198 lines
8.3 KiB
Python
198 lines
8.3 KiB
Python
from inspect import cleandoc
|
|
|
|
import torch
|
|
|
|
from .helper_functions import getIndexTensorAlongDim, comonLazy, parse_expr, eval_tensor_expr_with_tree, make_zero_like
|
|
|
|
from comfy_api.latest import io
|
|
|
|
import comfy.nested_tensor as _nested_tensor_module
|
|
|
|
|
|
|
|
class NoiseMathNode(io.ComfyNode):
|
|
"""
|
|
This node enables the use of math expressions on Latents.
|
|
inputs:
|
|
a, b, c, d:
|
|
Noise generators, bound to variables with the same name. Defaults to zero latent if not provided.
|
|
w, x, y, z:
|
|
Floats, bound to variables of the expression. Defaults to 0.0 if not provided.
|
|
Latent expression:
|
|
String, describing expression to mix noise.
|
|
|
|
outputs:
|
|
LATENT:
|
|
Returns a LATENT object that contains the result of the math expression applied to the input conditionings.
|
|
"""
|
|
def __init__(self):
|
|
pass
|
|
@classmethod
|
|
def define_schema(cls) -> io.Schema:
|
|
"""
|
|
"""
|
|
return io.Schema(
|
|
node_id="mrmth_NoiseMathNode",
|
|
category="More math",
|
|
display_name="Noise math",
|
|
inputs=[
|
|
io.Noise.Input(id="a"),
|
|
io.Noise.Input(id="b", optional=True,lazy=True),
|
|
io.Noise.Input(id="c", optional=True,lazy=True),
|
|
io.Noise.Input(id="d", optional=True,lazy=True),
|
|
io.Float.Input(id="w", default=0.0,optional=True,lazy=True, force_input=True),
|
|
io.Float.Input(id="x", default=0.0,optional=True,lazy=True, force_input=True),
|
|
io.Float.Input(id="y", default=0.0,optional=True,lazy=True, force_input=True),
|
|
io.Float.Input(id="z", default=0.0,optional=True,lazy=True, force_input=True),
|
|
io.String.Input(id="Noise", default="a*(1-w)+b*w", tooltip="Expression to apply on input noise generators"),
|
|
],
|
|
outputs=[
|
|
io.Noise.Output(),
|
|
],
|
|
)
|
|
#RETURN_NAMES = ("image_output_name",)
|
|
tooltip = cleandoc(__doc__)
|
|
#OUTPUT_NODE = False
|
|
#OUTPUT_TOOLTIPS = ("",) # Tooltips for the output node
|
|
|
|
CATEGORY = "More math"
|
|
@classmethod
|
|
def check_lazy_status(cls, Noise, a, b=[], c=[], d=[],w=0,x=0,y=0,z=0):
|
|
return comonLazy(Noise, a, b, c, d,w,x,y,z)
|
|
@classmethod
|
|
def execute(cls, Noise, a, b=None, c=None, d=None, w=0.0, x=0.0, y=0.0, z=0.0):
|
|
return (NoiseExecutor(a, b, c, d, w, x, y, z, Noise),)
|
|
|
|
"""
|
|
The node will always be re executed if any of the inputs change but
|
|
this method can be used to force the node to execute again even when the inputs don't change.
|
|
You can make this node return a number or a string. This value will be compared to the one returned the last time the node was
|
|
executed, if it is different the node will be executed again.
|
|
This method is used in the core repo for the LoadImage node where they return the image hash as a string, if the image hash
|
|
changes between executions the LoadImage node is executed again.
|
|
"""
|
|
#@classmethod
|
|
#def IS_CHANGED(s, image, string_field, int_field, float_field, print_to_screen):
|
|
# return ""
|
|
|
|
|
|
|
|
|
|
class NoiseExecutor():
|
|
def __init__(self,a,b,c,d,w,x,y,z, Noise):
|
|
self.a = a
|
|
self.b = b
|
|
self.c = c
|
|
self.d = d
|
|
self.w = w
|
|
self.x = x
|
|
self.y = y
|
|
self.z = z
|
|
self.expr = Noise
|
|
self.tree = parse_expr(Noise)
|
|
seed = -1;
|
|
def generate_noise(self, input_latent:torch.Tensor) -> torch.Tensor:
|
|
samples = input_latent["samples"]
|
|
|
|
a_val = self.a.generate_noise(input_latent) if self.a is not None else make_zero_like(samples)
|
|
b_val = self.b.generate_noise(input_latent) if self.b is not None else make_zero_like(samples)
|
|
c_val = self.c.generate_noise(input_latent) if self.c is not None else make_zero_like(samples)
|
|
d_val = self.d.generate_noise(input_latent) if self.d is not None else make_zero_like(samples)
|
|
|
|
# helper to convert a returned value into a list matching ref_list
|
|
def to_list(val, ref_list):
|
|
if val is None:
|
|
return [make_zero_like(r) for r in ref_list]
|
|
# If val is a NestedTensor-like, return underlying list
|
|
if hasattr(val, 'is_nested') and getattr(val, 'is_nested'):
|
|
return val.unbind()
|
|
if isinstance(val, list) or isinstance(val, tuple):
|
|
return list(val)
|
|
# If val is a single tensor that encodes multiple subtensors along batch dim,
|
|
# try to split it into pieces that match ref_list batch sizes.
|
|
if torch.is_tensor(val):
|
|
try:
|
|
sizes = [r.shape[0] for r in ref_list]
|
|
if val.shape[0] == sum(sizes):
|
|
return list(val.split(sizes, dim=0))
|
|
except Exception:
|
|
pass
|
|
# single tensor broadcast
|
|
return [val for _ in ref_list]
|
|
|
|
# nested case: merge subtensors, evaluate once, split back
|
|
if hasattr(samples, 'is_nested') and getattr(samples, 'is_nested'):
|
|
sample_list = samples.unbind()
|
|
sizes = [t.shape[0] for t in sample_list]
|
|
merged_samples = torch.cat(sample_list, dim=0)
|
|
|
|
def merge_to_tensor(val, ref):
|
|
if val is None:
|
|
return make_zero_like(ref)
|
|
if hasattr(val, 'is_nested') and getattr(val, 'is_nested'):
|
|
lst = val.unbind()
|
|
return torch.cat(lst, dim=0)
|
|
if isinstance(val, (list, tuple)):
|
|
return torch.cat(list(val), dim=0)
|
|
if torch.is_tensor(val):
|
|
if val.shape == ref.shape:
|
|
return val
|
|
if val.shape[0] == sum(sizes):
|
|
return val
|
|
# if val has per-subtensor batches, try to split and cat
|
|
try:
|
|
if val.shape[0] == len(sample_list):
|
|
return torch.cat([val[i].unsqueeze(0).expand(sample_list[i].shape[0], *val.shape[1:]) for i in range(len(sample_list))], dim=0)
|
|
except Exception:
|
|
pass
|
|
return make_zero_like(ref)
|
|
|
|
merged_a = merge_to_tensor(a_val, merged_samples)
|
|
merged_b = merge_to_tensor(b_val, merged_samples)
|
|
merged_c = merge_to_tensor(c_val, merged_samples)
|
|
merged_d = merge_to_tensor(d_val, merged_samples)
|
|
|
|
else:
|
|
merged_samples = samples
|
|
merged_a = a_val
|
|
merged_b = b_val
|
|
merged_c = c_val
|
|
merged_d = d_val
|
|
|
|
# evaluate once
|
|
ndim = merged_samples.ndim
|
|
batch_dim = 0
|
|
channel_dim = -3
|
|
height_dim = -2
|
|
width_dim = -1
|
|
time_dim = None
|
|
if ndim >= 5:
|
|
time_dim = -4
|
|
|
|
frame_count = merged_samples.shape[time_dim] if time_dim is not None else merged_samples.shape[batch_dim]
|
|
|
|
B = getIndexTensorAlongDim(merged_samples, batch_dim)
|
|
W = getIndexTensorAlongDim(merged_samples, width_dim)
|
|
H = getIndexTensorAlongDim(merged_samples, height_dim)
|
|
C = getIndexTensorAlongDim(merged_samples, channel_dim)
|
|
variables = {
|
|
'a': merged_a, 'b': merged_b, 'c': merged_c, 'd': merged_d,
|
|
'w': self.w, 'x': self.x, 'y': self.y, 'z': self.z,
|
|
'B': B, 'X': W, 'Y': H, 'C': C,
|
|
'W': merged_samples.shape[width_dim], 'H': merged_samples.shape[height_dim],
|
|
'I': merged_samples, 'T': frame_count, 'N': merged_samples.shape[channel_dim],
|
|
'batch': B, 'width': merged_samples.shape[width_dim], 'height': merged_samples.shape[height_dim], 'channel': C,
|
|
'batch_count': merged_samples.shape[0], 'channel_count': merged_samples.shape[1], 'input_latent': merged_samples,
|
|
}
|
|
if time_dim is not None:
|
|
F = getIndexTensorAlongDim(merged_samples, time_dim)
|
|
variables.update({'frame': F, 'frame_count': frame_count})
|
|
|
|
merged_result = eval_tensor_expr_with_tree(self.tree, variables, variables['a'].shape)
|
|
|
|
if hasattr(samples, 'is_nested') and getattr(samples, 'is_nested'):
|
|
split_results = list(merged_result.split(sizes, dim=0))
|
|
return _nested_tensor_module.NestedTensor(split_results)
|
|
|
|
return merged_result
|