Search found 17 matches
- Sat Nov 12, 2022 11:11 am
- Forum: Training Discussion
- Topic: Experiments in pretraining
- Replies: 16
- Views: 5064
Re: Perfect 0.0 Loss value achieved !
I am not blasting you. Just pointing out Dunning Kruger effect. The simple matter is that I don't have to train to 100 billion to know that FaceSwap can never become an "instant swapper" for the same reason that I know that a car cannot fly if you give it high enough octane fuel. It's not...
- Sat Nov 12, 2022 9:18 am
- Forum: Training Discussion
- Topic: Experiments in pretraining
- Replies: 16
- Views: 5064
Re: Perfect 0.0 Loss value achieved !
Perfect loss is mathematically impossible. It's not just an infeasible thing, it's complete nonsense. Like claiming that some invention creates more power than it takes in -- It's simply impossible. That said, what you're doing is even worse in many ways. Effectively you're massively overtraining y...
- Sat Nov 12, 2022 8:05 am
- Forum: Training Discussion
- Topic: Experiments in pretraining
- Replies: 16
- Views: 5064
Re: Perfect 0.0 Loss value achieved !
So, I have one DNY512 no edits stock model that I've struggled with NaNs, etc to get to over 1.11 million. I thought for giggles I'd try it out as a pre-training for another model with a completely different face structure. The two models couldn't be more different. And I was shocked to literally w...
- Fri Nov 11, 2022 4:26 am
- Forum: Training Discussion
- Topic: Experiments in pretraining
- Replies: 16
- Views: 5064
Re: Perfect 0.0 Loss value achieved !
Interesting. So In a nutshell, you're proposing that by training a model against itself for approx. 2M iterations will lead to a near perfect model that can then be used to train another model to near perfection? Did I get this right? Yes you got 1 thing right from my long post i.e a near perfect m...
- Wed Nov 09, 2022 9:04 pm
- Forum: Training Discussion
- Topic: Experiments in pretraining
- Replies: 16
- Views: 5064
Experiments in pretraining
This was one of the questions on perfect 0.0 loss : - https://forum.faceswap.dev/viewtopic.php?p=5140 that I asked about a year ago when I was known as ugramund here. I think a perfect 0.0 loss is not possible while swapping faces but it is possible when you do something called as Image Recreation. ...
- Fri Oct 07, 2022 2:33 pm
- Forum: Hardware
- Topic: Faceswap Model/Training folder on NVMe SSD - Great performance but with massive writes
- Replies: 4
- Views: 7540
Re: Faceswap Model/Training folder on NVMe SSD - Great performance but with massive writes
Aren't you worried that such writing will ruin your NVME? I am using my Model's folder on a 120GB Kingston HyperXfury SSD which I had no use & it was lying around. It has a tolerance of 354 TBW which I don't think any 120GB SSD has (I think this model has been discontinued). I have not even wri...
- Fri Oct 07, 2022 2:27 am
- Forum: Hardware
- Topic: Faceswap Model/Training folder on NVMe SSD - Great performance but with massive writes
- Replies: 4
- Views: 7540
Re: Faceswap Model/Training folder on NVMe SSD - Great performance but with massive writes
Sorry for the late reply, I was kinda of busy. I have already founded the solution for this. I just increased the SAVE INTERVAL from default 250 to 10000. To do that, click on the SAVE INTERVAL & enter the desired value manually. My RTX 3080 Ti achieves 10k iterations in around 15-20 minutes on ...
- Sun Sep 11, 2022 8:11 am
- Forum: Training Discussion
- Topic: Is there a way to make Disney DNY 384 model ?
- Replies: 1
- Views: 896
Is there a way to make Disney DNY 384 model ?
Is there a way to make a Disney DNY 384 model ? I tried by increasing the output size to 384 & changing the encoder scaling but that didn't work. Also tried to change the encoder to EfficientnetV2_S & changing the encoder scaling but that also didn't work. I tested both the above settings in...
- Fri Sep 09, 2022 5:15 am
- Forum: Training Discussion
- Topic: Loading weights from a lower resolution model to a higher resolution model within the same model format
- Replies: 1
- Views: 773
Loading weights from a lower resolution model to a higher resolution model within the same model format
Is it possible to load weights from a DNY 256 model to DNY 512 model because when i try to do so, I get the following warning :- 09/09/2022 10:35:10 WARNING The weights for layer 'fs_dny_enc_convblk_6_1_conv2d' are of incompatible shapes. Skipping. 09/09/2022 10:35:10 WARNING The weights file 'E:\Ex...
- Thu Sep 08, 2022 6:52 pm
- Forum: Training Discussion
- Topic: What are the original Input sizes for Disney(DNY) models?
- Replies: 1
- Views: 888
What are the original Input sizes for Disney(DNY) models?
In the DNY models, what are the original input sizes? I mean I want to know at what input resolution it trains the model? Should I train them with input size of half the output size or equal-to the output size? Right now I am working on a project with DNY 512 model & I am using input size of 256...
- Wed Sep 07, 2022 10:45 am
- Forum: Training Discussion
- Topic: Can somebody help me in using Efficientnet in Disney(DNY) models?
- Replies: 3
- Views: 1429
Re: Can somebody help me in using Efficientnet in Disney(DNY) models?
Thanks a million for explaining this math. I am sure this post will help many people who are trying to experiment with Efficientnet encoder. I applied your EfficientnetV2_S scaling settings on DNY 256 model & it woked straight away. I always train all my models at learning rate of 3e & so it...
- Mon Sep 05, 2022 7:05 am
- Forum: Training Discussion
- Topic: Can somebody help me in using Efficientnet in Disney(DNY) models?
- Replies: 3
- Views: 1429
Can somebody help me in using Efficientnet in Disney(DNY) models?
I don't understand the deep learning fundamentals so when opening the Phaze A model, I always get blown away by the options I see. I have read that Efficientnet v1/v2 are great trainers with high accuracy, faster training & lower model size. Has somebody here used the Efficientnet v1/v2 with Dis...
- Mon Sep 05, 2022 5:19 am
- Forum: Hardware
- Topic: Faceswap Model/Training folder on NVMe SSD - Great performance but with massive writes
- Replies: 4
- Views: 7540
Faceswap Model/Training folder on NVMe SSD - Great performance but with massive writes
Just last week I completed building my deep learning rig featuring the beast RTX 3080 ti with 2 NVMe Samsung SSDs & 1 HDD. Right now I testing two of my most interested models Realface & Disney 256 plugin under Phaze A. What I have noticed is when using my Model & Training images folder ...
- Mon Aug 29, 2022 7:09 am
- Forum: General Discussion
- Topic: Software Feature request
- Replies: 1
- Views: 1369
Software Feature request
@torzdf is it possible for you to add the Disney models (dny256, dny512, dny1024) outside of Phaze-A with simple settings like we have for other models?
- Mon Aug 29, 2022 6:20 am
- Forum: Training Discussion
- Topic: Is Progressive Image resize training supported in Phaze-A?
- Replies: 1
- Views: 721
Is Progressive Image resize training supported in Phaze-A?
I have not yet understood Phaze-A but is Progressive Image resize training supported in Phaze-A? I have read about it here : - https://towardsdatascience.com/explained-a-style-based-generator-architecture-for-gans-generating-and-tuning-realistic-6cb2be0f431 Is it possible that in Faceswap we can slo...
- Thu Aug 25, 2022 12:15 pm
- Forum: Training Discussion
- Topic: Input size vs Output size - Which one causes more VRAM usage?
- Replies: 1
- Views: 708
Input size vs Output size - Which one causes more VRAM usage?
While training a model, which one is the main cause of VRAM usage -> Input size or Output size? For eg: I am willing to train a RealFace model, so which of the following case will use more VRAM :- Case A) Input size = 64px Output size = 128px Case B) Input size = 128px Output size = 64px And is this...
- Fri Aug 19, 2022 5:25 am
- Forum: Hardware
- Topic: Is the Nvidia RTX A4000 GPU supported by the Faceswap software?
- Replies: 1
- Views: 6635
Is the Nvidia RTX A4000 GPU supported by the Faceswap software?
Hi everyone, I am new to this forum. I am making a PC purely for deep learning purpose with no intentions of gaming. My budget allows me to buy a RTX A4000 which in my country is cheaper than a RTX 3070. Here is the official page of the card :- Nvidia RTX A4000 Is this GPU supported by the Faceswap ...