[SUPPORT] GRTGBLN - DOCKER TEMPLATES

[grtgbln]

1 hour ago, UnraidTobias said:

Hi @grtgbln can you please provide a correct support-forum link in your gpt4all container and add an example on how to set it up properly? With a LLM as an example and a Frontend? That would be so helpful, probably for many of us,.... thanks so much in advance!

The support link on the template points to the GitHub page for GPT4ALL, which contains documentation about the project. Please direct your project-specific questions to those locations.

[Aegisnir]

Hello.  Trying out your Automatic1111 container but I can't find where the webui.sh file is located.  I can see it inside the container via command line, but not inside of the appdata directory when I navigate through the unraid file system or a file browser.  Can you please advise where this and the rest of the data is located?  I assume its writing directly into docker.img directly without using the appdata folder...?

[UnraidTobias]

On 7/29/2024 at 10:16 PM, grtgbln said:

The support link on the template points to the GitHub page for GPT4ALL, which contains documentation about the project. Please direct your project-specific questions to those locations.

Hi, i found out, the issue on my side is, that the nvidia card i have is to old. So nvidia drivers do not get detected....but according to gpt4all github readme, a gpu is not needed at all.

 

Is there an option you could provide the container also without the nvidia/gpu requirment? Or optional? Or can you elaborate a bit about the background why this is necessary, please? Thank you

Edited August 4 by UnraidTobias

[Krakout]

Can someone help me install LocalAI with AMD support?

 

I don't understand the instructions about adding devices and variables...

 

" For AMD GPU support, add "/dev/kfd" and "/dev/dri" each as a Device and add the required Variables: https://localai.io/features/gpu-acceleration/#setup-example-dockercontainerd"

 

Thanks!

[bmartino1]

10 hours ago, Krakout said:

Can someone help me install LocalAI with AMD support?

 

I don't understand the instructions about adding devices and variables...

 

" For AMD GPU support, add "/dev/kfd" and "/dev/dri" each as a Device and add the required Variables: https://localai.io/features/gpu-acceleration/#setup-example-dockercontainerd"

 

Thanks!

not possible. the models won't use the amd gpu.

I recommend using the AIO gpu hib version:

localai/localai:latest-aio-cpu

to add the gpu if any for amd to a docker.

 

--device=/dev/dri --device=/dev/kfd

add the above to the extra parameter option:

 

I see they finaly updateed and fixed it.

Edited August 10 by bmartino1
redaction.

[EvilOni]

Hi, i managed to get the anythingallm docker image running. It works fine for RAG tasks, however i have an issue with attching URL's via the bulk web scraper. I see the following error in the log.

[collector] info: Discovering links...
[collector] error: Failed to get page links from https://learn.microsoft.com/en-us/azure/well-architected/reliability. Error: Failed to launch the browser process!
[1006:1006:0810/184605.831237:FATAL:zygote_host_impl_linux.cc(127)] No usable sandbox! Update your kernel or see https://chromium.googlesource.com/chromium/src/+/main/docs/linux/suid_sandbox_development.md for more information on developing with the SUID sandbox. If you want to live dangerously and need an immediate workaround, you can try using --no-sandbox.

Any help would be appreciated

[bmartino1]

14 minutes ago, EvilOni said:

Hi, i managed to get the anythingallm docker image running. It works fine for RAG tasks, however i have an issue with attching URL's via the bulk web scraper. I see the following error in the log.

[collector] info: Discovering links...
[collector] error: Failed to get page links from https://learn.microsoft.com/en-us/azure/well-architected/reliability. Error: Failed to launch the browser process!
[1006:1006:0810/184605.831237:FATAL:zygote_host_impl_linux.cc(127)] No usable sandbox! Update your kernel or see https://chromium.googlesource.com/chromium/src/+/main/docs/linux/suid_sandbox_development.md for more information on developing with the SUID sandbox. If you want to live dangerously and need an immediate workaround, you can try using --no-sandbox.

Any help would be appreciated

For ameren PSP(power smart pricing) i ran a python script for HA to scrape the table data for the price at that hour.

You may be hitting an issue with the website you are scraping that need to be rendered in a browser, as its a script that needs ran to then scrape....

Any way, per the error,
as it appears that the Browser Error at launching chromium...

You can review similar python scraping here:

I have a constant discord message bot that successfully renders the page for me and sends me this data in this example.

I have once lxc setup for myself and one setup for a friend/solar panel business...

While I'm not familiar with what the anythingallm docker is nor its capabilities, if its web scraping, you may need to do something similar to collect the data.

I found it easier to use python  selenium with chrome installed
 

from selenium import webdriver
from selenium.webdriver.common.by import By
from selenium.webdriver.support.ui import WebDriverWait
from selenium.webdriver.support import expected_conditions as EC

and called:
 

 chrome_options = webdriver.ChromeOptions()
    chrome_options.add_argument('--headless')
    chrome_options.add_argument('--no-sandbox')
    chrome_options.add_argument('--disable-dev-shm-usage')

to launch the browser.

You may need similar settings...

Edited August 10 by bmartino1
spelling/gramar

[EvilOni]

@bmartino1

 

Ok i figured out the problem. You need to add an argument to extra parameters of the docker container settings.

--cap-add SYS_ADMIN

 

This is referenced in the AnythingLLM documentation, so the author should probably include this as part of the instructions.

 

 

https://docs.anythingllm.com/installation/self-hosted/local-docker#recommend-way-to-run-dockerized-anythingllm

Edited August 13 by EvilOni

[grtgbln]

58 minutes ago, EvilOni said:

@bmartino1

 

Ok i figured out the problem. You need to add an argument to extra parameters of the docker container settings.

--cap-add SYS_ADMIN

 

This is referenced in the AnythingLLM documentation, so the author should probably include this as part of the instructions.

 

 

https://docs.anythingllm.com/installation/self-hosted/local-docker#recommend-way-to-run-dockerized-anythingllm

Thanks for diagnosing the issue, I have updated the template accordingly.

[TheFullTimer]

For the Invoke-AI container the path to models and outputs can be mapped elsewhere. Assuming your storage is fast or you have extra NVMe laying around.

The paths are:
- Container Path: /invokeai_root/outputs
- Container Path: /invokeai_root/models

Edited August 18 by TheFullTimer
removed markdown formatting

[disco4000]

Hi,

the Invoke-AI container is pulling 4,5GB with every update almost every next day.
Do all these big files really change everytime there is an update or ist this a misconfiguration in the docker build file?

 

Cheers, d.

[grtgbln]

3 hours ago, disco4000 said:

Hi,

the Invoke-AI container is pulling 4,5GB with every update almost every next day.
Do all these big files really change everytime there is an update or ist this a misconfiguration in the docker build file?

 

Cheers, d.

I doubt the underlying dependencies are actually changing, but part of the Dockerfile build file is importing things like NVIDIA SMI and PyTorch on every build, without caching these layers, which unfortunately leads to having to redownload these base layers every time the image updates.

 

https://github.com/invoke-ai/InvokeAI/blob/5eb919f6020118764b8b5ece4f7660bd31f52472/docker/Dockerfile#L36
 

Edited September 8 by grtgbln

[MrTroll]

On 9/9/2024 at 2:45 AM, grtgbln said:

I doubt the underlying dependencies are actually changing, but part of the Dockerfile build file is importing things like NVIDIA SMI and PyTorch on every build, without caching these layers, which unfortunately leads to having to redownload these base layers every time the image updates.

 

https://github.com/invoke-ai/InvokeAI/blob/5eb919f6020118764b8b5ece4f7660bd31f52472/docker/Dockerfile#L36
 

Its also moved to the Pre-release branch now, is this expected behavior?

[grtgbln]

On 9/12/2024 at 12:48 AM, MrTroll said:

Its also moved to the Pre-release branch now, is this expected behavior?

It's set up to pull whatever the InvokeAI team deems the "latest" Docker image: https://github.com/invoke-ai/InvokeAI/pkgs/container/invokeai/273561071?tag=latest