Use Your Own Images Video - Step 17 - Reprise!
Related to: AI Unlimited
Date: 10/03/2023 08:24
User: Tony Dunne
Awards:
Hi Jorge,
This is a follow up to a previous Q & A under the above title, which was closed before I got the chance to advise you of my progress.
Sorry for the delay in offering feedback, but I've had so much trouble trying to navigate my way through the different steps that it's taken far longer than I expected.
Anyway, I thought this last time it was going to work, only to find that, at the end of the Training sequence on Step 19 I got a 'Something went wrong' message.
So again, I've not been able to make the progress I hoped for.
There's no indication what went wrong, so I've no idea what I have to do to correct it.
Any thoughts?
Many thanks.
Tony
Date: 02/02/2023 15:45
User: Jorge Vila
Awards:
No problem, I'm here to help!
Which process are you following? 1.5, 2.1 or generating with your own image?
Let me know so I can help you out!
Date: 02/02/2023 17:00
User: Tony Dunne
Awards:
Hi Jorge,
I'm following the Use Your Own Images instructions.
I got to Step 19 - Test the Model. It completed the training section, but didn't provide me with a link to take me any further. Instead I got a 'Something went wrong' message.
The process took about 40 minutes.
Hope you can help.
Thanks.
Tony
Date: 02/02/2023 17:12
User: Jorge Vila
Awards:
OK, let's try one thing, now your CHKPT fuile must be created, so, go to 1.5 option and when choosing the path for the Stable Diffussion, use the one you have just created.
Let me know!
Jorge
Date: 03/02/2023 16:25
User: Tony Dunne
Awards:
I'm not sure what you're advising me to do.
Is this instead of pasting in a token from Hugging Face?
Could I ask you to be a bit clearer?
Sorry, but this is all new to me.
Thanks for your patience.
Tony
Date: 03/02/2023 16:43
User: Jorge Vila
Awards:
Go to 1.5 option in the left menu, follow the steps and yes, instead of using the token from HFace just paste the route of your previously generated CKPT.
Let me know!
Have a nice weekend!
Date: 06/02/2023 17:07
User: Tony Dunne
Awards:
Hi Jorge,
I tried your last suggestion and the same thing happened again. At Step 6 I got a 'Something went wrong' message.
Whilst I can't make sense of the coding that came up in the box, there is a section that reads as follows:
OSError: We couldn't connect to 'https://huggingface.co' to load this model, couldn't find it in the cached files and it looks like /content/stable-diffusion-v1-5 is not the path to a directory containing a model_index.json file.
Checkout your internet connection or see how to run the library in offline mode at 'https://huggingface.co/docs/diffusers/installation#offline-mode'.
I promise you this is the last time I'll bother you with this. If I can't follow your advice the next time around I'll just have to admit defeat.
In any case, thanks for your help.
Tony
Date: 06/02/2023 18:12
User: Jorge Vila
Awards:
No way!
We're going to make this work for you!
Bother me as many times you need, in fact, i f you have some time, we cna connect through Zoom and I will help you out in real time.
It's just a matter of picking the right path for the generated file.
Send me an email to jorge@jorgevila.com so we can schedule!
Cheers
Jorge
Date: 10/02/2023 12:20
User: Tony Dunne
Awards:
Hi Jorge,
Step 6 finished in about 35 minutes and I got the same result as last time. A 'something went wrong' message.
The 'Test the Model' section produced this message:
'The model doesn't exist on you Gdrive, use the file explorer to get the path : '
I'll resize the images that are too big.
Should I then rerun what we did earlier?
Thanks for your patience.
Tony
Date: 10/02/2023 13:12
User: Jorge Vila
Awards:
Check first if the CKPT file has been created in your GDrive. If it's there you can go to option 1.5 in the left menu, using this already generated CKPT file and use that path.
Yep, retraining with the proper sizes would be great, for sure!
Cheers
Jorge
Date: 14/02/2023 12:10
User: Tony Dunne
Awards:
Hi Jorge,
I've tried again (several times) the process we went through the other day, with the same result. A message saying there's a conversion error.
Would it help if I deleted the existing file on my Google Drive and downloaded it again from scratch?
Failing that I don't know what to do. Probably give up on the idea of using my own photos and learn to create new AI images instead. Still an exciting prospect, just not what I was most interested in doing.
Anyway, thanks again for your patience.
Tony
Date: 14/02/2023 12:41
User: Jorge Vila
Awards:
You just need to use the CKPT file created and go through the steps of 1.5
Send me an email, we can make another Zoom call to help you out!
Date: 14/02/2023 14:31
User: Tony Dunne
Awards:
Sorry about this but I'm just about to go away for a few days.
Can I catch up with you on Friday so we can go through it then?
Thanks.
Tony
Date: 14/02/2023 15:22
User: Jorge Vila
Awards:
Sure, contact me when you are ready!
Cheers
Jorge
Date: 17/02/2023 12:49
User: Tony Dunne
Awards:
Hi Jorge,
Sorry about this, but it's just finished processing and I got the same message yet again - 'Something went wrong'.
Amongst the coding that appeared was the following:
'Could not load dynamic library 'libnvinfer.so.7'; dlerror: libnvinfer.so.7: cannot open shared object file: No such file or directory ..."
That line appeared twice and there was another similar line of coding.
If you can shed any light, that would be welcome.
Thanks again.
Tony
Date: 17/02/2023 16:19
User: Jorge Vila
Awards:
OK, I've changed the Colab version, so please go to the "Use your own images" in the training and do the process again (UPDATED FEB 2023), it's basically the same, if you still having issues, send me an email to jorge@jorgevila.com
Date: 20/02/2023 10:19
User: Tony Dunne
Awards:
Hi Jorge,
Thanks for your above response to the problems I've been having.
And I can now tell you that I've had some success. The process finished and I've been experimenting with my photos.
Not very good so far, but the important thing is it worked.
So, once again, thanks for your patience.
Tony
Date: 10/03/2023 08:24
User: Jorge Vila
Awards:
Respuesta Aprobada
Wow!!! nice!!!!!
So glad to read that!
Keep working on the prompts and you'll get better eventually!
Happy to help!
Jorge