* introduce new abstraction layer for GPU devices
* add unit test for device abstraction
* fix ruff
* convert TorchDeviceSelect into a stateless class
* move logic to select context-specific execution device into context API
* add mock hardware environments to pytest
* remove dangling mocker fixture
* fix unit test for running on non-CUDA systems
* remove unimplemented get_execution_device() call
* remove autocast precision
* Multiple changes:
1. Remove TorchDeviceSelect.get_execution_device(), as well as calls to
context.models.get_execution_device().
2. Rename TorchDeviceSelect to TorchDevice
3. Added back the legacy public API defined in `invocation_api`, including
choose_precision().
4. Added a config file migration script to accommodate removal of precision=autocast.
* add deprecation warnings to choose_torch_device() and choose_precision()
* fix test crash
* remove app_config argument from choose_torch_device() and choose_torch_dtype()
---------
Co-authored-by: Lincoln Stein <lstein@gmail.com>
Currently translated at 98.4% (1122 of 1140 strings)
translationBot(ui): update translation (Italian)
Currently translated at 98.4% (1120 of 1138 strings)
translationBot(ui): update translation (Italian)
Currently translated at 98.4% (1115 of 1133 strings)
Co-authored-by: Riccardo Giovanetti <riccardo.giovanetti@gmail.com>
Translate-URL: https://hosted.weblate.org/projects/invokeai/web-ui/it/
Translation: InvokeAI/Web UI
When using refiner with a mask (i.e. inpainting), we don't have noise provided as an input to the node.
This situation uniquely hits a code path that wasn't reviewed when gradient denoising was implemented.
That code path does two things wrong:
- It lerp'd the input latents. This was fixed in 5a1f4cb1ce.
- It added noise to the latents an extra time. This is fixed in this change.
We don't need to add noise in `latents_from_embeddings` because we do it just a lines later in `AddsMaskGuidance`.
- Remove the extraneous call to `add_noise`
- Make `seed` a required arg. We never call the function without seed anyways. If we refactor this in the future, it will be clearer that we need to look at how seed is handled.
- Move the call to create the noise to a deeper conditional, just before we call `AddsMaskGuidance`. The created noise tensor is now only used in that function, no need to create it every time.
Note: Whether or not having both noise and latents as inputs on the node is correct is a separate conversation. This change just fixes the issue with the current setup.
`LatentsField` objects have an optional `seed` field. This should only be populated when the latents are noise, generated from a seed.
`DenoiseLatentsInvocation` needs a seed value for scheduler initialization. It's used in a few places, and there is some logic for determining the seed to use with a series of fallbacks:
- Use the seed from the noise (a `LatentsField` object)
- Use the seed from the latents (a `LatentsField` object - normally it won't have a seed)
- Use `0` as a final fallback
In `DenoisLatentsInvocation`, we set the seed in the `LatentsOutput`, even though the output latents are not noise.
This is normally fine, but when we use refiner, we re-use the those same latents for the refiner denoise. This causes that characteristic same-seed-fried look on the refiner pass.
Simple fix - do not set the field in the output latents.
Handful of intertwined fixes.
- Create and use helper function to reset staging area.
- Clear staging area when queue items are canceled, failed, cleared, etc. Fixes a bug where the bbox ends up offset and images are put into the wrong spot.
- Fix a number of similar bugs where canvas would "forget" it had pending generations, but they continued to generate. Canvas needs to track batches that should be displayed in it using `state.canvas.batchIds`, and this was getting cleared without actually canceling those batches.
- Disable the `discard current image` button on canvas if there is only one image. Prevents accidentally canceling all canvas batches if you spam the button.
This is intended for debug usage, so it's hidden away in the workflow library `...` menu. Hold shift to see the button for it.
- Paste a graph (from a network request, for example) and then click the convert button to convert it to a workflow.
- Disable auto layout to stack the nodes with an offset (try it out). If you change this, you must re-convert to get the changes.
- Edit the workflow JSON if you need to tweak something before loading it.
- Allow user-defined precision on MPS.
- Use more explicit logic to handle all possible cases.
- Add comments.
- Remove the app_config args (they were effectively unused, just get the config using the singleton getter util)
This data is already in the template but it wasn't ever used.
One big place where this improves UX is the noise node. Previously, the UI let you change width and height in increments of 1, despite the template requiring a multiple of 8. It now works in multiples of 8.
Retrieving the DTO happens as part of the metadata parsing, not recall. This way, we don't show the option to recall a nonexistent image.
This matches the flow for other metadata entities like models - we don't show the model recall button if the model isn't available.
Currently translated at 73.3% (826 of 1126 strings)
Co-authored-by: Alexander Eichhorn <pfannkuchensack@einfach-doof.de>
Translate-URL: https://hosted.weblate.org/projects/invokeai/web-ui/de/
Translation: InvokeAI/Web UI
The previous algorithm errored if the image wasn't divisible by the tile size. I've reimplemented it from scratch to mitigate this issue.
The new algorithm is simpler. We create a pool of tiles, then use them to create an image composed completely of tiles. If there is any awkwardly sized space on the edge of the image, the tiles are cropped to fit.
Finally, paste the original image over the tile image.
I've added a jupyter notebook to do a smoke test of infilling methods, and 10 test images.
The other infill algorithms can be easily tested with the notebook on the same images, though I didn't set that up yet.
Tested and confirmed this gives results just as good as the earlier infill, though of course they aren't the same due to the change in the algorithm.
We have had a few bugs with v4 related to file encodings, especially on Windows.
Windows uses its own character encodings instead of `utf-8`, often `cp1252`. Some characters cannot be decoded using `utf-8`, causing `UnicodeDecodeError`.
There are a couple places where this can cause problems:
- In the installer bootstrap, we install or upgrade `pip` and decode the result, using `subprocess`.
The input to this includes the user's home dir. In #6105, the user had one of the problematic characters in their username. `subprocess` attempts and fails to decode the username, which crashes the installer.
To fix this, we need to use `locale.getpreferredencoding()` when executing the command.
- Similarly, in the model install service and config class, we attempt to load a yaml config file. If a problematic character is in the path to the file (which often includes the user's home dir), we can get the same error.
One example is #6129 in which the models.yaml migration fails.
To fix this, we need to open the file with `locale.getpreferredencoding()`.
Compare the installed paths to determine if the model is already installed. Fixes an issue where installed models showed up as uninstalled or vice-versa. Related to relative vs absolute path handling.
Renaming the model file to the model name introduces unnecessary contraints on model names.
For example, a model name can technically be any length, but a model _filename_ cannot be too long.
There are also constraints on valid characters for filenames which shouldn't be applied to model record names.
I believe the old behaviour is a holdover from the old system.
Setting to 'auto' works only for InvokeAI config and auto detects the SD model but will override if user explicitly sets it. If auto used with checkpoint models, we raise an error. Checkpoints will always need to set to non-auto.
The valid values for this parameter changed when inpainting changed to gradient denoise. The generation slice's redux migration wasn't updated, resulting in a generation error until you change the setting or reset web UI.
- Add and use more performant `deepClone` method for deep copying throughout the UI.
Benchmarks indicate the Really Fast Deep Clone library (`rfdc`) is the best all-around way to deep-clone large objects.
This is particularly relevant in canvas. When drawing or otherwise manipulating canvas objects, we need to do a lot of deep cloning of the canvas layer state objects.
Previously, we were using lodash's `cloneDeep`.
I did some fairly realistic benchmarks with a handful of deep-cloning algorithms/libraries (including the native `structuredClone`). I used a snapshot of the canvas state as the data to be copied:
On Chromium, `rfdc` is by far the fastest, over an order of magnitude faster than `cloneDeep`.
On FF, `fastest-json-copy` and `recursiveDeepCopy` are even faster, but are rather limited in data types. `rfdc`, while only half as fast as the former 2, is still nearly an order of magnitude faster than `cloneDeep`.
On Safari, `structuredClone` is the fastest, about 2x as fast as `cloneDeep`. `rfdc` is only 30% faster than `cloneDeep`.
`rfdc`'s peak memory usage is about 10% more than `cloneDeep` on Chrome. I couldn't get memory measurements from FF and Safari, but let's just assume the memory usage is similar relative to the other algos.
Overall, `rfdc` is the best choice for a single algo for all browsers. It's definitely the best for Chromium, by far the most popular desktop browser and thus our primary target.
A future enhancement might be to detect the browser and use that to determine which algorithm to use.
There were two ways the canvas history could grow too large (past the `MAX_HISTORY` setting):
- Sometimes, when pushing to history, we didn't `shift` an item out when we exceeded the max history size.
- If the max history size was exceeded by more than one item, we still only `shift`, which removes one item.
These issue could appear after an extended canvas session, resulting in a memory leak and recurring major GCs/browser performance issues.
To fix these issues, a helper function is added for both past and future layer states, which uses slicing to ensure history never grows too large.
Previously, exceptions raised as custom nodes are initialized were fatal errors, causing the app to exit.
With this change, any error on import is caught and the error message printed. App continues to start up without the node.
For example, a custom node that isn't updated for v4.0.0 may raise an error on import if it is attempting to import things that no longer exist.
Currently translated at 98.3% (1106 of 1124 strings)
translationBot(ui): update translation (Italian)
Currently translated at 98.3% (1104 of 1122 strings)
Co-authored-by: Riccardo Giovanetti <riccardo.giovanetti@gmail.com>
Translate-URL: https://hosted.weblate.org/projects/invokeai/web-ui/it/
Translation: InvokeAI/Web UI
