Merge ea5775c620 into 77e2ed5e01

comfy/cli_args.py

@@ -110,13 +110,11 @@ parser.add_argument("--preview-method", type=LatentPreviewMethod, default=Latent
 
 parser.add_argument("--preview-size", type=int, default=512, help="Sets the maximum preview size for sampler nodes.")
 
-CACHE_RAM_AUTO_GB = -1.0
-
 cache_group = parser.add_mutually_exclusive_group()
+cache_group.add_argument("--cache-ram", nargs='*', type=float, default=[], metavar="GB", help="Use RAM pressure caching with the specified headroom thresholds. This is the default caching mode. The first value sets the active-cache threshold; the optional second value sets the inactive-cache/pin threshold. Defaults when no values are provided: active 25%% of system RAM (min 4GB, max 32GB), inactive 75%% of system RAM (min 12GB, max 96GB).")
 cache_group.add_argument("--cache-classic", action="store_true", help="Use the old style (aggressive) caching.")
 cache_group.add_argument("--cache-lru", type=int, default=0, help="Use LRU caching with a maximum of N node results cached. May use more RAM/VRAM.")
 cache_group.add_argument("--cache-none", action="store_true", help="Reduced RAM/VRAM usage at the expense of executing every node for each run.")
-cache_group.add_argument("--cache-ram", nargs='?', const=CACHE_RAM_AUTO_GB, type=float, default=0, help="Use RAM pressure caching with the specified headroom threshold. If available RAM drops below the threshold the cache removes large items to free RAM. Default (when no value is provided): 25%% of system RAM (min 4GB, max 32GB).")
 
 attn_group = parser.add_mutually_exclusive_group()
 attn_group.add_argument("--use-split-cross-attention", action="store_true", help="Use the split cross attention optimization. Ignored when xformers is used.")
@@ -246,6 +244,9 @@ if comfy.options.args_parsing:
 else:
     args = parser.parse_args([])
 
+if args.cache_ram is not None and len(args.cache_ram) > 2:
+    parser.error("--cache-ram accepts at most two values: active GB and inactive GB")
+
 if args.windows_standalone_build:
     args.auto_launch = True
 

comfy/lora.py

@@ -484,16 +484,23 @@ def calculate_weight(patches, weight, key, intermediate_dtype=torch.float32, ori
 
     return weight
 
-def prefetch_prepared_value(value, allocate_buffer, stream):
+def prefetch_prepared_value(value, counter, destination, stream, copy):
     if isinstance(value, torch.Tensor):
-        dest = allocate_buffer(comfy.memory_management.vram_aligned_size(value))
+        size = comfy.memory_management.vram_aligned_size(value)
-        comfy.model_management.cast_to_gathered([value], dest, non_blocking=True, stream=stream)
+        offset = counter[0]
+        counter[0] += size
+        if destination is None:
+            return value
+
+        dest = destination[offset:offset + size]
+        if copy:
+            comfy.model_management.cast_to_gathered([value], dest, non_blocking=True, stream=stream)
         return comfy.memory_management.interpret_gathered_like([value], dest)[0]
     elif isinstance(value, weight_adapter.WeightAdapterBase):
-        return type(value)(value.loaded_keys, prefetch_prepared_value(value.weights, allocate_buffer, stream))
+        return type(value)(value.loaded_keys, prefetch_prepared_value(value.weights, counter, destination, stream, copy))
     elif isinstance(value, tuple):
-        return tuple(prefetch_prepared_value(item, allocate_buffer, stream) for item in value)
+        return tuple(prefetch_prepared_value(item, counter, destination, stream, copy) for item in value)
     elif isinstance(value, list):
-        return [prefetch_prepared_value(item, allocate_buffer, stream) for item in value]
+        return [prefetch_prepared_value(item, counter, destination, stream, copy) for item in value]
 
     return value

comfy/memory_management.py

@@ -48,6 +48,12 @@ def read_tensor_file_slice_into(tensor, destination):
     if info.size == 0:
         return True
 
+    hostbuf = getattr(destination.untyped_storage(), "_comfy_hostbuf", None)
+    if hostbuf is not None:
+        hostbuf.read_file_slice(file_obj, info.offset, info.size,
+                                offset=destination.data_ptr() - hostbuf.get_raw_address())
+        return True
+
     buf_type = ctypes.c_ubyte * info.size
     view = memoryview(buf_type.from_address(destination.data_ptr()))
 
@@ -151,7 +157,7 @@ def set_ram_cache_release_state(callback, headroom):
     extra_ram_release_callback = callback
     RAM_CACHE_HEADROOM = max(0, int(headroom))
 
-def extra_ram_release(target):
+def extra_ram_release(target, free_active=False):
     if extra_ram_release_callback is None:
         return 0
-    return extra_ram_release_callback(target)
+    return extra_ram_release_callback(target, free_active=free_active)

comfy/model_management.py

@@ -31,6 +31,7 @@ from contextlib import nullcontext
 import comfy.memory_management
 import comfy.utils
 import comfy.quant_ops
+import comfy_aimdo.host_buffer
 import comfy_aimdo.vram_buffer
 
 class VRAMState(Enum):
@@ -495,6 +496,14 @@ except:
 
 current_loaded_models = []
 
+DIRTY_MMAPS = set()
+
+PIN_PRESSURE_HYSTERESIS = 256 * 1024 * 1024
+
+#Freeing registerables on pressure does imply a GPU sync, so go big on
+#the hysteresis so each expensive sync gives us back a good chunk.
+REGISTERABLE_PIN_HYSTERESIS = 768 * 1024 * 1024
+
 def module_size(module):
     module_mem = 0
     sd = module.state_dict()
@@ -503,27 +512,44 @@ def module_size(module):
         module_mem += t.nbytes
     return module_mem
 
-def module_mmap_residency(module, free=False):
+def mark_mmap_dirty(storage):
-    mmap_touched_mem = 0
+    mmap_refs = getattr(storage, "_comfy_tensor_mmap_refs", None)
-    module_mem = 0
+    if mmap_refs is not None:
-    bounced_mmaps = set()
+        DIRTY_MMAPS.add(mmap_refs[0])
-    sd = module.state_dict()
+
-    for k in sd:
+def free_pins(size, evict_active=False):
-        t = sd[k]
+    if size <= 0:
-        module_mem += t.nbytes
+        return
-        storage = t._qdata.untyped_storage() if isinstance(t, comfy.quant_ops.QuantizedTensor) else t.untyped_storage()
+
-        if not getattr(storage, "_comfy_tensor_mmap_touched", False):
+    for loaded_model in reversed(current_loaded_models):
-            continue
+        model = loaded_model.model
-        mmap_touched_mem += t.nbytes
+        if model is not None and model.is_dynamic() and (evict_active or not model.model.dynamic_pins[model.load_device]["active"]):
-        if not free:
+            size -= model.partially_unload_ram(size)
-            continue
+            if size <= 0:
-        storage._comfy_tensor_mmap_touched = False
+                break
-        mmap_obj = storage._comfy_tensor_mmap_refs[0]
+
-        if mmap_obj in bounced_mmaps:
+def ensure_pin_budget(size, evict_active=False):
-            continue
+    if MAX_MODEL_MEMORY <= 0:
-        mmap_obj.bounce()
+        return
-        bounced_mmaps.add(mmap_obj)
+
-    return mmap_touched_mem, module_mem
+    shortfall = TOTAL_MODEL_MEMORY + size - MAX_MODEL_MEMORY
+    if shortfall <= 0:
+        return
+
+    free_pins(shortfall + PIN_PRESSURE_HYSTERESIS, evict_active=evict_active)
+
+def ensure_pin_registerable(size, evict_active=False):
+    shortfall = TOTAL_PINNED_MEMORY + size - MAX_PINNED_MEMORY
+    if MAX_PINNED_MEMORY <= 0 or shortfall <= 0:
+        return
+
+    shortfall += REGISTERABLE_PIN_HYSTERESIS
+    for loaded_model in reversed(current_loaded_models):
+        model = loaded_model.model
+        if model is not None and model.is_dynamic() and (evict_active or not model.model.dynamic_pins[model.load_device]["active"]):
+            shortfall -= model.unregister_inactive_pins(shortfall)
+            if shortfall <= 0:
+                return
 
 class LoadedModel:
     def __init__(self, model):
@@ -553,9 +579,6 @@ class LoadedModel:
     def model_memory(self):
         return self.model.model_size()
 
-    def model_mmap_residency(self, free=False):
-        return self.model.model_mmap_residency(free=free)
-
     def model_loaded_memory(self):
         return self.model.loaded_size()
 
@@ -635,15 +658,9 @@ WINDOWS = any(platform.win32_ver())
 
 EXTRA_RESERVED_VRAM = 400 * 1024 * 1024
 if WINDOWS:
-    import comfy.windows
     EXTRA_RESERVED_VRAM = 600 * 1024 * 1024 #Windows is higher because of the shared vram issue
     if total_vram > (15 * 1024):  # more extra reserved vram on 16GB+ cards
         EXTRA_RESERVED_VRAM += 100 * 1024 * 1024
-    def get_free_ram():
-        return comfy.windows.get_free_ram()
-else:
-    def get_free_ram():
-        return psutil.virtual_memory().available
 
 if args.reserve_vram is not None:
     EXTRA_RESERVED_VRAM = args.reserve_vram * 1024 * 1024 * 1024
@@ -657,7 +674,6 @@ def minimum_inference_memory():
 
 def free_memory(memory_required, device, keep_loaded=[], for_dynamic=False, pins_required=0, ram_required=0):
     cleanup_models_gc()
-    comfy.memory_management.extra_ram_release(max(pins_required, ram_required))
     unloaded_model = []
     can_unload = []
     unloaded_models = []
@@ -673,11 +689,9 @@ def free_memory(memory_required, device, keep_loaded=[], for_dynamic=False, pins
     for x in can_unload_sorted:
         i = x[-1]
         memory_to_free = 1e32
-        pins_to_free = 1e32
+        if current_loaded_models[i].model.is_dynamic() and (not DISABLE_SMART_MEMORY or device is None):
-        if not DISABLE_SMART_MEMORY or device is None:
             memory_to_free = 0 if device is None else memory_required - get_free_memory(device)
-            pins_to_free = pins_required - get_free_ram()
+            if for_dynamic:
-            if current_loaded_models[i].model.is_dynamic() and for_dynamic:
                 #don't actually unload dynamic models for the sake of other dynamic models
                 #as that works on-demand.
                 memory_required -= current_loaded_models[i].model.loaded_size()
@@ -685,18 +699,6 @@ def free_memory(memory_required, device, keep_loaded=[], for_dynamic=False, pins
         if memory_to_free > 0 and current_loaded_models[i].model_unload(memory_to_free):
             logging.debug(f"Unloading {current_loaded_models[i].model.model.__class__.__name__}")
             unloaded_model.append(i)
-        if pins_to_free > 0:
-            logging.debug(f"PIN Unloading {current_loaded_models[i].model.model.__class__.__name__}")
-            current_loaded_models[i].model.partially_unload_ram(pins_to_free)
-
-    for x in can_unload_sorted:
-        i = x[-1]
-        ram_to_free = ram_required - psutil.virtual_memory().available
-        if ram_to_free <= 0 and i not in unloaded_model:
-            continue
-        resident_memory, _ = current_loaded_models[i].model_mmap_residency(free=True)
-        if resident_memory > 0:
-            logging.debug(f"RAM Unloading {current_loaded_models[i].model.model.__class__.__name__}")
 
     for i in sorted(unloaded_model, reverse=True):
         unloaded_models.append(current_loaded_models.pop(i))
@@ -762,29 +764,16 @@ def load_models_gpu(models, memory_required=0, force_patch_weights=False, minimu
             model_to_unload.model.detach(unpatch_all=False)
             model_to_unload.model_finalizer.detach()
 
-
     total_memory_required = {}
-    total_pins_required = {}
-    total_ram_required = {}
     for loaded_model in models_to_load:
         device = loaded_model.device
         total_memory_required[device] = total_memory_required.get(device, 0) + loaded_model.model_memory_required(device)
-        resident_memory, model_memory = loaded_model.model_mmap_residency()
-        pinned_memory = loaded_model.model.pinned_memory_size()
-        #FIXME: This can over-free the pins as it budgets to pin the entire model. We should
-        #make this JIT to keep as much pinned as possible.
-        pins_required = model_memory - pinned_memory
-        ram_required = model_memory - resident_memory
-        total_pins_required[device] = total_pins_required.get(device, 0) + pins_required
-        total_ram_required[device] = total_ram_required.get(device, 0) + ram_required
 
     for device in total_memory_required:
         if device != torch.device("cpu"):
             free_memory(total_memory_required[device] * 1.1 + extra_mem,
                         device,
-                        for_dynamic=free_for_dynamic,
+                        for_dynamic=free_for_dynamic)
-                        pins_required=total_pins_required[device],
-                        ram_required=total_ram_required[device])
 
     for device in total_memory_required:
         if device != torch.device("cpu"):
@@ -1180,6 +1169,7 @@ STREAM_CAST_BUFFERS = {}
 LARGEST_CASTED_WEIGHT = (None, 0)
 STREAM_AIMDO_CAST_BUFFERS = {}
 LARGEST_AIMDO_CASTED_WEIGHT = (None, 0)
+STREAM_PIN_BUFFERS = {}
 
 DEFAULT_AIMDO_CAST_BUFFER_RESERVATION_SIZE = 16 * 1024 ** 3
 
@@ -1220,21 +1210,69 @@ def get_aimdo_cast_buffer(offload_stream, device):
     if cast_buffer is None:
         cast_buffer = comfy_aimdo.vram_buffer.VRAMBuffer(DEFAULT_AIMDO_CAST_BUFFER_RESERVATION_SIZE, device.index)
         STREAM_AIMDO_CAST_BUFFERS[offload_stream] = cast_buffer
-
     return cast_buffer
+
+def get_pin_buffer(offload_stream):
+    pin_buffer = STREAM_PIN_BUFFERS.get(offload_stream, None)
+    if pin_buffer is None:
+        pin_buffer = comfy_aimdo.host_buffer.HostBuffer(0)
+        STREAM_PIN_BUFFERS[offload_stream] = pin_buffer
+    elif offload_stream is not None:
+        event = getattr(pin_buffer, "_comfy_event", None)
+        if event is not None:
+            event.synchronize()
+            delattr(pin_buffer, "_comfy_event")
+    return pin_buffer
+
+def resize_pin_buffer(pin_buffer, size):
+    global TOTAL_MODEL_MEMORY, TOTAL_PINNED_MEMORY
+    old_size = pin_buffer.size
+    if size <= old_size:
+        return True
+    growth = size - old_size
+    comfy.memory_management.extra_ram_release(comfy.memory_management.RAM_CACHE_HEADROOM, free_active=True)
+    ensure_pin_budget(growth, evict_active=True)
+    ensure_pin_registerable(growth, evict_active=True)
+    try:
+        pin_buffer.extend(size=size, reallocate=True)
+    except RuntimeError:
+        return False
+    TOTAL_MODEL_MEMORY += pin_buffer.size - old_size
+    TOTAL_PINNED_MEMORY += pin_buffer.size - old_size
+    return True
+
 def reset_cast_buffers():
+    global TOTAL_MODEL_MEMORY, TOTAL_PINNED_MEMORY
     global LARGEST_CASTED_WEIGHT
     global LARGEST_AIMDO_CASTED_WEIGHT
 
     LARGEST_CASTED_WEIGHT = (None, 0)
     LARGEST_AIMDO_CASTED_WEIGHT = (None, 0)
-    for offload_stream in set(STREAM_CAST_BUFFERS) | set(STREAM_AIMDO_CAST_BUFFERS):
+    for offload_stream in set(STREAM_CAST_BUFFERS) | set(STREAM_AIMDO_CAST_BUFFERS) | set(STREAM_PIN_BUFFERS):
         if offload_stream is not None:
             offload_stream.synchronize()
     synchronize()
 
+    for mmap_obj in DIRTY_MMAPS:
+        mmap_obj.bounce()
+    DIRTY_MMAPS.clear()
+
+    for pin_buffer in STREAM_PIN_BUFFERS.values():
+        TOTAL_MODEL_MEMORY -= pin_buffer.size
+        TOTAL_PINNED_MEMORY -= pin_buffer.size
+    TOTAL_MODEL_MEMORY = max(0, TOTAL_MODEL_MEMORY)
+    TOTAL_PINNED_MEMORY = max(0, TOTAL_PINNED_MEMORY)
+
+    for loaded_model in current_loaded_models:
+        model = loaded_model.model
+        if model is not None and model.is_dynamic():
+            model.model.dynamic_pins[model.load_device]["active"] = False
+            model.partially_unload_ram(1e30, subsets=[ "patches" ])
+            model.model.dynamic_pins[model.load_device]["patches"] = (comfy_aimdo.host_buffer.HostBuffer(0, 8 * 1024 * 1024), [], [-1])
+
     STREAM_CAST_BUFFERS.clear()
     STREAM_AIMDO_CAST_BUFFERS.clear()
+    STREAM_PIN_BUFFERS.clear()
     soft_empty_cache()
 
 def get_offload_stream(device):
@@ -1296,8 +1334,7 @@ def cast_to_gathered(tensors, r, non_blocking=False, stream=None):
             if comfy.memory_management.read_tensor_file_slice_into(tensor, dest_view):
                 continue
             storage = tensor._qdata.untyped_storage() if isinstance(tensor, comfy.quant_ops.QuantizedTensor) else tensor.untyped_storage()
-            if hasattr(storage, "_comfy_tensor_mmap_touched"):
+            mark_mmap_dirty(storage)
-                storage._comfy_tensor_mmap_touched = True
             dest_view.copy_(tensor, non_blocking=non_blocking)
 
 
@@ -1335,14 +1372,18 @@ def cast_to_device(tensor, device, dtype, copy=False):
 
 
 PINNED_MEMORY = {}
+TOTAL_MODEL_MEMORY = 0
 TOTAL_PINNED_MEMORY = 0
+MAX_MODEL_MEMORY = -1
 MAX_PINNED_MEMORY = -1
 if not args.disable_pinned_memory:
     if is_nvidia() or is_amd():
+        ram = get_total_memory(torch.device("cpu"))
+        MAX_MODEL_MEMORY = min(ram - 4 * 1024 * 1024 * 1024, ram * 0.90)
         if WINDOWS:
-            MAX_PINNED_MEMORY = get_total_memory(torch.device("cpu")) * 0.40  # Windows limit is apparently 50%
+            MAX_PINNED_MEMORY = ram * 0.40  # Windows limit is apparently 50%
         else:
-            MAX_PINNED_MEMORY = get_total_memory(torch.device("cpu")) * 0.90
+            MAX_PINNED_MEMORY = ram * 0.90
         logging.info("Enabled pinned memory {}".format(MAX_PINNED_MEMORY // (1024 * 1024)))
 
 PINNING_ALLOWED_TYPES = set(["Tensor", "Parameter", "QuantizedTensor"])
@@ -1378,8 +1419,8 @@ def pin_memory(tensor):
         return False
 
     size = tensor.nbytes
-    if (TOTAL_PINNED_MEMORY + size) > MAX_PINNED_MEMORY:
+    comfy.memory_management.extra_ram_release(comfy.memory_management.RAM_CACHE_HEADROOM)
-        return False
+    ensure_pin_registerable(size)
 
     ptr = tensor.data_ptr()
     if ptr == 0:
@@ -1416,7 +1457,8 @@ def unpin_memory(tensor):
         return False
 
     if torch.cuda.cudart().cudaHostUnregister(ptr) == 0:
-        TOTAL_PINNED_MEMORY -= PINNED_MEMORY.pop(ptr)
+        size = PINNED_MEMORY.pop(ptr)
+        TOTAL_PINNED_MEMORY -= size
         return True
     else:
         logging.warning("Unpin error.")

comfy/model_patcher.py

@@ -35,6 +35,7 @@ import comfy.model_management
 import comfy.ops
 import comfy.patcher_extension
 import comfy.utils
+import comfy_aimdo.host_buffer
 from comfy.comfy_types import UnetWrapperFunction
 from comfy.quant_ops import QuantizedTensor
 from comfy.patcher_extension import CallbacksMP, PatcherInjection, WrappersMP
@@ -117,6 +118,8 @@ def string_to_seed(data):
     return comfy.utils.string_to_seed(data)
 
 class LowVramPatch:
+    is_lowvram_patch = True
+
     def __init__(self, key, patches, convert_func=None, set_func=None):
         self.key = key
         self.patches = patches
@@ -124,11 +127,21 @@ class LowVramPatch:
         self.set_func = set_func
         self.prepared_patches = None
 
-    def prepare(self, allocate_buffer, stream):
+    def memory_required(self):
-        self.prepared_patches = [
+        counter = [0]
-            (patch[0], comfy.lora.prefetch_prepared_value(patch[1], allocate_buffer, stream), patch[2], patch[3], patch[4])
+        for patch in self.patches[self.key]:
+            comfy.lora.prefetch_prepared_value(patch[1], counter, None, None, False)
+        return counter[0]
+
+    def prepare(self, destination, stream, copy=True, commit=True):
+        counter = [0]
+        prepared_patches = [
+            (patch[0], comfy.lora.prefetch_prepared_value(patch[1], counter, destination, stream, copy), patch[2], patch[3], patch[4])
             for patch in self.patches[self.key]
         ]
+        if commit:
+            self.prepared_patches = prepared_patches
+        return prepared_patches
 
     def clear_prepared(self):
         self.prepared_patches = None
@@ -341,9 +354,6 @@ class ModelPatcher:
         self.size = comfy.model_management.module_size(self.model)
         return self.size
 
-    def model_mmap_residency(self, free=False):
-        return comfy.model_management.module_mmap_residency(self.model, free=free)
-
     def loaded_size(self):
         return self.model.model_loaded_weight_memory
 
@@ -1119,7 +1129,7 @@ class ModelPatcher:
         return 0
 
     def partially_unload_ram(self, ram_to_unload):
-        pass
+        return 0
 
     def detach(self, unpatch_all=True):
         self.eject_model()
@@ -1543,6 +1553,15 @@ class ModelPatcherDynamic(ModelPatcher):
         super().__init__(model, load_device, offload_device, size, weight_inplace_update)
         if not hasattr(self.model, "dynamic_vbars"):
             self.model.dynamic_vbars = {}
+        if not hasattr(self.model, "dynamic_pins"):
+            self.model.dynamic_pins = {}
+        if self.load_device not in self.model.dynamic_pins:
+            self.model.dynamic_pins[self.load_device] = {
+                "weights": (comfy_aimdo.host_buffer.HostBuffer(0, 64 * 1024 * 1024), [], [-1]),
+                "patches": (comfy_aimdo.host_buffer.HostBuffer(0, 8 * 1024 * 1024), [], [-1]),
+                "failed": False,
+                "active": False,
+            }
         self.non_dynamic_delegate_model = None
         assert load_device is not None
 
@@ -1604,6 +1623,9 @@ class ModelPatcherDynamic(ModelPatcher):
             self.unpatch_hooks()
 
             vbar = self._vbar_get(create=True)
+            pin_state = self.model.dynamic_pins[self.load_device]
+            pin_state["failed"] = False
+            pin_state["active"] = True
             if vbar is not None:
                 vbar.prioritize()
 
@@ -1629,7 +1651,9 @@ class ModelPatcherDynamic(ModelPatcher):
                     if key in self.patches:
                         if comfy.lora.calculate_shape(self.patches[key], weight, key) != weight.shape:
                             return (True, 0)
-                        setattr(m, param_key + "_lowvram_function", LowVramPatch(key, self.patches))
+                        lowvram_patch = LowVramPatch(key, self.patches)
+                        lowvram_patch._pin_state = pin_state
+                        setattr(m, param_key + "_lowvram_function", lowvram_patch)
                         num_patches += 1
                     else:
                         setattr(m, param_key + "_lowvram_function", None)
@@ -1655,8 +1679,8 @@ class ModelPatcherDynamic(ModelPatcher):
 
                 if hasattr(m, "comfy_cast_weights"):
                     m.comfy_cast_weights = True
-                    m.pin_failed = False
                     m.seed_key = n
+                    m._pin_state = pin_state
                     set_dirty(m, dirty)
 
                     force_load, v_weight_size = setup_param(self, m, n, "weight")
@@ -1734,22 +1758,52 @@ class ModelPatcherDynamic(ModelPatcher):
         return freed
 
     def pinned_memory_size(self):
-        total = 0
+        return (self.model.dynamic_pins[self.load_device]["weights"][0].size +
-        loading = self._load_list(for_dynamic=True)
+                self.model.dynamic_pins[self.load_device]["patches"][0].size)
-        for x in loading:
-            _, _, _, _, m, _ = x
-            pin = comfy.pinned_memory.get_pin(m)
-            if pin is not None:
-                total += pin.numel() * pin.element_size()
-        return total
 
-    def partially_unload_ram(self, ram_to_unload):
+    def unregister_inactive_pins(self, ram_to_unload, subsets=[ "weights", "patches" ]):
-        loading = self._load_list(for_dynamic=True, default_device=self.offload_device)
+        freed = 0
-        for x in loading:
+        pin_state = self.model.dynamic_pins[self.load_device]
-            *_, m, _ = x
+        for subset in subsets:
-            ram_to_unload -= comfy.pinned_memory.unpin_memory(m)
+            hostbuf, stack, stack_split = pin_state[subset]
-            if ram_to_unload <= 0:
+            split = stack_split[0]
-                return
+            while split >= 0:
+                module, offset = stack[split]
+                split -= 1
+                stack_split[0] = split
+                if not module._pin_registered:
+                    continue
+                size = module._pin.numel() * module._pin.element_size()
+                if torch.cuda.cudart().cudaHostUnregister(module._pin.data_ptr()) != 0:
+                    comfy.model_management.discard_cuda_async_error()
+                    continue
+                module._pin_registered = False
+                comfy.model_management.TOTAL_PINNED_MEMORY = max(0, comfy.model_management.TOTAL_PINNED_MEMORY - size)
+                freed += size
+                ram_to_unload -= size
+                if ram_to_unload <= 0:
+                    return freed
+        return freed
+
+    def partially_unload_ram(self, ram_to_unload, subsets=[ "weights", "patches" ]):
+        freed = 0
+        pin_state = self.model.dynamic_pins[self.load_device]
+        for subset in subsets:
+            hostbuf, stack, stack_split = pin_state[subset]
+            while len(stack) > 0:
+                module, offset = stack.pop()
+                size = module._pin.numel() * module._pin.element_size()
+                del module._pin
+                hostbuf.truncate(offset, do_unregister=module._pin_registered)
+                stack_split[0] = min(stack_split[0], len(stack) - 1)
+                comfy.model_management.TOTAL_MODEL_MEMORY = max(0, comfy.model_management.TOTAL_MODEL_MEMORY - size)
+                if module._pin_registered:
+                    comfy.model_management.TOTAL_PINNED_MEMORY = max(0, comfy.model_management.TOTAL_PINNED_MEMORY - size)
+                freed += size
+                ram_to_unload -= size
+                if ram_to_unload <= 0:
+                    return freed
+        return freed
 
     def patch_model(self, device_to=None, lowvram_model_memory=0, load_weights=True, force_patch_weights=False):
         #This isn't used by the core at all and can only be to load a model out of

comfy/ops.py

@@ -75,6 +75,8 @@ except:
 
 cast_to = comfy.model_management.cast_to #TODO: remove once no more references
 
+STREAM_PIN_BUFFER_HEADROOM = 8 * 1024 * 1024
+
 def cast_to_input(weight, input, non_blocking=False, copy=True):
     return comfy.model_management.cast_to(weight, input.dtype, input.device, non_blocking=non_blocking, copy=copy)
 
@@ -91,6 +93,9 @@ def cast_modules_with_vbar(comfy_modules, dtype, device, bias_dtype, non_blockin
     offload_stream = None
     cast_buffer = None
     cast_buffer_offset = 0
+    stream_pin_hostbuf = None
+    stream_pin_offset = 0
+    stream_pin_queue = []
 
     def ensure_offload_stream(module, required_size, check_largest):
         nonlocal offload_stream
@@ -124,6 +129,22 @@ def cast_modules_with_vbar(comfy_modules, dtype, device, bias_dtype, non_blockin
         cast_buffer_offset += buffer_size
         return buffer
 
+    def get_stream_pin_buffer_offset(buffer_size):
+        nonlocal stream_pin_hostbuf
+        nonlocal stream_pin_offset
+
+        if buffer_size == 0 or offload_stream is None:
+            return None
+
+        if stream_pin_hostbuf is None:
+            stream_pin_hostbuf = comfy.model_management.get_pin_buffer(offload_stream)
+            if stream_pin_hostbuf is None:
+                return None
+
+        offset = stream_pin_offset
+        stream_pin_offset += buffer_size
+        return offset
+
     for s in comfy_modules:
         signature = comfy_aimdo.model_vbar.vbar_fault(s._v)
         resident = comfy_aimdo.model_vbar.vbar_signature_compare(signature, s._v_signature)
@@ -162,23 +183,45 @@ def cast_modules_with_vbar(comfy_modules, dtype, device, bias_dtype, non_blockin
         if xfer_dest is None:
             xfer_dest = get_cast_buffer(dest_size)
 
-        if signature is None and pin is None:
+        def cast_maybe_lowvram_patch(xfer_source, xfer_dest, stream):
-            comfy.pinned_memory.pin_memory(s)
+            if xfer_source is not None:
-            pin = comfy.pinned_memory.get_pin(s)
+                if getattr(xfer_source, "is_lowvram_patch", False):
-        else:
+                    xfer_source.prepare(xfer_dest, stream, copy=True, commit=False)
-            pin = None
+                else:
+                    comfy.model_management.cast_to_gathered(xfer_source, xfer_dest, non_blocking=non_blocking, stream=stream)
 
-        if pin is not None:
+        def handle_pin_miss(m, source, dest, subset="weights", size=None):
-            comfy.model_management.cast_to_gathered(xfer_source, pin)
+            pin = None
-            xfer_source = [ pin ]
+            if signature is None:
-        #send it over
+                comfy.pinned_memory.pin_memory(m, subset=subset, size=size)
-        comfy.model_management.cast_to_gathered(xfer_source, xfer_dest, non_blocking=non_blocking, stream=offload_stream)
+                pin = comfy.pinned_memory.get_pin(m, subset=subset)
+                if pin is not None:
+                    cast_maybe_lowvram_patch(source, pin, None)
+                    return [ pin ]
+            if pin is None:
+                pin_offset = get_stream_pin_buffer_offset(size)
+                if pin_offset is not None:
+                    stream_pin_queue.append((source, pin_offset, size, dest))
+                    return None
+            return source
+
+        if pin is None:
+            xfer_source = handle_pin_miss(s, xfer_source, xfer_dest, size=dest_size)
+
+        cast_maybe_lowvram_patch(xfer_source, xfer_dest, offload_stream)
 
         for param_key in ("weight", "bias"):
-            lowvram_fn = getattr(s, param_key + "_lowvram_function", None)
+            lowvram_source = getattr(s, param_key + "_lowvram_function", None)
-            if lowvram_fn is not None:
+            if lowvram_source is not None:
                 ensure_offload_stream(s, cast_buffer_offset, False)
-                lowvram_fn.prepare(lambda size: get_cast_buffer(size), offload_stream)
+                lowvram_size = lowvram_source.memory_required()
+                lowvram_dest = get_cast_buffer(lowvram_size)
+                lowvram_source.prepare(lowvram_dest, None, copy=False, commit=True)
+
+                pin = comfy.pinned_memory.get_pin(lowvram_source, subset="patches")
+                lowvram_source = handle_pin_miss(lowvram_source, lowvram_source, lowvram_dest, subset="patches", size=lowvram_size) if pin is None else [ pin ]
+
+                cast_maybe_lowvram_patch(lowvram_source, lowvram_dest, offload_stream)
 
         prefetch["xfer_dest"] = xfer_dest
         prefetch["cast_dest"] = cast_dest
@@ -186,6 +229,20 @@ def cast_modules_with_vbar(comfy_modules, dtype, device, bias_dtype, non_blockin
         prefetch["needs_cast"] = needs_cast
         s._prefetch = prefetch
 
+    if stream_pin_offset > 0:
+        if stream_pin_hostbuf.size < stream_pin_offset:
+            if not comfy.model_management.resize_pin_buffer(stream_pin_hostbuf, stream_pin_offset + STREAM_PIN_BUFFER_HEADROOM):
+                for xfer_source, _, _, xfer_dest in stream_pin_queue:
+                    cast_maybe_lowvram_patch(xfer_source, xfer_dest, offload_stream)
+                return offload_stream
+        stream_pin_tensor = comfy_aimdo.torch.hostbuf_to_tensor(stream_pin_hostbuf)
+        stream_pin_tensor.untyped_storage()._comfy_hostbuf = stream_pin_hostbuf
+        for xfer_source, pin_offset, pin_size, xfer_dest in stream_pin_queue:
+            pin = stream_pin_tensor[pin_offset:pin_offset + pin_size]
+            cast_maybe_lowvram_patch(xfer_source, pin, None)
+            comfy.model_management.cast_to_gathered([ pin ], xfer_dest, non_blocking=non_blocking, stream=offload_stream)
+        stream_pin_hostbuf._comfy_event = offload_stream.record_event()
+
     return offload_stream
 
 

comfy/pinned_memory.py

@@ -2,42 +2,55 @@ import comfy.model_management
 import comfy.memory_management
 import comfy_aimdo.host_buffer
 import comfy_aimdo.torch
+import torch
 
 from comfy.cli_args import args
 
-def get_pin(module):
+def get_pin(module, subset="weights"):
     return getattr(module, "_pin", None)
 
-def pin_memory(module):
+def pin_memory(module, subset="weights", size=None):
-    if module.pin_failed or args.disable_pinned_memory or get_pin(module) is not None:
+    pin_state = module._pin_state
+    if pin_state["failed"] or args.disable_pinned_memory:
         return
 
-    size = comfy.memory_management.vram_aligned_size([ module.weight, module.bias ])
+    hostbuf, stack, stack_split = pin_state[subset]
+    pin = get_pin(module, subset)
+    if pin is not None:
+        if module._pin_registered:
+            return
 
-    if comfy.model_management.MAX_PINNED_MEMORY <= 0 or (comfy.model_management.TOTAL_PINNED_MEMORY + size) > comfy.model_management.MAX_PINNED_MEMORY:
+        size = module._pin.nbytes
-        module.pin_failed = True
+        comfy.model_management.ensure_pin_registerable(size)
-        return False
+
+        if torch.cuda.cudart().cudaHostRegister(module._pin.data_ptr(), size, 1) != 0:
+            comfy.model_management.discard_cuda_async_error()
+            return False
+        module._pin_registered = True
+        stack_split[0] = max(stack_split[0], module._pin_stack_index)
+        comfy.model_management.TOTAL_PINNED_MEMORY += size
+        return True
+
+    if size is None:
+        size = comfy.memory_management.vram_aligned_size([ module.weight, module.bias ])
+    offset = hostbuf.size
+
+    comfy.memory_management.extra_ram_release(comfy.memory_management.RAM_CACHE_HEADROOM)
+    comfy.model_management.ensure_pin_budget(size)
+    comfy.model_management.ensure_pin_registerable(size)
 
     try:
-        hostbuf = comfy_aimdo.host_buffer.HostBuffer(size)
+        hostbuf.extend(size=size)
     except RuntimeError:
-        module.pin_failed = True
+        pin_state["failed"] = True
         return False
 
-    module._pin = comfy_aimdo.torch.hostbuf_to_tensor(hostbuf)
+    module._pin = comfy_aimdo.torch.hostbuf_to_tensor(hostbuf)[offset:offset + size]
-    module._pin_hostbuf = hostbuf
+    module._pin.untyped_storage()._comfy_hostbuf = hostbuf
+    stack.append((module, offset))
+    module._pin_registered = True
+    module._pin_stack_index = len(stack) - 1
+    stack_split[0] = max(stack_split[0], module._pin_stack_index)
+    comfy.model_management.TOTAL_MODEL_MEMORY += size
     comfy.model_management.TOTAL_PINNED_MEMORY += size
     return True
-
-def unpin_memory(module):
-    if get_pin(module) is None:
-        return 0
-    size = module._pin.numel() * module._pin.element_size()
-
-    comfy.model_management.TOTAL_PINNED_MEMORY -= size
-    if comfy.model_management.TOTAL_PINNED_MEMORY < 0:
-        comfy.model_management.TOTAL_PINNED_MEMORY = 0
-
-    del module._pin
-    del module._pin_hostbuf
-    return size

comfy/utils.py

@@ -113,7 +113,6 @@ def load_safetensors(ckpt):
                         "_comfy_tensor_file_slice",
                         comfy.memory_management.TensorFileSlice(f, threading.get_ident(), data_base_offset + start, end - start))
                 setattr(storage, "_comfy_tensor_mmap_refs", (model_mmap, mv))
-                setattr(storage, "_comfy_tensor_mmap_touched", False)
                 sd[name] = tensor
 
     return sd, header.get("__metadata__", {}),
@@ -1451,4 +1450,3 @@ def deepcopy_list_dict(obj, memo=None):
 
     memo[obj_id] = res
     return res
-

comfy/windows.py

@@ -1,52 +0,0 @@
-import ctypes
-import logging
-import psutil
-from ctypes import wintypes
-
-import comfy_aimdo.control
-
-psapi = ctypes.WinDLL("psapi")
-kernel32 = ctypes.WinDLL("kernel32")
-
-class PERFORMANCE_INFORMATION(ctypes.Structure):
-    _fields_ = [
-        ("cb", wintypes.DWORD),
-        ("CommitTotal", ctypes.c_size_t),
-        ("CommitLimit", ctypes.c_size_t),
-        ("CommitPeak", ctypes.c_size_t),
-        ("PhysicalTotal", ctypes.c_size_t),
-        ("PhysicalAvailable", ctypes.c_size_t),
-        ("SystemCache", ctypes.c_size_t),
-        ("KernelTotal", ctypes.c_size_t),
-        ("KernelPaged", ctypes.c_size_t),
-        ("KernelNonpaged", ctypes.c_size_t),
-        ("PageSize", ctypes.c_size_t),
-        ("HandleCount", wintypes.DWORD),
-        ("ProcessCount", wintypes.DWORD),
-        ("ThreadCount", wintypes.DWORD),
-    ]
-
-def get_free_ram():
-    #Windows is way too conservative and chalks recently used uncommitted model RAM
-    #as "in-use". So, calculate free RAM for the sake of general use as the greater of:
-    #
-    #1: What psutil says
-    #2: Total Memory - (Committed Memory - VRAM in use)
-    #
-    #We have to subtract VRAM in use from the comitted memory as WDDM creates a naked
-    #commit charge for all VRAM used just incase it wants to page it all out. This just
-    #isn't realistic so "overcommit" on our calculations by just subtracting it off.
-
-    pi = PERFORMANCE_INFORMATION()
-    pi.cb = ctypes.sizeof(pi)
-
-    if not psapi.GetPerformanceInfo(ctypes.byref(pi), pi.cb):
-        logging.warning("WARNING: Failed to query windows performance info. RAM usage may be sub optimal")
-        return psutil.virtual_memory().available
-
-    committed = pi.CommitTotal * pi.PageSize
-    total = pi.PhysicalTotal * pi.PageSize
-
-    return max(psutil.virtual_memory().available,
-               total - (committed - comfy_aimdo.control.get_total_vram_usage()))
-

execution.py

@@ -2,6 +2,7 @@ import copy
 import heapq
 import inspect
 import logging
+import psutil
 import sys
 import threading
 import time
@@ -727,6 +728,7 @@ class PromptExecutor:
 
         self._notify_prompt_lifecycle("start", prompt_id)
         ram_headroom = int(self.cache_args["ram"] * (1024 ** 3))
+        ram_inactive_headroom = int(self.cache_args["ram_inactive"] * (1024 ** 3))
         ram_release_callback = self.caches.outputs.ram_release if self.cache_type == CacheType.RAM_PRESSURE else None
         comfy.memory_management.set_ram_cache_release_state(ram_release_callback, ram_headroom)
 
@@ -780,7 +782,9 @@ class PromptExecutor:
                         execution_list.complete_node_execution()
 
                     if self.cache_type == CacheType.RAM_PRESSURE:
-                        comfy.model_management.free_memory(0, None, pins_required=ram_headroom, ram_required=ram_headroom)
+                        ram_release_callback(ram_inactive_headroom)
+                        ram_shortfall = ram_headroom - psutil.virtual_memory().available
+                        comfy.model_management.free_pins(ram_shortfall)
                         ram_release_callback(ram_headroom, free_active=True)
                 else:
                     # Only execute when the while-loop ends without break

main.py

@@ -283,19 +283,25 @@ def _collect_output_absolute_paths(history_result: dict) -> list[str]:
 
 def prompt_worker(q, server_instance):
     current_time: float = 0.0
-    cache_ram = args.cache_ram
+    cache_ram = 0
-    if cache_ram < 0:
+    cache_ram_inactive = 0
+    if not args.cache_classic and not args.cache_none and args.cache_lru <= 0:
         cache_ram = min(32.0, max(4.0, comfy.model_management.total_ram * 0.25 / 1024.0))
+        cache_ram_inactive = min(96.0, max(12.0, comfy.model_management.total_ram * 0.75 / 1024.0))
+        if len(args.cache_ram) > 0:
+            cache_ram = args.cache_ram[0]
+        if len(args.cache_ram) > 1:
+            cache_ram_inactive = args.cache_ram[1]
 
-    cache_type = execution.CacheType.CLASSIC
+    cache_type = execution.CacheType.RAM_PRESSURE
-    if args.cache_lru > 0:
+    if args.cache_classic:
+        cache_type = execution.CacheType.CLASSIC
+    elif args.cache_lru > 0:
         cache_type = execution.CacheType.LRU
-    elif cache_ram > 0:
-        cache_type = execution.CacheType.RAM_PRESSURE
     elif args.cache_none:
         cache_type = execution.CacheType.NONE
 
-    e = execution.PromptExecutor(server_instance, cache_type=cache_type, cache_args={ "lru" : args.cache_lru, "ram" : cache_ram } )
+    e = execution.PromptExecutor(server_instance, cache_type=cache_type, cache_args={ "lru" : args.cache_lru, "ram" : cache_ram, "ram_inactive" : cache_ram_inactive } )
     last_gc_collect = 0
     need_gc = False
     gc_collect_interval = 10.0

requirements.txt

@@ -23,7 +23,7 @@ SQLAlchemy>=2.0.0
 filelock
 av>=14.2.0
 comfy-kitchen>=0.2.8
-comfy-aimdo==0.3.0
+comfy-aimdo==0.4.1
 requests
 simpleeval>=1.0.0
 blake3