[SHARING] DFL 2.0 SAEHD Models

borgie · Jan 7, 2022

Works for me without error. DeepFaceLab_DirectX12 from 18.08.2021 (date of batch files) said:
borgie said:

Liae-ud res 128 , wf. Works with GTX 1660 TI 6 gb Vram
== ==
== Model name: wf_SAEHD ==
== ==
== Current iteration: 2440920 ==
== ==
==------------------ Model Options ------------------==
== ==
== resolution: 128 ==
== face_type: wf ==
== models_opt_on_gpu: True ==
== archi: liae-ud ==
== ae_dims: 256 ==
== e_dims: 64 ==
== d_dims: 64 ==
== d_mask_dims: 22 ==
== masked_training: True ==
== eyes_mouth_prio: True ==
== uniform_yaw: True ==
== adabelief: True ==
== lr_dropout: n ==
== random_warp: False ==
== true_face_power: 0.0 ==
== face_style_power: 0.0 ==
== bg_style_power: 0.0 ==
== ct_mode: rct ==
== clipgrad: False ==
== pretrain: True ==
== autobackup_hour: 1 ==
== write_preview_history: False ==
== target_iter: 0 ==
== random_src_flip: False ==
== random_dst_flip: True ==
== batch_size: 8 ==
== gan_power: 0.0 ==
== gan_patch_size: 192 ==
== gan_dims: 32 ==
== ==
==------------------- Running On --------------------==
== ==
== Device index: 0 ==
== Name: NVIDIA GeForce GTX 1660 Ti ==
== VRAM: 5.10GB ==
== ==
=======================================================

2.0 , after May 2021 -

Click to expand...

This model creates a Python error when I run it.

My DFL works fine with my own models.

I suspect this person put some malicious code into the model...and the overflow bug he is trying to abuse, or what it is, has been fixed in the meantime.

That is my guess.

I double virus scanned the model first ofc.

--------------------------------------------------------------------------------------------------------------------------------------

The first error was a memory error + python error with 10 GB RAM free with such a little model on pretrain data...using default settings (the same as the creator used)

2d and 3d time with 15 GB free RAM it only gave me python error.

I looked at the model files and it looks correct.

My GPU is newer/more powerful than his so mine for sure can run it.

My DFL is version 2, but should work even if his was version 1 it should not matter from what I have read.

************************
>> Be Aware! <<
************************

Maybe you can fix your problem if you raise the amount of virtual workspace of your hd.
sysdm.cpl
Set the amount of your workspace to 64000 MB

""

Give it a try, you can set it back if this don't work for you...

DrAstronaut · Jan 9, 2022

[font='Open Sans', sans-serif]DF-UDT[/font][font='Open Sans', sans-serif] - res: 256, AB, dims: 300/80/64/22, 600.000 iterations, Batch size 8, Whole Face [/font][font='Open Sans', sans-serif](SFW - FFHQ - Pretrain)[/font]

Groggy4 · Jan 14, 2022

Added 1800 or so more faces in WF faceset. Also noticed the link on front page is NOT THE ONE YOU ARE LOOKING FOR

Here is g-drive for both head/wf:

Groggy4 · Jan 24, 2022

DF-UDT - res: 416, AB, dims: 320/80/80/26, 1.364.859 iterations, Whole Face (NSFW DST - Female Only - Specific Celebrity Trained (Taylor Swift) + 4 different celebs - XSeg Masked)

Groggy4 · Jan 24, 2022

DF-UDT - res: 320, AB, dims: 352/88/88/28, 1.480.100 iterations, Full Face (NSFW - Female Only - Specific Celebrity Trained (Taylor Swift) - XSeg Masked)

Valance · Jan 27, 2022

[font='Open Sans', sans-serif]LIAE-UD[/font][font='Open Sans', sans-serif] - res: 256, AB, dims: 256/80/80/22, 971874 iterations, Whole Face [/font][font='Open Sans', sans-serif](SFW - FFHQ - Pretrain + Masktrain)[/font]

[font='Open Sans', sans-serif]

Code:

================== Model Summary ==================
==                                               ==
==            Model name: SaeHD liae-ud256_SAEHD ==
==                                               ==
==     Current iteration: 971874                 ==
==                                               ==
==---------------- Model Options ----------------==
==                                               ==
==            resolution: 256                    ==
==             face_type: wf                     ==
==     models_opt_on_gpu: False                  ==
==                 archi: liae-ud                ==
==               ae_dims: 256                    ==
==                e_dims: 80                     ==
==                d_dims: 80                     ==
==           d_mask_dims: 22                     ==
==       masked_training: True                   ==
==       eyes_mouth_prio: True                   ==
==           uniform_yaw: True                   ==
==         blur_out_mask: True                   ==
==             adabelief: True                   ==
==            lr_dropout: n                      ==
==           random_warp: False                  ==
==      random_hsv_power: 0.0                    ==
==       true_face_power: 0.0                    ==
==      face_style_power: 0.0                    ==
==        bg_style_power: 0.0                    ==
==               ct_mode: none                   ==
==              clipgrad: False                  ==
==              pretrain: True                   ==
==       autobackup_hour: 0                      ==
== write_preview_history: True                   ==
==           target_iter: 0                      ==
==       random_src_flip: False                  ==
==       random_dst_flip: False                  ==
==            batch_size: 10                     ==
==             gan_power: 0.0                    ==
==        gan_patch_size: 32                     ==
==              gan_dims: 16                     ==
==                                               ==
==----------------- Running On ------------------==
==                                               ==
==          Device index: 0                      ==
==                  Name: Tesla P100-PCIE-16GB   ==
==                  VRAM: 14.61GB                ==
==                                               ==
===================================================

Groggy4 · Feb 9, 2022

DF-UDT - res: 256, AB, dims: 256/64/64/22, 447,515 iterations, Full Face (NSFW - Female Only - Dst/Dst Trained - XSeg Masked)

Roy_B · Mar 6, 2022

[font='Open Sans', sans-serif]

[/font]

Code:

================== Model Summary ===================
==                                                ==
==            Model name: LUDT-224-WF_SAEHD       ==
==                                                ==
==     Current iteration: 1000000                 ==
==                                                ==
==---------------- Model Options -----------------==
==                                                ==
==            resolution: 224                     ==
==             face_type: wf                      ==
==     models_opt_on_gpu: True                    ==
==                 archi: liae-udt                ==
==               ae_dims: 256                     ==
==                e_dims: 64                      ==
==                d_dims: 64                      ==
==           d_mask_dims: 22                      ==
==       masked_training: True                    ==
==       eyes_mouth_prio: True                    ==
==           uniform_yaw: True                    ==
==         blur_out_mask: False                   ==
==             adabelief: True                    ==
==            lr_dropout: n                       ==
==           random_warp: False                   ==
==      random_hsv_power: 0.0                     ==
==       true_face_power: 0.0                     ==
==      face_style_power: 0.0                     ==
==        bg_style_power: 0.0                     ==
==               ct_mode: none                    ==
==              clipgrad: False                   ==
==              pretrain: True                    ==
==       autobackup_hour: 0                       ==
== write_preview_history: False                   ==
==           target_iter: 1000000                 ==
==       random_src_flip: False                   ==
==       random_dst_flip: True                    ==
==            batch_size: 8                       ==
==             gan_power: 0.0                     ==
==        gan_patch_size: 28                      ==
==              gan_dims: 16                      ==
==                                                ==
==------------------ Running On ------------------==
==                                                ==
==          Device index: 0                       ==
==                  Name: NVIDIA GeForce RTX 3070 ==
==                  VRAM: 5.34GB                  ==
==                                                ==
====================================================

Roy_B · Mar 23, 2022

Method used:
+ 500k default settings
+ 150k mouth priority ON
+ 150k mouth priority OFF, eyes piority ON
+ 100k eyes priority OFF, retrain high loss samples ON
+ 100k with settings restored to default

Code:

================== Model Summary ===================
==                                                ==
==            Model name: LUDT224WFRTM_SAEHD      ==
==                                                ==
==     Current iteration: 1000000                 ==
==                                                ==
==---------------- Model Options -----------------==
==                                                ==
==            resolution: 224                     ==
==             face_type: wf                      ==
==     models_opt_on_gpu: True                    ==
==                 archi: liae-udt                ==
==               ae_dims: 256                     ==
==                e_dims: 64                      ==
==                d_dims: 64                      ==
==           d_mask_dims: 22                      ==
==       masked_training: True                    ==
==       eyes_mouth_prio: False                   ==
==           uniform_yaw: True                    ==
==         blur_out_mask: False                   ==
==             adabelief: True                    ==
==            lr_dropout: n                       ==
==           random_warp: False                   ==
==      random_hsv_power: 0.0                     ==
==       true_face_power: 0.0                     ==
==      face_style_power: 0.0                     ==
==        bg_style_power: 0.0                     ==
==               ct_mode: none                    ==
==              clipgrad: False                   ==
==              pretrain: True                    ==
==       autobackup_hour: 0                       ==
== write_preview_history: False                   ==
==           target_iter: 1000000                 ==
==       random_src_flip: False                   ==
==       random_dst_flip: True                    ==
==            batch_size: 8                       ==
==             gan_power: 0.0                     ==
==        gan_patch_size: 28                      ==
==              gan_dims: 16                      ==
==              use_fp16: False                   ==
==             eyes_prio: False                   ==
==            mouth_prio: False                   ==
==         loss_function: SSIM                    ==
==     random_downsample: False                   ==
==          random_noise: False                   ==
==           random_blur: False                   ==
==           random_jpeg: False                   ==
==         random_shadow: none                    ==
==      background_power: 0.0                     ==
==          random_color: False                   ==
==               cpu_cap: 8                       ==
==       preview_samples: 4                       ==
==    force_full_preview: False                   ==
==                    lr: 5e-05                   ==
==          session_name: fork                    ==
==     maximum_n_backups: 2                       ==
==    retraining_samples: False                   ==
==         gan_smoothing: 0.1                     ==
==             gan_noise: 0.0                     ==
==                                                ==
==------------------ Running On ------------------==
==                                                ==
==          Device index: 0                       ==
==                  Name: NVIDIA GeForce RTX 3070 ==
==                  VRAM: 5.34GB                  ==
==                                                ==
====================================================

TMBDF · Mar 26, 2022

Reminder: models should be uploaded loosely in a folder to free cloud storage, zip/rar and especially password protected ones are not allowed, neither is uploading to paid cloud storage or one that requires an account to view and download the files. Remember to include all the model files, including a summary.txt of the model, use the template and correctly describe your model, it's type, training workflow, type of SRC and DST datasets and whether XSeg masking was applied for training or not, if you want to share a dataset used with model do so in the same message ("Groggy4" is exepction because he is a cool guy

) Image of the preview is optional but welcome, if you want to share RTM model do so in the RTM model sharing thread (link in the first post) but if you want you can post it in both places, just not only here.

Groggy4 · Mar 26, 2022

I'm so damn cool that I made a new g-mail/drive if anyone need it for model sharing. Then you will almost be as cool as me.

Pass: GroggyTheCoolGuy

Roy_B · Apr 7, 2022

Method used:
+ 400k default settings
+ 200k retrain high loss samples ON
+ 400k default settings (high loss samples OFF)

TheAfkAaron · Jun 13, 2022

Groggy4 said:
DF-UDT - res: 256, AB, dims: 256/64/64/22, 447,515 iterations, Full Face (NSFW - Female Only - Dst/Dst Trained - XSeg Masked)

Hey so with this here, excuse me for im am really new to all this, so i know when you train over a pretrained it will slowly weed out the old trained images with the new ones, now does that also apply for the xseg? and if it does what would i need to do to either maintain the edits you did or to make my own

thanks!

Groggy4 · Jun 14, 2022

TheAfkAaron said:
Hey so with this here, excuse me for im am really new to all this, so i know when you train over a pretrained it will slowly weed out the old trained images with the new ones, now does that also apply for the xseg? and if it does what would i need to do to either maintain the edits you did or to make my own

thanks!

When you train an xseg model on new labeled faces, it will start to forget the old data yeah. It's always good to have an own collection

[SHARING] DFL 2.0 SAEHD Models

borgie

New member

DrAstronaut

New member

Groggy4

NotSure

Groggy4

NotSure

Groggy4

NotSure

Valance

New member

Groggy4

NotSure

Roy_B

Member

Roy_B

Member

TMBDF

Moderator | Deepfake Creator | Guide maintainer

Groggy4

NotSure

Roy_B

Member

TheAfkAaron

New member

Groggy4

NotSure