[SHARING] DFL 2.0 ReadyToTrain model

[DeliveryBoy]

i saw message i got banned so i disabled access to my model

I got really ugly result when i enabled blur out mask.

The admins of this site would Ban you for sharing Your RTM?

i tried to login but it kept telling me i am banned ....


I am making a new 250 Liae-UDT .... this one is going to take me a while to make ....

 

[ikke21]

You know you can use any other resolution? 224 is kinda low, especially for WF, it's equivalent of like 160FF, unless your GPU is a limiting factor you really shouldn't be using 224 becasue you are not gonna get very detailed results, max out your GPU.

On my first go, I mostly just copied what Iperov did. His model was 224. I might make another model, I'll probably try a higher resolution then. What would you recommend? Also for the related settings, like other dimensions? I have a pretty decent GPU, an RTX 3090, but I don't really know what I'm doing.

 

[Groggy4]

The admins of this site would Ban you for sharing Your RTM?

Everyone got a ban message due to server errors at that time. Some took it more serious then others. It happens occasionally.

 

[DeliveryBoy]

You know you can use any other resolution? 224 is kinda low, especially for WF, it's equivalent of like 160FF, unless your GPU is a limiting factor you really shouldn't be using 224 becasue you are not gonna get very detailed results, max out your GPU.

On my first go, I mostly just copied what Iperov did. His model was 224. I might make another model, I'll probably try a higher resolution then. What would you recommend? Also for the related settings, like other dimensions? I have a pretty decent GPU, an RTX 3090, but I don't really know what I'm doing.

Try make RTM 320 Liae-UDT 



At the moment, i am making a new RTM 256 Liae-UDT model for you guys. Please give me a couple of weeks as this version i am making is quite taxing on my GPU.

Resolution 256, ae_dim 512, e_dim 128, d_dim 128, and d_mask_dim 22.

================== Model Summary ===================
==                                                ==
==            Model name: 256-liah-udt_SAEHD      ==
==                                                ==
==     Current iteration: 97513                   ==
==                                                ==
==---------------- Model Options -----------------==
==                                                ==
==            resolution: 256                     ==
==             face_type: wf                      ==
==     models_opt_on_gpu: True                    ==
==                 archi: liae-udt                ==
==               ae_dims: 512                     ==
==                e_dims: 128                     ==
==                d_dims: 128                     ==
==           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: lct                     ==
==              clipgrad: False                   ==
==              pretrain: True                    ==
==       autobackup_hour: 0                       ==
== write_preview_history: False                   ==
==           target_iter: 0                       ==
==       random_src_flip: False                   ==
==       random_dst_flip: True                    ==
==            batch_size: 4                       ==
==             gan_power: 0.0                     ==
==        gan_patch_size: 32                      ==
==              gan_dims: 16                      ==
==                                                ==
==------------------ Running On ------------------==
==                                                ==
==          Device index: 0                       ==
==                  Name: NVIDIA GeForce RTX 3090 ==
==                  VRAM: 21.17GB                 ==
==                                                ==
====================================================
Starting. Press "Enter" to stop training and save model.
[22:12:43][#100132][0463ms][0.7410][0.7752]

 

[TIDF]

You're welcome it should be finished in a couple of days when i reach 3,800,000 iterations.


For those that can't wait for the 320 RTM to reach 3,800,000 iterations here is the download link for the 320 RTM model that is very close to 3,700,000 iterations.

Download Link:

Just for information, did you ever finish the 320 RTM model, or is it best to use your latest one (3615968 iterations) ?

Thanks again for all your work!

 

[DeliveryBoy]

You're welcome it should be finished in a couple of days when i reach 3,800,000 iterations.


For those that can't wait for the 320 RTM to reach 3,800,000 iterations here is the download link for the 320 RTM model that is very close to 3,700,000 iterations.

Download Link:

Just for information, did you ever finish the 320 RTM model, or is it best to use your latest one (3615968 iterations) ?

Thanks again for all your work!

Here is the latest version 320 RTM Liae-UD

 

[AdamW]

You know you can use any other resolution? 224 is kinda low, especially for WF, it's equivalent of like 160FF, unless your GPU is a limiting factor you really shouldn't be using 224 becasue you are not gonna get very detailed results, max out your GPU.

On my first go, I mostly just copied what Iperov did. His model was 224. I might make another model, I'll probably try a higher resolution then. What would you recommend? Also for the related settings, like other dimensions? I have a pretty decent GPU, an RTX 3090, but I don't really know what I'm doing.

Try make RTM 320 Liae-UDT 



At the moment, i am making a new RTM 256 Liae-UDT model for you guys. Please give me a couple of weeks as this version i am making is quite taxing on my GPU.

Resolution 256, ae_dim 512, e_dim 128, d_dim 128, and d_mask_dim 22.

================== Model Summary ===================
==                                                ==
==            Model name: 256-liah-udt_SAEHD      ==
==                                                ==
==     Current iteration: 97513                   ==
==                                                ==
==---------------- Model Options -----------------==
==                                                ==
==            resolution: 256                     ==
==             face_type: wf                      ==
==     models_opt_on_gpu: True                    ==
==                 archi: liae-udt                ==
==               ae_dims: 512                     ==
==                e_dims: 128                     ==
==                d_dims: 128                     ==
==           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: lct                     ==
==              clipgrad: False                   ==
==              pretrain: True                    ==
==       autobackup_hour: 0                       ==
== write_preview_history: False                   ==
==           target_iter: 0                       ==
==       random_src_flip: False                   ==
==       random_dst_flip: True                    ==
==            batch_size: 4                       ==
==             gan_power: 0.0                     ==
==        gan_patch_size: 32                      ==
==              gan_dims: 16                      ==
==                                                ==
==------------------ Running On ------------------==
==                                                ==
==          Device index: 0                       ==
==                  Name: NVIDIA GeForce RTX 3090 ==
==                  VRAM: 21.17GB                 ==
==                                                ==
====================================================
Starting. Press "Enter" to stop training and save model.
[22:12:43][#100132][0463ms][0.7410][0.7752]

Cool! I was looking for a model that can works with Colab free (K80 sucks...) and I found your message. When you release it, I will try it

 

[DeliveryBoy]

You know you can use any other resolution? 224 is kinda low, especially for WF, it's equivalent of like 160FF, unless your GPU is a limiting factor you really shouldn't be using 224 becasue you are not gonna get very detailed results, max out your GPU.

On my first go, I mostly just copied what Iperov did. His model was 224. I might make another model, I'll probably try a higher resolution then. What would you recommend? Also for the related settings, like other dimensions? I have a pretty decent GPU, an RTX 3090, but I don't really know what I'm doing.

Try make RTM 320 Liae-UDT 



At the moment, i am making a new RTM 256 Liae-UDT model for you guys. Please give me a couple of weeks as this version i am making is quite taxing on my GPU.

Resolution 256, ae_dim 512, e_dim 128, d_dim 128, and d_mask_dim 22.

================== Model Summary ===================
==                                                ==
==            Model name: 256-liah-udt_SAEHD      ==
==                                                ==
==     Current iteration: 97513                   ==
==                                                ==
==---------------- Model Options -----------------==
==                                                ==
==            resolution: 256                     ==
==             face_type: wf                      ==
==     models_opt_on_gpu: True                    ==
==                 archi: liae-udt                ==
==               ae_dims: 512                     ==
==                e_dims: 128                     ==
==                d_dims: 128                     ==
==           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: lct                     ==
==              clipgrad: False                   ==
==              pretrain: True                    ==
==       autobackup_hour: 0                       ==
== write_preview_history: False                   ==
==           target_iter: 0                       ==
==       random_src_flip: False                   ==
==       random_dst_flip: True                    ==
==            batch_size: 4                       ==
==             gan_power: 0.0                     ==
==        gan_patch_size: 32                      ==
==              gan_dims: 16                      ==
==                                                ==
==------------------ Running On ------------------==
==                                                ==
==          Device index: 0                       ==
==                  Name: NVIDIA GeForce RTX 3090 ==
==                  VRAM: 21.17GB                 ==
==                                                ==
====================================================
Starting. Press "Enter" to stop training and save model.
[22:12:43][#100132][0463ms][0.7410][0.7752]

Cool! I was looking for a model that can works with Colab free (K80 sucks...) and I found your message. When you release it, I will try it

I am doing 320 RTM Liae-UDT at the moment because i couldn't find pretrained model for 256 Liae-UDT.

Here is my progress

================== Model Summary ===================
==                                                ==
==            Model name: new_SAEHD              ==
==                                                ==
==    Current iteration: 525912                  ==
==                                                ==
==---------------- Model Options -----------------==
==                                                ==
==            resolution: 320                    ==
==            face_type: wf                      ==
==    models_opt_on_gpu: True                    ==
==                archi: liae-udt                ==
==              ae_dims: 256                    ==
==                e_dims: 72                      ==
==                d_dims: 72                      ==
==          d_mask_dims: 32                      ==
==      masked_training: True                    ==
==      eyes_mouth_prio: True                    ==
==          uniform_yaw: True                    ==
==        blur_out_mask: True                    ==
==            adabelief: True                    ==
==            lr_dropout: n                      ==
==          random_warp: True                    ==
==      true_face_power: 0.0                    ==
==      face_style_power: 0.0                    ==
==        bg_style_power: 0.0                    ==
==              ct_mode: lct                    ==
==              clipgrad: False                  ==
==              pretrain: False                  ==
==      autobackup_hour: 0                      ==
== write_preview_history: False                  ==
==          target_iter: 0                      ==
==      random_src_flip: True                    ==
==      random_dst_flip: True                    ==
==            batch_size: 8                      ==
==            gan_power: 0.0                    ==
==        gan_patch_size: 40                      ==
==              gan_dims: 16                      ==
==      random_hsv_power: 0.0                    ==
==                                                ==
==------------------ Running On ------------------==
==                                                ==
==          Device index: 0                      ==
==                  Name: NVIDIA GeForce RTX 3090 ==
==                  VRAM: 21.17GB                ==
==                                                ==
====================================================
Starting. Press "Enter" to stop training and save model.
[06:30:20][#528365][0521ms][0.4881][1.0065]
[06:55:15][#531120][0528ms][0.4904][1.0079]
[07:20:15][#533816][0531ms][0.4872][1.0079]
[07:45:15][#536546][0541ms][0.4868][1.0068]
[08:10:15][#539279][0526ms][0.4997][1.0174]
[08:35:15][#542006][0508ms][0.4847][1.0043]
[09:00:16][#544775][0878ms][0.4856][1.0031]
[09:16:17][#546507][0986ms][0.4853][1.0070]
[09:25:15][#547471][0512ms][0.4856][1.0023]
[09:50:15][#550176][0878ms][0.4861][1.0015]
[10:15:15][#552895][0896ms][0.4835][1.0026]
[10:40:15][#555613][0529ms][0.4832][0.9997]
[11:05:15][#558332][0911ms][0.4810][1.0003]
[11:30:15][#561041][0549ms][0.4825][0.9998]
[11:55:15][#563750][0531ms][0.4805][0.9965]
[12:20:15][#566457][0948ms][0.4821][1.0009]
[12:45:15][#569176][0908ms][0.4823][0.9968]
[13:10:15][#571895][0551ms][0.4809][0.9957]
[13:12:51][#572152][0667ms][0.4808][0.9898]
[13:35:15][#574543][0499ms][0.4806][0.9956]
[14:00:15][#577233][0503ms][0.4779][0.9933]
[14:25:15][#580043][0513ms][0.4833][0.9991]
[14:50:15][#582742][0596ms][0.4748][0.9880]
[15:15:15][#585486][0856ms][0.4765][0.9938]
[15:40:15][#588299][0559ms][0.4750][0.9927]
[16:05:15][#590991][0614ms][0.4756][0.9942]
[16:30:15][#593675][0566ms][0.4848][0.9989]
[16:55:15][#596354][0539ms][0.4736][0.9913]
[17:20:15][#599032][0541ms][0.4749][0.9907]
[17:29:11][#599998][0542ms][0.4097][1.0450]

 

[AdamW]

You know you can use any other resolution? 224 is kinda low, especially for WF, it's equivalent of like 160FF, unless your GPU is a limiting factor you really shouldn't be using 224 becasue you are not gonna get very detailed results, max out your GPU.

On my first go, I mostly just copied what Iperov did. His model was 224. I might make another model, I'll probably try a higher resolution then. What would you recommend? Also for the related settings, like other dimensions? I have a pretty decent GPU, an RTX 3090, but I don't really know what I'm doing.

Try make RTM 320 Liae-UDT 



At the moment, i am making a new RTM 256 Liae-UDT model for you guys. Please give me a couple of weeks as this version i am making is quite taxing on my GPU.

Resolution 256, ae_dim 512, e_dim 128, d_dim 128, and d_mask_dim 22.

================== Model Summary ===================
==                                                ==
==            Model name: 256-liah-udt_SAEHD      ==
==                                                ==
==     Current iteration: 97513                   ==
==                                                ==
==---------------- Model Options -----------------==
==                                                ==
==            resolution: 256                     ==
==             face_type: wf                      ==
==     models_opt_on_gpu: True                    ==
==                 archi: liae-udt                ==
==               ae_dims: 512                     ==
==                e_dims: 128                     ==
==                d_dims: 128                     ==
==           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: lct                     ==
==              clipgrad: False                   ==
==              pretrain: True                    ==
==       autobackup_hour: 0                       ==
== write_preview_history: False                   ==
==           target_iter: 0                       ==
==       random_src_flip: False                   ==
==       random_dst_flip: True                    ==
==            batch_size: 4                       ==
==             gan_power: 0.0                     ==
==        gan_patch_size: 32                      ==
==              gan_dims: 16                      ==
==                                                ==
==------------------ Running On ------------------==
==                                                ==
==          Device index: 0                       ==
==                  Name: NVIDIA GeForce RTX 3090 ==
==                  VRAM: 21.17GB                 ==
==                                                ==
====================================================
Starting. Press "Enter" to stop training and save model.
[22:12:43][#100132][0463ms][0.7410][0.7752]

Cool! I was looking for a model that can works with Colab free (K80 sucks...) and I found your message. When you release it, I will try it

I am doing 320 RTM Liae-UDT at the moment because i couldn't find pretrained model for 256 Liae-UDT.

Here is my progress

================== Model Summary ===================
==                                                ==
==            Model name: new_SAEHD              ==
==                                                ==
==    Current iteration: 525912                  ==
==                                                ==
==---------------- Model Options -----------------==
==                                                ==
==            resolution: 320                    ==
==            face_type: wf                      ==
==    models_opt_on_gpu: True                    ==
==                archi: liae-udt                ==
==              ae_dims: 256                    ==
==                e_dims: 72                      ==
==                d_dims: 72                      ==
==          d_mask_dims: 32                      ==
==      masked_training: True                    ==
==      eyes_mouth_prio: True                    ==
==          uniform_yaw: True                    ==
==        blur_out_mask: True                    ==
==            adabelief: True                    ==
==            lr_dropout: n                      ==
==          random_warp: True                    ==
==      true_face_power: 0.0                    ==
==      face_style_power: 0.0                    ==
==        bg_style_power: 0.0                    ==
==              ct_mode: lct                    ==
==              clipgrad: False                  ==
==              pretrain: False                  ==
==      autobackup_hour: 0                      ==
== write_preview_history: False                  ==
==          target_iter: 0                      ==
==      random_src_flip: True                    ==
==      random_dst_flip: True                    ==
==            batch_size: 8                      ==
==            gan_power: 0.0                    ==
==        gan_patch_size: 40                      ==
==              gan_dims: 16                      ==
==      random_hsv_power: 0.0                    ==
==                                                ==
==------------------ Running On ------------------==
==                                                ==
==          Device index: 0                      ==
==                  Name: NVIDIA GeForce RTX 3090 ==
==                  VRAM: 21.17GB                ==
==                                                ==
====================================================
Starting. Press "Enter" to stop training and save model.
[06:30:20][#528365][0521ms][0.4881][1.0065]
[06:55:15][#531120][0528ms][0.4904][1.0079]
[07:20:15][#533816][0531ms][0.4872][1.0079]
[07:45:15][#536546][0541ms][0.4868][1.0068]
[08:10:15][#539279][0526ms][0.4997][1.0174]
[08:35:15][#542006][0508ms][0.4847][1.0043]
[09:00:16][#544775][0878ms][0.4856][1.0031]
[09:16:17][#546507][0986ms][0.4853][1.0070]
[09:25:15][#547471][0512ms][0.4856][1.0023]
[09:50:15][#550176][0878ms][0.4861][1.0015]
[10:15:15][#552895][0896ms][0.4835][1.0026]
[10:40:15][#555613][0529ms][0.4832][0.9997]
[11:05:15][#558332][0911ms][0.4810][1.0003]
[11:30:15][#561041][0549ms][0.4825][0.9998]
[11:55:15][#563750][0531ms][0.4805][0.9965]
[12:20:15][#566457][0948ms][0.4821][1.0009]
[12:45:15][#569176][0908ms][0.4823][0.9968]
[13:10:15][#571895][0551ms][0.4809][0.9957]
[13:12:51][#572152][0667ms][0.4808][0.9898]
[13:35:15][#574543][0499ms][0.4806][0.9956]
[14:00:15][#577233][0503ms][0.4779][0.9933]
[14:25:15][#580043][0513ms][0.4833][0.9991]
[14:50:15][#582742][0596ms][0.4748][0.9880]
[15:15:15][#585486][0856ms][0.4765][0.9938]
[15:40:15][#588299][0559ms][0.4750][0.9927]
[16:05:15][#590991][0614ms][0.4756][0.9942]
[16:30:15][#593675][0566ms][0.4848][0.9989]
[16:55:15][#596354][0539ms][0.4736][0.9913]
[17:20:15][#599032][0541ms][0.4749][0.9907]
[17:29:11][#599998][0542ms][0.4097][1.0450]

Unfortunately the 320 resolution model is too heavy for the K80. I hope you will find a pretrained 256 resolution model because maybe it can work with the K80

 

[DeliveryBoy]

If someone donate to me 224 or 256 RTM Liae-udt pretrained model, i will make those RTM for you.

I have already made a 256 RTM Liae-ud if you want to use that it's on page 8. I will probably make a Liae-udt version in feb 2022 when someone is willing to share a pretrained model for the 256 Liae-udt pretrained model.

If you want to use my 256 RTM Liae-UD here is the link

 

[AdamW]

If someone donate to me 224 or 256 RTM Liae-udt pretrained model, i will make those RTM for you.

I have already made a 256 RTM Liae-ud if you want to use that it's on page 8. I will probably make a Liae-udt version in feb 2022 when someone is willing to share a pretrained model for the 256 Liae-udt pretrained model.

If you want to use my 256 RTM Liae-UD here is the link

I will try it, thanks!
I have a question. To use this RTM with my dst and src, do I have to delete the Inter_ab file or something else?

 

[TMBDF]

If someone donate to me 224 or 256 RTM Liae-udt pretrained model, i will make those RTM for you.

I have already made a 256 RTM Liae-ud if you want to use that it's on page 8. I will probably make a Liae-udt version in feb 2022 when someone is willing to share a pretrained model for the 256 Liae-udt pretrained model.

If you want to use my 256 RTM Liae-UD here is the link

I will try it, thanks!
I have a question. To use this RTM with my dst and src, do I have to delete the Inter_ab file or something else?

"[font='Open Sans', sans-serif]3) reused to train new fake from scratch. In this case you have to delete INTER_B.npy from the model folder. Also don't forget to set training settings to initial, such as RW(ON), LRD(off), GAN(0.0), etc... Don't enable/disable pretraining ![/font]

[font='Open Sans', sans-serif]4) reused to train new RTM model. In this case you have to delete INTER_AB.npy from the model folder."
[/font]
So if you just change DST part to use your own random DST or train a model the usual way you delete inter_b, if you change both SRC and DST to create a whole new RTM model for different celeb you delete inter_ab.

 

[perfalen]

How can I use these to retrain on a new destination ? Trainer won't start without images in the data_src/aligned folder, what should i put there ? just any one random image of the source subject ?

 

[TMBDF]

How can I use these to retrain on a new destination ? Trainer won't start without images in the data_src/aligned folder, what should i put there ? just any one random image of the source subject ?

A lot of random DST faces that cover all angles, expressions and lighting conditions, enable DST flip for even more variety, some CT of your choice, good SRC set and just train it really well. 30k faces should be enough as long as there is a good variety and not 29k of frontal faces and 1k of faces at an angle, after enabling random flip you get 60k of training samples, model should be able to swap almost any face as long as the shape is similar to SRC. Don't flip SRC faces.

 

[TMBDF]

Added RTM workflow to the guide, "iperov" I hope you can check it and point out any errors becasue your instructions aren't complete and there is no mention of whether LRD needs to be disabled when disabling RW and how exactly one should do the extra training on top of RTM, I assume RW: off, LRD off, then LRD and then LRD and GAN as with regular training but if LRD and GAN should be kept enabled and we only replace DST let me know so I can change that part of the guide.

 

[TIDF]

in the RTM workflow, at the end, "Extra training and reuse of LIAE RTM models:", you mention "3. To create a new RTM model for different celebrity delete inter_ab file, replace SRC with new one, add random DST set and proceed with the same workflow as when creating new RTM model."

Just for information, I'm using 's 256 RTM LIAE-UD model, and have never deleted the inter_ab file when changing the src/celeb.

When starting a new workspace / celeb, I just copy the original model to the workspace and start training it immediately, without deleting files in the model directory.

It seems to work for me.. is there an upside in deleting the inter_ab file ?

 

[1D0F4K35]

Is there an upside in deleting the inter_ab file ?

I'd like to know that as well. Tuts? Iperov? anyone who knows?

 

[honor]

Is there an upside in deleting the inter_ab file ?

I'd like to know that as well. Tuts? Iperov? anyone who knows?

not deleting inter_ab, doesn' stop you from running the model...

yet if you don't delete inter_ab, you are training over the already trained data ( all of it );

Hence the end result would be less like Src ( and more of a morph ) ( Vs. if you did delete Inter_ab ).
(Having said that give it a try, anyway, you like, if you are happy with the results, then perhaps post a link to video you made, so all of us, can check it out )

 

[NoWayMan]

Hello, im looking my first view in DFLIVE. I have some questions, i hope someone can help me:

1) i want to swap me using an italian celebs.

I have to train it (copying celeb file in src dir ofc) starting from a pretrain RTM generic model and then continue with my celeb src file? 


2) wich is the correct distance from me and camera? I don't want to very near. This can work?

<a href=""><img src="" alt="Whats-App-Image-2022-01-11-at-11-26-06" border="0"></a><br /><a target='_blank' href=''>photos high</a><br />

 

[TIDF]

Ok, I know I also commented (my mistake), but not to further anger the forum masters, please use this thread only to post / share RTM models.

If you have questions, create a new post/thread, or use an existing thread that is for questions.