Closed MattiaRigi97 closed 2 years ago
There is sth. weird with JAX installation. Could you do 'pip install -U jax[cuda] flax' and then run it again?
Enviroment: docker image with 'Linux 3b017f71c659 4.15.0-88-generic' after running 'pip install -U jax[cuda] flax', then run 'jina flow --uses flow.yml', I get error:
dalle flow is untested on CPU, also not recommend to run on CPU, it is too computational demanding.
Your error suggests you are trying to run JAX on a CPU machine
Actually I'm running in a GPU machine, although I'm running in a docker container, which I run with argument 'docker run --gpus all'
---Original--- From: "Han @.> Date: Sat, May 7, 2022 22:12 PM To: @.>; Cc: @.**@.>; Subject: Re: [jina-ai/dalle-flow] Failed to run my own server (Issue #1)
dalle flow is untested on CPU, also not recommend to run on CPU, it is too computational demanding.
Your error suggests you are trying to run JAX on a CPU machine
— Reply to this email directly, view it on GitHub, or unsubscribe. You are receiving this because you commented.Message ID: @.***>
maybe jax is too old? try pip install -U jax in docker?
CPU definitely doesn't work. I can attest to that :)
There are some caveats you may encounter when deploying your own server:
Hi, I have 8 GPU, and files and folders have been structured as described. After I do ' pip install -U jax' and upgrade jaxlab to 0.3.7, it generated error: I searched this error in stackflow and I got this: isin error
I installed with 'pip install jax==0.3.4 jaxlib==0.3.2' but it didn't help, it generated this error:
So I wonder which version of jax, jaxlib, flax, scipy you are using.
i see, let me work on an official Dockerfile today.
Thanks for your help
I ran into just about the same sequence of errors trying to run this on a Debian 11 GCloud Compute Engine VM.
To get the PrecisionLike
import working (after install the jax/jaxlib versions mentioned in https://github.com/jina-ai/dalle-flow/issues/1#issuecomment-1120607507), I installed the latest flax version from Github with:
pip install --upgrade git+https://github.com/google/flax.git
And now the jina flow builds/starts successfuly!
install latest flax and jax version from github follewed by 'pip install --upgrade jaxlib' worked for me, thanks!
after I deployed my own server and I changed the server_url to 'grpc://0.0.0.0:51005' in my client.py like
server has received the request but the client.py was stuck like forever
after I pressed Ctrl+C to stop it, I get trace back information:
and I found that the output image was not saved as expected, whether my writing form of 'grpc://0.0.0.0:51005' is correct?
guys, im almost ready with my dockerfile, see #7 still last check.
yes! it works with CUDA 11.6 out of the box! I'm going to write some README and merge into master!
It's done! Give it try, open new issue if you have problems
Nope, still need the official help
---Original--- From: "Jinke @.> Date: Wed, May 11, 2022 20:31 PM To: @.>; Cc: @.**@.>; Subject: Re: [jina-ai/dalle-flow] Failed to run my own server (Issue #1)
@huanghaifeng1234 Hi, I am also having this issue. The server received and processed the request but the client seemed to get stuck. Was this solved?
— Reply to this email directly, view it on GitHub, or unsubscribe. You are receiving this because you were mentioned.Message ID: @.***>
Hi, i'm trying to run my own server following the steps in the repo but I've the following error:
I'm on AWS Sagemaker console.. I tried with a ml.g4dn.16xlarge to avoid having resource issues as I thought it was related to that.
Can you help me to solve this issue?