diff --git a/comfy/model_management.py b/comfy/model_management.py
index fb93c8acc..ee4191929 100644
--- a/comfy/model_management.py
+++ b/comfy/model_management.py
@@ -1674,6 +1674,8 @@ def synchronize():
         torch.cuda.synchronize()
 
 def soft_empty_cache(force=False):
+    if cpu_mode():
+        return
     global cpu_state
     if cpu_state == CPUState.CPU:
         return
diff --git a/comfy/ops.py b/comfy/ops.py
index 8275dd0a5..3e19cd1b6 100644
--- a/comfy/ops.py
+++ b/comfy/ops.py
@@ -660,23 +660,29 @@ class fp8_ops(manual_cast):
 
 CUBLAS_IS_AVAILABLE = False
 try:
-    from cublas_ops import CublasLinear
+    from cublas_ops import CublasLinear, cublas_half_matmul
     CUBLAS_IS_AVAILABLE = True
 except ImportError:
     pass
 
 if CUBLAS_IS_AVAILABLE:
-    class cublas_ops(disable_weight_init):
-        class Linear(CublasLinear, disable_weight_init.Linear):
+    class cublas_ops(manual_cast):
+        class Linear(CublasLinear, manual_cast.Linear):
             def reset_parameters(self):
                 return None
 
             def forward_comfy_cast_weights(self, input):
-                return super().forward(input)
+                weight, bias, offload_stream = cast_bias_weight(self, input, offloadable=True)
+                x = cublas_half_matmul(input, weight, bias, self._epilogue_str, self.has_bias)
+                uncast_bias_weight(self, weight, bias, offload_stream)
+                return x
 
             def forward(self, *args, **kwargs):
-                return super().forward(*args, **kwargs)
-
+                run_every_op()
+                if self.comfy_cast_weights or len(self.weight_function) > 0 or len(self.bias_function) > 0:
+                    return self.forward_comfy_cast_weights(*args, **kwargs)
+                else:
+                    return super().forward(*args, **kwargs)
 
 # ==============================================================================
 # Mixed Precision Operations
diff --git a/comfyui_version.py b/comfyui_version.py
index 6a35c6de3..0aea18d3a 100644
--- a/comfyui_version.py
+++ b/comfyui_version.py
@@ -1,3 +1,3 @@
 # This file is automatically generated by the build process when version is
 # updated in pyproject.toml.
-__version__ = "0.15.1"
+__version__ = "0.16.0"
diff --git a/pyproject.toml b/pyproject.toml
index 1b2318273..f2133d99c 100644
--- a/pyproject.toml
+++ b/pyproject.toml
@@ -1,6 +1,6 @@
 [project]
 name = "ComfyUI"
-version = "0.15.1"
+version = "0.16.0"
 readme = "README.md"
 license = { file = "LICENSE" }
 requires-python = ">=3.10"
diff --git a/requirements.txt b/requirements.txt
index dae46d873..866818e08 100644
--- a/requirements.txt
+++ b/requirements.txt
@@ -1,5 +1,5 @@
 comfyui-frontend-package==1.39.19
-comfyui-workflow-templates==0.9.5
+comfyui-workflow-templates==0.9.7
 comfyui-embedded-docs==0.4.3
 torch
 torchsde