Running Docker image with customized yaml doesnt work
pasa13142 opened this issue · comments
Hi, Thanks for this great repo!
If i try to run clip as service in docker with
docker run -p 51009:51000 -v $HOME/.cache:/home/cas/.cache --gpus all jinaai/clip-server:master-onnx everything is okay. It works on gpu with 1 replica.
I just would like to run this proper docker with more replicas and created custom yaml:
python -m clip_server onnx-flow-custom.yml
and tried:
cat onnx-flow-custom.yml | docker run -i -p 51009:51000 -v $HOME/.cache:/home/cas/.cache --gpus all jinaai/clip-server:master-onnx -i
and i just want to add more replicas in it instead of 1.
jtype: Flow
version: '1'
with:
port: 51000
executors:
- name: clip_t
replicas: 4
uses:
jtype: CLIPEncoder
with:
device: cuda
metas:
py_modules:
- clip_server.executors.clip_onnx
But I try with above yaml or even with original yamls like:
jtype: Flow
version: '1'
with:
port: 51000
executors:
- name: clip_o
uses:
jtype: CLIPEncoder
metas:
py_modules:
- clip_server.executors.clip_onnx
timeout_ready: 3000000
replicas: 1
they dont work. no gpu, no even replicas. Whatever i do, it throws that warning message with a full of page of it: Default docker image run command doesnt throw this deprecationWarning message.
DeprecationWarning: tostring() is deprecated. Use tobytes() instead. (raised from /usr/local/lib/python3.8/dist-packages/onnxconverter_common/float16.py:95)
DeprecationWarning: The binary mode of fromstring is deprecated, as it behaves surprisingly on unicode inputs. Use frombuffer instead (raised from /usr/local/lib/python3.8/dist-packages/onnxconverter_common/float16.py:91)
I just wanted to use original master-onnx image with more replicas. Can anyone help?
Could you try this:
docker run -i -p 51009:51000 -v $HOME/.cache:/home/cas/.cache -v $PWD:/cas/ --gpus all jinaai/clip-server:master-onnx onnx-flow-custom.yml
Didnt work again. Throws full of a page of this error as same as i explained methods above. Default docker image run command doesnt throw this deprecationWarning message.
DeprecationWarning: tostring() is deprecated. Use tobytes() instead. (raised from /usr/local/lib/python3.8/dist-packages/onnxconverter_common/float16.py:95)
DeprecationWarning: The binary mode of fromstring is deprecated, as it behaves surprisingly on unicode inputs. Use frombuffer instead (raised from /usr/local/lib/python3.8/dist-packages/onnxconverter_common/float16.py:91)
@pasa13142 There is a issue with the clip_server onnx docker image, we are fixing it and will let you know for any progress. Sorry for the inconvenience! You might use clip_server torch image or serve with source code to unblock your work
Hi @pasa13142
The issue is fixed and the pr is merged. Please pull the latest jinaai/clip-server:master-onnx docker image and try again.
Notice that we also updated the instructions on how to serve the docker container here. The correct way to run a onnx docker container using the default config is like
docker run -p 51009:51000 -v $HOME/.cache:/home/cas/.cache --gpus all jinaai/clip-server:master-onnx onnx-flow.yml (If you don't add onnx-flow.yml it will use the default yaml which runs in pytorch mode)
And to use a custom config is the same
cat my.yml | docker run -i -p 51009:51000 -v $HOME/.cache:/home/cas/.cache --gpus all jinaai/clip-server:master-onnx -i
Thanks for reporting this issue 🍻
Hello, Can I have clip-server:master-onnx dockerfile?
It's in the repo:
Dockerfiles/server.Dockerfile
after built image with :
git clone https://github.com/jina-ai/clip-as-service.git
docker build . -f Dockerfiles/server.Dockerfile --build-arg GROUP_ID=$(id -g ${USER}) --build-arg USER_ID=$(id -u ${USER}) -t jinaai/clip-server
and run with both :
cat server/clip_server/onnx-flow.yml | docker run -i -p 51009:51000 -v $HOME/.cache:/home/cas/.cache --gpus all clip/clon:latest -i
and
docker run -p 51009:51000 -v $HOME/.cache:/home/cas/.cache --gpus all clip/clon:latest onnx-flow.yml
They throw the same error:
ERROR clip_o/rep-0@24 fail to load file dependency [04/13/23 16:22:41]
ERROR clip_o/rep-0@24 FileNotFoundError('can not find [04/13/23 16:22:41]
clip_server.executors.clip_onnx') during <class
'jina.serve.runtimes.worker.WorkerRuntime'>
initialization
add "--quiet-error" to suppress the exception
details
Traceback (most recent call last):
File
"/usr/local/lib/python3.8/dist-packages/jina/orchest…
line 79, in run
runtime = runtime_cls(
File
"/usr/local/lib/python3.8/dist-packages/jina/serve/r…
line 41, in __init__
super().__init__(args, **kwargs)
File
"/usr/local/lib/python3.8/dist-packages/jina/serve/r…
line 78, in __init__
self._loop.run_until_complete(self.async_setup())
File "/usr/lib/python3.8/asyncio/base_events.py",
line 616, in run_until_complete
return future.result()
File
"/usr/local/lib/python3.8/dist-packages/jina/serve/r…
line 106, in async_setup
self._request_handler = WorkerRequestHandler(
File
"/usr/local/lib/python3.8/dist-packages/jina/serve/r…
line 53, in __init__
self._load_executor(
File
"/usr/local/lib/python3.8/dist-packages/jina/serve/r…
line 203, in _load_executor
self._executor: BaseExecutor =
BaseExecutor.load_config(
File
"/usr/local/lib/python3.8/dist-packages/jina/jaml/__…
line 742, in load_config
load_py_modules(
File
"/usr/local/lib/python3.8/dist-packages/jina/jaml/he…
line 285, in load_py_modules
PathImporter.add_modules(*mod)
File
"/usr/local/lib/python3.8/dist-packages/jina/importe…
line 161, in add_modules
_path_import(complete_path(m))
File
"/usr/local/lib/python3.8/dist-packages/jina/jaml/he…
line 229, in complete_path
raise FileNotFoundError(f'can not find {path}')
FileNotFoundError: can not find
clip_server.executors.clip_onnx
ERROR Flow@ 1 Flow is aborted due to ['clip_o'] can not be [04/13/23 16:22:41]
started.
Traceback (most recent call last):
File "/usr/lib/python3.8/runpy.py", line 194, in _run_module_as_main
return _run_code(code, main_globals, None,
File "/usr/lib/python3.8/runpy.py", line 87, in _run_code
exec(code, run_globals)
File "/usr/local/lib/python3.8/dist-packages/clip_server/__main__.py", line 25, in <module>
with f:
File "/usr/local/lib/python3.8/dist-packages/jina/orchestrate/orchestrator.py", line 14, in __enter__
return self.start()
File "/usr/local/lib/python3.8/dist-packages/jina/orchestrate/flow/builder.py", line 33, in arg_wrapper
return func(self, *args, **kwargs)
File "/usr/local/lib/python3.8/dist-packages/jina/orchestrate/flow/base.py", line 1788, in start
self._wait_until_all_ready()
File "/usr/local/lib/python3.8/dist-packages/jina/orchestrate/flow/base.py", line 1927, in _wait_until_all_ready
raise RuntimeFailToStart
jina.excepts.RuntimeFailToStart