Skip to content

The most powerful and modular diffusion model GUI, api and backend with a graph/nodes interface. The fastest local inference engine in the world.

Use the lowvram cast_to function for everything.

comfyanonymous67158994a435
diff --git a/comfy/model_management.py b/comfy/model_management.py
index 2346d4acb..0d5b07305 100644
--- a/comfy/model_management.py
+++ b/comfy/model_management.py
@@ -840,27 +840,21 @@ def force_channels_last():
#TODO
return False
-def cast_to_device(tensor, device, dtype, copy=False):
- device_supports_cast = False
- if tensor.dtype == torch.float32 or tensor.dtype == torch.float16:
- device_supports_cast = True
- elif tensor.dtype == torch.bfloat16:
- if hasattr(device, 'type') and device.type.startswith("cuda"):
- device_supports_cast = True
- elif is_intel_xpu():
- device_supports_cast = True
+def cast_to(weight, dtype=None, device=None, non_blocking=False, copy=False):
+ if device is None or weight.device == device:
+ if not copy:
+ if dtype is None or weight.dtype == dtype:
+ return weight
+ return weight.to(dtype=dtype, copy=copy)
- non_blocking = device_should_use_non_blocking(device)
+ r = torch.empty_like(weight, dtype=dtype, device=device)
+ r.copy_(weight, non_blocking=non_blocking)
+ return r
+
+def cast_to_device(tensor, device, dtype, copy=False):
+ non_blocking = device_supports_non_blocking(device)
+ return cast_to(tensor, dtype=dtype, device=device, non_blocking=non_blocking, copy=copy)
- if device_supports_cast:
- if copy:
- if tensor.device == device:
- return tensor.to(dtype, copy=copy, non_blocking=non_blocking)
- return tensor.to(device, copy=copy, non_blocking=non_blocking).to(dtype, non_blocking=non_blocking)
- else:
- return tensor.to(device, non_blocking=non_blocking).to(dtype, non_blocking=non_blocking)
- else:
- return tensor.to(device, dtype, copy=copy, non_blocking=non_blocking)
def xformers_enabled():
global directml_enabled
diff --git a/comfy/ops.py b/comfy/ops.py
index c90e25ead..a8bfe1ea7 100644
--- a/comfy/ops.py
+++ b/comfy/ops.py
@@ -20,19 +20,10 @@ import torch
import comfy.model_management
from comfy.cli_args import args
-def cast_to(weight, dtype=None, device=None, non_blocking=False, copy=False):
- if device is None or weight.device == device:
- if not copy:
- if dtype is None or weight.dtype == dtype:
- return weight
- return weight.to(dtype=dtype, copy=copy)
-
- r = torch.empty_like(weight, dtype=dtype, device=device)
- r.copy_(weight, non_blocking=non_blocking)
- return r
+cast_to = comfy.model_management.cast_to #TODO: remove once no more references
def cast_to_input(weight, input, non_blocking=False, copy=True):
- return cast_to(weight, input.dtype, input.device, non_blocking=non_blocking, copy=copy)
+ return comfy.model_management.cast_to(weight, input.dtype, input.device, non_blocking=non_blocking, copy=copy)
def cast_bias_weight(s, input=None, dtype=None, device=None, bias_dtype=None):
if input is not None:
@@ -47,12 +38,12 @@ def cast_bias_weight(s, input=None, dtype=None, device=None, bias_dtype=None):
non_blocking = comfy.model_management.device_supports_non_blocking(device)
if s.bias is not None:
has_function = s.bias_function is not None
- bias = cast_to(s.bias, bias_dtype, device, non_blocking=non_blocking, copy=has_function)
+ bias = comfy.model_management.cast_to(s.bias, bias_dtype, device, non_blocking=non_blocking, copy=has_function)
if has_function:
bias = s.bias_function(bias)
has_function = s.weight_function is not None
- weight = cast_to(s.weight, dtype, device, non_blocking=non_blocking, copy=has_function)
+ weight = comfy.model_management.cast_to(s.weight, dtype, device, non_blocking=non_blocking, copy=has_function)
if has_function:
weight = s.weight_function(weight)
return weight, bias