Commit Graph

1490 Commits

Author SHA1 Message Date
psychedelicious
cb0b389b4b tidy(nodes): clarify comment 2024-02-15 17:30:03 +11:00
psychedelicious
8892df1d97 Revert "feat(nodes): use LATENT_SCALE_FACTOR const in tensor output builders"
This reverts commit ef18fc546560277302f3886e456da9a47e8edce0.
2024-02-15 17:30:03 +11:00
psychedelicious
bc5f356390 feat(nodes): use LATENT_SCALE_FACTOR const in tensor output builders 2024-02-15 17:30:03 +11:00
psychedelicious
bcb85e100d tests: fix broken tests 2024-02-15 17:30:03 +11:00
psychedelicious
1f27ddc07d tidy(nodes): minor spelling correction 2024-02-15 17:30:03 +11:00
psychedelicious
83ddcc5f3a feat(nodes): allow _delete_all in obj serializer to be called at any time
`_delete_all` logged how many items it deleted, and had to be called _after_ service start bc it needed access to logger.

Move the logger call to the startup method and return the the deleted stats from `_delete_all`. This lets `_delete_all` be called at any time.
2024-02-15 17:30:03 +11:00
psychedelicious
55fa785561 tidy(nodes): remove object serializer on_saved
It's unused.
2024-02-15 17:30:03 +11:00
psychedelicious
06429028c8 revert(nodes): revert making tensors/conditioning use item storage
Turns out they are just different enough in purpose that the implementations would be rather unintuitive. I've made a separate ObjectSerializer service to handle tensors and conditioning.

Refined the class a bit too.
2024-02-15 17:30:03 +11:00
psychedelicious
8b6e322697 feat(nodes): support custom exception in ephemeral disk storage 2024-02-15 17:30:03 +11:00
psychedelicious
54a67459bf feat(nodes): support custom save and load functions in ItemStorageEphemeralDisk 2024-02-15 17:30:03 +11:00
psychedelicious
7fe5283e74 feat(nodes): create helper function to generate the item ID 2024-02-15 17:30:03 +11:00
psychedelicious
fe0391c86b feat(nodes): use ItemStorageABC for tensors and conditioning
Turns out `ItemStorageABC` was almost identical to `PickleStorageBase`. Instead of maintaining separate classes, we can use `ItemStorageABC` for both.

There's only one change needed - the `ItemStorageABC.set` method must return the newly stored item's ID. This allows us to let the service handle the responsibility of naming the item, but still create the requisite output objects during node execution.

The naming implementation is improved here. It extracts the name of the generic and appends a UUID to that string when saving items.
2024-02-15 17:30:03 +11:00
psychedelicious
25386a76ef tidy(nodes): do not refer to files as latents in PickleStorageTorch (again) 2024-02-15 17:30:03 +11:00
psychedelicious
fd30cb4d90 feat(nodes): ItemStorageABC typevar no longer bound to pydantic.BaseModel
This bound is totally unnecessary. There's no requirement for any implementation of `ItemStorageABC` to work only on pydantic models.
2024-02-15 17:30:03 +11:00
psychedelicious
0266946d3d fix(nodes): add super init to PickleStorageTorch 2024-02-15 17:30:03 +11:00
psychedelicious
a7f91b3e01 tidy(nodes): do not refer to files as latents in PickleStorageTorch 2024-02-15 17:30:03 +11:00
psychedelicious
de0b72528c feat(nodes): replace latents service with tensors and conditioning services
- New generic class `PickleStorageBase`, implements the same API as `LatentsStorageBase`, use for storing non-serializable data via pickling
- Implementation `PickleStorageTorch` uses `torch.save` and `torch.load`, same as `LatentsStorageDisk`
- Add `tensors: PickleStorageBase[torch.Tensor]` to `InvocationServices`
- Add `conditioning: PickleStorageBase[ConditioningFieldData]` to `InvocationServices`
- Remove `latents` service and all `LatentsStorage` classes
- Update `InvocationContext` and all usage of old `latents` service to use the new services/context wrapper methods
2024-02-15 17:30:03 +11:00
psychedelicious
2932652787 tidy(nodes): delete onnx.py
It doesn't work and keeping it updated to prevent the app from starting was getting tedious. Deleted.
2024-02-15 17:30:03 +11:00
psychedelicious
db6bc7305a fix(nodes): rearrange fields.py to avoid needing forward refs 2024-02-15 17:30:02 +11:00
psychedelicious
a5db204629 tidy(nodes): remove unnecessary, shadowing class attr declarations 2024-02-15 17:30:02 +11:00
psychedelicious
c16eba78ab feat(nodes): add WithBoard field helper class
This class works the same way as `WithMetadata` - it simply adds a `board` field to the node. The context wrapper function is able to pull the board id from this. This allows image-outputting nodes to get a board field "for free", and have their outputs automatically saved to it.

This is a breaking change for node authors who may have a field called `board`, because it makes `board` a reserved field name. I'll look into how to avoid this - maybe by naming this invoke-managed field `_board` to avoid collisions?

Supporting changes:
- `WithBoard` is added to all image-outputting nodes, giving them the ability to save to board.
- Unused, duplicate `WithMetadata` and `WithWorkflow` classes are deleted from `baseinvocation.py`. The "real" versions are in `fields.py`.
- Remove `LinearUIOutputInvocation`. Now that all nodes that output images also have a `board` field by default, this node is no longer necessary. See comment here for context: https://github.com/invoke-ai/InvokeAI/pull/5491#discussion_r1480760629
- Without `LinearUIOutputInvocation`, the `ImagesInferface.update` method is no longer needed, and removed.

Note: This commit does not bump all node versions. I will ensure that is done correctly before merging the PR of which this commit is a part.

Note: A followup commit will implement the frontend changes to support this change.
2024-02-15 17:30:02 +11:00
psychedelicious
1a191c4655 remove unused configdict import 2024-02-15 17:30:02 +11:00
psychedelicious
b1ba18b3d1 fix(nodes): do not freeze or cache config in context wrapper
- The config is already cached by the config class's `get_config()` method.
- The config mutates itself in its `root_path` property getter. Freezing the class makes any attempt to grab a path from the config error. Unfortunately this means we cannot easily freeze the class without fiddling with the inner workings of `InvokeAIAppConfig`, which is outside the scope here.
2024-02-15 17:30:02 +11:00
psychedelicious
aff46759f9 feat(nodes): context.data -> context._data 2024-02-15 17:30:02 +11:00
psychedelicious
d7b7dcc7fe feat(nodes): context.__services -> context._services 2024-02-15 17:30:02 +11:00
psychedelicious
889a26c5b6 feat(nodes): cache invocation interface config 2024-02-15 17:30:02 +11:00
psychedelicious
b4c774896a feat(nodes): do not hide services in invocation context interfaces 2024-02-15 17:30:02 +11:00
psychedelicious
afbe889d35 fix(nodes): restore missing context type annotations 2024-02-15 17:30:02 +11:00
psychedelicious
3f5ab02da9 chore(nodes): add comments for ConfigInterface 2024-02-15 17:30:02 +11:00
psychedelicious
e52434cb99 feat(nodes): add boards interface to invocation context 2024-02-15 17:30:02 +11:00
psychedelicious
483bdbcb9f fix(nodes): restore type annotations for InvocationContext 2024-02-15 17:30:02 +11:00
psychedelicious
ae421fb4ab feat(nodes): do not freeze InvocationContextData, prevents it from being subclassesd 2024-02-15 17:30:02 +11:00
psychedelicious
a7e23af9c6 feat(nodes): create invocation_api.py
This is the public API for invocations.

Everything a custom node might need should be re-exported from this file.
2024-02-15 17:30:02 +11:00
psychedelicious
3de4390711 feat(nodes): move ConditioningFieldData to conditioning_data.py 2024-02-15 17:30:02 +11:00
psychedelicious
5c7ed24aab feat(nodes): restore previous invocation context methods with deprecation warnings 2024-02-15 17:30:02 +11:00
psychedelicious
183c9c4799 chore: ruff 2024-02-15 17:30:02 +11:00
psychedelicious
8baf3f78a2 feat(nodes): tidy invocation_context.py, improve comments 2024-02-15 17:30:02 +11:00
psychedelicious
7e5ba2795e feat(nodes): update all invocations to use new invocation context
Update all invocations to use the new context. The changes are all fairly simple, but there are a lot of them.

Supporting minor changes:
- Patch bump for all nodes that use the context
- Update invocation processor to provide new context
- Minor change to `EventServiceBase` to accept a node's ID instead of the dict version of a node
- Minor change to `ModelManagerService` to support the new wrapped context
- Fanagling of imports to avoid circular dependencies
2024-02-15 17:30:02 +11:00
psychedelicious
f0e60a4ba2 feat(nodes): restricts invocation context power
Creates a low-power `InvocationContext` with simplified methods and data.

See `invocation_context.py` for detailed comments.
2024-02-15 17:30:02 +11:00
psychedelicious
aa089e8108 tidy(nodes): move all field things to fields.py
Unfortunately, this is necessary to prevent circular imports at runtime.
2024-02-15 17:30:02 +11:00
psychedelicious
fc278c5cb1 fix(images_default): correct get_metadata error message
The error was misleading, indicating an issue with getting the image DTO, when it was actually an issue with getting metadata.
2024-02-14 16:21:39 -05:00
psychedelicious
3726293258 feat(nodes): improve types in graph.py
Methods `get_node` and `complete` were typed as returning a dynamically created unions `InvocationsUnion` and `InvocationOutputsUnion`, respectively.

Static type analysers cannot work with dynamic objects, so these methods end up as effectively un-annotated, returning `Unknown`.

They now return `BaseInvocation` and `BaseInvocationOutput`, respectively, which are the superclasses of all members of each union. This gives us the best type annotation that is possible.

Note: the return types of these methods are never introspected, so it doesn't really matter what they are at runtime.
2024-02-14 07:56:10 +11:00
blessedcoolant
e82c21b5ba chore: rename DWPose to DW Openpose 2024-02-12 11:12:45 -05:00
blessedcoolant
50b93992cf cleanup: Remove Openpose Image Processor 2024-02-12 11:12:45 -05:00
blessedcoolant
67daf1751c fix: lint erros 2024-02-12 11:12:45 -05:00
blessedcoolant
7d80261d47 chore: Add code attribution for the DWPoseDetector 2024-02-12 11:12:45 -05:00
blessedcoolant
67cbfeb33d feat: Add output image resizing for DWPose 2024-02-12 11:12:45 -05:00
blessedcoolant
0a27b0379f feat: Initial implementation of DWPoseDetector 2024-02-12 11:12:45 -05:00
psychedelicious
c5f069a255 feat(backend): remove dependency on basicsr
`basicsr` has a hard dependency on torchvision <= 0.16 and is unmaintained. Extract the code we need from it and remove the dep.

Closes #5108
2024-02-11 08:34:54 +11:00
psychedelicious
d20f98fb4f fix(nodes): deep copy graph inputs
The change to memory session storage brings a subtle behaviour change.

Previously, we serialized and deserialized everything (e.g. field state, invocation outputs, etc) constantly. The meant we were effectively working with deep-copied objects at all time. We could mutate objects freely without worrying about other references to the object.

With memory storage, objects are now passed around by reference, and we cannot handle them in the same way.

This is problematic for nodes that mutate their own inputs. There are two ways this causes a problem:

- An output is used as input for multiple nodes. If the first node mutates the output object while `invoke`ing, the next node will get the mutated object.
- The invocation cache stores live python objects. When a node mutates an output pulled from the cache, the next node that uses the cached object will get the mutated object.

The solution is to deep-copy a node's inputs as they are set, effectively reproducing the same behaviour as we had with the SQLite session storage. Nodes can safely mutate their inputs and those changes never leave the node's scope.

Closes  #5665
2024-02-09 21:17:32 +11:00