0.5.4

README.md

@@ -12,7 +12,7 @@ This model is SOTA for text-to-image for now.
 
 Please join <a href="https://discord.gg/xBPBXfcFHd"><img alt="Join us on Discord" src="https://img.shields.io/discord/823813159592001537?color=5865F2&logo=discord&logoColor=white"></a> if you are interested in helping out with the replication with the <a href="https://laion.ai/">LAION</a> community | <a href="https://www.youtube.com/watch?v=AIOE1l1W0Tw">Yannic Interview</a>
 
-There was enough interest for a <a href="https://github.com/lucidrains/dalle2-jax">Jax version</a>. I will also eventually extend this to <a href="https://github.com/lucidrains/dalle2-video">text to video</a>, once the repository is in a good place.
+As of 5/23/22, it is no longer SOTA. SOTA will be <a href="https://github.com/lucidrains/imagen-pytorch">here</a>. Jax versions as well as text-to-video project will be shifted towards the Imagen architecture, as it is way simpler.
 
 ## Status
 
@@ -26,7 +26,7 @@ There was enough interest for a <a href="https://github.com/lucidrains/dalle2-ja
 
 ## Pre-Trained Models
 - LAION is training prior models. Checkpoints are available on <a href="https://huggingface.co/zenglishuci/conditioned-prior">🤗huggingface</a> and the training statistics are available on <a href="https://wandb.ai/nousr_laion/conditioned-prior/reports/LAION-DALLE2-PyTorch-Prior--VmlldzoyMDI2OTIx">🐝WANDB</a>.
-- Decoder 🚧
+- Decoder - <a href="https://wandb.ai/veldrovive/dalle2_train_decoder/runs/jkrtg0so?workspace=user-veldrovive">In-progress test run</a> 🚧
 - DALL-E 2 🚧
 
 ## Install
@@ -1195,4 +1195,12 @@ This library would not have gotten to this working state without the help of
 }
 ```
 
+```bibtex
+@misc{Saharia2022,
+    title   = {Imagen: unprecedented photorealism × deep level of language understanding},
+    author  = {Chitwan Saharia*, William Chan*, Saurabh Saxena†, Lala Li†, Jay Whang†, Emily Denton, Seyed Kamyar Seyed Ghasemipour, Burcu Karagol Ayan, S. Sara Mahdavi, Rapha Gontijo Lopes, Tim Salimans, Jonathan Ho†, David Fleet†, Mohammad Norouzi*},
+    year    = {2022}
+}
+```
+
 *Creating noise from data is easy; creating data from noise is generative modeling.* - <a href="https://arxiv.org/abs/2011.13456">Yang Song's paper</a>

configs/train_prior_config.example.json

@@ -0,0 +1,70 @@
+{
+    "prior": {
+        "clip": {
+            "make": "x-clip",
+            "model": "ViT-L/14",
+            "base_model_kwargs": {
+                "dim_text": 768,
+                "dim_image": 768,
+                "dim_latent": 768
+            }
+        },
+        "net": {
+            "dim": 768,
+            "depth": 12,
+            "num_timesteps": 1000,
+            "num_time_embeds": 1,
+            "num_image_embeds": 1,
+            "num_text_embeds": 1,
+            "dim_head": 64,
+            "heads": 12,
+            "ff_mult": 4,
+            "norm_out": true,
+            "attn_dropout": 0.0,
+            "ff_dropout": 0.0,
+            "final_proj": true,
+            "normformer": true,
+            "rotary_emb": true
+        },
+        "image_embed_dim": 768,
+        "image_size": 224,
+        "image_channels": 3,
+        "timesteps": 1000,
+        "cond_drop_prob": 0.1,
+        "loss_type": "l2",
+        "predict_x_start": true,
+        "beta_schedule": "cosine",
+        "condition_on_text_encodings": true
+    },
+    "data": {
+        "image_url": "https://mystic.the-eye.eu/public/AI/cah/laion5b/embeddings/laion2B-en/img_emb/",
+        "text_url": "https://mystic.the-eye.eu/public/AI/cah/laion5b/embeddings/laion2B-en/text_emb/",
+        "meta_url": "https://mystic.the-eye.eu/public/AI/cah/laion5b/embeddings/laion2B-en/laion2B-en-metadata/",
+        "batch_size": 256,
+        "splits": {
+            "train": 0.9,
+            "val": 1e-7,
+            "test": 0.0999999
+        }
+    },
+    "train": {
+        "epochs": 1,
+        "lr": 1.1e-4,
+        "wd": 6.02e-2,
+        "max_grad_norm": 0.5,
+        "use_ema": true,
+        "amp": false,
+        "save_every": 10000
+    },
+    "load": {
+        "source": null,
+        "resume": false
+    },
+    "tracker": {
+        "tracker_type": "wandb",
+        "data_path": "./prior_checkpoints",
+        "wandb_entity": "laion",
+        "wandb_project": "diffusion-prior",
+        "verbose": true
+    }
+}

dalle2_pytorch/dalle2_pytorch.py

@@ -890,6 +890,8 @@ class DiffusionPrior(BaseGaussianDiffusion):
         )
 
         if exists(clip):
+            assert image_channels == clip.image_channels, f'channels of image ({image_channels}) should be equal to the channels that CLIP accepts ({clip.image_channels})'
+
             if isinstance(clip, CLIP):
                 clip = XClipAdapter(clip, **clip_adapter_overrides)
             elif isinstance(clip, CoCa):
@@ -1105,13 +1107,20 @@ class Block(nn.Module):
         groups = 8
     ):
         super().__init__()
-        self.block = nn.Sequential(
+        self.project = nn.Conv2d(dim, dim_out, 3, padding = 1)
-            nn.Conv2d(dim, dim_out, 3, padding = 1),
+        self.norm = nn.GroupNorm(groups, dim_out)
-            nn.GroupNorm(groups, dim_out),
+        self.act = nn.SiLU()
-            nn.SiLU()
+
-        )
+    def forward(self, x, scale_shift = None):
-    def forward(self, x):
+        x = self.project(x)
-        return self.block(x)
+        x = self.norm(x)
+
+        if exists(scale_shift):
+            scale, shift = scale_shift
+            x = x * (scale + 1) + shift
+
+        x = self.act(x)
+        return x
 
 class ResnetBlock(nn.Module):
     def __init__(
@@ -1130,7 +1139,7 @@ class ResnetBlock(nn.Module):
         if exists(time_cond_dim):
             self.time_mlp = nn.Sequential(
                 nn.SiLU(),
-                nn.Linear(time_cond_dim, dim_out)
+                nn.Linear(time_cond_dim, dim_out * 2)
             )
 
         self.cross_attn = None
@@ -1150,11 +1159,14 @@ class ResnetBlock(nn.Module):
         self.res_conv = nn.Conv2d(dim, dim_out, 1) if dim != dim_out else nn.Identity()
 
     def forward(self, x, cond = None, time_emb = None):
-        h = self.block1(x)
 
+        scale_shift = None
         if exists(self.time_mlp) and exists(time_emb):
             time_emb = self.time_mlp(time_emb)
-            h = rearrange(time_emb, 'b c -> b c 1 1') + h
+            time_emb = rearrange(time_emb, 'b c -> b c 1 1')
+            scale_shift = time_emb.chunk(2, dim = 1)
+
+        h = self.block1(x, scale_shift = scale_shift)
 
         if exists(self.cross_attn):
             assert exists(cond)
@@ -1702,6 +1714,8 @@ class Decoder(BaseGaussianDiffusion):
         vb_loss_weight = 0.001,
         unconditional = False,
         auto_normalize_img = True,                  # whether to take care of normalizing the image from [0, 1] to [-1, 1] and back automatically - you can turn this off if you want to pass in the [-1, 1] ranged image yourself from the dataloader
+        use_dynamic_thres = False,                  # from the Imagen paper
+        dynamic_thres_percentile = 0.9
     ):
         super().__init__(
             beta_schedule = beta_schedule,
@@ -1710,12 +1724,19 @@ class Decoder(BaseGaussianDiffusion):
         )
 
         self.unconditional = unconditional
-        assert not (condition_on_text_encodings and unconditional), 'unconditional decoder image generation cannot be set to True if conditioning on text is present'
 
-        assert self.unconditional or (exists(clip) or exists(image_size) or exists(image_sizes)), 'either CLIP is supplied, or you must give the image_size and channels (usually 3 for RGB)'
+        # text conditioning
+
+        assert not (condition_on_text_encodings and unconditional), 'unconditional decoder image generation cannot be set to True if conditioning on text is present'
+        self.condition_on_text_encodings = condition_on_text_encodings
+
+        # clip
 
         self.clip = None
         if exists(clip):
+            assert not unconditional, 'clip must not be given if doing unconditional image training'
+            assert channels == clip.image_channels, f'channels of image ({channels}) should be equal to the channels that CLIP accepts ({clip.image_channels})'
+
             if isinstance(clip, CLIP):
                 clip = XClipAdapter(clip, **clip_adapter_overrides)
             elif isinstance(clip, CoCa):
@@ -1725,13 +1746,20 @@ class Decoder(BaseGaussianDiffusion):
             assert isinstance(clip, BaseClipAdapter)
 
             self.clip = clip
-            self.clip_image_size = clip.image_size
-            self.channels = clip.image_channels
-        else:
-            self.clip_image_size = default(image_size, lambda: image_sizes[-1])
-            self.channels = channels
 
-        self.condition_on_text_encodings = condition_on_text_encodings
+        # determine image size, with image_size and image_sizes taking precedence
+
+        if exists(image_size) or exists(image_sizes):
+            assert exists(image_size) ^ exists(image_sizes), 'only one of image_size or image_sizes must be given'
+            image_size = default(image_size, lambda: image_sizes[-1])
+        elif exists(clip):
+            image_size = clip.image_size
+        else:
+            raise Error('either image_size, image_sizes, or clip must be given to decoder')
+
+        # channels
+
+        self.channels = channels
 
         # automatically take care of ensuring that first unet is unconditional
         # while the rest of the unets are conditioned on the low resolution image produced by previous unet
@@ -1773,7 +1801,7 @@ class Decoder(BaseGaussianDiffusion):
 
         # unet image sizes
 
-        image_sizes = default(image_sizes, (self.clip_image_size,))
+        image_sizes = default(image_sizes, (image_size,))
         image_sizes = tuple(sorted(set(image_sizes)))
 
         assert len(self.unets) == len(image_sizes), f'you did not supply the correct number of u-nets ({len(self.unets)}) for resolutions {image_sizes}'
@@ -1810,7 +1838,13 @@ class Decoder(BaseGaussianDiffusion):
         self.clip_denoised = clip_denoised
         self.clip_x_start = clip_x_start
 
+        # dynamic thresholding settings, if clipping denoised during sampling
+
+        self.use_dynamic_thres = use_dynamic_thres
+        self.dynamic_thres_percentile = dynamic_thres_percentile
+
         # normalize and unnormalize image functions
+
         self.normalize_img = normalize_neg_one_to_one if auto_normalize_img else identity
         self.unnormalize_img = unnormalize_zero_to_one if auto_normalize_img else identity
 
@@ -1851,7 +1885,21 @@ class Decoder(BaseGaussianDiffusion):
             x_recon = self.predict_start_from_noise(x, t = t, noise = pred)
 
         if clip_denoised:
-            x_recon.clamp_(-1., 1.)
+            # s is the threshold amount
+            # static thresholding would just be s = 1
+            s = 1.
+            if self.use_dynamic_thres:
+                s = torch.quantile(
+                    rearrange(x_recon, 'b ... -> b (...)').abs(),
+                    self.dynamic_thres_percentile,
+                    dim = -1
+                )
+
+                s.clamp_(min = 1.)
+                s = s.view(-1, *((1,) * (x_recon.ndim - 1)))
+
+            # clip by threshold, depending on whether static or dynamic
+            x_recon = x_recon.clamp(-s, s) / s
 
         model_mean, posterior_variance, posterior_log_variance = self.q_posterior(x_start=x_recon, x_t=x, t=t)
 

dalle2_pytorch/optimizer.py

@@ -12,6 +12,7 @@ def get_optimizer(
     betas = (0.9, 0.999),
     eps = 1e-8,
     filter_by_requires_grad = False,
+    group_wd_params = True,
     **kwargs
 ):
     if filter_by_requires_grad:
@@ -20,12 +21,12 @@ def get_optimizer(
     if wd == 0:
         return Adam(params, lr = lr, betas = betas, eps = eps)
 
-    params = set(params)
+    if group_wd_params:
-    wd_params, no_wd_params = separate_weight_decayable_params(params)
+        wd_params, no_wd_params = separate_weight_decayable_params(params)
 
-    param_groups = [
+        params = [
-        {'params': list(wd_params)},
+            {'params': list(wd_params)},
-        {'params': list(no_wd_params), 'weight_decay': 0},
+            {'params': list(no_wd_params), 'weight_decay': 0},
-    ]
+        ]
 
-    return AdamW(param_groups, lr = lr, weight_decay = wd, betas = betas, eps = eps)
+    return AdamW(params, lr = lr, weight_decay = wd, betas = betas, eps = eps)

dalle2_pytorch/train_configs.py

@@ -3,7 +3,18 @@ from torchvision import transforms as T
 from pydantic import BaseModel, validator, root_validator
 from typing import List, Iterable, Optional, Union, Tuple, Dict, Any
 
-from dalle2_pytorch.dalle2_pytorch import Unet, Decoder, DiffusionPrior, DiffusionPriorNetwork
+from x_clip import CLIP as XCLIP
+from coca_pytorch import CoCa
+
+from dalle2_pytorch.dalle2_pytorch import (
+    CoCaAdapter,
+    OpenAIClipAdapter,
+    Unet,
+    Decoder,
+    DiffusionPrior,
+    DiffusionPriorNetwork,
+    XClipAdapter,
+)
 
 # helper functions
 
@@ -16,7 +27,44 @@ def default(val, d):
 def ListOrTuple(inner_type):
     return Union[List[inner_type], Tuple[inner_type]]
 
-# pydantic classes
+# general pydantic classes
+
+class TrainSplitConfig(BaseModel):
+    train: float = 0.75
+    val: float = 0.15
+    test: float = 0.1
+
+    @root_validator
+    def validate_all(cls, fields):
+        actual_sum = sum([*fields.values()])
+        if actual_sum != 1.:
+            raise ValueError(f'{fields.keys()} must sum to 1.0. Found: {actual_sum}')
+        return fields
+
+class TrackerConfig(BaseModel):
+    tracker_type: str = 'console'           # Decoder currently supports console and wandb
+    data_path: str = './models'             # The path where files will be saved locally
+    init_config: Dict[str, Any] = None
+    wandb_entity: str = ''                  # Only needs to be set if tracker_type is wandb
+    wandb_project: str = ''
+    verbose: bool = False                   # Whether to print console logging for non-console trackers
+
+# diffusion prior pydantic classes
+
+class AdapterConfig(BaseModel):
+    make: str = "openai"
+    model: str = "ViT-L/14"
+    base_model_kwargs: Dict[str, Any] = None
+
+    def create(self):
+        if self.make == "openai":
+            return OpenAIClipAdapter(self.model)
+        elif self.make == "x-clip":
+            return XClipAdapter(XCLIP(**self.base_model_kwargs))
+        elif self.make == "coca":
+            return CoCaAdapter(CoCa(**self.base_model_kwargs))
+        else:
+            raise AttributeError("No adapter with that name is available.")
 
 class DiffusionPriorNetworkConfig(BaseModel):
     dim: int
@@ -35,8 +83,12 @@ class DiffusionPriorNetworkConfig(BaseModel):
     normformer: bool = False
     rotary_emb: bool = True
 
+    def create(self):
+        kwargs = self.dict()
+        return DiffusionPriorNetwork(**kwargs)
+
 class DiffusionPriorConfig(BaseModel):
-    # only clip-less diffusion prior config for now
+    clip: AdapterConfig
     net: DiffusionPriorNetworkConfig
     image_embed_dim: int
     image_size: int
@@ -46,15 +98,52 @@ class DiffusionPriorConfig(BaseModel):
     loss_type: str = 'l2'
     predict_x_start: bool = True
     beta_schedule: str = 'cosine'
-
+    condition_on_text_encodings: bool = True
-    def create(self):
-        kwargs = self.dict()
-        diffusion_prior_network = DiffusionPriorNetwork(**kwargs.pop('net'))
-        return DiffusionPrior(net = diffusion_prior_network, **kwargs)
 
     class Config:
         extra = "allow"
 
+    def create(self):
+        kwargs = self.dict()
+        clip = AdapterConfig(**kwargs.pop('clip')).create()
+        diffusion_prior_network = DiffusionPriorNetworkConfig(**kwargs.pop('net')).create()
+        return DiffusionPrior(net = diffusion_prior_network, clip=clip, **kwargs)
+
+class DiffusionPriorTrainConfig(BaseModel):
+    epochs: int = 1
+    lr: float = 1.1e-4
+    wd: float = 6.02e-2
+    max_grad_norm: float = 0.5
+    use_ema: bool = True
+    ema_beta: float = 0.99
+    amp: bool = False
+    save_every: int = 10000 # what steps to save on
+
+class DiffusionPriorDataConfig(BaseModel):
+    image_url: str     # path to embeddings folder
+    meta_url: str      # path to metadata (captions) for images
+    splits: TrainSplitConfig
+    batch_size: int = 64
+
+class DiffusionPriorLoadConfig(BaseModel):
+    source: str = None
+    resume: bool = False
+
+class TrainDiffusionPriorConfig(BaseModel):
+    prior: DiffusionPriorConfig
+    data: DiffusionPriorDataConfig
+    train: DiffusionPriorTrainConfig
+    load: DiffusionPriorLoadConfig
+    tracker: TrackerConfig
+
+    @classmethod
+    def from_json_path(cls, json_path):
+        with open(json_path) as f:
+            config = json.load(f)
+        return cls(**config)
+
+# decoder pydantic classes
+
 class UnetConfig(BaseModel):
     dim: int
     dim_mults: ListOrTuple(int)
@@ -94,17 +183,6 @@ class DecoderConfig(BaseModel):
     class Config:
         extra = "allow"
 
-class TrainSplitConfig(BaseModel):
-    train: float = 0.75
-    val: float = 0.15
-    test: float = 0.1
-
-    @root_validator
-    def validate_all(cls, fields):
-        if sum([*fields.values()]) != 1.:
-            raise ValueError(f'{fields.keys()} must sum to 1.0')
-        return fields
-
 class DecoderDataConfig(BaseModel):
     webdataset_base_url: str     # path to a webdataset with jpg images
     embeddings_url: str          # path to .npy files with embeddings
@@ -160,14 +238,6 @@ class DecoderEvaluateConfig(BaseModel):
     KID: Dict[str, Any] = None
     LPIPS: Dict[str, Any] = None
 
-class TrackerConfig(BaseModel):
-    tracker_type: str = 'console'           # Decoder currently supports console and wandb
-    data_path: str = './models'             # The path where files will be saved locally
-    init_config: Dict[str, Any] = None
-    wandb_entity: str = ''                  # Only needs to be set if tracker_type is wandb
-    wandb_project: str = ''
-    verbose: bool = False                   # Whether to print console logging for non-console trackers
-
 class DecoderLoadConfig(BaseModel):
     source: str = None                      # Supports file and wandb
     run_path: str = ''                      # Used only if source is wandb

dalle2_pytorch/trainer.py

@@ -254,6 +254,7 @@ class DiffusionPriorTrainer(nn.Module):
         eps = 1e-6,
         max_grad_norm = None,
         amp = False,
+        group_wd_params = True,
         **kwargs
     ):
         super().__init__()
@@ -279,6 +280,7 @@ class DiffusionPriorTrainer(nn.Module):
             lr = lr,
             wd = wd,
             eps = eps,
+            group_wd_params = group_wd_params,
             **kwargs
         )
 
@@ -410,6 +412,7 @@ class DecoderTrainer(nn.Module):
         eps = 1e-8,
         max_grad_norm = 0.5,
         amp = False,
+        group_wd_params = True,
         **kwargs
     ):
         super().__init__()
@@ -435,6 +438,7 @@ class DecoderTrainer(nn.Module):
                 lr = unet_lr,
                 wd = unet_wd,
                 eps = unet_eps,
+                group_wd_params = group_wd_params,
                 **kwargs
             )
 

setup.py

@@ -10,7 +10,7 @@ setup(
       'dream = dalle2_pytorch.cli:dream'
     ],
   },
-  version = '0.4.10',
+  version = '0.5.4',
   license='MIT',
   description = 'DALL-E 2',
   author = 'Phil Wang',