.n_heads, ch * 3, length).split(ch, dim=1)
+#
+#        if encoder_out is not None:
+#            encoder_kv = self.encoder_kv(encoder_out)
+#            assert encoder_kv.shape[1] == self.n_heads * ch * 2
+#            ek, ev = encoder_kv.reshape(bs * self.n_heads, ch * 2, -1).split(ch, dim=1)
+#            k = torch.cat([ek, k], dim=-1)
+#            v = torch.cat([ev, v], dim=-1)
+#
+#        scale = 1 / math.sqrt(math.sqrt(ch))
+#        weight = torch.einsum("bct,bcs->bts", q * scale, k * scale)  # More stable with f16 than dividing afterwards
+#        weight = torch.softmax(weight.float(), dim=-1).type(weight.dtype)
+#
+#        a = torch.einsum("bts,bcs->bct", weight, v)
+#        h = a.reshape(bs, -1, length)
+#
+#        h = self.proj_out(h)
+#        h = h.reshape(b, c, *spatial)
+#
 #        return (x + h) / np.sqrt(2.0)
+
+
+# TODO(Patrick) - this can and should be removed
+def zero_module(module):
+    """
+    Zero out the parameters of a module and return it.
+    """
+    for p in module.parameters():
+        p.detach().zero_()
+    return module
+
+
+# TODO(Patrick) - remove once all weights have been converted -> not needed anymore then
+class NIN(nn.Module):
+    def __init__(self, in_dim, num_units, init_scale=0.1):
+        super().__init__()
+        self.W = nn.Parameter(torch.zeros(in_dim, num_units), requires_grad=True)
+        self.b = nn.Parameter(torch.zeros(num_units), requires_grad=True)

src/diffusers/models/embeddings.py

@@ -65,19 +65,3 @@ class GaussianFourierProjection(nn.Module):
     def forward(self, x):
         x_proj = x[:, None] * self.W[None, :] * 2 * np.pi
         return torch.cat([torch.sin(x_proj), torch.cos(x_proj)], dim=-1)
-
-
-# unet_rl.py - TODO(need test)
-class SinusoidalPosEmb(nn.Module):
-    def __init__(self, dim):
-        super().__init__()
-        self.dim = dim
-
-    def forward(self, x):
-        device = x.device
-        half_dim = self.dim // 2
-        emb = math.log(10000) / (half_dim - 1)
-        emb = torch.exp(torch.arange(half_dim, device=device) * -emb)
-        emb = x[:, None] * emb[None, :]
-        emb = torch.cat((emb.sin(), emb.cos()), dim=-1)
-        return emb

src/diffusers/models/unet.py

@@ -15,24 +15,14 @@
 
 # helpers functions
 
-import copy
-import math
-from pathlib import Path
-
 import torch
 from torch import nn
-from torch.cuda.amp import GradScaler, autocast
-from torch.optim import Adam
-from torch.utils import data
-
-from PIL import Image
-from tqdm import tqdm
 
 from ..configuration_utils import ConfigMixin
 from ..modeling_utils import ModelMixin
+from .attention import AttentionBlock
 from .embeddings import get_timestep_embedding
 from .resnet import Downsample, Upsample
-from .attention2d import AttentionBlock
 
 
 def nonlinearity(x):
@@ -219,11 +209,7 @@ class UNetModel(ModelMixin, ConfigMixin):
             for i_block in range(self.num_res_blocks):
                 h = self.down[i_level].block[i_block](hs[-1], temb)
                 if len(self.down[i_level].attn) > 0:
-#                    self.down[i_level].attn_2[i_block].set_weights(self.down[i_level].attn[i_block])
-#                    h = self.down[i_level].attn_2[i_block](h)
-
                     h = self.down[i_level].attn[i_block](h)
-#                    print("Result", (h - h_2).abs().sum())
                 hs.append(h)
             if i_level != self.num_resolutions - 1:
                 hs.append(self.down[i_level].downsample(hs[-1]))

src/diffusers/models/unet_glide.py

@@ -6,7 +6,7 @@ import torch.nn.functional as F
 
 from ..configuration_utils import ConfigMixin
 from ..modeling_utils import ModelMixin
-from .attention2d import AttentionBlock
+from .attention import AttentionBlock
 from .embeddings import get_timestep_embedding
 from .resnet import Downsample, Upsample
 

src/diffusers/models/unet_grad_tts.py

 import torch
-from numpy import pad
 
 from ..configuration_utils import ConfigMixin
 from ..modeling_utils import ModelMixin
+from .attention import LinearAttention
 from .embeddings import get_timestep_embedding
 from .resnet import Downsample, Upsample
 
@@ -54,32 +54,6 @@ class ResnetBlock(torch.nn.Module):
         return output
 
 
-class LinearAttention(torch.nn.Module):
-    def __init__(self, dim, heads=4, dim_head=32):
-        super(LinearAttention, self).__init__()
-        self.heads = heads
-        self.dim_head = dim_head
-        hidden_dim = dim_head * heads
-        self.to_qkv = torch.nn.Conv2d(dim, hidden_dim * 3, 1, bias=False)
-        self.to_out = torch.nn.Conv2d(hidden_dim, dim, 1)
-
-    def forward(self, x):
-        b, c, h, w = x.shape
-        qkv = self.to_qkv(x)
-        #        q, k, v = rearrange(qkv, "b (qkv heads c) h w -> qkv b heads c (h w)", heads=self.heads, qkv=3)
-        q, k, v = (
-            qkv.reshape(b, 3, self.heads, self.dim_head, h, w)
-            .permute(1, 0, 2, 3, 4, 5)
-            .reshape(3, b, self.heads, self.dim_head, -1)
-        )
-        k = k.softmax(dim=-1)
-        context = torch.einsum("bhdn,bhen->bhde", k, v)
-        out = torch.einsum("bhde,bhdn->bhen", context, q)
-        #        out = rearrange(out, "b heads c (h w) -> b (heads c) h w", heads=self.heads, h=h, w=w)
-        out = out.reshape(b, self.heads, self.dim_head, h, w).reshape(b, self.heads * self.dim_head, h, w)
-        return self.to_out(out)
-
-
 class Residual(torch.nn.Module):
     def __init__(self, fn):
         super(Residual, self).__init__()

src/diffusers/models/unet_ldm.py

@@ -9,7 +9,7 @@ import torch.nn.functional as F
 
 from ..configuration_utils import ConfigMixin
 from ..modeling_utils import ModelMixin
-from .attention2d import AttentionBlock
+from .attention import AttentionBlock
 from .embeddings import get_timestep_embedding
 from .resnet import Downsample, Upsample
 

src/diffusers/models/unet_sde_score_estimation.py

@@ -16,6 +16,7 @@
 # helpers functions
 
 import functools
+import math
 import string
 
 import numpy as np
@@ -25,6 +26,7 @@ import torch.nn.functional as F
 
 from ..configuration_utils import ConfigMixin
 from ..modeling_utils import ModelMixin
+from .attention import AttentionBlock
 from .embeddings import GaussianFourierProjection, get_timestep_embedding
 
 
@@ -414,37 +416,6 @@ class Combine(nn.Module):
             raise ValueError(f"Method {self.method} not recognized.")
 
 
-class AttnBlockpp(nn.Module):
-    """Channel-wise self-attention block. Modified from DDPM."""
-
-    def __init__(self, channels, skip_rescale=False, init_scale=0.0):
-        super().__init__()
-        self.GroupNorm_0 = nn.GroupNorm(num_groups=min(channels // 4, 32), num_channels=channels, eps=1e-6)
-        self.NIN_0 = NIN(channels, channels)
-        self.NIN_1 = NIN(channels, channels)
-        self.NIN_2 = NIN(channels, channels)
-        self.NIN_3 = NIN(channels, channels, init_scale=init_scale)
-        self.skip_rescale = skip_rescale
-
-    def forward(self, x):
-        B, C, H, W = x.shape
-        h = self.GroupNorm_0(x)
-        q = self.NIN_0(h)
-        k = self.NIN_1(h)
-        v = self.NIN_2(h)
-
-        w = torch.einsum("bchw,bcij->bhwij", q, k) * (int(C) ** (-0.5))
-        w = torch.reshape(w, (B, H, W, H * W))
-        w = F.softmax(w, dim=-1)
-        w = torch.reshape(w, (B, H, W, H, W))
-        h = torch.einsum("bhwij,bcij->bchw", w, v)
-        h = self.NIN_3(h)
-        if not self.skip_rescale:
-            return x + h
-        else:
-            return (x + h) / np.sqrt(2.0)
-
-
 class Upsample(nn.Module):
     def __init__(self, in_ch=None, out_ch=None, with_conv=False, fir=False, fir_kernel=(1, 3, 3, 1)):
         super().__init__()
@@ -756,8 +727,7 @@ class NCSNpp(ModelMixin, ConfigMixin):
             modules[-1].weight.data = default_init()(modules[-1].weight.shape)
             nn.init.zeros_(modules[-1].bias)
 
-        AttnBlock = functools.partial(AttnBlockpp, init_scale=init_scale, skip_rescale=skip_rescale)
-
+        AttnBlock = functools.partial(AttentionBlock, overwrite_linear=True, rescale_output_factor=math.sqrt(2.0))
         Up_sample = functools.partial(Upsample, with_conv=resamp_with_conv, fir=fir, fir_kernel=fir_kernel)
 
         if progressive == "output_skip":

tests/test_modeling_utils.py

@@ -859,7 +859,9 @@ class PipelineTesterMixin(unittest.TestCase):
         image_slice = image[0, -1, -3:, -3:].cpu()
 
         assert image.shape == (1, 3, 32, 32)
-        expected_slice = torch.tensor([-0.5712, -0.6215, -0.5953, -0.5438, -0.4775, -0.4539, -0.5172, -0.4872, -0.5105])
+        expected_slice = torch.tensor(
+            [-0.5712, -0.6215, -0.5953, -0.5438, -0.4775, -0.4539, -0.5172, -0.4872, -0.5105]
+        )
         assert (image_slice.flatten() - expected_slice).abs().max() < 1e-2
 
     @slow