Skip to content
GitLab
Menu
Projects
Groups
Snippets
Loading...
Help
Help
Support
Community forum
Keyboard shortcuts
?
Submit feedback
Contribute to GitLab
Sign in / Register
Toggle navigation
Menu
Open sidebar
chenpangpang
voicechat2
Commits
d948fb4e
Commit
d948fb4e
authored
Aug 12, 2024
by
chenpangpang
Browse files
feat:解决了很多bug
parent
fdec5b2a
Changes
3
Hide whitespace changes
Inline
Side-by-side
Showing
3 changed files
with
16 additions
and
6 deletions
+16
-6
Dockerfile
Dockerfile
+9
-4
README.md
README.md
+6
-1
启动器.ipynb
启动器.ipynb
+1
-1
No files found.
Dockerfile
View file @
d948fb4e
...
@@ -4,18 +4,23 @@ ARG IMAGE_UPPER=voicechat2
...
@@ -4,18 +4,23 @@ ARG IMAGE_UPPER=voicechat2
ARG
BRANCH=gpu
ARG
BRANCH=gpu
RUN
cd
/root
&&
git clone
-b
$BRANCH
http://developer.hpccube.com/codes/chenpangpang/
$IMAGE
.git
RUN
cd
/root
&&
git clone
-b
$BRANCH
http://developer.hpccube.com/codes/chenpangpang/
$IMAGE
.git
WORKDIR
/root/$IMAGE/$IMAGE_UPPER
WORKDIR
/root/$IMAGE/$IMAGE_UPPER
RUN
pip
install
-r
requirements.txt
#########
#########
# Prod #
# Prod #
#########
#########
FROM
image.sourcefind.cn:5000/gpu/admin/base/jupyterlab-pytorch:2.2.0-py3.10-cuda12.1-ubuntu22.04-devel
FROM
image.sourcefind.cn:5000/gpu/admin/base/jupyterlab-pytorch:2.2.0-py3.10-cuda12.1-ubuntu22.04-devel
COPY
chenyh/$IMAGE/Meta-Llama-3-8B-Instruct-Q4_K_M.gguf /root/$IMAGE/$IMAGE_UPPER/llama.cpp/Meta-Llama-3-8B-Instruct-Q4_K_M.gguf
ARG
IMAGE=voicechat2
ARG
IMAGE_UPPER=voicechat2
COPY
chenyh/$IMAGE/Meta-Llama-3-8B-Instruct-Q4_K_M.gguf /root/$IMAGE_UPPER/llama.cpp/Meta-Llama-3-8B-Instruct-Q4_K_M.gguf
COPY
chenyh/$IMAGE/tts_models--en--vctk--vits /root/.local/share/tts/tts_models--en--vctk--vits
COPY
chenyh/$IMAGE/tts_models--en--vctk--vits /root/.local/share/tts/tts_models--en--vctk--vits
COPY
chenyh/$IMAGE/distil-whisper/large-v2 /root/$IMAGE
/$IMAGE
_UPPER/distil-whisper/large-v2
COPY
chenyh/$IMAGE/distil-whisper/large-v2 /root/$IMAGE_UPPER/distil-whisper/large-v2
RUN
apt-get update
&&
apt-get
install
-y
build-essential byobu curl wget espeak-ng ffmpeg libopus0 libopus-dev
RUN
apt-get update
&&
apt-get
install
-y
build-essential byobu curl wget espeak-ng ffmpeg libopus0 libopus-dev
COPY
--from=base /opt/conda/lib/python3.10/site-packages /opt/conda/lib/python3.10/site-packages
COPY
--from=base /root/$IMAGE/$IMAGE_UPPER /root/$IMAGE_UPPER
COPY
--from=base /root/$IMAGE/$IMAGE_UPPER /root/$IMAGE_UPPER
WORKDIR
/root/$IMAGE_UPPER
# 如果直接复制库,则缺少可执行程序,故在子镜像中安装
RUN
pip
install
-r
requirements.txt
COPY
--from=base /root/$IMAGE/启动器.ipynb /root/$IMAGE/start.sh /root/
COPY
--from=base /root/$IMAGE/启动器.ipynb /root/$IMAGE/start.sh /root/
COPY
--from=base /root/$IMAGE/assets /root/assets
# 需要在同种型号gpu上编译
# 需要在同种型号gpu上编译
# RUN cd llama.cpp && make GGML_CUDA=1 -j
# RUN cd llama.cpp && make GGML_CUDA=1 -j
README.md
View file @
d948fb4e
# voicechat2
# voicechat2
## 需要的模型
-https://huggingface.co/bartowski/Meta-Llama-3-8B-Instruct-GGUF/resolve/main/Meta-Llama-3-8B-Instruct-Q4_K_M.gguf
-https://huggingface.co/bartowski/Meta-Llama-3-8B-Instruct-GGUF/resolve/main/Meta-Llama-3-8B-Instruct-Q4_K_M.gguf
-
distil-whisper/distil-large-v2
-
distil-whisper/distil-large-v2
-
https://coqui.gateway.scarf.sh/v0.6.1_models/tts_models--en--vctk--vits.zip
-
https://coqui.gateway.scarf.sh/v0.6.1_models/tts_models--en--vctk--vits.zip
\ No newline at end of file
## 创建方式
1.
Dockerfile构建镜像
2.
平台上编译:
`cd llama.cpp && make GGML_CUDA=1 -j`
3.
打开notebook页面,运行成功后,清空页面,导出镜像
\ No newline at end of file
启动器.ipynb
View file @
d948fb4e
...
@@ -9,7 +9,7 @@
...
@@ -9,7 +9,7 @@
"source": [
"source": [
"## 说明\n",
"## 说明\n",
"- 启动需要加载模型,需要1分钟左右的时间\n",
"- 启动需要加载模型,需要1分钟左右的时间\n",
"- 启动和重启 Notebook 点上方工具栏中的「重启并运行所有单元格」\n",
"- 启动和重启 Notebook 点上方工具栏中的「重启并运行所有单元格」
,出现`Uvicorn running on http://0.0.0.0:8003 (Press CTRL+C to quit)`就成功,可以开启页面了
\n",
"- 通过以下方式开启页面:\n",
"- 通过以下方式开启页面:\n",
" - 控制台打开「自定义服务」了,访问自定义服务端口号设置为8000\n",
" - 控制台打开「自定义服务」了,访问自定义服务端口号设置为8000\n",
"\n",
"\n",
...
...
Write
Preview
Markdown
is supported
0%
Try again
or
attach a new file
.
Attach a file
Cancel
You are about to add
0
people
to the discussion. Proceed with caution.
Finish editing this message first!
Cancel
Please
register
or
sign in
to comment