refactor so that the causal transformer in the diffusion prior network can be conditioned without text encodings (for Laions parallel efforts, although it seems from the paper it is needed)

README.md

@@ -523,6 +523,7 @@ Once built, images will be saved to the same directory the command is invoked
 - [x] offload unets not being trained on to CPU for memory efficiency (for training each resolution unets separately)
 - [x] build out latent diffusion architecture, with the vq-reg variant (vqgan-vae), make it completely optional and compatible with cascading ddpms
 - [x] for decoder, allow ability to customize objective (predict epsilon vs x0), in case latent diffusion does better with prediction of x0
+- [x] use attention-based upsampling https://arxiv.org/abs/2112.11435
 - [ ] spend one day cleaning up tech debt in decoder
 - [ ] become an expert with unets, cleanup unet code, make it fully configurable, port all learnings over to https://github.com/lucidrains/x-unet
 - [ ] copy the cascading ddpm code to a separate repo (perhaps https://github.com/lucidrains/denoising-diffusion-pytorch) as the main contribution of dalle2 really is just the prior network
@@ -531,7 +532,6 @@ Once built, images will be saved to the same directory the command is invoked
 - [ ] extend diffusion head to use diffusion-gan (potentially using lightweight-gan) to speed up inference
 - [ ] bring in tools to train vqgan-vae
 - [ ] bring in vit-vqgan https://arxiv.org/abs/2110.04627 for the latent diffusion
-- [ ] experiment with https://arxiv.org/abs/2112.11435 as upsampler, test in https://github.com/lucidrains/lightweight-gan first
 
 ## Citations
 
@@ -577,14 +577,4 @@ Once built, images will be saved to the same directory the command is invoked
 }
 ```
 
-```bibtex
-@article{Arar2021LearnedQF,
-    title   = {Learned Queries for Efficient Local Attention},
-    author  = {Moab Arar and Ariel Shamir and Amit H. Bermano},
-    journal = {ArXiv},
-    year    = {2021},
-    volume  = {abs/2112.11435}
-}
-```
-
 *Creating noise from data is easy; creating data from noise is generative modeling.* - Yang Song's <a href="https://arxiv.org/abs/2011.13456">paper</a>

dalle2_pytorch/attention.py

@@ -0,0 +1,125 @@
+import torch
+from torch import nn, einsum
+import torch.nn.functional as F
+
+from einops import rearrange, repeat
+
+class LayerNormChan(nn.Module):
+    def __init__(
+        self,
+        dim,
+        eps = 1e-5
+    ):
+        super().__init__()
+        self.eps = eps
+        self.gamma = nn.Parameter(torch.ones(1, dim, 1, 1))
+
+    def forward(self, x):
+        var = torch.var(x, dim = 1, unbiased = False, keepdim = True)
+        mean = torch.mean(x, dim = 1, keepdim = True)
+        return (x - mean) / (var + self.eps).sqrt() * self.gamma
+
+# attention-based upsampling
+# from https://arxiv.org/abs/2112.11435
+
+class QueryAndAttend(nn.Module):
+    def __init__(
+        self,
+        *,
+        dim,
+        num_queries = 1,
+        dim_head = 32,
+        heads = 8,
+        window_size = 3
+    ):
+        super().__init__()
+        self.scale = dim_head ** -0.5
+        inner_dim = dim_head * heads
+        self.heads = heads
+        self.dim_head = dim_head
+        self.window_size = window_size
+        self.num_queries = num_queries
+
+        self.rel_pos_bias = nn.Parameter(torch.randn(heads, num_queries, window_size * window_size, 1, 1))
+
+        self.queries = nn.Parameter(torch.randn(heads, num_queries, dim_head))
+        self.to_kv = nn.Conv2d(dim, dim_head * 2, 1, bias = False)
+        self.to_out = nn.Conv2d(inner_dim, dim, 1, bias = False)
+
+    def forward(self, x):
+        """
+        einstein notation
+        b - batch
+        h - heads
+        l - num queries
+        d - head dimension
+        x - height
+        y - width
+        j - source sequence for attending to (kernel size squared in this case)
+        """
+
+        wsz, heads, dim_head, num_queries = self.window_size, self.heads, self.dim_head, self.num_queries
+        batch, _, height, width = x.shape
+
+        is_one_query = self.num_queries == 1
+
+        # queries, keys, values
+
+        q = self.queries * self.scale
+        k, v = self.to_kv(x).chunk(2, dim = 1)
+
+        # similarities
+
+        sim = einsum('h l d, b d x y -> b h l x y', q, k)
+        sim = rearrange(sim, 'b ... x y -> b (...) x y')
+
+        # unfold the similarity scores, with float(-inf) as padding value
+
+        mask_value = -torch.finfo(sim.dtype).max
+        sim = F.pad(sim, ((wsz // 2,) * 4), value = mask_value)
+        sim = F.unfold(sim, kernel_size = wsz)
+        sim = rearrange(sim, 'b (h l j) (x y) -> b h l j x y', h = heads, l = num_queries, x = height, y = width)
+
+        # rel pos bias
+
+        sim = sim + self.rel_pos_bias
+
+        # numerically stable attention
+
+        sim = sim - sim.amax(dim = -3, keepdim = True).detach()
+        attn = sim.softmax(dim = -3)
+
+        # unfold values
+
+        v = F.pad(v, ((wsz // 2,) * 4), value = 0.)
+        v = F.unfold(v, kernel_size = wsz)
+        v = rearrange(v, 'b (d j) (x y) -> b d j x y', d = dim_head, x = height, y = width)
+
+        # aggregate values
+
+        out = einsum('b h l j x y, b d j x y -> b l h d x y', attn, v)
+
+        # combine heads
+
+        out = rearrange(out, 'b l h d x y -> (b l) (h d) x y')
+        out = self.to_out(out)
+        out = rearrange(out, '(b l) d x y -> b l d x y', b = batch)
+
+        # return original input if one query
+
+        if is_one_query:
+            out = rearrange(out, 'b 1 ... -> b ...')
+
+        return out
+
+class QueryAttnUpsample(nn.Module):
+    def __init__(self, dim, **kwargs):
+        super().__init__()
+        self.norm = LayerNormChan(dim)
+        self.qna = QueryAndAttend(dim = dim, num_queries = 4, **kwargs)
+
+    def forward(self, x):
+        x = self.norm(x)
+        out = self.qna(x)
+        out = rearrange(out, 'b (w1 w2) c h w -> b c (h w1) (w w2)', w1 = 2, w2 = 2)
+        return out

dalle2_pytorch/dalle2_pytorch.py

@@ -17,6 +17,7 @@ from kornia.filters import gaussian_blur2d
 
 from dalle2_pytorch.tokenizer import tokenizer
 from dalle2_pytorch.vqgan_vae import NullVQGanVAE, VQGanVAE
+from dalle2_pytorch.attention import QueryAttnUpsample
 
 # use x-clip
 
@@ -420,25 +421,41 @@ class DiffusionPriorNetwork(nn.Module):
         image_embed,
         diffusion_timesteps,
         *,
-        text_encodings,
         text_embed,
+        text_encodings = None,
         mask = None,
         cond_drop_prob = 0.2
     ):
-        batch, text_enc_len, device = image_embed.shape[0], text_encodings.shape[-2], image_embed.device
+        batch, dim, device, dtype = *image_embed.shape, image_embed.device, image_embed.dtype
 
         # in section 2.2, last paragraph
         # "... consisting of encoded text, CLIP text embedding, diffusion timestep embedding, noised CLIP image embedding, final embedding for prediction"
 
         text_embed, image_embed = rearrange_many((text_embed, image_embed), 'b d -> b 1 d')
 
+        # make text encodings optional
+        # although the paper seems to suggest it is present <--
+
+        if not exists(text_encodings):
+            text_encodings = torch.empty((batch, 0, dim), device = device, dtype = dtype)
+
+        if not exists(mask):
+            mask = torch.ones((batch, text_encodings.shape[-2]), device = device, dtype = torch.bool)
+
+        # classifier free guidance
+
+        cond_prob_mask = prob_mask_like((batch,), cond_drop_prob, device = device)
+        cond_prob_mask = rearrange(cond_prob_mask, 'b -> b 1')
+
+        mask &= cond_prob_mask
+
+        # whether text embedding is masked or not depends on the classifier free guidance conditional masking
+
+        mask = torch.cat((mask, cond_prob_mask), dim = 1)
+
         # whether text embedding is used for conditioning depends on whether text encodings are available for attention (for classifier free guidance, even though it seems from the paper it was not used in the prior ddpm, as the objective is different)
         # but let's just do it right
 
-        if exists(mask):
-            not_all_masked_out = mask.any(dim = -1)
-            mask = torch.cat((mask, rearrange(not_all_masked_out, 'b -> b 1')), dim = 1)
-
         if exists(mask):
             mask = F.pad(mask, (0, 2), value = True) # extend mask for text embedding, noised image embedding, time step embedding, and learned query
 
@@ -454,16 +471,6 @@ class DiffusionPriorNetwork(nn.Module):
             learned_queries
         ), dim = -2)
 
-        # mask if it doesn't exist
-
-        if not exists(mask):
-            mask = torch.ones((batch, text_enc_len), device = device, dtype = torch.bool)
-
-        # classifier free guidance
-
-        cond_prob_mask = prob_mask_like((batch,), cond_drop_prob, device = device)
-        mask &= rearrange(cond_prob_mask, 'b -> b 1')
-
         # attend
 
         tokens = self.causal_transformer(tokens, mask = mask)
@@ -485,6 +492,7 @@ class DiffusionPrior(nn.Module):
         loss_type = "l1",
         predict_x_start = True,
         beta_schedule = "cosine",
+        condition_on_text_encodings = True, # the paper suggests this is needed, but you can turn it off for your CLIP preprocessed text embed -> image embed training
     ):
         super().__init__()
         assert isinstance(clip, CLIP)
@@ -495,7 +503,9 @@ class DiffusionPrior(nn.Module):
         self.image_embed_dim = clip.dim_latent
         self.channels = clip.image_channels
         self.image_size = clip.image_size
+
         self.cond_drop_prob = cond_drop_prob
+        self.condition_on_text_encodings = condition_on_text_encodings
 
         self.predict_x_start = predict_x_start
         # in paper, they do not predict the noise, but predict x0 directly for image embedding, claiming empirically better results. I'll just offer both.
@@ -560,6 +570,10 @@ class DiffusionPrior(nn.Module):
         text_cls, text_encodings = text_encodings[:, 0], text_encodings[:, 1:]
         text_embed = self.clip.to_text_latent(text_cls)
         text_embed = l2norm(text_embed)
+
+        if not self.condition_on_text_encodings:            
+            return dict(text_embed = text_embed)
+
         return dict(text_encodings = text_encodings, text_embed = text_embed, mask = text != 0)
 
     def q_mean_variance(self, x_start, t):
@@ -1116,7 +1130,7 @@ class Decoder(nn.Module):
         unet,
         *,
         clip,
-        vae = None,
+        vae = tuple(),
         timesteps = 1000,
         cond_drop_prob = 0.2,
         loss_type = 'l1',

dalle2_pytorch/train.py

@@ -0,0 +1,53 @@
+import copy
+import torch
+from torch import nn
+
+# exponential moving average wrapper
+
+class EMA(nn.Module):
+    def __init__(
+        self,
+        model,
+        beta = 0.99,
+        ema_update_after_step = 1000,
+        ema_update_every = 10,
+    ):
+        super().__init__()
+        self.beta = beta
+        self.online_model = model
+        self.ema_model = copy.deepcopy(model)
+
+        self.ema_update_after_step = ema_update_after_step # only start EMA after this step number, starting at 0
+        self.ema_update_every = ema_update_every
+
+        self.register_buffer('initted', torch.Tensor([False]))
+        self.register_buffer('step', torch.tensor([0.]))
+
+    def update(self):
+        self.step += 1
+
+        if self.step <= self.ema_update_after_step or (self.step % self.ema_update_every) != 0:
+            return
+
+        if not self.initted:
+            self.ema_model.state_dict(self.online_model.state_dict())
+            self.initted.data.copy_(torch.Tensor([True]))
+
+        self.update_moving_average(self.ema_model, self.online_model)
+
+    def update_moving_average(ma_model, current_model):
+        def calculate_ema(beta, old, new):
+            if not exists(old):
+                return new
+            return old * beta + (1 - beta) * new
+
+        for current_params, ma_params in zip(current_model.parameters(), ma_model.parameters()):
+            old_weight, up_weight = ma_params.data, current_params.data
+            ma_params.data = calculate_ema(self.beta, old_weight, up_weight)
+
+        for current_buffer, ma_buffer in zip(current_model.buffers(), ma_model.buffers()):
+            new_buffer_value = calculate_ema(self.beta, ma_buffer, current_buffer)
+            ma_buffer.copy_(new_buffer_value)
+
+    def __call__(self, *args, **kwargs):
+        return self.ema_model(*args, **kwargs)

dalle2_pytorch/vqgan_vae.py

@@ -13,6 +13,8 @@ import torchvision
 
 from einops import rearrange, reduce, repeat
 
+from dalle2_pytorch.attention import QueryAttnUpsample
+
 # constants
 
 MList = nn.ModuleList
@@ -243,111 +245,6 @@ class ResBlock(nn.Module):
     def forward(self, x):
         return self.net(x) + x
 
-# attention-based upsampling
-# from https://arxiv.org/abs/2112.11435
-
-class QueryAndAttend(nn.Module):
-    def __init__(
-        self,
-        *,
-        dim,
-        num_queries = 1,
-        dim_head = 32,
-        heads = 8,
-        window_size = 3
-    ):
-        super().__init__()
-        self.scale = dim_head ** -0.5
-        inner_dim = dim_head * heads
-        self.heads = heads
-        self.dim_head = dim_head
-        self.window_size = window_size
-        self.num_queries = num_queries
-
-        self.rel_pos_bias = nn.Parameter(torch.randn(heads, num_queries, window_size * window_size, 1, 1))
-
-        self.queries = nn.Parameter(torch.randn(heads, num_queries, dim_head))
-        self.to_kv = nn.Conv2d(dim, dim_head * 2, 1, bias = False)
-        self.to_out = nn.Conv2d(inner_dim, dim, 1, bias = False)
-
-    def forward(self, x):
-        """
-        einstein notation
-        b - batch
-        h - heads
-        l - num queries
-        d - head dimension
-        x - height
-        y - width
-        j - source sequence for attending to (kernel size squared in this case)
-        """
-
-        wsz, heads, dim_head, num_queries = self.window_size, self.heads, self.dim_head, self.num_queries
-        batch, _, height, width = x.shape
-
-        is_one_query = self.num_queries == 1
-
-        # queries, keys, values
-
-        q = self.queries * self.scale
-        k, v = self.to_kv(x).chunk(2, dim = 1)
-
-        # similarities
-
-        sim = einsum('h l d, b d x y -> b h l x y', q, k)
-        sim = rearrange(sim, 'b ... x y -> b (...) x y')
-
-        # unfold the similarity scores, with float(-inf) as padding value
-
-        mask_value = -torch.finfo(sim.dtype).max
-        sim = F.pad(sim, ((wsz // 2,) * 4), value = mask_value)
-        sim = F.unfold(sim, kernel_size = wsz)
-        sim = rearrange(sim, 'b (h l j) (x y) -> b h l j x y', h = heads, l = num_queries, x = height, y = width)
-
-        # rel pos bias
-
-        sim = sim + self.rel_pos_bias
-
-        # numerically stable attention
-
-        sim = sim - sim.amax(dim = -3, keepdim = True).detach()
-        attn = sim.softmax(dim = -3)
-
-        # unfold values
-
-        v = F.pad(v, ((wsz // 2,) * 4), value = 0.)
-        v = F.unfold(v, kernel_size = wsz)
-        v = rearrange(v, 'b (d j) (x y) -> b d j x y', d = dim_head, x = height, y = width)
-
-        # aggregate values
-
-        out = einsum('b h l j x y, b d j x y -> b l h d x y', attn, v)
-
-        # combine heads
-
-        out = rearrange(out, 'b l h d x y -> (b l) (h d) x y')
-        out = self.to_out(out)
-        out = rearrange(out, '(b l) d x y -> b l d x y', b = batch)
-
-        # return original input if one query
-
-        if is_one_query:
-            out = rearrange(out, 'b 1 ... -> b ...')
-
-        return out
-
-class QueryAttnUpsample(nn.Module):
-    def __init__(self, dim, **kwargs):
-        super().__init__()
-        self.norm = LayerNormChan(dim)
-        self.qna = QueryAndAttend(dim = dim, num_queries = 4, **kwargs)
-
-    def forward(self, x):
-        x = self.norm(x)
-        out = self.qna(x)
-        out = rearrange(out, 'b (w1 w2) c h w -> b c (h w1) (w w2)', w1 = 2, w2 = 2)
-        return out
-
 # vqgan attention layer
 class VQGanAttention(nn.Module):
     def __init__(
@@ -481,7 +378,7 @@ class VQGanVAE(nn.Module):
 
         for layer_index, (dim_in, dim_out), layer_num_resnet_blocks, layer_use_attn in zip(range(layers), dim_pairs, num_resnet_blocks, use_attn):
             append(self.encoders, nn.Sequential(nn.Conv2d(dim_in, dim_out, 4, stride = 2, padding = 1), leaky_relu()))
-            prepend(self.decoders, nn.Sequential(QueryAttnUpsample(dim_out), nn.Conv2d(dim_out, dim_in, 3, padding = 1), leaky_relu()))
+            prepend(self.decoders, nn.Sequential(nn.ConvTranspose2d(dim_out, dim_in, 4, 2, 1), leaky_relu()))
 
             if layer_use_attn:
                 prepend(self.decoders, VQGanAttention(dim = dim_out, heads = attn_heads, dim_head = attn_dim_head, dropout = attn_dropout))

setup.py

@@ -10,7 +10,7 @@ setup(
       'dream = dalle2_pytorch.cli:dream'
     ],
   },
-  version = '0.0.44',
+  version = '0.0.47',
   license='MIT',
   description = 'DALL-E 2',
   author = 'Phil Wang',