Merge branch 'main' into main

invokeai/app/invocations/compel.py

@@ -108,13 +108,14 @@ class CompelInvocation(BaseInvocation):
                 print(f'Warn: trigger: "{trigger}" not found')
 
         with (
-            ModelPatcher.apply_lora_text_encoder(text_encoder_info.context.model, _lora_loader()),
             ModelPatcher.apply_ti(tokenizer_info.context.model, text_encoder_info.context.model, ti_list) as (
                 tokenizer,
                 ti_manager,
             ),
             ModelPatcher.apply_clip_skip(text_encoder_info.context.model, self.clip.skipped_layers),
             text_encoder_info as text_encoder,
+            # Apply the LoRA after text_encoder has been moved to its target device for faster patching.
+            ModelPatcher.apply_lora_text_encoder(text_encoder, _lora_loader()),
         ):
             compel = Compel(
                 tokenizer=tokenizer,
@@ -229,13 +230,14 @@ class SDXLPromptInvocationBase:
                 print(f'Warn: trigger: "{trigger}" not found')
 
         with (
-            ModelPatcher.apply_lora(text_encoder_info.context.model, _lora_loader(), lora_prefix),
             ModelPatcher.apply_ti(tokenizer_info.context.model, text_encoder_info.context.model, ti_list) as (
                 tokenizer,
                 ti_manager,
             ),
             ModelPatcher.apply_clip_skip(text_encoder_info.context.model, clip_field.skipped_layers),
             text_encoder_info as text_encoder,
+            # Apply the LoRA after text_encoder has been moved to its target device for faster patching.
+            ModelPatcher.apply_lora(text_encoder, _lora_loader(), lora_prefix),
         ):
             compel = Compel(
                 tokenizer=tokenizer,

invokeai/app/invocations/latent.py

@@ -710,9 +710,10 @@ class DenoiseLatentsInvocation(BaseInvocation):
             )
             with (
                 ExitStack() as exit_stack,
-                ModelPatcher.apply_lora_unet(unet_info.context.model, _lora_loader()),
                 set_seamless(unet_info.context.model, self.unet.seamless_axes),
                 unet_info as unet,
+                # Apply the LoRA after unet has been moved to its target device for faster patching.
+                ModelPatcher.apply_lora_unet(unet, _lora_loader()),
             ):
                 latents = latents.to(device=unet.device, dtype=unet.dtype)
                 if noise is not None:

invokeai/app/services/config/config_default.py

@@ -45,6 +45,7 @@ InvokeAI:
     ram: 13.5
     vram: 0.25
     lazy_offload: true
+    log_memory_usage: false
   Device:
     device: auto
     precision: auto
@@ -261,6 +262,7 @@ class InvokeAIAppConfig(InvokeAISettings):
     ram                 : float = Field(default=7.5, gt=0, description="Maximum memory amount used by model cache for rapid switching (floating point number, GB)", json_schema_extra=Categories.ModelCache, )
     vram                : float = Field(default=0.25, ge=0, description="Amount of VRAM reserved for model storage (floating point number, GB)", json_schema_extra=Categories.ModelCache, )
     lazy_offload        : bool = Field(default=True, description="Keep models in VRAM until their space is needed", json_schema_extra=Categories.ModelCache, )
+    log_memory_usage    : bool = Field(default=False, description="If True, a memory snapshot will be captured before and after every model cache operation, and the result will be logged (at debug level). There is a time cost to capturing the memory snapshots, so it is recommended to only enable this feature if you are actively inspecting the model cache's behaviour.", json_schema_extra=Categories.ModelCache)
 
     # DEVICE
     device              : Literal["auto", "cpu", "cuda", "cuda:1", "mps"] = Field(default="auto", description="Generation device", json_schema_extra=Categories.Device)

invokeai/backend/model_management/lora.py

@@ -1,6 +1,6 @@
 from __future__ import annotations
 
-import copy
+import pickle
 from contextlib import contextmanager
 from pathlib import Path
 from typing import Any, Dict, List, Optional, Tuple, Union
@@ -54,24 +54,6 @@ class ModelPatcher:
 
         return (module_key, module)
 
-    @staticmethod
-    def _lora_forward_hook(
-        applied_loras: List[Tuple[LoRAModel, float]],
-        layer_name: str,
-    ):
-        def lora_forward(module, input_h, output):
-            if len(applied_loras) == 0:
-                return output
-
-            for lora, weight in applied_loras:
-                layer = lora.layers.get(layer_name, None)
-                if layer is None:
-                    continue
-                output += layer.forward(module, input_h, weight)
-            return output
-
-        return lora_forward
-
     @classmethod
     @contextmanager
     def apply_lora_unet(
@@ -129,21 +111,40 @@ class ModelPatcher:
                         if not layer_key.startswith(prefix):
                             continue
 
+                        # TODO(ryand): A non-negligible amount of time is currently spent resolving LoRA keys. This
+                        # should be improved in the following ways:
+                        # 1. The key mapping could be more-efficiently pre-computed. This would save time every time a
+                        #    LoRA model is applied.
+                        # 2. From an API perspective, there's no reason that the `ModelPatcher` should be aware of the
+                        #    intricacies of Stable Diffusion key resolution. It should just expect the input LoRA
+                        #    weights to have valid keys.
                         module_key, module = cls._resolve_lora_key(model, layer_key, prefix)
+
+                        # All of the LoRA weight calculations will be done on the same device as the module weight.
+                        # (Performance will be best if this is a CUDA device.)
+                        device = module.weight.device
+                        dtype = module.weight.dtype
+
                         if module_key not in original_weights:
                             original_weights[module_key] = module.weight.detach().to(device="cpu", copy=True)
 
-                        # enable autocast to calc fp16 loras on cpu
-                        # with torch.autocast(device_type="cpu"):
-                        layer.to(dtype=torch.float32)
                         layer_scale = layer.alpha / layer.rank if (layer.alpha and layer.rank) else 1.0
-                        layer_weight = layer.get_weight(original_weights[module_key]) * lora_weight * layer_scale
+
+                        # We intentionally move to the target device first, then cast. Experimentally, this was found to
+                        # be significantly faster for 16-bit CPU tensors being moved to a CUDA device than doing the
+                        # same thing in a single call to '.to(...)'.
+                        layer.to(device=device)
+                        layer.to(dtype=torch.float32)
+                        # TODO(ryand): Using torch.autocast(...) over explicit casting may offer a speed benefit on CUDA
+                        # devices here. Experimentally, it was found to be very slow on CPU. More investigation needed.
+                        layer_weight = layer.get_weight(module.weight) * (lora_weight * layer_scale)
+                        layer.to(device="cpu")
 
                         if module.weight.shape != layer_weight.shape:
                             # TODO: debug on lycoris
                             layer_weight = layer_weight.reshape(module.weight.shape)
 
-                        module.weight += layer_weight.to(device=module.weight.device, dtype=module.weight.dtype)
+                        module.weight += layer_weight.to(dtype=dtype)
 
             yield  # wait for context manager exit
 
@@ -164,7 +165,13 @@ class ModelPatcher:
         new_tokens_added = None
 
         try:
-            ti_tokenizer = copy.deepcopy(tokenizer)
+            # HACK: The CLIPTokenizer API does not include a way to remove tokens after calling add_tokens(...). As a
+            # workaround, we create a full copy of `tokenizer` so that its original behavior can be restored after
+            # exiting this `apply_ti(...)` context manager.
+            #
+            # In a previous implementation, the deep copy was obtained with `ti_tokenizer = copy.deepcopy(tokenizer)`,
+            # but a pickle roundtrip was found to be much faster (1 sec vs. 0.05 secs).
+            ti_tokenizer = pickle.loads(pickle.dumps(tokenizer))
             ti_manager = TextualInversionManager(ti_tokenizer)
             init_tokens_count = text_encoder.resize_token_embeddings(None).num_embeddings
 
@@ -196,7 +203,9 @@ class ModelPatcher:
 
                     if model_embeddings.weight.data[token_id].shape != embedding.shape:
                         raise ValueError(
-                            f"Cannot load embedding for {trigger}. It was trained on a model with token dimension {embedding.shape[0]}, but the current model has token dimension {model_embeddings.weight.data[token_id].shape[0]}."
+                            f"Cannot load embedding for {trigger}. It was trained on a model with token dimension"
+                            f" {embedding.shape[0]}, but the current model has token dimension"
+                            f" {model_embeddings.weight.data[token_id].shape[0]}."
                         )
 
                     model_embeddings.weight.data[token_id] = embedding.to(
@@ -257,7 +266,8 @@ class TextualInversionModel:
         if "string_to_param" in state_dict:
             if len(state_dict["string_to_param"]) > 1:
                 print(
-                    f'Warn: Embedding "{file_path.name}" contains multiple tokens, which is not supported. The first token will be used.'
+                    f'Warn: Embedding "{file_path.name}" contains multiple tokens, which is not supported. The first'
+                    " token will be used."
                 )
 
             result.embedding = next(iter(state_dict["string_to_param"].values()))
@@ -435,7 +445,13 @@ class ONNXModelPatcher:
         orig_embeddings = None
 
         try:
-            ti_tokenizer = copy.deepcopy(tokenizer)
+            # HACK: The CLIPTokenizer API does not include a way to remove tokens after calling add_tokens(...). As a
+            # workaround, we create a full copy of `tokenizer` so that its original behavior can be restored after
+            # exiting this `apply_ti(...)` context manager.
+            #
+            # In a previous implementation, the deep copy was obtained with `ti_tokenizer = copy.deepcopy(tokenizer)`,
+            # but a pickle roundtrip was found to be much faster (1 sec vs. 0.05 secs).
+            ti_tokenizer = pickle.loads(pickle.dumps(tokenizer))
             ti_manager = TextualInversionManager(ti_tokenizer)
 
             def _get_trigger(ti_name, index):
@@ -470,7 +486,9 @@ class ONNXModelPatcher:
 
                     if embeddings[token_id].shape != embedding.shape:
                         raise ValueError(
-                            f"Cannot load embedding for {trigger}. It was trained on a model with token dimension {embedding.shape[0]}, but the current model has token dimension {embeddings[token_id].shape[0]}."
+                            f"Cannot load embedding for {trigger}. It was trained on a model with token dimension"
+                            f" {embedding.shape[0]}, but the current model has token dimension"
+                            f" {embeddings[token_id].shape[0]}."
                         )
 
                     embeddings[token_id] = embedding

invokeai/backend/model_management/memory_snapshot.py

@@ -64,7 +64,7 @@ class MemorySnapshot:
         return cls(process_ram, vram, malloc_info)
 
 
-def get_pretty_snapshot_diff(snapshot_1: MemorySnapshot, snapshot_2: MemorySnapshot) -> str:
+def get_pretty_snapshot_diff(snapshot_1: Optional[MemorySnapshot], snapshot_2: Optional[MemorySnapshot]) -> str:
     """Get a pretty string describing the difference between two `MemorySnapshot`s."""
 
     def get_msg_line(prefix: str, val1: int, val2: int):
@@ -73,6 +73,9 @@ def get_pretty_snapshot_diff(snapshot_1: MemorySnapshot, snapshot_2: MemorySnaps
 
     msg = ""
 
+    if snapshot_1 is None or snapshot_2 is None:
+        return msg
+
     msg += get_msg_line("Process RAM", snapshot_1.process_ram, snapshot_2.process_ram)
 
     if snapshot_1.malloc_info is not None and snapshot_2.malloc_info is not None:

invokeai/backend/model_management/model_cache.py

@@ -117,6 +117,7 @@ class ModelCache(object):
         lazy_offloading: bool = True,
         sha_chunksize: int = 16777216,
         logger: types.ModuleType = logger,
+        log_memory_usage: bool = False,
     ):
         """
         :param max_cache_size: Maximum size of the RAM cache [6.0 GB]
@@ -126,6 +127,10 @@ class ModelCache(object):
         :param lazy_offloading: Keep model in VRAM until another model needs to be loaded
         :param sequential_offload: Conserve VRAM by loading and unloading each stage of the pipeline sequentially
         :param sha_chunksize: Chunksize to use when calculating sha256 model hash
+        :param log_memory_usage: If True, a memory snapshot will be captured before and after every model cache
+            operation, and the result will be logged (at debug level). There is a time cost to capturing the memory
+            snapshots, so it is recommended to disable this feature unless you are actively inspecting the model cache's
+            behaviour.
         """
         self.model_infos: Dict[str, ModelBase] = dict()
         # allow lazy offloading only when vram cache enabled
@@ -137,6 +142,7 @@ class ModelCache(object):
         self.storage_device: torch.device = storage_device
         self.sha_chunksize = sha_chunksize
         self.logger = logger
+        self._log_memory_usage = log_memory_usage
 
         # used for stats collection
         self.stats = None
@@ -144,6 +150,11 @@ class ModelCache(object):
         self._cached_models = dict()
         self._cache_stack = list()
 
+    def _capture_memory_snapshot(self) -> Optional[MemorySnapshot]:
+        if self._log_memory_usage:
+            return MemorySnapshot.capture()
+        return None
+
     def get_key(
         self,
         model_path: str,
@@ -223,10 +234,10 @@ class ModelCache(object):
 
             # Load the model from disk and capture a memory snapshot before/after.
             start_load_time = time.time()
-            snapshot_before = MemorySnapshot.capture()
+            snapshot_before = self._capture_memory_snapshot()
             with skip_torch_weight_init():
                 model = model_info.get_model(child_type=submodel, torch_dtype=self.precision)
-            snapshot_after = MemorySnapshot.capture()
+            snapshot_after = self._capture_memory_snapshot()
             end_load_time = time.time()
 
             self_reported_model_size_after_load = model_info.get_size(submodel)
@@ -275,9 +286,9 @@ class ModelCache(object):
             return
 
         start_model_to_time = time.time()
-        snapshot_before = MemorySnapshot.capture()
+        snapshot_before = self._capture_memory_snapshot()
         cache_entry.model.to(target_device)
-        snapshot_after = MemorySnapshot.capture()
+        snapshot_after = self._capture_memory_snapshot()
         end_model_to_time = time.time()
         self.logger.debug(
             f"Moved model '{key}' from {source_device} to"
@@ -286,7 +297,12 @@ class ModelCache(object):
             f"{get_pretty_snapshot_diff(snapshot_before, snapshot_after)}"
         )
 
-        if snapshot_before.vram is not None and snapshot_after.vram is not None:
+        if (
+            snapshot_before is not None
+            and snapshot_after is not None
+            and snapshot_before.vram is not None
+            and snapshot_after.vram is not None
+        ):
             vram_change = abs(snapshot_before.vram - snapshot_after.vram)
 
             # If the estimated model size does not match the change in VRAM, log a warning.
@@ -422,12 +438,17 @@ class ModelCache(object):
         self.logger.debug(f"Before unloading: cached_models={len(self._cached_models)}")
 
         pos = 0
+        models_cleared = 0
         while current_size + bytes_needed > maximum_size and pos < len(self._cache_stack):
             model_key = self._cache_stack[pos]
             cache_entry = self._cached_models[model_key]
 
             refs = sys.getrefcount(cache_entry.model)
 
+            # HACK: This is a workaround for a memory-management issue that we haven't tracked down yet. We are directly
+            # going against the advice in the Python docs by using `gc.get_referrers(...)` in this way:
+            # https://docs.python.org/3/library/gc.html#gc.get_referrers
+
             # manualy clear local variable references of just finished function calls
             # for some reason python don't want to collect it even by gc.collect() immidiately
             if refs > 2:
@@ -453,15 +474,16 @@ class ModelCache(object):
                 f" refs: {refs}"
             )
 
-            # 2 refs:
+            # Expected refs:
             # 1 from cache_entry
             # 1 from getrefcount function
             # 1 from onnx runtime object
-            if not cache_entry.locked and refs <= 3 if "onnx" in model_key else 2:
+            if not cache_entry.locked and refs <= (3 if "onnx" in model_key else 2):
                 self.logger.debug(
                     f"Unloading model {model_key} to free {(model_size/GIG):.2f} GB (-{(cache_entry.size/GIG):.2f} GB)"
                 )
                 current_size -= cache_entry.size
+                models_cleared += 1
                 if self.stats:
                     self.stats.cleared += 1
                 del self._cache_stack[pos]
@@ -471,7 +493,20 @@ class ModelCache(object):
             else:
                 pos += 1
 
-        gc.collect()
+        if models_cleared > 0:
+            # There would likely be some 'garbage' to be collected regardless of whether a model was cleared or not, but
+            # there is a significant time cost to calling `gc.collect()`, so we want to use it sparingly. (The time cost
+            # is high even if no garbage gets collected.)
+            #
+            # Calling gc.collect(...) when a model is cleared seems like a good middle-ground:
+            # - If models had to be cleared, it's a signal that we are close to our memory limit.
+            # - If models were cleared, there's a good chance that there's a significant amount of garbage to be
+            #   collected.
+            #
+            # Keep in mind that gc is only responsible for handling reference cycles. Most objects should be cleaned up
+            # immediately when their reference count hits 0.
+            gc.collect()
+
         torch.cuda.empty_cache()
         if choose_torch_device() == torch.device("mps"):
             mps.empty_cache()
@@ -491,7 +526,6 @@ class ModelCache(object):
                 vram_in_use = torch.cuda.memory_allocated()
                 self.logger.debug(f"{(vram_in_use/GIG):.2f}GB VRAM used for models; max allowed={(reserved/GIG):.2f}GB")
 
-        gc.collect()
         torch.cuda.empty_cache()
         if choose_torch_device() == torch.device("mps"):
             mps.empty_cache()

invokeai/backend/model_management/model_load_optimizations.py

@@ -17,7 +17,7 @@ def skip_torch_weight_init():
     completely unnecessary if the intent is to load checkpoint weights from disk for the layer. This context manager
     monkey-patches common torch layers to skip the weight initialization step.
     """
-    torch_modules = [torch.nn.Linear, torch.nn.modules.conv._ConvNd]
+    torch_modules = [torch.nn.Linear, torch.nn.modules.conv._ConvNd, torch.nn.Embedding]
     saved_functions = [m.reset_parameters for m in torch_modules]
 
     try:

invokeai/backend/model_management/model_manager.py

@@ -351,6 +351,7 @@ class ModelManager(object):
             precision=precision,
             sequential_offload=sequential_offload,
             logger=logger,
+            log_memory_usage=self.app_config.log_memory_usage,
         )
 
         self._read_models(config)

invokeai/backend/model_management/models/lora.py

@@ -440,33 +440,19 @@ class IA3Layer(LoRALayerBase):
 class LoRAModelRaw:  # (torch.nn.Module):
     _name: str
     layers: Dict[str, LoRALayer]
-    _device: torch.device
-    _dtype: torch.dtype
 
     def __init__(
         self,
         name: str,
         layers: Dict[str, LoRALayer],
-        device: torch.device,
-        dtype: torch.dtype,
     ):
         self._name = name
-        self._device = device or torch.cpu
-        self._dtype = dtype or torch.float32
         self.layers = layers
 
     @property
     def name(self):
         return self._name
 
-    @property
-    def device(self):
-        return self._device
-
-    @property
-    def dtype(self):
-        return self._dtype
-
     def to(
         self,
         device: Optional[torch.device] = None,
@@ -475,8 +461,6 @@ class LoRAModelRaw:  # (torch.nn.Module):
         # TODO: try revert if exception?
         for key, layer in self.layers.items():
             layer.to(device=device, dtype=dtype)
-        self._device = device
-        self._dtype = dtype
 
     def calc_size(self) -> int:
         model_size = 0
@@ -557,8 +541,6 @@ class LoRAModelRaw:  # (torch.nn.Module):
             file_path = Path(file_path)
 
         model = cls(
-            device=device,
-            dtype=dtype,
             name=file_path.stem,  # TODO:
             layers=dict(),
         )

invokeai/frontend/merge/merge_diffusers.py

@@ -274,9 +274,10 @@ class mergeModelsForm(npyscreen.FormMultiPageAction):
         else:
             interp = self.interpolations[self.merge_method.value[0]]
 
+        bases = ["sd-1", "sd-2", "sdxl"]
         args = dict(
             model_names=models,
-            base_model=tuple(BaseModelType)[self.base_select.value[0]],
+            base_model=BaseModelType(bases[self.base_select.value[0]]),
             alpha=self.alpha.value,
             interp=interp,
             force=self.force.value,

invokeai/frontend/web/public/locales/en.json

@@ -722,7 +722,9 @@
         "noMatchingModels": "No matching Models",
         "noModelsAvailable": "No models available",
         "selectLoRA": "Select a LoRA",
-        "selectModel": "Select a Model"
+        "selectModel": "Select a Model",
+        "noLoRAsInstalled": "No LoRAs installed",
+        "noRefinerModelsInstalled": "No SDXL Refiner models installed"
     },
     "nodes": {
         "addNode": "Add Node",

invokeai/frontend/web/src/features/dynamicPrompts/components/ParamDynamicPromptsCollapse.tsx

@@ -16,15 +16,13 @@ const ParamDynamicPromptsCollapse = () => {
     () =>
       createSelector(stateSelector, ({ dynamicPrompts }) => {
         const count = dynamicPrompts.prompts.length;
-        if (count === 1) {
-          return t('dynamicPrompts.promptsWithCount_one', {
-            count,
-          });
-        } else {
+        if (count > 1) {
           return t('dynamicPrompts.promptsWithCount_other', {
             count,
           });
         }
+
+        return;
       }),
     [t]
   );

invokeai/frontend/web/src/features/lora/components/ParamLoraSelect.tsx

@@ -10,6 +10,7 @@ import { loraAdded } from 'features/lora/store/loraSlice';
 import { MODEL_TYPE_MAP } from 'features/parameters/types/constants';
 import { forEach } from 'lodash-es';
 import { memo, useCallback, useMemo } from 'react';
+import { useTranslation } from 'react-i18next';
 import { useGetLoRAModelsQuery } from 'services/api/endpoints/models';
 
 const selector = createSelector(
@@ -24,7 +25,7 @@ const ParamLoRASelect = () => {
   const dispatch = useAppDispatch();
   const { loras } = useAppSelector(selector);
   const { data: loraModels } = useGetLoRAModelsQuery();
-
+  const { t } = useTranslation();
   const currentMainModel = useAppSelector(
     (state: RootState) => state.generation.model
   );
@@ -79,7 +80,7 @@ const ParamLoRASelect = () => {
     return (
       <Flex sx={{ justifyContent: 'center', p: 2 }}>
         <Text sx={{ fontSize: 'sm', color: 'base.500', _dark: 'base.700' }}>
-          No LoRAs Loaded
+          {t('models.noLoRAsInstalled')}
         </Text>
       </Flex>
     );

invokeai/frontend/web/src/features/parameters/components/Parameters/Advanced/ParamAdvancedCollapse.tsx

@@ -28,9 +28,7 @@ export default function ParamAdvancedCollapse() {
   const activeLabel = useMemo(() => {
     const activeLabel: string[] = [];
 
-    if (shouldUseCpuNoise) {
-      activeLabel.push(t('parameters.cpuNoise'));
-    } else {
+    if (!shouldUseCpuNoise) {
       activeLabel.push(t('parameters.gpuNoise'));
     }
 

invokeai/frontend/web/src/features/parameters/components/Parameters/HighResFix/ParamHrfCollapse.tsx

@@ -4,12 +4,13 @@ import { RootState, stateSelector } from 'app/store/store';
 import { useAppSelector } from 'app/store/storeHooks';
 import { defaultSelectorOptions } from 'app/store/util/defaultMemoizeOptions';
 import IAICollapse from 'common/components/IAICollapse';
+import { useFeatureStatus } from 'features/system/hooks/useFeatureStatus';
 import { useMemo } from 'react';
+import { useTranslation } from 'react-i18next';
+import ParamHrfHeight from './ParamHrfHeight';
 import ParamHrfStrength from './ParamHrfStrength';
 import ParamHrfToggle from './ParamHrfToggle';
 import ParamHrfWidth from './ParamHrfWidth';
-import ParamHrfHeight from './ParamHrfHeight';
-import { useFeatureStatus } from 'features/system/hooks/useFeatureStatus';
 
 const selector = createSelector(
   stateSelector,
@@ -22,15 +23,14 @@ const selector = createSelector(
 );
 
 export default function ParamHrfCollapse() {
+  const { t } = useTranslation();
   const isHRFFeatureEnabled = useFeatureStatus('hrf').isFeatureEnabled;
   const { hrfEnabled } = useAppSelector(selector);
   const activeLabel = useMemo(() => {
     if (hrfEnabled) {
-      return 'On';
-    } else {
-      return 'Off';
+      return t('common.on');
     }
-  }, [hrfEnabled]);
+  }, [t, hrfEnabled]);
 
   if (!isHRFFeatureEnabled) {
     return null;

invokeai/frontend/web/src/features/sdxl/components/ParamSDXLRefinerCollapse.tsx

@@ -1,4 +1,4 @@
-import { Flex } from '@chakra-ui/react';
+import { Flex, Text } from '@chakra-ui/react';
 import { createSelector } from '@reduxjs/toolkit';
 import { stateSelector } from 'app/store/store';
 import { useAppSelector } from 'app/store/storeHooks';
@@ -14,6 +14,7 @@ import ParamSDXLRefinerStart from './SDXLRefiner/ParamSDXLRefinerStart';
 import ParamSDXLRefinerSteps from './SDXLRefiner/ParamSDXLRefinerSteps';
 import ParamUseSDXLRefiner from './SDXLRefiner/ParamUseSDXLRefiner';
 import { useTranslation } from 'react-i18next';
+import { useIsRefinerAvailable } from 'services/api/hooks/useIsRefinerAvailable';
 
 const selector = createSelector(
   stateSelector,
@@ -31,6 +32,19 @@ const selector = createSelector(
 const ParamSDXLRefinerCollapse = () => {
   const { activeLabel, shouldUseSliders } = useAppSelector(selector);
   const { t } = useTranslation();
+  const isRefinerAvailable = useIsRefinerAvailable();
+
+  if (!isRefinerAvailable) {
+    return (
+      <IAICollapse label={t('sdxl.refiner')} activeLabel={activeLabel}>
+        <Flex sx={{ justifyContent: 'center', p: 2 }}>
+          <Text sx={{ fontSize: 'sm', color: 'base.500', _dark: 'base.700' }}>
+            {t('models.noRefinerModelsInstalled')}
+          </Text>
+        </Flex>
+      </IAICollapse>
+    );
+  }
 
   return (
     <IAICollapse label={t('sdxl.refiner')} activeLabel={activeLabel}>

tests/backend/model_management/test_lora.py

@@ -0,0 +1,102 @@
+# test that if the model's device changes while the lora is applied, the weights can still be restored
+
+# test that LoRA patching works on both CPU and CUDA
+
+import pytest
+import torch
+
+from invokeai.backend.model_management.lora import ModelPatcher
+from invokeai.backend.model_management.models.lora import LoRALayer, LoRAModelRaw
+
+
+@pytest.mark.parametrize(
+    "device",
+    [
+        "cpu",
+        pytest.param("cuda", marks=pytest.mark.skipif(not torch.cuda.is_available(), reason="requires CUDA device")),
+    ],
+)
+@torch.no_grad()
+def test_apply_lora(device):
+    """Test the basic behavior of ModelPatcher.apply_lora(...). Check that patching and unpatching produce the correct
+    result, and that model/LoRA tensors are moved between devices as expected.
+    """
+
+    linear_in_features = 4
+    linear_out_features = 8
+    lora_dim = 2
+    model = torch.nn.ModuleDict(
+        {"linear_layer_1": torch.nn.Linear(linear_in_features, linear_out_features, device=device, dtype=torch.float16)}
+    )
+
+    lora_layers = {
+        "linear_layer_1": LoRALayer(
+            layer_key="linear_layer_1",
+            values={
+                "lora_down.weight": torch.ones((lora_dim, linear_in_features), device="cpu", dtype=torch.float16),
+                "lora_up.weight": torch.ones((linear_out_features, lora_dim), device="cpu", dtype=torch.float16),
+            },
+        )
+    }
+    lora = LoRAModelRaw("lora_name", lora_layers)
+
+    lora_weight = 0.5
+    orig_linear_weight = model["linear_layer_1"].weight.data.detach().clone()
+    expected_patched_linear_weight = orig_linear_weight + (lora_dim * lora_weight)
+
+    with ModelPatcher.apply_lora(model, [(lora, lora_weight)], prefix=""):
+        # After patching, all LoRA layer weights should have been moved back to the cpu.
+        assert lora_layers["linear_layer_1"].up.device.type == "cpu"
+        assert lora_layers["linear_layer_1"].down.device.type == "cpu"
+
+        # After patching, the patched model should still be on its original device.
+        assert model["linear_layer_1"].weight.data.device.type == device
+
+        torch.testing.assert_close(model["linear_layer_1"].weight.data, expected_patched_linear_weight)
+
+    # After unpatching, the original model weights should have been restored on the original device.
+    assert model["linear_layer_1"].weight.data.device.type == device
+    torch.testing.assert_close(model["linear_layer_1"].weight.data, orig_linear_weight)
+
+
+@pytest.mark.skipif(not torch.cuda.is_available(), reason="requires CUDA device")
+@torch.no_grad()
+def test_apply_lora_change_device():
+    """Test that if LoRA patching is applied on the CPU, and then the patched model is moved to the GPU, unpatching
+    still behaves correctly.
+    """
+    linear_in_features = 4
+    linear_out_features = 8
+    lora_dim = 2
+    # Initialize the model on the CPU.
+    model = torch.nn.ModuleDict(
+        {"linear_layer_1": torch.nn.Linear(linear_in_features, linear_out_features, device="cpu", dtype=torch.float16)}
+    )
+
+    lora_layers = {
+        "linear_layer_1": LoRALayer(
+            layer_key="linear_layer_1",
+            values={
+                "lora_down.weight": torch.ones((lora_dim, linear_in_features), device="cpu", dtype=torch.float16),
+                "lora_up.weight": torch.ones((linear_out_features, lora_dim), device="cpu", dtype=torch.float16),
+            },
+        )
+    }
+    lora = LoRAModelRaw("lora_name", lora_layers)
+
+    orig_linear_weight = model["linear_layer_1"].weight.data.detach().clone()
+
+    with ModelPatcher.apply_lora(model, [(lora, 0.5)], prefix=""):
+        # After patching, all LoRA layer weights should have been moved back to the cpu.
+        assert lora_layers["linear_layer_1"].up.device.type == "cpu"
+        assert lora_layers["linear_layer_1"].down.device.type == "cpu"
+
+        # After patching, the patched model should still be on the CPU.
+        assert model["linear_layer_1"].weight.data.device.type == "cpu"
+
+        # Move the model to the GPU.
+        assert model.to("cuda")
+
+    # After unpatching, the original model weights should have been restored on the GPU.
+    assert model["linear_layer_1"].weight.data.device.type == "cuda"
+    torch.testing.assert_close(model["linear_layer_1"].weight.data, orig_linear_weight, check_device=False)

tests/backend/model_management/test_memory_snapshot.py

@@ -13,10 +13,11 @@ def test_memory_snapshot_capture():
 
 
 snapshots = [
-    MemorySnapshot(process_ram=1.0, vram=2.0, malloc_info=Struct_mallinfo2()),
-    MemorySnapshot(process_ram=1.0, vram=2.0, malloc_info=None),
-    MemorySnapshot(process_ram=1.0, vram=None, malloc_info=Struct_mallinfo2()),
-    MemorySnapshot(process_ram=1.0, vram=None, malloc_info=None),
+    MemorySnapshot(process_ram=1, vram=2, malloc_info=Struct_mallinfo2()),
+    MemorySnapshot(process_ram=1, vram=2, malloc_info=None),
+    MemorySnapshot(process_ram=1, vram=None, malloc_info=Struct_mallinfo2()),
+    MemorySnapshot(process_ram=1, vram=None, malloc_info=None),
+    None,
 ]
 
 
@@ -26,10 +27,12 @@ def test_get_pretty_snapshot_diff(snapshot_1, snapshot_2):
     """Test that get_pretty_snapshot_diff() works with various combinations of missing MemorySnapshot fields."""
     msg = get_pretty_snapshot_diff(snapshot_1, snapshot_2)
 
-    expected_lines = 1
-    if snapshot_1.vram is not None and snapshot_2.vram is not None:
+    expected_lines = 0
+    if snapshot_1 is not None and snapshot_2 is not None:
         expected_lines += 1
-    if snapshot_1.malloc_info is not None and snapshot_2.malloc_info is not None:
-        expected_lines += 5
+        if snapshot_1.vram is not None and snapshot_2.vram is not None:
+            expected_lines += 1
+        if snapshot_1.malloc_info is not None and snapshot_2.malloc_info is not None:
+            expected_lines += 5
 
     assert len(msg.splitlines()) == expected_lines

tests/backend/model_management/test_model_load_optimization.py

@@ -11,6 +11,7 @@ from invokeai.backend.model_management.model_load_optimizations import _no_op, s
         (torch.nn.Conv1d, {"in_channels": 10, "out_channels": 20, "kernel_size": 3}),
         (torch.nn.Conv2d, {"in_channels": 10, "out_channels": 20, "kernel_size": 3}),
         (torch.nn.Conv3d, {"in_channels": 10, "out_channels": 20, "kernel_size": 3}),
+        (torch.nn.Embedding, {"num_embeddings": 10, "embedding_dim": 10}),
     ],
 )
 def test_skip_torch_weight_init_linear(torch_module, layer_args):
@@ -36,12 +37,14 @@ def test_skip_torch_weight_init_linear(torch_module, layer_args):
     # Check that reset_parameters is skipped while `skip_torch_weight_init()` is active.
     assert reset_params_fn_during == _no_op
     assert not torch.allclose(layer_before.weight, layer_during.weight)
-    assert not torch.allclose(layer_before.bias, layer_during.bias)
+    if hasattr(layer_before, "bias"):
+        assert not torch.allclose(layer_before.bias, layer_during.bias)
 
     # Check that the original behavior is restored after `skip_torch_weight_init()` ends.
     assert reset_params_fn_before is reset_params_fn_after
     assert torch.allclose(layer_before.weight, layer_after.weight)
-    assert torch.allclose(layer_before.bias, layer_after.bias)
+    if hasattr(layer_before, "bias"):
+        assert torch.allclose(layer_before.bias, layer_after.bias)
 
 
 def test_skip_torch_weight_init_restores_base_class_behavior():