This PR attempts to fix `--free_gpu_mem` option that was not working in
CKPT-based diffuser model after #1583.
I noticed that the memory usage after #1583 did not decrease after
generating an image when `--free_gpu_mem` option was enabled.
It turns out that the option was not propagated into `Generator`
instance, hence the generation will always run without the memory saving
procedure.
This PR also related to #2326. Initially, I was trying to make
`--free_gpu_mem` works on 🤗 diffuser model as well.
In the process, I noticed that InvokeAI will raise an exception when
`--free_gpu_mem` is enabled.
I tried to quickly fix it by simply ignoring the exception and produce a
warning message to user's console.
This PR adds `scripts/merge_fe.py`, which will merge any 2-3 diffusers
models registered in InvokeAI's `models.yaml`, producing a new merged
model that will be registered as well.
Currently this script will only work if all models to be merged are
known by their repo_ids. Local models, including those converted from
ckpt files, will cause a crash due to a bug in the diffusers
`checkpoint_merger.py` code. I have made a PR against
huggingface/diffusers which fixes this:
https://github.com/huggingface/diffusers/pull/2060
I've written up the install procedure for xFormers on Linux systems.
I need help with the Windows install; I don't know what the build
dependencies (compiler, etc) are. This section of the docs is currently
empty.
Please see `docs/installation/070_INSTALL_XFORMERS.md`
Starting `invoke.py` with --no-xformers will disable
memory-efficient-attention support if xformers is installed.
For symmetry, `--xformers` will enable support, but this is already the
default if xformers is available.
This commit suppresses a few irrelevant warning messages that the
diffusers module produces:
1. The warning that turning off the NSFW detector makes you an
irresponsible person.
2. Warnings about running fp16 models stored in CPU (we are not running
them in CPU, just caching them in CPU RAM)
This commit suppresses a few irrelevant warning messages that the
diffusers module produces:
1. The warning that turning off the NSFW detector makes you an
irresponsible person.
2. Warnings about running fp16 models stored in CPU (we are not running
them in CPU, just caching them in CPU RAM)
Starting `invoke.py` with --no-xformers will disable
memory-efficient-attention support if xformers is installed.
--xformers will enable support, but this is already the
default.
- During trigger token processing, emit better status messages indicating
which triggers were found.
- Suppress message "<token> is not known to HuggingFace library, when
token is in fact a local embed.
