* Bump diffusers to 0.21.2.
* Add T2IAdapterInvocation boilerplate.
* Add T2I-Adapter model to model-management.
* (minor) Tidy prepare_control_image(...).
* Add logic to run the T2I-Adapter models at the start of the DenoiseLatentsInvocation.
* Add logic for applying T2I-Adapter weights and accumulating.
* Add T2IAdapter to MODEL_CLASSES map.
* yarn typegen
* Add model probes for T2I-Adapter models.
* Add all of the frontend boilerplate required to use T2I-Adapter in the nodes editor.
* Add T2IAdapterModel.convert_if_required(...).
* Fix errors in T2I-Adapter input image sizing logic.
* Fix bug with handling of multiple T2I-Adapters.
* black / flake8
* Fix typo
* yarn build
* Add num_channels param to prepare_control_image(...).
* Link to upstream diffusers bugfix PR that currently requires a workaround.
* feat: Add Color Map Preprocessor
Needed for the color T2I Adapter
* feat: Add Color Map Preprocessor to Linear UI
* Revert "feat: Add Color Map Preprocessor"
This reverts commit a1119a00bf.
* Revert "feat: Add Color Map Preprocessor to Linear UI"
This reverts commit bd8a9b82d8.
* Fix T2I-Adapter field rendering in workflow editor.
* yarn build, yarn typegen
---------
Co-authored-by: blessedcoolant <54517381+blessedcoolant@users.noreply.github.com>
Co-authored-by: psychedelicious <4822129+psychedelicious@users.noreply.github.com>
This is fired when the dnd image is moved over the 'none' board. Weren't defaulting to 'none' for the image's board_id, resulting in it being possible to drag a 'none' image onto 'none'.
Selections were not being `uniqBy()`'d, or were `uniqBy()`'d without a proper iteratee. This results in duplicate images in selections in certain situations.
Add correct `uniqBy()` to the reducer to prevent this in the future.
This caused a crapload of network requests any time an image was generated.
The counts are necessary to handle the logic for inserting images into existing image list caches; we have to keep track of the counts.
Replace tag invalidation with manual cache updates in all cases, except the initial request (which is necessary to get the initial image counts).
One subtle change is to make the counts an object instead of a number. This is required for `immer` to handle draft states. This should be raised as a bug with RTK Query, as no error is thrown when attempting to update a primitive immer draft.
* feat(ui): max upscale pixels config
Add `maxUpscalePixels: number` to the app config. The number should be the *total* number of pixels eg `maxUpscalePixels: 4096 * 4096`.
If not provided, any size image may be upscaled.
If the config is provided, users will see be advised if their image is too large for either model, or told to switch to an x2 model if it's only too large for x4.
The message is via tooltip in the popover and via toast if the user uses the hotkey to upscale.
* feat(ui): "mayUpscale" -> "isAllowedToUpscale"
* Initial commit of edge drag feature.
* Fixed build warnings
* code cleanup and drag to existing node
* improved isValidConnection check
* fixed build issues, removed cyclic dependency
* edge created nodes now spawn at cursor
* Add Node popover will no longer show when using drag to delete an edge.
* Fixed collection handling, added priority for handles matching name of source handle, removed current image/notes nodes from filtered list
* Fixed not properly clearing startParams when closing the Add Node popover
* fix(ui): do not allow Collect -> Iterate connection
This can be removed when #3956 is resolved
* feat(ui): use existing node validation logic in add-node-on-drop
This logic handles a number of special cases
---------
Co-authored-by: Millun Atluri <Millu@users.noreply.github.com>
Co-authored-by: psychedelicious <4822129+psychedelicious@users.noreply.github.com>
* add control net to useRecallParams
* got recall controlnets working
* fix metadata viewer controlnet
* fix type errors
* fix controlnet metadata viewer
* add ip adapter to metadata
* added ip adapter to recall parameters
* got ip adapter recall working, still need to fix type errors
* fix type issues
* clean up logs
* python formatting
* cleanup
* fix(ui): only store `image_name` as ip adapter image
* fix(ui): use nullish coalescing operator for numbers
Need to use the nullish coalescing operator `??` instead of false-y coalescing operator `||` when the value being check is a number. This prevents unintended coalescing when the value is zero and therefore false-y.
* feat(ui): fall back on default values for ip adapter metadata
* fix(ui): remove unused schema
* feat(ui): re-use existing schemas in metadata schema
* fix(ui): do not disable invocationCache
---------
Co-authored-by: psychedelicious <4822129+psychedelicious@users.noreply.github.com>
This hook was rerendering any time anything changed. Moved it to a logical component, put its useEffects inside the component. This reduces the effect of the rerenders to just that tiny always-null component.
* add control net to useRecallParams
* got recall controlnets working
* fix metadata viewer controlnet
* fix type errors
* fix controlnet metadata viewer
* set control image and use correct processor type and node
* clean up logs
* recall processor using substring
* feat(ui): enable controlNet when recalling one
---------
Co-authored-by: psychedelicious <4822129+psychedelicious@users.noreply.github.com>
- Current image number & total are displayed
- Left/right wrap around instead of stopping on first/last image
- Disable the left/right/number buttons when showing base layer
- improved translations
- Drag the end of an edge away from its handle to disconnect it
- Drop in empty space to delete the edge
- Drop on valid handle to reconnect it
- Update connection logic slightly to allow edge updates
* feat(ui): add error handling for enqueueBatch route, remove sessions
This re-implements the handling for the session create/invoke errors, but for batches.
Also remove all references to the old sessions routes in the UI.
* feat(ui): improve canvas image error UI
* make canvas error state gray instead of red
---------
Co-authored-by: Mary Hipp <maryhipp@Marys-MacBook-Air.local>
* add skeleton loading state for queue lit
* hide use cache checkbox if cache is disabled
* undo accidental add
* feat(ui): hide node footer entirely if nothing to show there
---------
Co-authored-by: Mary Hipp <maryhipp@Marys-MacBook-Air.local>
Co-authored-by: psychedelicious <4822129+psychedelicious@users.noreply.github.com>
Skeletons are for when we know the number of specific content items that are loading. When the queue is loading, we don't know how many items there are, or how many will load, so the whole list should be replaced with loading state.
The previous behaviour rendered a static number of skeletons. That number would rarely be the right number - the app shouldn't say "I'm loading 7 queue items", then load none, or load 50.
A future enhancement could use the queue item skeleton component and go by the total number of queue items, as reported by the queue status. I tried this but had some layout jankiness, not worth the effort right now.
The queue item skeleton component's styling was updated to support this future enhancement, making it exactly the same size as a queue item (it was a bit smaller before).
This is actually a platform-specific issue. `madge` is complaining about a circular dependency on a single file - `invokeai/frontend/web/src/features/queue/store/nanoStores.ts`. In that file, we import from the `nanostores` package. Very similar name to the file itself.
The error only appears on Windows and macOS, I imagine because those systems both resolve `nanostores` to itself before resolving to the package.
The solution is simple - rename `nanoStores.ts`. It's now `queueNanoStore.ts`.
- Change translations to use arrays of paragraphs instead of a single paragraph.
- Change component to accept a `feature` prop to identify the feature which the popover describes.
- Add optional `wrapperProps`: passed to the wrapper element, allowing more flexibility when using the popover
- Add optional `popoverProps`: passed to the `<Popover />` component, allowing for overriding individual instances of the popover's props
- Move definitions of features and popover settings to `invokeai/frontend/web/src/common/components/IAIInformationalPopover/constants.ts`
- Add some type safety to the `feature` prop
- Edit `POPOVER_DATA` to provide `image`, `href`, `buttonLabel`, and any popover props. The popover props are applied to all instances of the popover for the given feature. Note that the component prop `popoverProps` will override settings here.
- Remove the popover's arrow. Because the popover is wrapping groups of components, sometimes the error ends up pointing to nothing, which looks kinda janky. I've just removed the arrow entirely, but feel free to add it back if you think it looks better.
- Use a `link` variant button with external link icon to better communicate that clicking the button will open a new tab.
- Default the link button label to "Learn More" (if a label is provided, that will be used instead)
- Make default position `top`, but set manually set some to `right` - namely, anything with a dropdown. This prevents the popovers from obscuring or being obscured by the dropdowns.
- Do a bit more restructuring of the Popover component itself, and how it is integrated with other components
- More ref forwarding
- Make the open delay 1s
- Set the popovers to use lazy mounting (eg do not mount until the user opens the thing)
- Update the verbiage for many popover items and add missing dynamic prompts stuff
- No longer need to make network request to add image to board after it's finished - removed
- Update linear graphs & upscale graph to save image to the board
- Update autoSwitch logic so when image is generated we still switch to the right board
- Remove the add-to-board node
- Create `BoardField` field type & add it to `save_image` node
- Add UI for `BoardField`
- Tighten up some loose types
- Make `save_image` node, in workflow editor, default to not intermediate
- Patch bump `save_image`
* break out separate functions for preselected images, remove recallAllParameters dep as it causes circular logic with model being set
* lint
---------
Co-authored-by: Mary Hipp <maryhipp@Marys-MacBook-Air.local>
- New routes to clear, enable, disable and get the status of the cache
- Status includes hits, misses, size, max size, enabled
- Add client cache queries and mutations, abstracted into hooks
- Add invocation cache status area (next to queue status) w/ buttons
* Initial commit. Feature works, but code might need some cleanup
* Cleaned up diff
* Made mousePosition a XYPosition again so its nicely typed
* Fixed yarn issues
* Paste now properly takes node width/height into account when pasting
* feat(ui): use react's types in the `onMouseMove` `reactflow` handler
* feat(ui): use refs to access `reactflow`'s DOM elements
* feat(ui): use a ref to store cursor position in nodes
---------
Co-authored-by: psychedelicious <4822129+psychedelicious@users.noreply.github.com>
Polymorphic fields now render the appropriate input component for their base type.
For example, float polymorphics will render the number input box.
You no longer need to specify ui_type to force it to display.
TODO: The UI *may* break if a list is provided as the default value for a polymorphic field.
* feat(ui): tweak queue UI components
* fix(ui): manually dispatch queue status query on queue item status change
RTK Query occasionally aborts the query that occurs when the tag is invalidated, especially if multples of them fire in rapid succession.
This resulted in the queue status and progress bar sometimes not reseting when the queue finishes its last item.
Manually dispatch the query now to get around this. Eventually should probably move this to a socket so we don't need to keep responding to socket with HTTP requests. Just send ti directly via socket
* chore(ui): remove errant console.logs
* fix(ui): do not accumulate node outputs in outputs area
* fix(ui): fix merge issue
---------
Co-authored-by: Kent Keirsey <31807370+hipsterusername@users.noreply.github.com>
Add `batch_id` to outbound events. This necessitates adding it to both `InvocationContext` and `InvocationQueueItem`. This allows the canvas to receive images.
When the user enqueues a batch on the canvas, it is expected that all images from that batch are directed to the canvas.
The simplest, most flexible solution is to add the `batch_id` to the invocation context-y stuff. Then everything knows what batch it came from, and we can have the canvas pick up images associated with its list of canvas `batch_id`s.
* fix(config): fix typing issues in `config/`
`config/invokeai_config.py`:
- use `Optional` for things that are optional
- fix typing of `ram_cache_size()` and `vram_cache_size()`
- remove unused and incorrectly typed method `autoconvert_path`
- fix types and logic for `parse_args()`, in which `InvokeAIAppConfig.initconf` *must* be a `DictConfig`, but function would allow it to be set as a `ListConfig`, which presumably would cause issues elsewhere
`config/base.py`:
- use `cls` for first arg of class methods
- use `Optional` for things that are optional
- fix minor type issue related to setting of `env_prefix`
- remove unused `add_subparser()` method, which calls `add_parser()` on an `ArgumentParser` (method only available on the `_SubParsersAction` object, which is returned from ArgumentParser.add_subparsers()`)
* feat: queued generation and batches
Due to a very messy branch with broad addition of `isort` on `main` alongside it, some git surgery was needed to get an agreeable git history. This commit represents all of the work on queued generation. See PR for notes.
* chore: flake8, isort, black
* fix(nodes): fix incorrect service stop() method
* fix(nodes): improve names of a few variables
* fix(tests): fix up tests after changes to batches/queue
* feat(tests): add unit tests for session queue helper functions
* feat(ui): dynamic prompts is always enabled
* feat(queue): add queue_status_changed event
* feat(ui): wip queue graphs
* feat(nodes): move cleanup til after invoker startup
* feat(nodes): add cancel_by_batch_ids
* feat(ui): wip batch graphs & UI
* fix(nodes): remove `Batch.batch_id` from required
* fix(ui): cleanup and use fixedCacheKey for all mutations
* fix(ui): remove orphaned nodes from canvas graphs
* fix(nodes): fix cancel_by_batch_ids result count
* fix(ui): only show cancel batch tooltip when batches were canceled
* chore: isort
* fix(api): return `[""]` when dynamic prompts generates no prompts
Just a simple fallback so we always have a prompt.
* feat(ui): dynamicPrompts.combinatorial is always on
There seems to be little purpose in using the combinatorial generation for dynamic prompts. I've disabled it by hiding it from the UI and defaulting combinatorial to true. If we want to enable it again in the future it's straightforward to do so.
* feat: add queue_id & support logic
* feat(ui): fix upscale button
It prepends the upscale operation to queue
* feat(nodes): return queue item when enqueuing a single graph
This facilitates one-off graph async workflows in the client.
* feat(ui): move controlnet autoprocess to queue
* fix(ui): fix non-serializable DOMRect in redux state
* feat(ui): QueueTable performance tweaks
* feat(ui): update queue list
Queue items expand to show the full queue item. Just as JSON for now.
* wip threaded session_processor
* feat(nodes,ui): fully migrate queue to session_processor
* feat(nodes,ui): add processor events
* feat(ui): ui tweaks
* feat(nodes,ui): consolidate events, reduce network requests
* feat(ui): cleanup & abstract queue hooks
* feat(nodes): optimize batch permutation
Use a generator to do only as much work as is needed.
Previously, though we only ended up creating exactly as many queue items as was needed, there was still some intermediary work that calculated *all* permutations. When that number was very high, the system had a very hard time and used a lot of memory.
The logic has been refactored to use a generator. Additionally, the batch validators are optimized to return early and use less memory.
* feat(ui): add seed behaviour parameter
This dynamic prompts parameter allows the seed to be randomized per prompt or per iteration:
- Per iteration: Use the same seed for all prompts in a single dynamic prompt expansion
- Per prompt: Use a different seed for every single prompt
"Per iteration" is appropriate for exploring a the latents space with a stable starting noise, while "Per prompt" provides more variation.
* fix(ui): remove extraneous random seed nodes from linear graphs
* fix(ui): fix controlnet autoprocess not working when queue is running
* feat(queue): add timestamps to queue status updates
Also show execution time in queue list
* feat(queue): change all execution-related events to use the `queue_id` as the room, also include `queue_item_id` in InvocationQueueItem
This allows for much simpler handling of queue items.
* feat(api): deprecate sessions router
* chore(backend): tidy logging in `dependencies.py`
* fix(backend): respect `use_memory_db`
* feat(backend): add `config.log_sql` (enables sql trace logging)
* feat: add invocation cache
Supersedes #4574
The invocation cache provides simple node memoization functionality. Nodes that use the cache are memoized and not re-executed if their inputs haven't changed. Instead, the stored output is returned.
## Results
This feature provides anywhere some significant to massive performance improvement.
The improvement is most marked on large batches of generations where you only change a couple things (e.g. different seed or prompt for each iteration) and low-VRAM systems, where skipping an extraneous model load is a big deal.
## Overview
A new `invocation_cache` service is added to handle the caching. There's not much to it.
All nodes now inherit a boolean `use_cache` field from `BaseInvocation`. This is a node field and not a class attribute, because specific instances of nodes may want to opt in or out of caching.
The recently-added `invoke_internal()` method on `BaseInvocation` is used as an entrypoint for the cache logic.
To create a cache key, the invocation is first serialized using pydantic's provided `json()` method, skipping the unique `id` field. Then python's very fast builtin `hash()` is used to create an integer key. All implementations of `InvocationCacheBase` must provide a class method `create_key()` which accepts an invocation and outputs a string or integer key.
## In-Memory Implementation
An in-memory implementation is provided. In this implementation, the node outputs are stored in memory as python classes. The in-memory cache does not persist application restarts.
Max node cache size is added as `node_cache_size` under the `Generation` config category.
It defaults to 512 - this number is up for discussion, but given that these are relatively lightweight pydantic models, I think it's safe to up this even higher.
Note that the cache isn't storing the big stuff - tensors and images are store on disk, and outputs include only references to them.
## Node Definition
The default for all nodes is to use the cache. The `@invocation` decorator now accepts an optional `use_cache: bool` argument to override the default of `True`.
Non-deterministic nodes, however, should set this to `False`. Currently, all random-stuff nodes, including `dynamic_prompt`, are set to `False`.
The field name `use_cache` is now effectively a reserved field name and possibly a breaking change if any community nodes use this as a field name. In hindsight, all our reserved field names should have been prefixed with underscores or something.
## One Gotcha
Leaf nodes probably want to opt out of the cache, because if they are not cached, their outputs are not saved again.
If you run the same graph multiple times, you only end up with a single image output, because the image storage side-effects are in the `invoke()` method, which is bypassed if we have a cache hit.
## Linear UI
The linear graphs _almost_ just work, but due to the gotcha, we need to be careful about the final image-outputting node. To resolve this, a `SaveImageInvocation` node is added and used in the linear graphs.
This node is similar to `ImagePrimitive`, except it saves a copy of its input image, and has `use_cache` set to `False` by default.
This is now the leaf node in all linear graphs, and is the only node in those graphs with `use_cache == False` _and_ the only node with `is_intermedate == False`.
## Workflow Editor
All nodes now have a footer with a new `Use Cache [ ]` checkbox. It defaults to the value set by the invocation in its python definition, but can be changed by the user.
The workflow/node validation logic has been updated to migrate old workflows to use the new default values for `use_cache`. Users may still want to review the settings that have been chosen. In the event of catastrophic failure when running this migration, the default value of `True` is applied, as this is correct for most nodes.
Users should consider saving their workflows after loading them in and having them updated.
## Future Enhancements - Callback
A future enhancement would be to provide a callback to the `use_cache` flag that would be run as the node is executed to determine, based on its own internal state, if the cache should be used or not.
This would be useful for `DynamicPromptInvocation`, where the deterministic behaviour is determined by the `combinatorial: bool` field.
## Future Enhancements - Persisted Cache
Similar to how the latents storage is backed by disk, the invocation cache could be persisted to the database or disk. We'd need to be very careful about deserializing outputs, but it's perhaps worth exploring in the future.
* fix(ui): fix queue list item width
* feat(nodes): do not send the whole node on every generator progress
* feat(ui): strip out old logic related to sessions
Things like `isProcessing` are no longer relevant with queue. Removed them all & updated everything be appropriate for queue. May be a few little quirks I've missed...
* feat(ui): fix up param collapse labels
* feat(ui): click queue count to go to queue tab
* tidy(queue): update comment, query format
* feat(ui): fix progress bar when canceling
* fix(ui): fix circular dependency
* feat(nodes): bail on node caching logic if `node_cache_size == 0`
* feat(nodes): handle KeyError on node cache pop
* feat(nodes): bypass cache codepath if caches is disabled
more better no do thing
* fix(ui): reset api cache on connect/disconnect
* feat(ui): prevent enqueue when no prompts generated
* feat(ui): add queue controls to workflow editor
* feat(ui): update floating buttons & other incidental UI tweaks
* fix(ui): fix missing/incorrect translation keys
* fix(tests): add config service to mock invocation services
invoking needs access to `node_cache_size` to occur
* optionally remove pause/resume buttons from queue UI
* option to disable prepending
* chore(ui): remove unused file
* feat(queue): remove `order_id` entirely, `item_id` is now an autoinc pk
---------
Co-authored-by: Mary Hipp <maryhipp@Marys-MacBook-Air.local>
The immutable and serializable checks for redux can cause substantial performance issues. The immutable check in particular is pretty heavy. It's only run in dev mode, but this and really slow down the already-slower performance of dev mode.
The most important one for us is serializable, which has far less of a performance impact.
The immutable check is largely redundant because we use immer-backed RTK for everything and immer gives us confidence there.
Disable the immutable check, leaving serializable in.
A few weeks back, we changed how the canvas scales in response to changes in window/panel size.
This introduced a bug where if we the user hadn't already clicked the canvas tab once to initialize the stage elements, the stage's dimensions were zero, then the calculation of the stage's scale ends up zero, then something is divided by that zero and Konva dies.
This is only a problem on Chromium browsers - somehow Firefox handles it gracefully.
Now, when calculating the stage scale, never return a 0 - if it's a zero, return 1 instead. This is enough to fix the crash, but the image ends up centered on the top-left corner of the stage (the origin of the canvas).
Because the canvas elements are not initialized at this point (we haven't switched tabs yet), the stage dimensions fall back to (0,0). This means the center of the stage is also (0,0) - so the image is centered on (0,0), the top-left corner of the stage.
To fix this, we need to ensure we:
- Change to the canvas tab before actually setting the image, so the stage elements are able to initialize
- Use `flushSync` to flush DOM updates for this tab change so we actually have DOM elements to work with
- Update the stage dimensions once on first load of it (so in the effect that sets up the resize observer, we update the stage dimensions)
The result now is the expected behaviour - images sent to canvas do not crash and end up in the center of the canvas.
JSX is not serializable, so it cannot be in redux. Non-serializable global state may be put into `nanostores`.
- Use `nanostores` for `customStarUI`
- Use `nanostores` for `headerComponent`
- Re-enable the serializable & immutable check redux middlewares
This maps values to labels for multiple-choice fields.
This allows "enum" fields (i.e. `Literal["val1", "val2", ...]` fields) to use code-friendly string values for choices, but present this to the UI as human-friendly labels.
This simply hides nodes from the workflow editor. The nodes will still work if an API request is made with them. For example, you could hide `iterate` nodes from the workflow editor, but if the Linear UI makes use of those nodes, they will still function.
- Update `AppConfig` with optional property `nodesDenylist: string[]`
- If provided, nodes are filtered out by `type` in the workflow editor
* Consolidated saturation/luminosity adjust.
Now allows increasing and inverting.
Accepts any color PIL format and channel designation.
* Updated docs/nodes/defaultNodes.md
* shortened tags list to channel types only
* fix typo in mode list
* split features into offset and multiply nodes
* Updated documentation
* Change invert to discrete boolean.
Previous math was unclear and had issues with 0 values.
* chore: black
* chore(ui): typegen
---------
Co-authored-by: psychedelicious <4822129+psychedelicious@users.noreply.github.com>
