[pull] master from comfyanonymous:master

comfy/k_diffusion/sampling.py

@@ -1447,27 +1447,34 @@ def post_cfg_function(args):
         old_d = d
     return x
 
+
 @torch.no_grad()
 def sample_gradient_estimation_cfg_pp(model, x, sigmas, extra_args=None, callback=None, disable=None, ge_gamma=2.):
     return sample_gradient_estimation(model, x, sigmas, extra_args=extra_args, callback=callback, disable=disable, ge_gamma=ge_gamma, cfg_pp=True)
 
+
 @torch.no_grad()
-def sample_er_sde(model, x, sigmas, extra_args=None, callback=None, disable=None, s_noise=1., noise_sampler=None, noise_scaler=None, max_stage=3):
-    """
-    Extended Reverse-Time SDE solver (VE ER-SDE-Solver-3). Arxiv: https://arxiv.org/abs/2309.06169.
+def sample_er_sde(model, x, sigmas, extra_args=None, callback=None, disable=None, s_noise=1.0, noise_sampler=None, noise_scaler=None, max_stage=3):
+    """Extended Reverse-Time SDE solver (VP ER-SDE-Solver-3). arXiv: https://arxiv.org/abs/2309.06169.
     Code reference: https://github.com/QinpengCui/ER-SDE-Solver/blob/main/er_sde_solver.py.
     """
     extra_args = {} if extra_args is None else extra_args
     seed = extra_args.get("seed", None)
     noise_sampler = default_noise_sampler(x, seed=seed) if noise_sampler is None else noise_sampler
     s_in = x.new_ones([x.shape[0]])
 
-    def default_noise_scaler(sigma):
-        return sigma * ((sigma ** 0.3).exp() + 10.0)
-    noise_scaler = default_noise_scaler if noise_scaler is None else noise_scaler
+    def default_er_sde_noise_scaler(x):
+        return x * ((x ** 0.3).exp() + 10.0)
+
+    noise_scaler = default_er_sde_noise_scaler if noise_scaler is None else noise_scaler
     num_integration_points = 200.0
     point_indice = torch.arange(0, num_integration_points, dtype=torch.float32, device=x.device)
 
+    model_sampling = model.inner_model.model_patcher.get_model_object("model_sampling")
+    sigmas = offset_first_sigma_for_snr(sigmas, model_sampling)
+    half_log_snrs = sigma_to_half_log_snr(sigmas, model_sampling)
+    er_lambdas = half_log_snrs.neg().exp()  # er_lambda_t = sigma_t / alpha_t
+
     old_denoised = None
     old_denoised_d = None
 
@@ -1478,32 +1485,36 @@ def default_noise_scaler(sigma):
         stage_used = min(max_stage, i + 1)
         if sigmas[i + 1] == 0:
             x = denoised
-        elif stage_used == 1:
-            r = noise_scaler(sigmas[i + 1]) / noise_scaler(sigmas[i])
-            x = r * x + (1 - r) * denoised
         else:
-            r = noise_scaler(sigmas[i + 1]) / noise_scaler(sigmas[i])
-            x = r * x + (1 - r) * denoised
-
-            dt = sigmas[i + 1] - sigmas[i]
-            sigma_step_size = -dt / num_integration_points
-            sigma_pos = sigmas[i + 1] + point_indice * sigma_step_size
-            scaled_pos = noise_scaler(sigma_pos)
-
-            # Stage 2
-            s = torch.sum(1 / scaled_pos) * sigma_step_size
-            denoised_d = (denoised - old_denoised) / (sigmas[i] - sigmas[i - 1])
-            x = x + (dt + s * noise_scaler(sigmas[i + 1])) * denoised_d
-
-            if stage_used >= 3:
-                # Stage 3
-                s_u = torch.sum((sigma_pos - sigmas[i]) / scaled_pos) * sigma_step_size
-                denoised_u = (denoised_d - old_denoised_d) / ((sigmas[i] - sigmas[i - 2]) / 2)
-                x = x + ((dt ** 2) / 2 + s_u * noise_scaler(sigmas[i + 1])) * denoised_u
-            old_denoised_d = denoised_d
-
-        if s_noise != 0 and sigmas[i + 1] > 0:
-            x = x + noise_sampler(sigmas[i], sigmas[i + 1]) * s_noise * (sigmas[i + 1] ** 2 - sigmas[i] ** 2 * r ** 2).sqrt().nan_to_num(nan=0.0)
+            er_lambda_s, er_lambda_t = er_lambdas[i], er_lambdas[i + 1]
+            alpha_s = sigmas[i] / er_lambda_s
+            alpha_t = sigmas[i + 1] / er_lambda_t
+            r_alpha = alpha_t / alpha_s
+            r = noise_scaler(er_lambda_t) / noise_scaler(er_lambda_s)
+
+            # Stage 1 Euler
+            x = r_alpha * r * x + alpha_t * (1 - r) * denoised
+
+            if stage_used >= 2:
+                dt = er_lambda_t - er_lambda_s
+                lambda_step_size = -dt / num_integration_points
+                lambda_pos = er_lambda_t + point_indice * lambda_step_size
+                scaled_pos = noise_scaler(lambda_pos)
+
+                # Stage 2
+                s = torch.sum(1 / scaled_pos) * lambda_step_size
+                denoised_d = (denoised - old_denoised) / (er_lambda_s - er_lambdas[i - 1])
+                x = x + alpha_t * (dt + s * noise_scaler(er_lambda_t)) * denoised_d
+
+                if stage_used >= 3:
+                    # Stage 3
+                    s_u = torch.sum((lambda_pos - er_lambda_s) / scaled_pos) * lambda_step_size
+                    denoised_u = (denoised_d - old_denoised_d) / ((er_lambda_s - er_lambdas[i - 2]) / 2)
+                    x = x + alpha_t * ((dt ** 2) / 2 + s_u * noise_scaler(er_lambda_t)) * denoised_u
+                old_denoised_d = denoised_d
+
+            if s_noise > 0:
+                x = x + alpha_t * noise_sampler(sigmas[i], sigmas[i + 1]) * s_noise * (er_lambda_t ** 2 - er_lambda_s ** 2 * r ** 2).sqrt().nan_to_num(nan=0.0)
         old_denoised = denoised
     return x
 

comfy_extras/nodes_custom_sampler.py

@@ -2,6 +2,7 @@
 import comfy.samplers
 import comfy.sample
 from comfy.k_diffusion import sampling as k_diffusion_sampling
+from comfy.comfy_types import IO, ComfyNodeABC, InputTypeDict
 import latent_preview
 import torch
 import comfy.utils
@@ -480,6 +481,46 @@ def get_sampler(self, order, rtol, atol, h_init, pcoeff, icoeff, dcoeff, accept_
                                                               "s_noise":s_noise })
         return (sampler, )
 
+
+class SamplerER_SDE(ComfyNodeABC):
+    @classmethod
+    def INPUT_TYPES(cls) -> InputTypeDict:
+        return {
+            "required": {
+                "solver_type": (IO.COMBO, {"options": ["ER-SDE", "Reverse-time SDE", "ODE"]}),
+                "max_stage": (IO.INT, {"default": 3, "min": 1, "max": 3}),
+                "eta": (
+                    IO.FLOAT,
+                    {"default": 1.0, "min": 0.0, "max": 100.0, "step": 0.01, "round": False, "tooltip": "Stochastic strength of reverse-time SDE.\nWhen eta=0, it reduces to deterministic ODE. This setting doesn't apply to ER-SDE solver type."},
+                ),
+                "s_noise": (IO.FLOAT, {"default": 1.0, "min": 0.0, "max": 100.0, "step": 0.01, "round": False}),
+            }
+        }
+
+    RETURN_TYPES = (IO.SAMPLER,)
+    CATEGORY = "sampling/custom_sampling/samplers"
+
+    FUNCTION = "get_sampler"
+
+    def get_sampler(self, solver_type, max_stage, eta, s_noise):
+        if solver_type == "ODE" or (solver_type == "Reverse-time SDE" and eta == 0):
+            eta = 0
+            s_noise = 0
+
+        def reverse_time_sde_noise_scaler(x):
+            return x ** (eta + 1)
+
+        if solver_type == "ER-SDE":
+            # Use the default one in sample_er_sde()
+            noise_scaler = None
+        else:
+            noise_scaler = reverse_time_sde_noise_scaler
+
+        sampler_name = "er_sde"
+        sampler = comfy.samplers.ksampler(sampler_name, {"s_noise": s_noise, "noise_scaler": noise_scaler, "max_stage": max_stage})
+        return (sampler,)
+
+
 class Noise_EmptyNoise:
     def __init__(self):
         self.seed = 0
@@ -787,6 +828,7 @@ def add_noise(self, model, noise, sigmas, latent_image):
     "SamplerDPMPP_SDE": SamplerDPMPP_SDE,
     "SamplerDPMPP_2S_Ancestral": SamplerDPMPP_2S_Ancestral,
     "SamplerDPMAdaptative": SamplerDPMAdaptative,
+    "SamplerER_SDE": SamplerER_SDE,
     "SplitSigmas": SplitSigmas,
     "SplitSigmasDenoise": SplitSigmasDenoise,
     "FlipSigmas": FlipSigmas,

comfy_extras/nodes_perpneg.py

@@ -4,6 +4,7 @@
 import comfy.samplers
 import comfy.utils
 import node_helpers
+import math
 
 def perp_neg(x, noise_pred_pos, noise_pred_neg, noise_pred_nocond, neg_scale, cond_scale):
     pos = noise_pred_pos - noise_pred_nocond
@@ -69,6 +70,12 @@ def predict_noise(self, x, timestep, model_options={}, seed=None):
         negative_cond = self.conds.get("negative", None)
         empty_cond = self.conds.get("empty_negative_prompt", None)
 
+        if model_options.get("disable_cfg1_optimization", False) == False:
+            if math.isclose(self.neg_scale, 0.0):
+                negative_cond = None
+                if math.isclose(self.cfg, 1.0):
+                    empty_cond = None
+
         conds = [positive_cond, negative_cond, empty_cond]
 
         out = comfy.samplers.calc_cond_batch(self.inner_model, conds, x, timestep, model_options)

comfy_extras/nodes_tcfg.py

@@ -0,0 +1,71 @@
+# TCFG: Tangential Damping Classifier-free Guidance - (arXiv: https://arxiv.org/abs/2503.18137)
+
+import torch
+
+from comfy.comfy_types import IO, ComfyNodeABC, InputTypeDict
+
+
+def score_tangential_damping(cond_score: torch.Tensor, uncond_score: torch.Tensor) -> torch.Tensor:
+    """Drop tangential components from uncond score to align with cond score."""
+    # (B, 1, ...)
+    batch_num = cond_score.shape[0]
+    cond_score_flat = cond_score.reshape(batch_num, 1, -1).float()
+    uncond_score_flat = uncond_score.reshape(batch_num, 1, -1).float()
+
+    # Score matrix A (B, 2, ...)
+    score_matrix = torch.cat((uncond_score_flat, cond_score_flat), dim=1)
+    try:
+        _, _, Vh = torch.linalg.svd(score_matrix, full_matrices=False)
+    except RuntimeError:
+        # Fallback to CPU
+        _, _, Vh = torch.linalg.svd(score_matrix.cpu(), full_matrices=False)
+
+    # Drop the tangential components
+    v1 = Vh[:, 0:1, :].to(uncond_score_flat.device)  # (B, 1, ...)
+    uncond_score_td = (uncond_score_flat @ v1.transpose(-2, -1)) * v1
+    return uncond_score_td.reshape_as(uncond_score).to(uncond_score.dtype)
+
+
+class TCFG(ComfyNodeABC):
+    @classmethod
+    def INPUT_TYPES(cls) -> InputTypeDict:
+        return {
+            "required": {
+                "model": (IO.MODEL, {}),
+            }
+        }
+
+    RETURN_TYPES = (IO.MODEL,)
+    RETURN_NAMES = ("patched_model",)
+    FUNCTION = "patch"
+
+    CATEGORY = "advanced/guidance"
+    DESCRIPTION = "TCFG – Tangential Damping CFG (2503.18137)\n\nRefine the uncond (negative) to align with the cond (positive) for improving quality."
+
+    def patch(self, model):
+        m = model.clone()
+
+        def tangential_damping_cfg(args):
+            #  Assume [cond, uncond, ...]
+            x = args["input"]
+            conds_out = args["conds_out"]
+            if len(conds_out) <= 1 or None in args["conds"][:2]:
+                # Skip when either cond or uncond is None
+                return conds_out
+            cond_pred = conds_out[0]
+            uncond_pred = conds_out[1]
+            uncond_td = score_tangential_damping(x - cond_pred, x - uncond_pred)
+            uncond_pred_td = x - uncond_td
+            return [cond_pred, uncond_pred_td] + conds_out[2:]
+
+        m.set_model_sampler_pre_cfg_function(tangential_damping_cfg)
+        return (m,)
+
+
+NODE_CLASS_MAPPINGS = {
+    "TCFG": TCFG,
+}
+
+NODE_DISPLAY_NAME_MAPPINGS = {
+    "TCFG": "Tangential Damping CFG",
+}

nodes.py

@@ -2283,6 +2283,7 @@ def init_builtin_extra_nodes():
         "nodes_string.py",
         "nodes_camera_trajectory.py",
         "nodes_edit_model.py",
+        "nodes_tcfg.py"
     ]
 
     import_failed = []