fix a bug with numerical stability in attention, sorry! 🐛

* add eval/train toggles

* train/eval flags

* shift train toggle

Co-authored-by: nousr <z@localhost.com>

dalle2_pytorch/dalle2_pytorch.py

@@ -677,7 +677,7 @@ class Attention(nn.Module):
 
         # attention
 
-        sim = sim - sim.amax(dim = -1, keepdim = True)
+        sim = sim - sim.amax(dim = -1, keepdim = True).detach()
         attn = sim.softmax(dim = -1)
         attn = self.dropout(attn)
 
@@ -1204,7 +1204,7 @@ class CrossAttention(nn.Module):
             mask = rearrange(mask, 'b j -> b 1 1 j')
             sim = sim.masked_fill(~mask, max_neg_value)
 
-        sim = sim - sim.amax(dim = -1, keepdim = True)
+        sim = sim - sim.amax(dim = -1, keepdim = True).detach()
         attn = sim.softmax(dim = -1)
 
         out = einsum('b h i j, b h j d -> b h i d', attn, v)

setup.py

@@ -10,7 +10,7 @@ setup(
       'dream = dalle2_pytorch.cli:dream'
     ],
   },
-  version = '0.2.1',
+  version = '0.2.2',
   license='MIT',
   description = 'DALL-E 2',
   author = 'Phil Wang',

train_diffusion_prior.py

@@ -42,6 +42,7 @@ def eval_model(model,device,image_reader,text_reader,start,end,batch_size,loss_t
         wandb.log({f'{phase} {loss_type}': avg_loss})
 
 def report_cosine_sims(diffusion_prior,image_reader,text_reader,train_set_size,NUM_TEST_EMBEDDINGS,device):
+    diffusion_prior.eval()
 
     cos = nn.CosineSimilarity(dim=1, eps=1e-6)
 
@@ -119,6 +120,7 @@ def train(image_embed_dim,
           learning_rate=0.001,
           max_grad_norm=0.5,
           weight_decay=0.01,
+          dropout=0.05,
           amp=False):
 
     # DiffusionPriorNetwork 
@@ -127,6 +129,8 @@ def train(image_embed_dim,
             depth = dpn_depth, 
             dim_head = dpn_dim_head, 
             heads = dpn_heads,
+            attn_dropout = dropout,
+            ff_dropout = dropout,
             normformer = dp_normformer).to(device)
     
     # DiffusionPrior with text embeddings and image embeddings pre-computed
@@ -167,10 +171,12 @@ def train(image_embed_dim,
     eval_start = train_set_size
 
     for _ in range(epochs):
-        diffusion_prior.train()
 
         for emb_images,emb_text in zip(image_reader(batch_size=batch_size, start=0, end=train_set_size),
                 text_reader(batch_size=batch_size, start=0, end=train_set_size)):
+
+            diffusion_prior.train()
+            
             emb_images_tensor = torch.tensor(emb_images[0]).to(device)
             emb_text_tensor = torch.tensor(emb_text[0]).to(device)
 
@@ -244,6 +250,7 @@ def main():
     # Hyperparameters
     parser.add_argument("--learning-rate", type=float, default=1.1e-4)
     parser.add_argument("--weight-decay", type=float, default=6.02e-2)
+    parser.add_argument("--dropout", type=float, default=5e-2)
     parser.add_argument("--max-grad-norm", type=float, default=0.5)
     parser.add_argument("--batch-size", type=int, default=10**4)
     parser.add_argument("--num-epochs", type=int, default=5)
@@ -261,7 +268,6 @@ def main():
     # DiffusionPrior(dp) parameters
     parser.add_argument("--dp-condition-on-text-encodings", type=bool, default=False)
     parser.add_argument("--dp-timesteps", type=int, default=100)
-    parser.add_argument("--dp-l2norm-output", type=bool, default=False)
     parser.add_argument("--dp-normformer", type=bool, default=False)
     parser.add_argument("--dp-cond-drop-prob", type=float, default=0.1)
     parser.add_argument("--dp-loss-type", type=str, default="l2")
@@ -340,6 +346,7 @@ def main():
           args.learning_rate,
           args.max_grad_norm,
           args.weight_decay,
+          args.dropout,
           args.amp)
 
 if __name__ == "__main__":