298977887/ComfyUI
1
0
Fork 0
mirror of https://github.com/comfyanonymous/ComfyUI.git synced 2025-03-12 22:02:14 +00:00
Code Issues Packages Projects Releases Wiki Activity

Add noise augmentation setting to unCLIPConditioning.

Browse Source
This commit is contained in:
comfyanonymous 2023-04-03 13:50:29 -04:00
parent 72f9235a49
commit f50b1fec69
2 changed files with 16 additions and 5 deletions
Show Stats Download Patch File Download Diff File Expand all files Collapse all files

16
comfy/samplers.py
View File

@ -348,17 +348,27 @@ def encode_adm(noise_augmentor, conds, batch_size, device):
if 'adm' in x[1]: if 'adm' in x[1]:
adm_inputs = [] adm_inputs = []
weights = [] weights = []
noise_aug = []
adm_in = x[1]["adm"] adm_in = x[1]["adm"]
for adm_c in adm_in: for adm_c in adm_in:
adm_cond = adm_c[0].image_embeds adm_cond = adm_c[0].image_embeds
weight = adm_c[1] weight = adm_c[1]
c_adm, noise_level_emb = noise_augmentor(adm_cond.to(device), noise_level=torch.tensor([0], device=device)) noise_augment = adm_c[2]
noise_level = round((noise_augmentor.max_noise_level - 1) * noise_augment)
c_adm, noise_level_emb = noise_augmentor(adm_cond.to(device), noise_level=torch.tensor([noise_level], device=device))
adm_out = torch.cat((c_adm, noise_level_emb), 1) * weight adm_out = torch.cat((c_adm, noise_level_emb), 1) * weight
weights.append(weight) weights.append(weight)
noise_aug.append(noise_augment)
adm_inputs.append(adm_out) adm_inputs.append(adm_out)
adm_out = torch.stack(adm_inputs).sum(0) if len(noise_aug) > 1:
#TODO: Apply Noise to Embedding Mix adm_out = torch.stack(adm_inputs).sum(0)
#TODO: add a way to control this
noise_augment = 0.05
noise_level = round((noise_augmentor.max_noise_level - 1) * noise_augment)
print(noise_level)
c_adm, noise_level_emb = noise_augmentor(adm_out[:, :noise_augmentor.time_embed.dim], noise_level=torch.tensor([noise_level], device=device))
adm_out = torch.cat((c_adm, noise_level_emb), 1)
else: else:
adm_out = torch.zeros((1, noise_augmentor.time_embed.dim * 2), device=device) adm_out = torch.zeros((1, noise_augmentor.time_embed.dim * 2), device=device)
x[1] = x[1].copy() x[1] = x[1].copy()

5
nodes.py
View File

@ -445,17 +445,18 @@ class unCLIPConditioning:
return {"required": {"conditioning": ("CONDITIONING", ), return {"required": {"conditioning": ("CONDITIONING", ),
"clip_vision_output": ("CLIP_VISION_OUTPUT", ), "clip_vision_output": ("CLIP_VISION_OUTPUT", ),
"strength": ("FLOAT", {"default": 1.0, "min": -10.0, "max": 10.0, "step": 0.01}), "strength": ("FLOAT", {"default": 1.0, "min": -10.0, "max": 10.0, "step": 0.01}),
"noise_augmentation": ("FLOAT", {"default": 0.0, "min": 0.0, "max": 1.0, "step": 0.01}),
}} }}
RETURN_TYPES = ("CONDITIONING",) RETURN_TYPES = ("CONDITIONING",)
FUNCTION = "apply_adm" FUNCTION = "apply_adm"
CATEGORY = "_for_testing/unclip" CATEGORY = "_for_testing/unclip"
def apply_adm(self, conditioning, clip_vision_output, strength): def apply_adm(self, conditioning, clip_vision_output, strength, noise_augmentation):
c = [] c = []
for t in conditioning: for t in conditioning:
o = t[1].copy() o = t[1].copy()
x = (clip_vision_output, strength) x = (clip_vision_output, strength, noise_augmentation)
if "adm" in o: if "adm" in o:
o["adm"] = o["adm"][:] + [x] o["adm"] = o["adm"][:] + [x]
else: else: