koboldcpp.sh - The Mamba Multitool (LostRuins#554)

* .sh script V1

* koboldcpp.sh polish

* koboldcpp.sh dist generator

* Include html's in dist

* RWKV in Linux Dist

* Lower dependency requirements

* Eliminate wget dependency

* More distinct binary name

I know its technically amd64, but I don't want to cause confusion among nvidia users.

* Use System OpenCL

Unsure how this will behave in the pyinstaller build, but pocl ended up CPU only. With a bit of luck the pyinstaller uses the one from the actual system if compiled in a system without opencl, while conda now includes it for that specific system.

* Add cblas dependency

Missing this causes compile failures on some system's

* ICD workaround

Ideally we find a better solution, but conda forces ICD and needs this for the successful compile. However, pyinstaller then embeds the ICD causing it to be limited to the system it was compiled for. By temporarily removing the ICD pyinstaller can't find it and everything remains functional. Ideally we do this on a pyinstaller level, but I could not find any good options to do so yet.

---------

Co-authored-by: root <root@DESKTOP-DQ1QRAG>

Author: henk717

.gitignore

@@ -118,5 +118,8 @@ hipblas.dll
 koboldcpp_hipblas.so
 koboldcpp_hipblas.dll
 
+bin/
+conda/
+
 # Jetbrains idea folder
 .idea/

environment.yaml

@@ -0,0 +1,24 @@
+name: koboldcpp
+channels:
+  - nvidia/label/cuda-11.5.0
+  - conda-forge
+  - defaults
+dependencies:
+  - python=3.8
+  - cuda-nvcc=11.5
+  - cuda-libraries-dev=11.5
+  - cxx-compiler
+  - gxx=10
+  - pip
+  - git=2.35.1
+  - libopenblas
+  - openblas
+  - clblast
+  - ninja
+  - make
+  - packaging
+  - pyinstaller
+  - libcblas
+  - ocl-icd-system
+  - pip:
+    - customtkinter

koboldcpp.sh

@@ -0,0 +1,22 @@
+#!/bin/bash
+if [ ! -f "bin/micromamba" ]; then
+	curl -Ls https://anaconda.org/conda-forge/micromamba/1.5.3/download/linux-64/micromamba-1.5.3-0.tar.bz2 | tar -xvj bin/micromamba
+fi
+
+if [[ ! -f "conda/envs/linux/bin/python" || $1 == "rebuild" ]]; then
+	bin/micromamba create --no-shortcuts -r conda -n linux -f environment.yaml -y
+	bin/micromamba create --no-shortcuts -r conda -n linux -f environment.yaml -y
+	bin/micromamba run -r conda -n linux make clean
+fi
+
+bin/micromamba run -r conda -n linux make LLAMA_OPENBLAS=1 LLAMA_CLBLAST=1 LLAMA_CUBLAS=1 LLAMA_PORTABLE=1
+
+if [[ $1 == "rebuild" ]]; then
+	echo Rebuild complete, you can now try to launch Koboldcpp.
+elif [[ $1 == "dist" ]]; then
+	bin/micromamba remove -r conda -n linux --force ocl-icd -y
+	bin/micromamba run -r conda -n linux pyinstaller --noconfirm --onefile --collect-all customtkinter --add-data='./koboldcpp_default.so:.' --add-data='./koboldcpp_cublas.so:.' --add-data='./koboldcpp_openblas.so:.' --add-data='./koboldcpp_clblast.so:.' --add-data='./koboldcpp_clblast_noavx2.so:.' --add-data='./klite.embd:.' --add-data='./kcpp_docs.embd:.' --add-data='./rwkv_vocab.embd:.' --add-data='./rwkv_world_vocab.embd:.' --clean --console koboldcpp.py -n "koboldcpp-linux-x64"
+	bin/micromamba install -r conda -n linux ocl-icd -c conda-forge -y
+ else
+	bin/micromamba run -r conda -n linux python koboldcpp.py $*
+fi