v0.12.0 #824
Merged
v0.12.0 #824
Conversation
This file contains bidirectional Unicode text that may be interpreted or compiled differently than what appears below. To review, open the file in an editor that reveals hidden Unicode characters.
Learn more about bidirectional Unicode characters
joeyballentine
force-pushed
the
v0.12.0
branch
from
7ac6a8e
to
8d47bfa
Compare
update react-flow
* add swinir, initial * kinda working * Fix "classical_sr" swinir models not working with fp16 * Backend linting fixes * Add links to supported arches in README (#817) * update docs
* Paranoid incomplete initial commit * Another fuse function * More fuse functions * THE MONSTER * Final Fusion * Some more main and bug fixes * Creating layers * Finished adding code needed for ESRGAN * Param now printing correctly for ESRGAN test * Bin now writing correctly for ESRGAN * Convert -> Save NCNN working in chaiNNer * Added fp16 to converter * FpMode type * Model structure changes and convert additions; untested * Fixed bugs caused by previous commit * More ops + some refactoring * More ops * More ops and bugfixes * MemoryData and bugfixes * More ops and bugfixes * Added rest of ops except hard ones (I think) * Minor bugfix * Added an incomplete op * Added interpolation * fp16/fp32 can now interpolate * Load Model and Upscale Image now working * Renamed a couple files * More tweaks and fixes, eliminated major slowdowns * Fixed 1-channel bug * opt/non-opt now interpolable * Removed some unnecessary code * Cleaned up code and fixed bugs * Removed __main__ * Minor bugfix * linting bugfixes * more linting bugfixes * Removed unnecessary parse_ncnn_param.py * Fixed category error and linting issues * Fixed some pyright complaints * Fixed a fix * PR changes * Removed unused ncnn_parsers.py * Updated save test * Minor fix for commit before last * Changed NCNN model loading again * Removed garbage auto-import * PR comment * Implemented static factory method * Removed references to convertmodel.com * Added missing weightOrders to param schema * Fixed bug and added load prelu weight * Added error messages for param key errors * Added FP Mode TextOutput * Optimize onnx pre-conversion to for torch 1.12 compatibility * Compare weight shape instead of just size * Add onnxoptimizer to onnx requirements * Added onnxoptimizer size estimate * Added onnxoptimizer to requirements * Trigger CI * Fixed raise statements * Typing attempt * Add links to supported arches in README (#817) Co-authored-by: Joey Ballentine <34788790+joeyballentine@users.noreply.github.com>
* Added ONNX interpolate node * Added save test * Fixed linting errors
* Paranoid incomplete initial commit * Another fuse function * More fuse functions * THE MONSTER * Final Fusion * Some more main and bug fixes * Creating layers * Finished adding code needed for ESRGAN * Param now printing correctly for ESRGAN test * Bin now writing correctly for ESRGAN * Convert -> Save NCNN working in chaiNNer * Added fp16 to converter * FpMode type * Model structure changes and convert additions; untested * Fixed bugs caused by previous commit * More ops + some refactoring * More ops * More ops and bugfixes * MemoryData and bugfixes * More ops and bugfixes * Added rest of ops except hard ones (I think) * Minor bugfix * Added an incomplete op * Added interpolation * fp16/fp32 can now interpolate * Load Model and Upscale Image now working * Renamed a couple files * More tweaks and fixes, eliminated major slowdowns * Fixed 1-channel bug * opt/non-opt now interpolable * Removed some unnecessary code * Cleaned up code and fixed bugs * Removed __main__ * Minor bugfix * linting bugfixes * more linting bugfixes * Removed unnecessary parse_ncnn_param.py * Fixed category error and linting issues * Fixed some pyright complaints * Fixed a fix * PR changes * Removed unused ncnn_parsers.py * Updated save test * Minor fix for commit before last * Changed NCNN model loading again * Removed garbage auto-import * PR comment * Implemented static factory method * Updates NCNN upscale channel handling * Transparency hack for 1-channel models * Unsqueeze 1D arrays to work with pytorch * Removed references to convertmodel.com * Added missing weightOrders to param schema * Fixed bug and added load prelu weight * Added error messages for param key errors * Added FP Mode TextOutput * Optimize onnx pre-conversion to for torch 1.12 compatibility * Compare weight shape instead of just size * Add onnxoptimizer to onnx requirements * Added onnxoptimizer size estimate * Added onnxoptimizer to requirements * Trigger CI * Fixed raise statements * Typing attempt * Add links to supported arches in README (#817) Co-authored-by: Joey Ballentine <34788790+joeyballentine@users.noreply.github.com>
Fixed convenient upscale bugs
* Replace tile size target with new dropdown * Change to "number of tiles" * Update descriptions * make tile numbers correct * add migration * remove debug log
* Compile cache options before running * wip broken caching * working cache reduction * logging/linting fixes
* Allow numbers for all string inputs * fix text append type
Make all iterator indexes numbers
* Add direct pytorch to ncnn conversion * none
* Caching optimization fixes * garbage collection * manually delete and garbage collect auto splitting * use lazy init * remove some logging * Revert "use lazy init" This reverts commit 9129db8.
* Interpolate noop and onnx output bugfix * Fixed interpolate tile mode issue
* Replace the ncnn logo with something closer to the official one * replace with better ncnn icon
* Added Canny Edge Detection node (@jumpyjacko) * Add Canny Edge Detection node (@jumpyjacko) Added a newline to end of file for consistency between the files. * Add changes suggested in #869 * Fix bad grammar in node (#869) * Added Canny Edge Detection Node (#869)
* Add toast warning when modifying the chain during an execution * use id and make it subtle * put in middle
* Allow model save nodes to be used without onnx or ncnn installed * better return * fix import * remove dividing line
* Use opencv imwrite when possible to save ram * use rename and temp path
* Model file iterator * Update description
* Added toggle for checking for update on startup * Fix linting errors
Disable FP16 processing for swinir Found more models that don't support it, so disabling until I figure out why that is.
Fix lifecycle event handlers
* rely on estimated split depth * possibly more optimizations * oops * Fixes * Fix CPU
* Add favorites section to node selection context menu * Hide favorites if none
* Show execution time on nodes * Refactor into separate component * ts fixes
* Fixed blend shape bugs * Removed unused import Blend fix 2 electric blendaloo (#888) Shifted copy
joeyballentine
force-pushed
the
v0.12.0
branch
from
8d47bfa
to
4c846f3
Compare
This pull request was closed.
Sign up for free
to join this conversation on GitHub.
Already have an account?
Sign in to comment
Add this suggestion to a batch that can be applied as a single commit.
This suggestion is invalid because no changes were made to the code.
Suggestions cannot be applied while the pull request is closed.
Suggestions cannot be applied while viewing a subset of changes.
Only one suggestion per line can be applied in a batch.
Add this suggestion to a batch that can be applied as a single commit.
Applying suggestions on deleted lines is not supported.
You must change the existing code in this line in order to create a valid suggestion.
Outdated suggestions cannot be applied.
This suggestion has been applied or marked resolved.
Suggestions cannot be applied from pending reviews.
Suggestions cannot be applied on multi-line comments.
Suggestions cannot be applied while the pull request is queued to merge.
Suggestion cannot be applied right now. Please check back later.
Closes #235
Closes #883
Has the potential to close both #783 as well as #826, but needs verification