v3.7.2 release with Windows installer
v3.7.2 release with Windows installer.
Due to the size of the packaged python environment, the installer is within a multi-part zip file.
The multipart zip can be extracted using 7-Zip: https://www.7-zip.org/
Download both dgenerate_installer.zip.001
and dgenerate_installer.zip.002
to a folder.
Unzip dgenerate_installer.zip.001
to a directory (Right click, 7-Zip -> Extract to "dgenerate_installer") and then run dgenerate_installer\dgenerate.msi
to install.
dgenerate will be installed under C:\Program Files\dgenerate
by default with an isolated python environment provided.
The install directory will be added to PATH, and dgenerate will be available from the command line.
Portable Install
A portable install is provided via dgenerate_portable.zip.001
and dgenerate_portable.zip.002
, these contain
nothing but the dgenerate executable and a frozen python environment which can be placed anywhere.
3.7.2 Patch
Fixes missing value for --output-size
in the argument reconstruction output of --output-configs
and --output-metadata
that was occurring for prompt only generations and in some other circumstances.
Add --output-prefix
to the output of the options mentioned above (prefix on generated file name)
Remove --device
from the output of the mentioned options above, reproductions of the config that generated
an image or animation should be device agnostic and default to the first available GPU.
3.7.0 Features & Fixes
- Textual Inversion (prompt embeddings) support for Stable Diffusion XL,
--textual-inversions
now works with--model-type torch-sdxl
- Added
dgenerate --file
argument, which allows running configuration scripts from a file path instead of using a pipe or file redirect, this also allows you to use#! dgenerate --file
as the hash bang interpreter specification for bash or other unix shells in your config file - Any dgenerate argument which can accept a single file in a URI specification or otherwise, such as a
.safetensors
file for instance, can now be given a direct URL link to that model, dgenerate will download the file to its web cache and then use it. This allows for the use of direct links to CivitAI models in most cases provided that you append your CivitAI token to the URL arguments. See: The SDXL Textual Inversion Example and Utilizing CivitAI links and Other Hosted Models \download
directive anddownload()
template function for config, allows downloading models / images to dgenerates web cache or arbitrary paths. Simple caching behavior is implemented for downloads to arbitrary paths. Caching can be disabled. See:\download --help
and documentation- Adopted the file extension
.dgen
for dgenerate config / script files. - Windows installer can now create an optional file association for the extension
.dgen
which launches the dgenerate Console UI when clicking on a dgenerate config file, the file is opened in the Console UI in multiline editing mode automatically. - Console UI, improved syntax highlighting and fixes for URL & File Path entities
- Console UI, context aware syntax highlighting for the
\import_plugins
directive, plugin name arguments including file paths highlight as keywords for contrast - Console UI,
Run
menu now features aDebug Mode
option, which relaunches the shell process with the-v
argument, this causes copious amounts of debug information to be output when dgenerate is running, including unhandled exceptions from directives - Console UI, fix for typing in certain printable characters not triggering a re-highlight of the input text on Linux
- Fixes to cache flush on file expiry, non existent files no longer cause an error when being flushed from the web cache
- Latest
diffusers
library, latestspandrel
, many dependencies updated - Additional documentation about writing dgenerate config files and the config language itself with table of contents