- This PR adds support for embedding files that contain a single key
"emb_params". The only example I know of this format is the
"EasyNegative" embedding on HuggingFace, but there are certainly others.
- This PR also adds support for loading embedding files that have been
saved in safetensors format.
- It also cleans up the code so that the logic of probing for and
selecting the right format parser is clear.
- This is the same as #3045, which is on the 2.3 branch.
- Commands, invocations and their parameters will now autocomplete using
introspection.
- Two types of parameter *arguments* will also autocomplete:
- --sampler_name will autocomplete the scheduler name
- --model will autocomplete the model name
- There don't seem to be commands for reading/writing image files yet,
so path autocompletion is not implemented
A long-standing issue with importing legacy checkpoints (both ckpt and
safetensors) is that the user has to identify the correct config file,
either by providing its path or by selecting which type of model the
checkpoint is (e.g. "v1 inpainting"). In addition, some users wish to
provide custom VAEs for use with the model. Currently this is done in
the WebUI by importing the model, editing it, and then typing in the
path to the VAE.
## Model configuration file selection
To improve the user experience, the model manager's `heuristic_import()`
method has been enhanced as follows:
1. When initially called, the caller can pass a config file path, in
which case it will be used.
2. If no config file provided, the method looks for a .yaml file in the
same directory as the model which bears the same basename. e.g.
```
my-new-model.safetensors
my-new-model.yaml
```
The yaml file is then used as the configuration file for importation and
conversion.
3. If no such file is found, then the method opens up the checkpoint and
probes it to determine whether it is V1, V1-inpaint or V2. If it is a V1
format, then the appropriate v1-inference.yaml config file is used.
Unfortunately there are two V2 variants that cannot be distinguished by
introspection.
4. If the probe algorithm is unable to determine the model type, then
its last-ditch effort is to execute an optional callback function that
can be provided by the caller. This callback, named
`config_file_callback` receives the path to the legacy checkpoint and
returns the path to the config file to use. The CLI uses to put up a
multiple choice prompt to the user. The WebUI **could** use this to
prompt the user to choose from a radio-button selection.
5. If the config file cannot be determined, then the import is
abandoned.
## Custom VAE Selection
The user can attach a custom VAE to the imported and converted model by
copying the desired VAE into the same directory as the file to be
imported, and giving it the same basename. E.g.:
```
my-new-model.safetensors
my-new-model.vae.pt
```
For this to work, the VAE must end with ".vae.pt", ".vae.ckpt", or
".vae.safetensors". The indicated VAE will be converted into diffusers
format and stored with the converted models file, so the ".pt" file can
be deleted after conversion.
No facility is currently provided to swap a diffusers VAE at import
time, but this can be done after the fact using the WebUI and CLI's
model editing functions.
Note that this is the same fix that was applied to the 2.3 branch in
#3043 . This applies to `main`.
## Enable the on-the-fly conversion of models based on SD 2.0/2.1 into
diffusers
This commit fixes bugs related to the on-the-fly conversion and loading
of legacy checkpoint models built on SD-2.0 base.
- When legacy checkpoints built on SD-2.0 models were converted
on-the-fly using --ckpt_convert, generation would crash with a precision
incompatibility error. This problem has been found and fixed.
This commit fixes bugs related to the on-the-fly conversion and loading of
legacy checkpoint models built on SD-2.0 base.
- When legacy checkpoints built on SD-2.0 models were converted
on-the-fly using --ckpt_convert, generation would crash with a
precision incompatibility error.
The Pytorch ROCm version in the documentation in outdated (`rocm5.2`)
which leads to errors during the installation of InvokeAI.
This PR updates the documentation with the latest Pytorch ROCm `5.4.2`
version.
A long-standing issue with importing legacy checkpoints (both ckpt and
safetensors) is that the user has to identify the correct config file,
either by providing its path or by selecting which type of model the
checkpoint is (e.g. "v1 inpainting"). In addition, some users wish to
provide custom VAEs for use with the model. Currently this is done in
the WebUI by importing the model, editing it, and then typing in the
path to the VAE.
To improve the user experience, the model manager's
`heuristic_import()` method has been enhanced as follows:
1. When initially called, the caller can pass a config file path, in
which case it will be used.
2. If no config file provided, the method looks for a .yaml file in the
same directory as the model which bears the same basename. e.g.
```
my-new-model.safetensors
my-new-model.yaml
```
The yaml file is then used as the configuration file for
importation and conversion.
3. If no such file is found, then the method opens up the checkpoint
and probes it to determine whether it is V1, V1-inpaint or V2.
If it is a V1 format, then the appropriate v1-inference.yaml config
file is used. Unfortunately there are two V2 variants that cannot be
distinguished by introspection.
4. If the probe algorithm is unable to determine the model type, then its
last-ditch effort is to execute an optional callback function that can
be provided by the caller. This callback, named `config_file_callback`
receives the path to the legacy checkpoint and returns the path to the
config file to use. The CLI uses to put up a multiple choice prompt to
the user. The WebUI **could** use this to prompt the user to choose
from a radio-button selection.
5. If the config file cannot be determined, then the import is abandoned.
The user can attach a custom VAE to the imported and converted model
by copying the desired VAE into the same directory as the file to be
imported, and giving it the same basename. E.g.:
```
my-new-model.safetensors
my-new-model.vae.pt
```
For this to work, the VAE must end with ".vae.pt", ".vae.ckpt", or
".vae.safetensors". The indicated VAE will be converted into diffusers
format and stored with the converted models file, so the ".pt" file
can be deleted after conversion.
No facility is currently provided to swap a diffusers VAE at import
time, but this can be done after the fact using the WebUI and CLI's
model editing functions.
- This PR adds support for embedding files that contain a single key
"emb_params". The only example I know of this format is the
"EasyNegative" embedding on HuggingFace, but there are certainly
others.
- This PR also adds support for loading embedding files that have been
saved in safetensors format.
- It also cleans up the code so that the logic of probing for and
selecting the right format parser is clear.
keeping `main` up to date with my api nodes branch:
- bd7e515290: [nodes] Add cancelation to
the API @Kyle0654
- 5fe38f7: fix(backend): simple typing fixes
- just picking some low-hanging fruit to improve IDE hinting
- c34ac91: fix(nodes): fix cancel; fix callback for img2img, inpaint
- makes nodes cancel immediate, use fix progress images on nodes, fix
callbacks for img2img/inpaint
- 4221cf7: fix(nodes): fix schema generation for output classes
- did this previously for some other class; needed to not have node
outputs be optional
Some schedulers report not only the noisy latents at the current
timestep, but also their estimate so far of what the de-noised latents
will be.
It makes for a more legible preview than the noisy latents do.
I think this is a huge improvement, but there are a few considerations:
- Need to not spook @JPPhoto by changing how previews look.
- Some schedulers (most notably **DPM Solver++**) don't provide this
data, and it falls back to the current behavior there. That's not
terrible, but seeing such a big difference in how _previews_ look from
one scheduler to the next might mislead people into thinking there's a
bigger difference in their overall effectiveness than there really is.
My fear of configuration-option-overwhelm leaves me inclined to _not_
add a configuration option for this, but we could.
- Commands, invocations and their parameters will now autocomplete
using introspection.
- Two types of parameter *arguments* will also autocomplete:
- --sampler_name will autocomplete the scheduler name
- --model will autocomplete the model name
- There don't seem to be commands for reading/writing image files yet, so
path autocompletion is not implemented
- resolve conflicts with generate.py invocation
- remove unused symbols that pyflakes complains about
- add **untested** code for passing intermediate latent image to the
step callback in the format expected.
This PR fixes#2951 and restores the step_callback argument in the
refactored generate() method. Note that this issue states that
"something is still wrong because steps and step are zero." However,
I think this is confusion over the call signature of the callback, which
since the diffusers merge has been `callback(state:PipelineIntermediateState)`
This is the test script that I used to determine that `step` is being passed
correctly:
```
from pathlib import Path
from invokeai.backend import ModelManager, PipelineIntermediateState
from invokeai.backend.globals import global_config_dir
from invokeai.backend.generator import Txt2Img
def my_callback(state:PipelineIntermediateState, total_steps:int):
print(f'callback(step={state.step}/{total_steps})')
def main():
manager = ModelManager(Path(global_config_dir()) / "models.yaml")
model = manager.get_model('stable-diffusion-1.5')
print ('=== TXT2IMG TEST ===')
steps=30
output = next(Txt2Img(model).generate(prompt='banana sushi',
iterations=None,
steps=steps,
step_callback=lambda x: my_callback(x,steps)
)
)
print(f'image={output.image}, seed={output.seed}, steps={output.params.steps}')
if __name__=='__main__':
main()
```
