handel inplace optimizer correctly for shared codebook under residual vq setting

Author: lucidrains

pyproject.toml

@@ -1,6 +1,6 @@
 [project]
 name = "vector-quantize-pytorch"
-version = "1.17.5"
+version = "1.17.6"
 description = "Vector Quantization - Pytorch"
 authors = [
     { name = "Phil Wang", email = "lucidrains@gmail.com" }

vector_quantize_pytorch/residual_vq.py

@@ -139,7 +139,8 @@ def __init__(
 
         if shared_codebook:
             vq_kwargs.update(
-                manual_ema_update = True
+                manual_ema_update = True,
+                manual_in_place_optimizer_update = True
             )
 
         self.layers = ModuleList([VectorQuantize(dim = codebook_dim, codebook_dim = codebook_dim, accept_image_fmap = accept_image_fmap, **vq_kwargs) for _ in range(num_quantizers)])
@@ -360,6 +361,7 @@ def forward(
 
         if self.shared_codebook:
             first(self.layers)._codebook.update_ema()
+            first(self.layers).update_in_place_optimizer()
 
         # project out, if needed
 

vector_quantize_pytorch/vector_quantize_pytorch.py

@@ -818,6 +818,7 @@ def __init__(
         manual_ema_update = False,
         learnable_codebook = False,
         in_place_codebook_optimizer: Callable[..., Optimizer] = None, # Optimizer used to update the codebook embedding if using learnable_codebook
+        manual_in_place_optimizer_update = False,
         affine_param = False,
         affine_param_batch_decay = 0.99,
         affine_param_codebook_decay = 0.9, 
@@ -913,6 +914,7 @@ def __init__(
         self._codebook = codebook_class(**codebook_kwargs)
 
         self.in_place_codebook_optimizer = in_place_codebook_optimizer(self._codebook.parameters()) if exists(in_place_codebook_optimizer) else None
+        self.manual_in_place_optimizer_update = manual_in_place_optimizer_update
 
         self.codebook_size = codebook_size
 
@@ -966,6 +968,13 @@ def get_output_from_indices(self, indices):
         codes = self.get_codes_from_indices(indices)
         return self.project_out(codes)
 
+    def update_in_place_optimizer(self):
+        if not exists(self.in_place_codebook_optimizer):
+            return
+
+        self.in_place_codebook_optimizer.step()
+        self.in_place_codebook_optimizer.zero_grad()
+
     def forward(
         self,
         x,
@@ -1057,8 +1066,9 @@ def forward(
                 loss = F.mse_loss(quantize, x.detach())
 
             loss.backward()
-            self.in_place_codebook_optimizer.step()
-            self.in_place_codebook_optimizer.zero_grad()
+
+            if not self.manual_in_place_optimizer_update:
+                self.update_in_place_optimizer()
 
             inplace_optimize_loss = loss