Allow different models to estimate memory usage differently.

comfy/model_base.py

@@ -157,6 +157,16 @@ def state_dict_for_saving(self, clip_state_dict, vae_state_dict):
     def set_inpaint(self):
         self.inpaint_model = True
 
+    def memory_required(self, input_shape):
+        area = input_shape[0] * input_shape[2] * input_shape[3]
+        if comfy.model_management.xformers_enabled() or comfy.model_management.pytorch_attention_flash_attention():
+            #TODO: this needs to be tweaked
+            return (area / 20) * (1024 * 1024)
+        else:
+            #TODO: this formula might be too aggressive since I tweaked the sub-quad and split algorithms to use less memory.
+            return (((area * 0.6) / 0.9) + 1024) * (1024 * 1024)
+
+
 def unclip_adm(unclip_conditioning, device, noise_augmentor, noise_augment_merge=0.0):
     adm_inputs = []
     weights = []

comfy/model_management.py

@@ -579,27 +579,6 @@ def get_free_memory(dev=None, torch_free_too=False):
     else:
         return mem_free_total
 
-def batch_area_memory(area):
-    if xformers_enabled() or pytorch_attention_flash_attention():
-        #TODO: these formulas are copied from maximum_batch_area below
-        return (area / 20) * (1024 * 1024)
-    else:
-        return (((area * 0.6) / 0.9) + 1024) * (1024 * 1024)
-
-def maximum_batch_area():
-    global vram_state
-    if vram_state == VRAMState.NO_VRAM:
-        return 0
-
-    memory_free = get_free_memory() / (1024 * 1024)
-    if xformers_enabled() or pytorch_attention_flash_attention():
-        #TODO: this needs to be tweaked
-        area = 20 * memory_free
-    else:
-        #TODO: this formula is because AMD sucks and has memory management issues which might be fixed in the future
-        area = ((memory_free - 1024) * 0.9) / (0.6)
-    return int(max(area, 0))
-
 def cpu_mode():
     global cpu_state
     return cpu_state == CPUState.CPU

comfy/model_patcher.py

@@ -52,6 +52,9 @@ def is_clone(self, other):
             return True
         return False
 
+    def memory_required(self, input_shape):
+        return self.model.memory_required(input_shape=input_shape)
+
     def set_model_sampler_cfg_function(self, sampler_cfg_function):
         if len(inspect.signature(sampler_cfg_function).parameters) == 3:
             self.model_options["sampler_cfg_function"] = lambda args: sampler_cfg_function(args["cond"], args["uncond"], args["cond_scale"]) #Old way

comfy/sample.py

@@ -83,7 +83,7 @@ def prepare_sampling(model, noise_shape, positive, negative, noise_mask):
 
     real_model = None
     models, inference_memory = get_additional_models(positive, negative, model.model_dtype())
-    comfy.model_management.load_models_gpu([model] + models, comfy.model_management.batch_area_memory(noise_shape[0] * noise_shape[2] * noise_shape[3]) + inference_memory)
+    comfy.model_management.load_models_gpu([model] + models, model.memory_required(noise_shape) + inference_memory)
     real_model = model.model
 
     return real_model, positive, negative, noise_mask, models

comfy/samplers.py

@@ -134,7 +134,7 @@ def cond_cat(c_list):
 
             return out
 
-        def calc_cond_uncond_batch(model, cond, uncond, x_in, timestep, max_total_area, model_options):
+        def calc_cond_uncond_batch(model, cond, uncond, x_in, timestep, model_options):
             out_cond = torch.zeros_like(x_in)
             out_count = torch.ones_like(x_in) * 1e-37
 
@@ -170,9 +170,11 @@ def calc_cond_uncond_batch(model, cond, uncond, x_in, timestep, max_total_area,
                 to_batch_temp.reverse()
                 to_batch = to_batch_temp[:1]
 
+                free_memory = model_management.get_free_memory(x_in.device)
                 for i in range(1, len(to_batch_temp) + 1):
                     batch_amount = to_batch_temp[:len(to_batch_temp)//i]
-                    if (len(batch_amount) * first_shape[0] * first_shape[2] * first_shape[3] < max_total_area):
+                    input_shape = [len(batch_amount) * first_shape[0]] + list(first_shape)[1:]
+                    if model.memory_required(input_shape) < free_memory:
                         to_batch = batch_amount
                         break
 
@@ -242,11 +244,10 @@ def calc_cond_uncond_batch(model, cond, uncond, x_in, timestep, max_total_area,
             return out_cond, out_uncond
 
 
-        max_total_area = model_management.maximum_batch_area()
         if math.isclose(cond_scale, 1.0):
             uncond = None
 
-        cond, uncond = calc_cond_uncond_batch(model, cond, uncond, x, timestep, max_total_area, model_options)
+        cond, uncond = calc_cond_uncond_batch(model, cond, uncond, x, timestep, model_options)
         if "sampler_cfg_function" in model_options:
             args = {"cond": x - cond, "uncond": x - uncond, "cond_scale": cond_scale, "timestep": timestep, "input": x, "sigma": timestep}
             return x - model_options["sampler_cfg_function"](args)