Please add fix to properly use own trained 512 model #246

New Issue

2022-04-26T10:10:53+02:00

netrunner-exe commented

2022-04-26 10:10:53 +02:00

(Migrated from github.com)

Now option 512 is set to use beta 512 only. I hope this will remove a lot of future questions about using own 512 model. If I understood correctly, my suggestion:
In test_video....py and test_wholeimage...py change

    if crop_size == 512:
        opt.which_epoch = 550000
        opt.name = '512'
        mode = 'ffhq'
    else:
        mode = 'None'

to

    if crop_size == 512:
        mode = 'ffhq'
    else:
        mode = 'None'

    if opt.name == '512':
        opt.which_epoch = 550000

And in Inference for image or video face swapping in example about 512:

--name people to --name 512

Thanks guys, you are the best! We are all looking forward to your new paper and SimSwap ++! Thanks!

Now option 512 is set to use beta 512 only. I hope this will remove a lot of future questions about using own 512 model. If I understood correctly, my suggestion: In `test_video....py` and` test_wholeimage...py` change ``` if crop_size == 512: opt.which_epoch = 550000 opt.name = '512' mode = 'ffhq' else: mode = 'None' ``` to ``` if crop_size == 512: mode = 'ffhq' else: mode = 'None' if opt.name == '512': opt.which_epoch = 550000 ``` And in [Inference for image or video face swapping](https://github.com/neuralchen/SimSwap/blob/main/docs/guidance/usage.md) in example about 512: `--name people` to `--name 512` Thanks guys, you are the best! We are all looking forward to your new paper and SimSwap ++! Thanks!

👍 1

papipulato commented

2022-04-26 11:16:57 +02:00

(Migrated from github.com)

are you have 512 model now sir ? how much epoch do you train ?

netrunner-exe commented

2022-04-26 12:10:19 +02:00

(Migrated from github.com)

are you have 512 model now sir ? how much epoch do you train ?

No, just try to train 512. I have a weak GPU, and in Colab it still takes a very long time to do this. Now I'm trying to make hq dataset before train 224

> are you have 512 model now sir ? how much epoch do you train ? No, just try to train 512. I have a weak GPU, and in Colab it still takes a very long time to do this. Now I'm trying to make hq dataset before train 224

Sign in to join this conversation.

1 Participants

Notifications

Due Date

No due date set.

Dependencies

No dependencies set.

Reference: CalvinBackup/neuralchen-SimSwap#246