- Adds mpmath==1.3.0 as a direct requirement to workaround a sympy
dependency issue.
- Change setup_venv.ps1 to install torchvision --nodeps. This avoids
a failure to find any common torch version dependency between the
pinned torch-mlir version and torchvision after a long search. This
also more closely matches what setup_venv.sh does.
* Shark-1.0: Update gradio to 4.19.2
- Move pin for gradio from 4.12 -> 4.19.2 on SHARK-1.0 branch
- Fix png import txt2img/txt2img_sdxl event handling for gradio 4.19.2
- Fix sendto txt2img/txt2img_sdxl event handling for gradio 4.19.2
* Shark-1.0: Reformat with black
- Apply the new (2024) black reformattings
* Add the @classmethod decorator to the base implementation of favored_base_models. This fixes running any SD pipeline that does
not have a specific implementation of favored_base_models
(everything but txt2img and txt2img_sdxl) break at the Sharkification
stage.
* Move IREE pins for windows, linux, and pip packaging files.
* Remove linux builds from 1.0 nightly workflow
It seems that the VMs used for these workflows are no longer available. Removing linux builds since publishing .exe is sufficient for the one-shot nightly workflows we trigger for SHARK-1.0
* SDXL Tab
* Filter VAEs in dropdown in the same manner as models
* Set default VAE selection to 'madebyollin/sdxl-vae-fp16-fix'
* Set default image size to 768x768 to match current Vulkan constraints
* SharkifySDModel Base Unet Model Determination
* Alway use the model_to_run as the base model for unet, if it is in
base_model.json, instead of potentially trying to compile for other base
models.
* Allow SharkSDPipelines to define a 'favor_base_models' @classmethod,
answering a list of sane base model names for the pipeline. Exclude base
models not in that list from compilation attempts when trying to determine
a base unet model.
* Add a 'favor_base_models' method for both Normal and SDXL Txt2Img
Pipelines. Define the method as answering 'None' in the base class.
* (Shark 1.0) UI: Upgrade to gradio 4.12.0 and fix breakage
* Upgrade Shark 1.0 to gradio 4.12.0
* Add javascript workaround for gradio currently ignoring @media rules in custom CSS. This fixes UI not showing at full width on desktop (>1536px width).
* (Shark 1.0) UI: Re-enable gallery download buttons
* Re-enable gallery download buttons, as this is now working again in Gradio 4.12
* Merges LoRA selection in the UI into a single selection, rather than
one for LoRAs under ./models and another for Hugging Face Id
* Add LoRA strength to UI and pipeline parameters.
* Add a `--lora_strength` command line argument.
* Bake LoRA strength into .vmfb naming when a LoRA is specified.
* Use LoRA embedded alpha values and (up tensor dimension *
LoRA strength) for final alpha when applying LoRA weights rather
than a hardcoded value of 0.75
* Adds additional cases to the LoRA weight application that are
present for weight application in the Kohya scripts.
* Include lora strength when reading and writing png metadata.
* Allow lora_strength to be set above 1.0 in the UI, so similar effects
to the prior (overdriven alpha) implementation can be obtained.
* Upgrade Gradio pin from 4.7.1 to 4.80.
* Make Nod AI logos visible again.
* Remove image toolbars from png import boxes.
* Set Input Images on img2img, outpaint and upscaler tabs to be upload
only.
* Change Image control to an ImageEditor control for masking on the
inpaint tab. Remove previous height restriction as this hides the
editing controls.
* Move Input Image/Masked Image on img2img, inpaint, outpaint and
upscaler tabs to be the first control on their tabs.
* Remove download buttons from all galleries as they download some
html rather the image (gradio issue #6595)
* Remove add new row and column from Output Gallery parameters
dataframe.
* Add partial workaround for not being able to select text in the Output
Gallery Gallery parameters dataframe (gradio issue #6086 )
* Fix uglified formatting of subdirectory selection dropown, refresh
button, and open folder buttons on the Output Gallery tab.
* Force Output Gallery to use the full width of the Gallery control
for the preview overlay when an image is selected, rather than
an overlay the width of the selected image.
* Fix sendto buttons.
* Reset Inpaint ImageEditor control with the Mask Layer after generation
is complete, as it gets lost if the image was sent to the tab from
another tab rather than being uploaded. Also rework queuing and
progress rendering along this codepath. This doesn't solve the
underlying problem of the Mask Layer being removed, but does get inpaint
fully working with the Gradio update.
* improved sharded performance and fixed issue with lmhead on rocm
* mmap shards + disable sharing of device arrays across devices
* fix device_idx for non-layer vmfbs
* fix time calc for sharded
---------
Co-authored-by: Elias Joseph <elias@nod-labs.com>
Co-authored-by: PhaneeshB <b.phaneesh@gmail.com>
* Fix some issues with defaults
Fixes to llama2 cpu compilation (turns off data tiling for old argmax
mode)
---------
Co-authored-by: Max Dawkins <max.dawkins@gmail.com>
* Update default CPU compilation flags.
c5a6cdc8dd52eb7e9b82
tweak CPU iree-compile flags to match upstream changes.
* Add an option for data tiling on SD models.
