Software Engineering Courses (SECourses)•3y ago

Wanna start by saying thanks for all the guides. Learned quite a bit from finding you on YouTube. I

Wanna start by saying thanks for all the guides. Learned quite a bit from finding you on YouTube. I wanted to ask if you had plans of doing a textual training guide for runpod. I've been trying to train an embedding for 3 days now and no matter what setup, I always get "CUDA out of memory". Figured its something I'm doing wrong on initial setup but there's not really any runpod guides for it.

�𝓙𝓪𝓵𝓮𝓲 Wanna start by saying thanks for all the guides. Learned quite a bit from findin...

Furkan Gözükara SECourses•5/31/23, 1:29 AM

textual inversion is the weakest link

Furkan Gözükara SECourses•5/31/23, 1:29 AM

why do you want it?

Furkan Gözükara SECourses•5/31/23, 1:29 AM

dreambooth best

�

𝓙𝓪𝓵𝓮𝓲OP•5/31/23, 1:30 AM

Dreambooth for a ckpt or to train a lora? I just wanted to be able to use my characters specific face and figured TI would be the easiest route.

�𝓙𝓪𝓵𝓮𝓲 Dreambooth for a ckpt or to train a lora? I just wanted to be able to use my cha...

Furkan Gözükara SECourses•5/31/23, 1:30 AM

dreambooth generate safetensors. yes it is the best one

Furkan Gözükara SECourses•5/31/23, 1:30 AM

most accurate

�

𝓙𝓪𝓵𝓮𝓲OP•5/31/23, 1:31 AM

Ok I'll revisit the video for that and look into it. Thanks for the reply

�𝓙𝓪𝓵𝓮𝓲 Ok I'll revisit the video for that and look into it. Thanks for the reply

Furkan Gözükara SECourses•5/31/23, 1:35 AM

you are welcome. i tested all textual inversion lora and dreambooth

Furkan Gözükara SECourses•5/31/23, 1:35 AM

dreambooth really the best

�

𝓙𝓪𝓵𝓮𝓲OP•5/31/23, 1:39 AM

I was hestitant doing dreambooth because my preferred model is meinaunreal. And I only have about 30 high quality pictures of the characters face.

�𝓙𝓪𝓵𝓮𝓲 I was hestitant doing dreambooth because my preferred model is meinaunreal. And ...

Furkan Gözükara SECourses•5/31/23, 1:40 AM

well i think it should work. i teach with 10-20 images a person

�

𝓙𝓪𝓵𝓮𝓲OP•5/31/23, 1:43 AM

Is the "RunPod Fix For DreamBooth & xFormers" the most up to date guide for dreambooth? Gonna give this a shot tonight.

�𝓙𝓪𝓵𝓮𝓲 Is the "RunPod Fix For DreamBooth & xFormers" the most up to date guide for drea...

Furkan Gözükara SECourses•5/31/23, 1:51 AM

this one still up to date : https://youtu.be/zA4LksIVas8

YouTubeSECourses

RunPod Fix For DreamBooth & xFormers - How To Use Automatic1111 Web...

RunPod: https://bit.ly/RunPodIO. Discord : https://bit.ly/SECoursesDiscord. How to install latest version of Automatic1111 Web UI and #DreamBooth on #RunPod. If I have been of assistance to you and you would like to show your support for my work, please consider becoming a patron on

https://www.patreon.com/SECourses

Playlist of #StableDiffusi...

�

𝓙𝓪𝓵𝓮𝓲OP•5/31/23, 1:53 AM

Thanks!

Ffelixsanz go here `chrome://settings/fonts`, and on the fixed-width font select monospace,...

Khang•5/31/23, 3:35 AM

Awesome, my chrome doesn't have monospace or inconsolata font so I ask ChatGPT and it told me to choose another monospace font named Consolas and it works. Thank you very much

Fiora•5/31/23, 7:42 AM

Why doesn't it access gpu? I use colab, but it doesn't access gpu.

Original message was deleted

Khang•5/31/23, 7:49 AM

That's awesome thank you so much Ashleyk I'm gonna test it out. Exactly what I wished for.

KKhang That's awesome thank you so much Ashleyk I'm gonna test it out. Exactly what I w...

ashley•5/31/23, 7:51 AM

Might take a little bit of time to pull all the layers of the image the first time a pod starts until it eventually gets cached on the host machine, but I would rather wait for the pod to start than mess around with configuring things manually each time.

ashley•5/31/23, 7:52 AM

I will add the ControlNet and Deforum extensions to the next version so that it will match the new RunPod version 8.0.0 configuration.

ashley•5/31/23, 7:53 AM

Their 8.0.0 version is behind though. They only have WebUI version 1.2.1 installed by default, but 1.3.0 is the latest release. My template installs 1.3.0.

ashley•5/31/23, 7:53 AM

And they don't install the Dreambooth extension at all in their 8.0.0 version.

Aashley Their 8.0.0 version is behind though. They only have WebUI version 1.2.1 instal...

Khang•5/31/23, 8:18 AM

Yeah I asked them and they said that DreamBooth tends to break things, but they are considering pre-installing it. I also have to configure things manually each time I create a new pod. And what do you mean by 8.0.0? Is it the Runpod SD template? I'm still using version 6.0.1 and have to do a 'git pull' and download everything each time.

Khang•5/31/23, 8:19 AM

The XYZ plot in their version doesn't work with models created by Kohya_ss. It shows the same images everytime

KKhang Yeah I asked them and they said that DreamBooth tends to break things, but they ...

ashley•5/31/23, 8:23 AM

They released 8.0.0 last night, but then rolled it back to 6.0.1 because they confused the hell out of everyone. Especially since it installed the WebUI to /stable-diffusion-webui instead of /workspace/stable-diffusion-webui.

8.0.0 still exists but its hidden from the UI. I can still start up a version 8.0.0 using my RunPod scripts. I have created the following RunPod scripts in Python but have not released them on Github yet.

cancel_dreambooth_job.py
create_on_demand_pod.py
create_spot_pod.py
create_template.py
get_bid_price.py
get_details.py
get_dreambooth_health.py
get_dreambooth_status.py
get_gpu_types.py
get_pod.py
get_pods.py
runpod.py
start_on_demand_pod.py
start_spot_pod.py
stop_pod.py
template.md
terminate_pod.py
train_dreambooth.py

cancel_dreambooth_job.py
create_on_demand_pod.py
create_spot_pod.py
create_template.py
get_bid_price.py
get_details.py
get_dreambooth_health.py
get_dreambooth_status.py
get_gpu_types.py
get_pod.py
get_pods.py
runpod.py
start_on_demand_pod.py
start_spot_pod.py
stop_pod.py
template.md
terminate_pod.py
train_dreambooth.py

If anyone is interested, I can make them publicly available on Github. RunPod have a Python SDK but I find it too difficult to use, and my scripts provide more functionality for managing pods than what runpodctl provides.

ashley•5/31/23, 8:26 AM

I would not recommend using the dreambooth scripts though, the RunPod Dreambooth endpoint is ridiculously expensive. It is MUCH cheaper just to train Dreambooth within your own pod.

Khang•5/31/23, 8:30 AM

Oh I see thank you very much. Hope that they will install DB and other stuff in the next version so you won't have to do all of that

KKhang Oh I see thank you very much. Hope that they will install DB and other stuff in ...

ashley•5/31/23, 8:39 AM

Yeah I am not sure why they said DreamBooth tends to break things, it hasn't broken things for me. Maybe they were using the dev branch which is almost always broken.

Fiora•5/31/23, 8:53 AM

How do I finish your latest video on colab? My colab is never GPU, why?

Casvseb•5/31/23, 9:00 AM

Hi, I was following your new Deepfake tutorial and ran in to the following issue "ModuleNotFoundError: No module named 'torch'". I have tried to look it up, but I am unable to find a fix or should I just reinstall python? Thanks in advance! And as always great vid!

CCasvseb Hi, I was following your new Deepfake tutorial and ran in to the following issue...

nikgrid•5/31/23, 9:03 AM

I'm no expert but can't you just type pip install torch?

CCasvseb Hi, I was following your new Deepfake tutorial and ran in to the following issue...

ashley•5/31/23, 9:08 AM

You failed to activate your Python venv correctly, as you can see in your screenshot, it says "active is not a recognised command", I am not familiar with Windows, but the command to activate Python venv is usually something like activateactivate and not "active". You should first activate your Python venv correctly and then install torch if the problem persists.

Original message was deleted

Enlil Wolf•5/31/23, 9:08 AM

Yeah, this should be pinned. Instead of redoing everything over & over again. This would be very helpful to anyone trying to jump right in. Massive thanks to you :thumbsup:

Wei•5/31/23, 9:15 AM

Hi, trying to do Studio quality photo, and have some issue

Wei•5/31/23, 9:15 AM

Wei•5/31/23, 9:16 AM

Error no kernel image is available for execution on the device at line 167 in file D:\ai\tool\bitsandbytes\csrc\ops.cu

Wei•5/31/23, 9:18 AM

Enlil Wolf•5/31/23, 9:21 AM

@ashleyk - So, I am trying to run it and it asked for a token, where do we find it?

EEnlil Wolf @ashleyk - So, I am trying to run it and it asked for a token, where do we find ...

ashley•5/31/23, 9:22 AM

Can you share a screenshot or something with me, I don't know what token it is asking for

Nnikgrid I'm no expert but can't you just type pip install torch?

Casvseb•5/31/23, 9:22 AM

Thanks that worked! Not sure what I did went wrong before, because i found similar command online that didn't work.

But I am already facing a new issue, maybe you know what to do next.
-----------------------------------------------------------------------------
Update: I ran "pip install opencv-python" and that worked I guess and the ran into a another no modlue named 'insightface', so I tried to install that and it did, but I got two errors:

Error: "Failed building wheel for insightface"

Error"Could not build wheels for insightface, which is required to install pyproject.toml-based projects"

ashley•5/31/23, 9:22 AM

You can DM me @EnlilWolf

Original message was deleted

Furkan Gözükara SECourses•5/31/23, 9:23 AM

nice. but xformers should be installed to speed up image generation after training :d

Aashley You failed to activate your Python venv correctly, as you can see in your screen...

Casvseb•5/31/23, 9:23 AM

Oh, I missed that..

FFiora Why doesn't it access gpu? I use colab, but it doesn't access gpu.

Furkan Gözükara SECourses•5/31/23, 9:23 AM

developers broken it. i updated colab please refresh and start again

FFurkan Gözükara SECourses nice. but xformers should be installed to speed up image generation after traini...

ashley•5/31/23, 9:23 AM

xformers is installed, just needs to be enabled in webui-user.sh

FFurkan Gözükara SECourses nice. but xformers should be installed to speed up image generation after traini...

ashley•5/31/23, 9:24 AM

Your xformers version from Huggingface gets installed in the image.

CCasvseb Hi, I was following your new Deepfake tutorial and ran in to the following issue...

Furkan Gözükara SECourses•5/31/23, 9:25 AM

ye it is because they are pushing without testing

Furkan Gözükara SECourses•5/31/23, 9:25 AM

 git checkout 251fde95968aecfebfc8f061e2cd7ebf8f2966d6 
and start
 with 
python run.py --keep-frames --keep-fps --max-cores 1

 git checkout 251fde95968aecfebfc8f061e2cd7ebf8f2966d6 
and start
 with 
python run.py --keep-frames --keep-fps --max-cores 1

WWei Click to see attachment

Furkan Gözükara SECourses•5/31/23, 9:26 AM

looks like installation error

Enlil Wolf•5/31/23, 9:27 AM

@ashleyk - I sent a DM :thumbsup:

Alamin H•5/31/23, 9:28 AM

why warning??

cancel_dreambooth_job.py create_on_demand_pod.py create_spot_pod.py create_template.py get_bid_price.py get_details.py get_dreambooth_health.py get_dreambooth_status.py get_gpu_types.py get_pod.py get_pods.py runpod.py start_on_demand_pod.py start_spot_pod.py stop_pod.py template.md terminate_pod.py train_dreambooth.py

Wanna start by saying thanks for all the guides. Learned quite a bit from finding you on YouTube. I

Similar Threads