add learned padding tokens, same strategy as dalle1, for diffusion prior, and get rid of masking in causal transformer

dalle2_pytorch/dalle2_pytorch.py

@@ -527,31 +527,25 @@ class NoiseScheduler(nn.Module):
 # diffusion prior
 
 class LayerNorm(nn.Module):
-    def __init__(self, dim, eps = 1e-5, stable = False):
+    def __init__(self, dim, eps = 1e-5):
         super().__init__()
         self.eps = eps
-        self.stable = stable
         self.g = nn.Parameter(torch.ones(dim))
 
     def forward(self, x):
-        if self.stable:
-            x = x / x.amax(dim = -1, keepdim = True).detach()
-
+        x = x / x.amax(dim = -1, keepdim = True).detach()
         var = torch.var(x, dim = -1, unbiased = False, keepdim = True)
         mean = torch.mean(x, dim = -1, keepdim = True)
         return (x - mean) * (var + self.eps).rsqrt() * self.g
 
 class ChanLayerNorm(nn.Module):
-    def __init__(self, dim, eps = 1e-5, stable = False):
+    def __init__(self, dim, eps = 1e-5):
         super().__init__()
         self.eps = eps
-        self.stable = stable
         self.g = nn.Parameter(torch.ones(1, dim, 1, 1))
 
     def forward(self, x):
-        if self.stable:
-            x = x / x.amax(dim = 1, keepdim = True).detach()
-
+        x = x / x.amax(dim = 1, keepdim = True).detach()
         var = torch.var(x, dim = 1, unbiased = False, keepdim = True)
         mean = torch.mean(x, dim = 1, keepdim = True)
         return (x - mean) * (var + self.eps).rsqrt() * self.g
@@ -675,7 +669,7 @@ class Attention(nn.Module):
         dropout = 0.,
         causal = False,
         rotary_emb = None,
-        pb_relax_alpha = 128
+        pb_relax_alpha = 32 ** 2
     ):
         super().__init__()
         self.pb_relax_alpha = pb_relax_alpha
@@ -766,7 +760,6 @@ class CausalTransformer(nn.Module):
         dim_head = 64,
         heads = 8,
         ff_mult = 4,
-        norm_in = False,
         norm_out = True,
         attn_dropout = 0.,
         ff_dropout = 0.,
@@ -775,8 +768,6 @@ class CausalTransformer(nn.Module):
         rotary_emb = True
     ):
         super().__init__()
-        self.init_norm = LayerNorm(dim) if norm_in else nn.Identity() # from latest BLOOM model and Yandex's YaLM
-
         self.rel_pos_bias = RelPosBias(heads = heads)
 
         rotary_emb = RotaryEmbedding(dim = min(32, dim_head)) if rotary_emb else None
@@ -788,14 +779,12 @@ class CausalTransformer(nn.Module):
                 FeedForward(dim = dim, mult = ff_mult, dropout = ff_dropout, post_activation_norm = normformer)
             ]))
 
-        self.norm = LayerNorm(dim, stable = True) if norm_out else nn.Identity()  # unclear in paper whether they projected after the classic layer norm for the final denoised image embedding, or just had the transformer output it directly: plan on offering both options
+        self.norm = LayerNorm(dim) if norm_out else nn.Identity()  # unclear in paper whether they projected after the classic layer norm for the final denoised image embedding, or just had the transformer output it directly: plan on offering both options
         self.project_out = nn.Linear(dim, dim, bias = False) if final_proj else nn.Identity()
 
     def forward(self, x):
         n, device = x.shape[1], x.device
 
-        x = self.init_norm(x)
-
         attn_bias = self.rel_pos_bias(n, n + 1, device = device)
 
         for attn, ff in self.layers:
@@ -895,7 +884,7 @@ class DiffusionPriorNetwork(nn.Module):
 
         if remainder > 0:
             text_encodings = F.pad(text_encodings, (0, 0, 0, remainder), value = 0.)
-            mask = F.pad(mask, (0, remainder), value = False)
+            mask = F.pad(mask, (0, remainder), value = 0.)
 
         null_text_embeds = self.null_text_embed.to(text_encodings.dtype)
 
@@ -2328,9 +2317,6 @@ class Decoder(nn.Module):
 
         img = torch.randn(shape, device = device)
 
-        if not is_latent_diffusion:
-            lowres_cond_img = maybe(self.normalize_img)(lowres_cond_img)
-
         for time, time_next in tqdm(time_pairs, desc = 'sampling loop time step'):
             alpha = alphas[time]
             alpha_next = alphas[time_next]

dalle2_pytorch/train_configs.py

@@ -137,7 +137,6 @@ class DiffusionPriorNetworkConfig(BaseModel):
     dim_head: int = 64
     heads: int = 8
     ff_mult: int = 4
-    norm_in: bool = False
     norm_out: bool = True
     attn_dropout: float = 0.
     ff_dropout: float = 0.

dalle2_pytorch/version.py

@@ -1 +1 @@
-__version__ = '0.23.4'
+__version__ = '0.23.0'

train_decoder.py

@@ -323,7 +323,7 @@ def train(
         last_snapshot = sample
 
         if next_task == 'train':
-            for i, (img, emb, txt) in enumerate(dataloaders["train"]):
+            for i, (img, emb, txt) in enumerate(trainer.train_loader):
                 # We want to count the total number of samples across all processes
                 sample_length_tensor[0] = len(img)
                 all_samples = accelerator.gather(sample_length_tensor)  # TODO: accelerator.reduce is broken when this was written. If it is fixed replace this.
@@ -358,7 +358,6 @@ def train(
                         else:
                             # Then we need to pass the text instead
                             tokenized_texts = tokenize(txt, truncate=True)
-                            assert tokenized_texts.shape[0] == len(img), f"The number of texts ({tokenized_texts.shape[0]}) should be the same as the number of images ({len(img)})"
                             forward_params['text'] = tokenized_texts
                     loss = trainer.forward(img, **forward_params, unet_number=unet)
                     trainer.update(unet_number=unet)
@@ -417,7 +416,7 @@ def train(
             timer = Timer()
             accelerator.wait_for_everyone()
             i = 0
-            for i, (img, emb, txt) in enumerate(dataloaders['val']):  # Use the accelerate prepared loader
+            for i, (img, emb, txt) in enumerate(trainer.val_loader):  # Use the accelerate prepared loader
                 val_sample_length_tensor[0] = len(img)
                 all_samples = accelerator.gather(val_sample_length_tensor)
                 total_samples = all_samples.sum().item()