Building wheels for collected packages: llama-cpp-python
678.1 Building wheel for llama-cpp-python (pyproject.toml): started
678.6 Building wheel for llama-cpp-python (pyproject.toml): finished with status 'error'
678.6 error: subprocess-exited-with-error
678.6
678.6 × Building wheel for llama-cpp-python (pyproject.toml) did not run successfully.
678.6 │ exit code: 1
678.6 ╰─> [16 lines of output]
678.6 *** scikit-build-core 0.11.4 using CMake 4.0.3 (wheel)
678.6 *** Configuring CMake...
678.6 loading initial cache file /tmp/tmp_rnn10qf/build/CMakeInit.txt
678.6 CMake Error at /tmp/pip-build-env-jv1jk7w_/normal/lib/python3.10/site-packages/cmake/data/share/cmake-4.0/Modules/CMakeDetermineCCompiler.cmake:49 (message):
678.6 Could not find compiler set in environment variable CC:
678.6
678.6 gcc-11.
678.6 Call Stack (most recent call first):
678.6 CMakeLists.txt:3 (project)
678.6
678.6
678.6 CMake Error: CMAKE_C_COMPILER not set, after EnableLanguage
678.6 CMake Error: CMAKE_CXX_COMPILER not set, after EnableLanguage
678.6 -- Configuring incomplete, errors occurred!
678.6
678.6 *** CMake configuration failed
678.6 [end of output]
678.6
678.6 note: This error originates from a subprocess, and is likely not a problem with pip.
678.6 ERROR: Failed building wheel for llama-cpp-python
678.6 Failed to build llama-cpp-python
678.7 ERROR: Failed to build installable wheels for some pyproject.toml based projects (llama-cpp-python)

failed to solve: process "/bin/sh -c cd /workspace && ./docker_build_script_ubuntu.sh" did not complete successfully: exit code: 1
ubuntu@ip-10-0-1-144:~/h2ogpt$ g++ --version

This is what my docker_build_script_ubuntu.sh looks like

#!/bin/bash
set -o pipefail
set -ex

export DEBIAN_FRONTEND=noninteractive
export PATH=/h2ogpt_conda/bin:$PATH
export HOME=/workspace
export PIP_EXTRA_INDEX_URL="https://download.pytorch.org/whl/cu121 https://huggingface.github.io/autogptq-index/whl/cu121"
export WOLFI_OS=0

Install linux dependencies

apt-get update && apt-get install -y
git
curl
wget
software-properties-common
pandoc
vim
libmagic-dev
poppler-utils
tesseract-ocr
libtesseract-dev
libreoffice
autoconf
libtool
docker.io
nodejs
npm
zip
unzip
htop
tree
tmux
jq
net-tools
nmap
ncdu
mtr
rsync
build-essential
parallel
bc
pv
expect
cron
at
screen
inotify-tools
jq
xmlstarlet
dos2unix
ssh

Run upgrades

apt-get upgrade -y

Install conda

wget https://repo.anaconda.com/miniconda/Miniconda3-latest-Linux-x86_64.sh &&
mkdir -p /h2ogpt_conda &&
bash ./Miniconda3-latest-Linux-x86_64.sh -b -u -p /h2ogpt_conda &&
conda update -n base conda &&
source /h2ogpt_conda/etc/profile.d/conda.sh &&
conda create -n h2ogpt -y &&
conda activate h2ogpt &&
conda install python=3.10 pygobject weasyprint -c conda-forge -y &&
echo "h2oGPT conda env: $CONDA_DEFAULT_ENV"

if building for CPU, would remove CMAKE_ARGS and avoid GPU image as base image

Choose llama_cpp_python ARGS for your system according to llama_cpp_python backend documentation, e.g. for CUDA:

export GGML_CUDA=0

export CMAKE_ARGS="-DGGML_BLAS=ON -DGGML_BLAS_VENDOR=OpenBLAS"

for Metal MAC M1/M2 comment out above two lines and uncomment out the below line

export CMAKE_ARGS="-DLLAMA_METAL=on"

export FORCE_CMAKE=1
export GPLOK=1
bash docs/linux_install.sh

chmod -R a+rwx /h2ogpt_conda

setup tiktoken cache

export TIKTOKEN_CACHE_DIR=/workspace/tiktoken_cache
python3.10 -c "
import tiktoken
from tiktoken_ext import openai_public

FakeTokenizer etc. needs tiktoken for general tasks

for enc in openai_public.ENCODING_CONSTRUCTORS:
encoding = tiktoken.get_encoding(enc)
model_encodings = [
'gpt-4',
'gpt-4-0314',
'gpt-4-32k',
'gpt-4-32k-0314',
'gpt-3.5-turbo',
'gpt-3.5-turbo-16k',
'gpt-3.5-turbo-0301',
'text-ada-001',
'ada',
'text-babbage-001',
'babbage',
'text-curie-001',
'curie',
'davinci',
'text-davinci-003',
'text-davinci-002',
'code-davinci-002',
'code-davinci-001',
'code-cushman-002',
'code-cushman-001'
]
for enc in model_encodings:
encoding = tiktoken.encoding_for_model(enc)
print('Done!')
"

Open Web UI

conda create -n open-webui -y
source /h2ogpt_conda/etc/profile.d/conda.sh
conda activate open-webui
conda install python=3.11 -y
echo "open-webui conda env: $CONDA_DEFAULT_ENV"

chmod -R a+rwx /h2ogpt_conda
pip install https://h2o-release.s3.amazonaws.com/h2ogpt/open_webui-0.3.8-py3-none-any.whl

Track build info

cp /workspace/build_info.txt /build_info.txt

mkdir -p /workspace/save
chmod -R a+rwx /workspace/save

Cleanup

                                                                                                                                        105,1         81%