bring in rotary embeddings for diffusion prior causal transformer (the most powerful relative positional encoding, used in PaLM) - 0.1.0 because of breaking change

README.md

@@ -967,7 +967,7 @@ Once built, images will be saved to the same directory the command is invoked
 - [x] make sure DDPMs can be run with traditional resnet blocks (but leave convnext as an option for experimentation)
 - [x] make sure for the latter unets in the cascade, one can train on crops for learning super resolution (constrain the unet to be only convolutions in that case, or allow conv-like attention with rel pos bias)
 - [ ] become an expert with unets, cleanup unet code, make it fully configurable, port all learnings over to https://github.com/lucidrains/x-unet (test out unet² in ddpm repo) - consider https://github.com/lucidrains/uformer-pytorch attention-based unet
-- [ ] make sure the cascading ddpm in the repository can be trained unconditionally, offer a one-line CLI tool for training on a folder of images
+- [ ] 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
 - [ ] transcribe code to Jax, which lowers the activation energy for distributed training, given access to TPUs
 - [ ] pull logic for training diffusion prior into a class DiffusionPriorTrainer, for eventual script based + CLI based training
 - [ ] train on a toy task, offer in colab
@@ -980,7 +980,6 @@ Once built, images will be saved to the same directory the command is invoked
 - [ ] interface out the vqgan-vae so a pretrained one can be pulled off the shelf to validate latent diffusion + DALL-E2
 - [ ] make sure FILIP works with DALL-E2 from x-clip https://arxiv.org/abs/2111.07783
 - [ ] make sure resnet hyperparameters can be configurable across unet depth (groups and expansion factor)
-- [ ] offer save / load methods on the trainer classes to automatically take care of state dicts for scalers / optimizers / saving versions and checking for breaking changes
 
 ## Citations
 

dalle2_pytorch/dalle2_pytorch.py

@@ -805,7 +805,6 @@ class DiffusionPrior(BaseGaussianDiffusion):
         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
         sampling_clamp_l2norm = False,
-        training_clamp_l2norm = False,
         image_embed_scale = None,           # this is for scaling the l2-normed image embedding, so it is more suitable for gaussian diffusion, as outlined by Katherine (@crowsonkb) https://github.com/lucidrains/DALLE2-pytorch/issues/60#issue-1226116132
         clip_adapter_overrides = dict()
     ):
@@ -843,7 +842,6 @@ class DiffusionPrior(BaseGaussianDiffusion):
 
         # whether to force an l2norm, similar to clipping denoised, when sampling
         self.sampling_clamp_l2norm = sampling_clamp_l2norm
-        self.training_clamp_l2norm = training_clamp_l2norm
 
     def p_mean_variance(self, x, t, text_cond, clip_denoised: bool):
         pred = self.net(x, t, **text_cond)
@@ -896,9 +894,6 @@ class DiffusionPrior(BaseGaussianDiffusion):
             **text_cond
         )
 
-        if self.predict_x_start and self.training_clamp_l2norm:
-            pred = l2norm(pred) * self.image_embed_scale
-
         target = noise if not self.predict_x_start else image_embed
 
         loss = self.loss_fn(pred, target)

setup.py

@@ -10,7 +10,7 @@ setup(
       'dream = dalle2_pytorch.cli:dream'
     ],
   },
-  version = '0.1.2',
+  version = '0.1.0',
   license='MIT',
   description = 'DALL-E 2',
   author = 'Phil Wang',
@@ -31,7 +31,6 @@ setup(
     'kornia>=0.5.4',
     'pillow',
     'resize-right>=0.0.2',
-    'rotary-embedding-torch',
     'torch>=1.10',
     'torchvision',
     'tqdm',