Add a TomePatchModel node to the _for_testing section.

Tome increases sampling speed at the expense of quality.

comfy/ldm/modules/attention.py

@@ -11,6 +11,7 @@ from .sub_quadratic_attention import efficient_dot_product_attention
 
 import model_management
 
+from . import tomesd
 
 if model_management.xformers_enabled():
     import xformers
@@ -508,8 +509,18 @@ class BasicTransformerBlock(nn.Module):
         return checkpoint(self._forward, (x, context, transformer_options), self.parameters(), self.checkpoint)
 
     def _forward(self, x, context=None, transformer_options={}):
-        x = self.attn1(self.norm1(x), context=context if self.disable_self_attn else None) + x
-        x = self.attn2(self.norm2(x), context=context) + x
+        n = self.norm1(x)
+        if "tomesd" in transformer_options:
+            m, u = tomesd.get_functions(x, transformer_options["tomesd"]["ratio"], transformer_options["original_shape"])
+            n = u(self.attn1(m(n), context=context if self.disable_self_attn else None))
+        else:
+            n = self.attn1(n, context=context if self.disable_self_attn else None)
+
+        x += n
+        n = self.norm2(x)
+        n = self.attn2(n, context=context)
+
+        x += n
         x = self.ff(self.norm3(x)) + x
         return x
 

comfy/ldm/modules/tomesd.py

@@ -0,0 +1,117 @@
+
+
+import torch
+from typing import Tuple, Callable
+import math
+
+def do_nothing(x: torch.Tensor, mode:str=None):
+    return x
+
+
+def bipartite_soft_matching_random2d(metric: torch.Tensor,
+                                     w: int, h: int, sx: int, sy: int, r: int,
+                                     no_rand: bool = False) -> Tuple[Callable, Callable]:
+    """
+    Partitions the tokens into src and dst and merges r tokens from src to dst.
+    Dst tokens are partitioned by choosing one randomy in each (sx, sy) region.
+
+    Args:
+     - metric [B, N, C]: metric to use for similarity
+     - w: image width in tokens
+     - h: image height in tokens
+     - sx: stride in the x dimension for dst, must divide w
+     - sy: stride in the y dimension for dst, must divide h
+     - r: number of tokens to remove (by merging)
+     - no_rand: if true, disable randomness (use top left corner only)
+    """
+    B, N, _ = metric.shape
+
+    if r <= 0:
+        return do_nothing, do_nothing
+    
+    with torch.no_grad():
+        
+        hsy, wsx = h // sy, w // sx
+
+        # For each sy by sx kernel, randomly assign one token to be dst and the rest src
+        idx_buffer = torch.zeros(1, hsy, wsx, sy*sx, 1, device=metric.device)
+
+        if no_rand:
+            rand_idx = torch.zeros(1, hsy, wsx, 1, 1, device=metric.device, dtype=torch.int64)
+        else:
+            rand_idx = torch.randint(sy*sx, size=(1, hsy, wsx, 1, 1), device=metric.device)
+        
+        idx_buffer.scatter_(dim=3, index=rand_idx, src=-torch.ones_like(rand_idx, dtype=idx_buffer.dtype))
+        idx_buffer = idx_buffer.view(1, hsy, wsx, sy, sx, 1).transpose(2, 3).reshape(1, N, 1)
+        rand_idx   = idx_buffer.argsort(dim=1)
+
+        num_dst = int((1 / (sx*sy)) * N)
+        a_idx = rand_idx[:, num_dst:, :] # src
+        b_idx = rand_idx[:, :num_dst, :] # dst
+
+        def split(x):
+            C = x.shape[-1]
+            src = x.gather(dim=1, index=a_idx.expand(B, N - num_dst, C))
+            dst = x.gather(dim=1, index=b_idx.expand(B, num_dst, C))
+            return src, dst
+
+        metric = metric / metric.norm(dim=-1, keepdim=True)
+        a, b = split(metric)
+        scores = a @ b.transpose(-1, -2)
+
+        # Can't reduce more than the # tokens in src
+        r = min(a.shape[1], r)
+
+        node_max, node_idx = scores.max(dim=-1)
+        edge_idx = node_max.argsort(dim=-1, descending=True)[..., None]
+
+        unm_idx = edge_idx[..., r:, :]  # Unmerged Tokens
+        src_idx = edge_idx[..., :r, :]  # Merged Tokens
+        dst_idx = node_idx[..., None].gather(dim=-2, index=src_idx)
+
+    def merge(x: torch.Tensor, mode="mean") -> torch.Tensor:
+        src, dst = split(x)
+        n, t1, c = src.shape
+        
+        unm = src.gather(dim=-2, index=unm_idx.expand(n, t1 - r, c))
+        src = src.gather(dim=-2, index=src_idx.expand(n, r, c))
+        dst = dst.scatter_reduce(-2, dst_idx.expand(n, r, c), src, reduce=mode)
+
+        return torch.cat([unm, dst], dim=1)
+
+    def unmerge(x: torch.Tensor) -> torch.Tensor:
+        unm_len = unm_idx.shape[1]
+        unm, dst = x[..., :unm_len, :], x[..., unm_len:, :]
+        _, _, c = unm.shape
+
+        src = dst.gather(dim=-2, index=dst_idx.expand(B, r, c))
+
+        # Combine back to the original shape
+        out = torch.zeros(B, N, c, device=x.device, dtype=x.dtype)
+        out.scatter_(dim=-2, index=b_idx.expand(B, num_dst, c), src=dst)
+        out.scatter_(dim=-2, index=a_idx.expand(B, a_idx.shape[1], 1).gather(dim=1, index=unm_idx).expand(B, unm_len, c), src=unm)
+        out.scatter_(dim=-2, index=a_idx.expand(B, a_idx.shape[1], 1).gather(dim=1, index=src_idx).expand(B, r, c), src=src)
+
+        return out
+
+    return merge, unmerge
+
+
+def get_functions(x, ratio, original_shape):
+    b, c, original_h, original_w = original_shape
+    original_tokens = original_h * original_w
+    downsample = int(math.sqrt(original_tokens // x.shape[1]))
+    stride_x = 2
+    stride_y = 2
+    max_downsample = 1
+
+    if downsample <= max_downsample:
+        w = original_w // downsample
+        h = original_h // downsample
+        r = int(x.shape[1] * ratio)
+        no_rand = True
+        m, u = bipartite_soft_matching_random2d(x, w, h, stride_x, stride_y, r, no_rand)
+        return m, u
+
+    nothing = lambda y: y
+    return nothing, nothing

comfy/samplers.py

@@ -104,7 +104,7 @@ def sampling_function(model_function, x, timestep, uncond, cond, cond_scale, con
                 out['c_concat'] = [torch.cat(c_concat)]
             return out
 
-        def calc_cond_uncond_batch(model_function, cond, uncond, x_in, timestep, max_total_area, cond_concat_in):
+        def calc_cond_uncond_batch(model_function, cond, uncond, x_in, timestep, max_total_area, cond_concat_in, model_options):
             out_cond = torch.zeros_like(x_in)
             out_count = torch.ones_like(x_in)/100000.0
 
@@ -195,7 +195,7 @@ def sampling_function(model_function, x, timestep, uncond, cond, cond_scale, con
 
 
         max_total_area = model_management.maximum_batch_area()
-        cond, uncond = calc_cond_uncond_batch(model_function, cond, uncond, x, timestep, max_total_area, cond_concat)
+        cond, uncond = calc_cond_uncond_batch(model_function, cond, uncond, x, timestep, max_total_area, cond_concat, model_options)
         return uncond + (cond - uncond) * cond_scale
 
 
@@ -212,8 +212,8 @@ class CFGNoisePredictor(torch.nn.Module):
         super().__init__()
         self.inner_model = model
         self.alphas_cumprod = model.alphas_cumprod
-    def apply_model(self, x, timestep, cond, uncond, cond_scale, cond_concat=None):
-        out = sampling_function(self.inner_model.apply_model, x, timestep, uncond, cond, cond_scale, cond_concat)
+    def apply_model(self, x, timestep, cond, uncond, cond_scale, cond_concat=None, model_options={}):
+        out = sampling_function(self.inner_model.apply_model, x, timestep, uncond, cond, cond_scale, cond_concat, model_options=model_options)
         return out
 
 
@@ -221,11 +221,11 @@ class KSamplerX0Inpaint(torch.nn.Module):
     def __init__(self, model):
         super().__init__()
         self.inner_model = model
-    def forward(self, x, sigma, uncond, cond, cond_scale, denoise_mask, cond_concat=None):
+    def forward(self, x, sigma, uncond, cond, cond_scale, denoise_mask, cond_concat=None, model_options={}):
         if denoise_mask is not None:
             latent_mask = 1. - denoise_mask
             x = x * denoise_mask + (self.latent_image + self.noise * sigma.reshape([sigma.shape[0]] + [1] * (len(self.noise.shape) - 1))) * latent_mask
-        out = self.inner_model(x, sigma, cond=cond, uncond=uncond, cond_scale=cond_scale, cond_concat=cond_concat)
+        out = self.inner_model(x, sigma, cond=cond, uncond=uncond, cond_scale=cond_scale, cond_concat=cond_concat, model_options=model_options)
         if denoise_mask is not None:
             out *= denoise_mask
 
@@ -333,7 +333,7 @@ class KSampler:
                 "lms", "dpm_fast", "dpm_adaptive", "dpmpp_2s_ancestral", "dpmpp_sde",
                 "dpmpp_2m", "ddim", "uni_pc", "uni_pc_bh2"]
 
-    def __init__(self, model, steps, device, sampler=None, scheduler=None, denoise=None):
+    def __init__(self, model, steps, device, sampler=None, scheduler=None, denoise=None, model_options={}):
         self.model = model
         self.model_denoise = CFGNoisePredictor(self.model)
         if self.model.parameterization == "v":
@@ -353,6 +353,7 @@ class KSampler:
         self.sigma_max=float(self.model_wrap.sigma_max)
         self.set_steps(steps, denoise)
         self.denoise = denoise
+        self.model_options = model_options
 
     def _calculate_sigmas(self, steps):
         sigmas = None
@@ -421,7 +422,7 @@ class KSampler:
         else:
             precision_scope = contextlib.nullcontext
 
-        extra_args = {"cond":positive, "uncond":negative, "cond_scale": cfg}
+        extra_args = {"cond":positive, "uncond":negative, "cond_scale": cfg, "model_options": self.model_options}
 
         cond_concat = None
         if hasattr(self.model, 'concat_keys'):

comfy/sd.py

@@ -1,5 +1,6 @@
 import torch
 import contextlib
+import copy
 
 import sd1_clip
 import sd2_clip
@@ -274,12 +275,20 @@ class ModelPatcher:
         self.model = model
         self.patches = []
         self.backup = {}
+        self.model_options = {"transformer_options":{}}
 
     def clone(self):
         n = ModelPatcher(self.model)
         n.patches = self.patches[:]
+        n.model_options = copy.deepcopy(self.model_options)
         return n
 
+    def set_model_tomesd(self, ratio):
+        self.model_options["transformer_options"]["tomesd"] = {"ratio": ratio}
+
+    def model_dtype(self):
+        return self.model.diffusion_model.dtype
+
     def add_patches(self, patches, strength=1.0):
         p = {}
         model_sd = self.model.state_dict()

nodes.py

@@ -254,6 +254,22 @@ class LoraLoader:
         model_lora, clip_lora = comfy.sd.load_lora_for_models(model, clip, lora_path, strength_model, strength_clip)
         return (model_lora, clip_lora)
 
+class TomePatchModel:
+    @classmethod
+    def INPUT_TYPES(s):
+        return {"required": { "model": ("MODEL",),
+                              "ratio": ("FLOAT", {"default": 0.3, "min": 0.0, "max": 1.0, "step": 0.01}),
+                              }}
+    RETURN_TYPES = ("MODEL",)
+    FUNCTION = "patch"
+
+    CATEGORY = "_for_testing"
+
+    def patch(self, model, ratio):
+        m = model.clone()
+        m.set_model_tomesd(ratio)
+        return (m, )
+
 class VAELoader:
     @classmethod
     def INPUT_TYPES(s):
@@ -646,7 +662,7 @@ def common_ksampler(model, seed, steps, cfg, sampler_name, scheduler, positive,
     model_management.load_controlnet_gpu(control_net_models)
 
     if sampler_name in comfy.samplers.KSampler.SAMPLERS:
-        sampler = comfy.samplers.KSampler(real_model, steps=steps, device=device, sampler=sampler_name, scheduler=scheduler, denoise=denoise)
+        sampler = comfy.samplers.KSampler(real_model, steps=steps, device=device, sampler=sampler_name, scheduler=scheduler, denoise=denoise, model_options=model.model_options)
     else:
         #other samplers
         pass
@@ -1016,6 +1032,7 @@ NODE_CLASS_MAPPINGS = {
     "CLIPVisionLoader": CLIPVisionLoader,
     "VAEDecodeTiled": VAEDecodeTiled,
     "VAEEncodeTiled": VAEEncodeTiled,
+    "TomePatchModel": TomePatchModel,
 }
 
 def load_custom_node(module_path):