allow for one to stop grouping out weight decayable parameters, to debug optimizer state dict problem

README.md

@@ -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>

dalle2_pytorch/dalle2_pytorch.py

@@ -1704,6 +1704,8 @@ 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,
@@ -1826,6 +1828,11 @@ 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
@@ -1868,7 +1875,21 @@ 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:
@@ -21,11 +22,13 @@ def get_optimizer(
         return Adam(params, lr = lr, betas = betas, eps = eps)
 
     params = set(params)
-    wd_params, no_wd_params = separate_weight_decayable_params(params)
 
-    param_groups = [
-        {'params': list(wd_params)},
-        {'params': list(no_wd_params), 'weight_decay': 0},
-    ]
+    if group_wd_params:
+        wd_params, no_wd_params = separate_weight_decayable_params(params)
 
-    return AdamW(param_groups, lr = lr, weight_decay = wd, betas = betas, eps = eps)
+        params = [
+            {'params': list(wd_params)},
+            {'params': list(no_wd_params), 'weight_decay': 0},
+        ]
+
+    return AdamW(params, lr = lr, weight_decay = wd, betas = betas, eps = eps)

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 = '0.4.14',
+  version = '0.5.1',
   license='MIT',
   description = 'DALL-E 2',
   author = 'Phil Wang',