Should i buy a 3070 ti 8gb or 2x 3060 12gb. It should be the same price. I heard more memory is better and would having 2 3060 12gb give you a total of 24gb of memory to play with. Thanks
Search found 9 matches
- Wed Apr 27, 2022 1:08 pm
- Forum: Hardware
- Topic: Advice on buying 2x 3060 12gb
- Replies: 1
- Views: 1931
- Mon Apr 18, 2022 8:42 am
- Forum: Extract Support
- Topic: Batch Editing Extracted Faces
- Replies: 2
- Views: 999
Batch Editing Extracted Faces
I tried batch editing the extracted faces (adjust exposure to make face much clearer) and i think it deleted the metadata. Cannot train because of this. my question is a. is this possible? b. if it is possible, what app do i need to use so that it will not delete the metadata Thanks again for all yo...
- Fri Mar 25, 2022 9:13 pm
- Forum: Training Support
- Topic: Screendoor like image when training
- Replies: 2
- Views: 962
Re: Screendoor like image when training
Houston we got a problem. It Nan'ed eventually.
Will be doing your suggestion, good thing it autosaves every 50k for me. Rolling back to 350k. Lowering the training rate and TRAINNN!!!
Thanks
- Wed Mar 23, 2022 6:36 am
- Forum: Cloud Discussion
- Topic: [Resource] Google Colab Notebook
- Replies: 32
- Views: 64351
Re: [Resource] Google Colab Notebook
Thanks for that, you are simply awesome. Also, why not place Phaze-A on as well since you will be implementing the configuration setup. Im planning to upgrade my free colab to colab pro for faster gpu. Im kinda a laptop guy so ur colab is a great help. Just want my mug to look good on my fav music v...
- Wed Mar 23, 2022 2:20 am
- Forum: Training Support
- Topic: Screendoor like image when training
- Replies: 2
- Views: 962
Screendoor like image when training
Hi,
Can you have a look at my screenshot below. There is a screendoor like red effect on top of my image when training. I invested a lot of hours(700k iterations) on this dfaker model. Is this model dead. Thank you
- Sun Mar 20, 2022 3:26 pm
- Forum: Cloud Discussion
- Topic: [Resource] Google Colab Notebook
- Replies: 32
- Views: 64351
Re: [Resource] Google Colab Notebook
Hi, Started training my model using this free google colab. My question is how do you change the parameters of the trainer (like loss, choice of trainer like adabelief, etc) since you can only change the main ones like iterations, batch size, save every.. etc BTW its much faster than my videocard lo...
- Sat Mar 12, 2022 9:07 pm
- Forum: Training Support
- Topic: Reusing original.h5 and original_state.json on Model Dir
- Replies: 3
- Views: 995
Re: Reusing original.h5 and original_state.json on Model Dir
Hi Torzdf, Thanks for that reply/ solution. Im trying the other trainers now since i want more than 64px. Does reusing the trained model .h5 and .json work on other trainers like dfaker, dfl-h128, dfl-sae, lae, fhaze-a, realface, unbalanced and villain. I will still use Bruno Mars on A and my face o...
- Tue Mar 08, 2022 5:50 pm
- Forum: Introductions
- Topic: Better late than ....
- Replies: 1
- Views: 6067
Better late than ....
Hi, Just new to this hobby, and i found this forums very helpful. I prefer this app than others (eg. deepfacelab) because of the GUI and easy to understand tutorials. I found this on youtube: https://www.youtube.com/watch?v=1leXXHfaOFw Very easy to understand, and you can jump to the forums to do yo...
- Tue Mar 08, 2022 5:24 am
- Forum: Training Support
- Topic: Reusing original.h5 and original_state.json on Model Dir
- Replies: 3
- Views: 995
Reusing original.h5 and original_state.json on Model Dir
Hi, Just started on this new hobby and its very good (not good for my electric bill lol) but worth it My question is Reusing original.h5 and original_state.json on Model Dir, on training new models to place my face in them I have trained my face to 2m iterations already and placed it on a music vide...