Releases · nod-ai/SHARK-Studio

Automatic snapshot release of nod.ai SHARK.

What's Changed

Macos fixes by @PhaneeshB in #1883
Add support for Falcon GPTQ by @vivekkhandelwal1 in #1884
Fix Falcon GPTQ Pipeline by @vivekkhandelwal1 in #1886
SD: Fix repeatable seeds consistency over different batch counts by @one-lithe-rune in #1889
Add matmul reassociation flags and fix CPU compile flag conditional by @monorimet in #1891
Add support for sharded Falcon model by @vivekkhandelwal1 in #1892
SD: Fix repeatable seeds when initial seed should be random by @one-lithe-rune in #1893
Fix Sharded Falcon-180b-GPTQ Pipeline by @vivekkhandelwal1 in #1897
Stringify path for ireec.compile_file by @monorimet in #1901
SharkInference: Fix various examples and README.md by @no1wudi in #1903
Add --additional_runtime_args option and use in OPT example. by @monorimet in #1855
SD/UI: Use a single model selection box on UI tabs by @one-lithe-rune in #1906
SD - Fix civitai download on Windows +other civitai download improvements by @one-lithe-rune in #1907
Fix MLIR Textual PassPipeline Error by @one-lithe-rune in #1910
LLaMa 2 Chat template fix by @xzuyn in #1913
Position Generate, Randomize Seed, & Stop Batch buttons to match txt2img tab by @xzuyn in #1915
Save hiresfix info if used by @xzuyn in #1914
SD: Fix unet untuned opt_flags by @one-lithe-rune in #1912
Updates to opt_causallm example by @monorimet in #1905
Add opt_causallm_samples.py. by @godot73 in #1916
update dependency on rocm/hip info command by @PhaneeshB in #1900
Add sharded Falcon support by @vivekkhandelwal1 in #1919
Use iree tools to get supported driver info by @PhaneeshB in #1815
Add smoothquant OPT to examples. by @monorimet in #1922
remove caching for iree rocm args by @PhaneeshB in #1923
Add ZoeDepth by @gpetters94 in #1834
Add sharded Falcon-GPTQ support by @vivekkhandelwal1 in #1927
Add 7800xt to target triples explicitly. by @monorimet in #1928
Fix issue in Falcon-GPTQ by @vivekkhandelwal1 in #1930
Add support for Falcon-40b-GPTQ by @vivekkhandelwal1 in #1940
[compile utils] Fix ROCM to not expect config.id as a default. by @raikonenfnu in #1939
[vicuna.py] Allow to pass extra arguments to iree-compile by @kuhar in #1935
SD/API: Add missing A1111 APIs to Shark to support koboldcpp image generation by @one-lithe-rune in #1924
(Studio) Disable SD tunings and sub-model downloads by @monorimet in #1944
UI: Fix some webui launch corner cases by @one-lithe-rune in #1952
fix vicuna cli vulkan path by @PhaneeshB in #1934
Fix wrong parameter index passed to 'compile_module_to_flatbuffer' by @no1wudi in #1921
Fix MacOS web execution flow by @PhaneeshB in #1899
Fix typo in the message; s/reponse/response by @godot73 in #1920
Bump pypdf from 3.12.2 to 3.17.0 in /apps/language_models/langchain by @dependabot in #1929
Bump langchain from 0.0.202 to 0.0.325 in /apps/language_models/langchain by @dependabot in #1926
Don't require vulkan installation to query devices. by @monorimet in #1953
fix rocm arch selection by @PhaneeshB in #1950
UI: Fix webui launch on non-Windows by @one-lithe-rune in #1963
Pin pydantic to 2.4.1 in requirements by @monorimet in #1967
Bump langchain from 0.0.325 to 0.0.329 in /apps/language_models/langchain by @dependabot in #1968
API/Docs: Fix an image link in koboldcpp doc by @one-lithe-rune in #1954
Update rocm device selection logic by @PhaneeshB in #1965
SD/UI Restrict hires fix/img2img resamplers/schedulers by @one-lithe-rune in #1955
(SHARK Studio) Add Turbine-based llm chatbot. by @monorimet in #1933
Add --no-index to setup_venv.sh runtime pip install. by @monorimet in #1971
fix iree version mismatch by @PhaneeshB in #1974
Update brevitas quant api by @jinchen62 in #1975
Re-enable SD tunings without matmuls. by @monorimet in #1976
UI: Display top tag frequencies and model trained against when a LoRA is selected by @one-lithe-rune in #1972
UI: Output galllery cleanups by @one-lithe-rune in #1959
Disable tunings and fix CI vulkan driver issues. by @monorimet in #1979
API/Docs: Fix incorrect cors arguments listing by @one-lithe-rune in #1983
Save intermediate values of controlnet by @gpetters94 in #1981
Small UI tweaks for chatbot, fix torchvision requirements by @monorimet in #1988
Use clean_device_info() by default and don't write .mlir to /tmp/ by @monorimet in #1984
quick fix rocm None device err by @PhaneeshB in #1990
Add .mlir to startup shark_tmp cleanup by @one-lithe-rune in #1991
Create specified dir if needed during save_mlir and fix vulkan device fetching without URI/ID by @monorimet in #1989
[vicuna.py] Rework benchmark statistics calculation by @kuhar in #1992
[vicuna.py] Add option to enable tracing by @kuhar in #1993
Add Falcon-GPTQ Support for 2-way sharding by @vivekkhandelwal1 in #1960
Fix typo in default_rocm_arch by @edrutte in #1998
[SD] Add flag to collapse reduction dims pre dispatch formation by @qedawkins in #1999
Cleanup xfails in pytest suite. by @monorimet in #1995
Fix unsharded Falcon pipeline by @vivekkhandelwal1 in #1997
Fix an importer issue on Linux by @gpetters94 in #1986
Update default CPU compilation flags. by @monorimet in #2000
Add control mode code by @gpetters94 in #1957
Add multicontrolnet by @gpetters94 in #1958
[SDXL] Add SDXL pipeline to SHARK by @Abhishek-Varma in #1941
updated sharded model to use input model path and pass token if needed by @Eliasj42 in #1982
(Studio) Update gradio and multicontrolnet UI. by @monorimet in #2001
(SD) Fix schedulers and multi-controlnet. by @monorimet in #2006
Fix zoedepth import and UI issues by @gpetters94 in #2010
Fixes to llama2 cpu compilation and studio UI, schedulers by @monorimet in #2013
improved sharded performance and fixed issue with lmhead on rocm by @Eliasj42 in #2008
finilized fixes for sharded llama2 by @Eliasj42 in #2018
[ui] Add UI for sharding by @Shukla-Gaurav in #2014
Fix SharkEulerDiscrete Scheduler by @monorimet in #2022
(Studio) fix nod logo by @monorimet in #2023
(vicuna.py) Move enable_tracy_tracing outside of BenchmarkRunInfo by @monorimet in #2011
#1843 - Add Export Default settings button by @RichardPastircak in #2016
Many UI fixes and controlnet impovements by @monorimet in #2025
Fix stencil pipline to use input image by @gpetters94 in #2027
(Studio) Fix controlnet switching. by @monorimet in #2026
vulkan device id fix by @PhaneeshB in #2028
Add support for StableLM-3B model by @vivekkhandelwal1 in https://github.com/nod-ai/SHARK/pul...