search

vipermu / StyleCLIP

Using CLIP and StyleGAN to generate faces from prompts.
MIT License
129 stars 28 forks source link

Error related to prompt #4

Open vettorazi opened 3 years ago

vettorazi commented 3 years ago

Hi, I'm trying to run the clip_generate.py, but I'm facing this error message every time I try to run the command: line 223, in makedirs mkdir(name, mode) OSError: [WinError 123] The filename, directory name, or volume label syntax is incorrect: './generations\\"The image of a woman with blonde hair and purple eyes"'

how can I fix it?

emwitchcraft commented 3 years ago

Go to line 62 in clip_generate.py and change the line to output_dir = os.path.join(output_path, f'{prompt}')

vipermu commented 3 years ago

Adding the fix! Thanks :)

emwitchcraft commented 3 years ago

How many iterations are you running?

On Thu, Sep 16, 2021, 5:04 AM Dave Jackson @.***> wrote:

hello, changing the prompt using --prompt "" doesn't actually change it. Similarly, editing the default prompt in clip_generate.py also doesn't change (still runs the default of blonde hair purple eyes). Am I missing something? cheers

— You are receiving this because you commented. Reply to this email directly, view it on GitHub https://github.com/vipermu/StyleCLIP/issues/4#issuecomment-920722782, or unsubscribe https://github.com/notifications/unsubscribe-auth/AORIL2SPHJQSSOCIWTI7RNLUCGXJ5ANCNFSM47LP2N7A . Triage notifications on the go with GitHub Mobile for iOS https://apps.apple.com/app/apple-store/id1477376905?ct=notification-email&mt=8&pt=524675 or Android https://play.google.com/store/apps/details?id=com.github.android&referrer=utm_campaign%3Dnotification-email%26utm_medium%3Demail%26utm_source%3Dgithub.

omnivore2000 commented 3 years ago

@emwitchcraft about 20-60. I've realised my question should have been about the ref_img as the starting point is always the same image, but it does end up in the right place after a number of iterations.

emwitchcraft commented 3 years ago

Change the random seed and you should see an at least slightly different starting image each time. In my experience it's pretty subtle in the first image, like one might have earrings and the next one doesn't. I set the random seed from int(time.time()) so every generation ends up differently even when the prompt is the same. Seems like the image to image difference levels off around image 800-1200.

On Thu, Sep 16, 2021, 9:57 AM Dave Jackson @.***> wrote:

@emwitchcraft https://github.com/emwitchcraft about 20-60. I've realised my question should have been about the ref_img as the starting point is always the same image, but it does end up in the right place after a number of iterations.

— You are receiving this because you were mentioned. Reply to this email directly, view it on GitHub https://github.com/vipermu/StyleCLIP/issues/4#issuecomment-920926447, or unsubscribe https://github.com/notifications/unsubscribe-auth/AORIL2Q5L23RCDL3PPWSY6LUCHZT7ANCNFSM47LP2N7A . Triage notifications on the go with GitHub Mobile for iOS https://apps.apple.com/app/apple-store/id1477376905?ct=notification-email&mt=8&pt=524675 or Android https://play.google.com/store/apps/details?id=com.github.android&referrer=utm_campaign%3Dnotification-email%26utm_medium%3Demail%26utm_source%3Dgithub.

omnivore2000 commented 3 years ago

@emwitchcraft thank you for taking the time to reply, I will play around with random seed!