hack around some inplace error, also make sure for openai clip text encoding, only tokens after eos_id is masked out

dalle2_pytorch/dalle2_pytorch.py

@@ -278,6 +278,7 @@ class OpenAIClipAdapter(BaseClipAdapter):
         import clip
         openai_clip, preprocess = clip.load(name)
         super().__init__(openai_clip)
+        self.eos_id = 49407 # for handling 0 being also '!'
 
         text_attention_final = self.find_layer('ln_final')
         self.handle = text_attention_final.register_forward_hook(self._hook)
@@ -316,7 +317,10 @@ class OpenAIClipAdapter(BaseClipAdapter):
     @torch.no_grad()
     def embed_text(self, text):
         text = text[..., :self.max_text_len]
-        text_mask = text != 0
+
+        is_eos_id = (text == self.eos_id)
+        text_mask_excluding_eos = is_eos_id.cumsum(dim = -1) == 0
+        text_mask = F.pad(text_mask_excluding_eos, (1, -1), value = True)
         assert not self.cleared
 
         text_embed = self.clip.encode_text(text)
@@ -900,7 +904,7 @@ class DiffusionPriorNetwork(nn.Module):
         null_text_embeds = self.null_text_embed.to(text_encodings.dtype)
 
         text_encodings = torch.where(
-            rearrange(mask, 'b n -> b n 1'),
+            rearrange(mask, 'b n -> b n 1').clone(),
             text_encodings,
             null_text_embeds
         )
@@ -1251,6 +1255,14 @@ class DiffusionPrior(nn.Module):
 
 # decoder
 
+def NearestUpsample(dim, dim_out = None):
+    dim_out = default(dim_out, dim)
+
+    return nn.Sequential(
+        nn.Upsample(scale_factor = 2, mode = 'nearest'),
+        nn.Conv2d(dim, dim_out, 3, padding = 1)
+    )
+
 class PixelShuffleUpsample(nn.Module):
     """
     code shared by @MalumaDev at DALLE2-pytorch for addressing checkboard artifacts
@@ -1657,7 +1669,7 @@ class Unet(nn.Module):
 
         # upsample klass
 
-        upsample_klass = ConvTransposeUpsample if not pixel_shuffle_upsample else PixelShuffleUpsample
+        upsample_klass = NearestUpsample if not pixel_shuffle_upsample else PixelShuffleUpsample
 
         # give memory efficient unet an initial resnet block
 
@@ -1946,6 +1958,7 @@ class LowresConditioner(nn.Module):
         self,
         downsample_first = True,
         downsample_mode_nearest = False,
+        blur_prob = 0.5,
         blur_sigma = 0.6,
         blur_kernel_size = 3,
         input_image_range = None
@@ -1956,6 +1969,7 @@ class LowresConditioner(nn.Module):
 
         self.input_image_range = input_image_range
 
+        self.blur_prob = blur_prob
         self.blur_sigma = blur_sigma
         self.blur_kernel_size = blur_kernel_size
 
@@ -1968,20 +1982,27 @@ class LowresConditioner(nn.Module):
         blur_sigma = None,
         blur_kernel_size = None
     ):
-        if self.training and self.downsample_first and exists(downsample_image_size):
+        if self.downsample_first and exists(downsample_image_size):
             cond_fmap = resize_image_to(cond_fmap, downsample_image_size, clamp_range = self.input_image_range, nearest = self.downsample_mode_nearest)
 
-        if self.training:
+        # blur is only applied 50% of the time
+        # section 3.1 in https://arxiv.org/abs/2106.15282
+
+        if random.random() < self.blur_prob:
+
             # when training, blur the low resolution conditional image
+
             blur_sigma = default(blur_sigma, self.blur_sigma)
             blur_kernel_size = default(blur_kernel_size, self.blur_kernel_size)
 
             # allow for drawing a random sigma between lo and hi float values
+
             if isinstance(blur_sigma, tuple):
                 blur_sigma = tuple(map(float, blur_sigma))
                 blur_sigma = random.uniform(*blur_sigma)
 
             # allow for drawing a random kernel size between lo and hi int values
+
             if isinstance(blur_kernel_size, tuple):
                 blur_kernel_size = tuple(map(int, blur_kernel_size))
                 kernel_size_lo, kernel_size_hi = blur_kernel_size
@@ -1990,7 +2011,6 @@ class LowresConditioner(nn.Module):
             cond_fmap = gaussian_blur2d(cond_fmap, cast_tuple(blur_kernel_size, 2), cast_tuple(blur_sigma, 2))
 
         cond_fmap = resize_image_to(cond_fmap, target_image_size, clamp_range = self.input_image_range)
-
         return cond_fmap
 
 class Decoder(nn.Module):
@@ -2014,6 +2034,7 @@ class Decoder(nn.Module):
         random_crop_sizes = None,                   # whether to random crop the image at that stage in the cascade (super resoluting convolutions at the end may be able to generalize on smaller crops)
         lowres_downsample_first = True,             # cascading ddpm - resizes to lower resolution, then to next conditional resolution + blur
         lowres_downsample_mode_nearest = False,     # cascading ddpm - whether to use nearest mode downsampling for lower resolution
+        blur_prob = 0.5,                            # cascading ddpm - when training, the gaussian blur is only applied 50% of the time
         blur_sigma = 0.6,                           # cascading ddpm - blur sigma
         blur_kernel_size = 3,                       # cascading ddpm - blur kernel size
         clip_denoised = True,
@@ -2162,9 +2183,12 @@ class Decoder(nn.Module):
         lowres_conditions = tuple(map(lambda t: t.lowres_cond, self.unets))
         assert lowres_conditions == (False, *((True,) * (len(self.unets) - 1))), 'the first unet must be unconditioned (by low resolution image), and the rest of the unets must have `lowres_cond` set to True'
 
+        self.lowres_downsample_mode_nearest = lowres_downsample_mode_nearest
+
         self.to_lowres_cond = LowresConditioner(
             downsample_first = lowres_downsample_first,
             downsample_mode_nearest = lowres_downsample_mode_nearest,
+            blur_prob = blur_prob,
             blur_sigma = blur_sigma,
             blur_kernel_size = blur_kernel_size,
             input_image_range = self.input_image_range
@@ -2328,6 +2352,9 @@ 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]
@@ -2480,7 +2507,7 @@ class Decoder(nn.Module):
                 shape = (batch_size, channel, image_size, image_size)
 
                 if unet.lowres_cond:
-                    lowres_cond_img = self.to_lowres_cond(img, target_image_size = image_size)
+                    lowres_cond_img = resize_image_to(img, target_image_size = image_size, clamp_range = self.input_image_range, nearest = self.lowres_downsample_mode_nearest)
 
                 is_latent_diffusion = isinstance(vae, VQGanVAE)
                 image_size = vae.get_encoded_fmap_size(image_size)

dalle2_pytorch/trainer.py

@@ -673,8 +673,14 @@ class DecoderTrainer(nn.Module):
     def sample(self, *args, **kwargs):
         distributed = self.accelerator.num_processes > 1
         base_decoder = self.accelerator.unwrap_model(self.decoder)
+
+        was_training = base_decoder.training
+        base_decoder.eval()
+
         if kwargs.pop('use_non_ema', False) or not self.use_ema:
-            return base_decoder.sample(*args, **kwargs, distributed = distributed)
+            out = base_decoder.sample(*args, **kwargs, distributed = distributed)
+            base_decoder.train(was_training)
+            return out
 
         trainable_unets = self.accelerator.unwrap_model(self.decoder).unets
         base_decoder.unets = self.unets                  # swap in exponential moving averaged unets for sampling
@@ -687,6 +693,7 @@ class DecoderTrainer(nn.Module):
         for ema in self.ema_unets:
             ema.restore_ema_model_device()
 
+        base_decoder.train(was_training)
         return output
 
     @torch.no_grad()

dalle2_pytorch/version.py

@@ -1 +1 @@
-__version__ = '0.23.3'
+__version__ = '0.23.8'