Traceback (most recent call last):
File "C:\Users\UserName\Desktop\AI\ai-toolkit\run.py", line 75, in
main()
File "C:\Users\UserName\Desktop\AI\ai-toolkit\run.py", line 71, in main
raise e
File "C:\Users\UserName\Desktop\AI\ai-toolkit\run.py", line 62, in main
job = get_job(config_file, args.name)
File "C:\Users\UserName\Desktop\AI\ai-toolkit\toolkit\job.py", line 20, in get_job
return TrainJob(config)
File "C:\Users\UserName\Desktop\AI\ai-toolkit\jobs\TrainJob.py", line 40, in init
self.load_processes(process_dict)
File "C:\Users\UserName\Desktop\AI\ai-toolkit\jobs\BaseJob.py", line 65, in load_processes
self.process.append(ProcessClass(i, self, process))
File "C:\Users\UserName\Desktop\AI\ai-toolkit\jobs\process\TrainSliderProcess.py", line 32, in init
self.slider_config = SliderConfig(**self.get_conf('slider', {}))
File "C:\Users\UserName\Desktop\AI\ai-toolkit\toolkit\config_modules.py", line 154, in init
target_permutations = get_slider_target_permutations(target)
File "C:\Users\UserName\Desktop\AI\ai-toolkit\toolkit\prompt_utils.py", line 260, in get_slider_target_permutations
SliderTargetConfig(
File "C:\Users\UserName\Desktop\AI\ai-toolkit\toolkit\config_modules.py", line 121, in init
self.target_class: str = kwargs.get('target_class', '')
MemoryError
It seems that if too many tags are put into "positive" and/or "negative" for your target, the program will eat up a very high amount of system ram.
Running with these prompts in the config led to the script trying to use every single bit of my 48gb of system ram it could possibly get.
Positive: "complex 3d render ultra detailed, 150 mm, beautiful studio soft light, rim light, vibrant details, hyperrealistic"
Negative: ""illustration, 3d, sepia, painting, cartoons, sketch, worst quality, low quality, normal quality"
However, ~4-8gb would be in use with a tags few pruned:
Positive: "Ultra detailed, 150 mm, beautiful studio soft light, rim light, vibrant details, hyperrealistic"
Negative: "illustration, sepia, painting, cartoons, sketch, worst quality, low quality, normal quality"
As an addendum, if "shuffle" is set to false, this does not happen,
Relevant error message:
Error running job:
======================================== Result:
1 failure
Traceback (most recent call last): File "C:\Users\UserName\Desktop\AI\ai-toolkit\run.py", line 75, in
main()
File "C:\Users\UserName\Desktop\AI\ai-toolkit\run.py", line 71, in main
raise e
File "C:\Users\UserName\Desktop\AI\ai-toolkit\run.py", line 62, in main
job = get_job(config_file, args.name)
File "C:\Users\UserName\Desktop\AI\ai-toolkit\toolkit\job.py", line 20, in get_job
return TrainJob(config)
File "C:\Users\UserName\Desktop\AI\ai-toolkit\jobs\TrainJob.py", line 40, in init
self.load_processes(process_dict)
File "C:\Users\UserName\Desktop\AI\ai-toolkit\jobs\BaseJob.py", line 65, in load_processes
self.process.append(ProcessClass(i, self, process))
File "C:\Users\UserName\Desktop\AI\ai-toolkit\jobs\process\TrainSliderProcess.py", line 32, in init
self.slider_config = SliderConfig(**self.get_conf('slider', {}))
File "C:\Users\UserName\Desktop\AI\ai-toolkit\toolkit\config_modules.py", line 154, in init
target_permutations = get_slider_target_permutations(target)
File "C:\Users\UserName\Desktop\AI\ai-toolkit\toolkit\prompt_utils.py", line 260, in get_slider_target_permutations
SliderTargetConfig(
File "C:\Users\UserName\Desktop\AI\ai-toolkit\toolkit\config_modules.py", line 121, in init
self.target_class: str = kwargs.get('target_class', '')
MemoryError
It seems that if too many tags are put into "positive" and/or "negative" for your target, the program will eat up a very high amount of system ram.
Running with these prompts in the config led to the script trying to use every single bit of my 48gb of system ram it could possibly get.
Positive: "complex 3d render ultra detailed, 150 mm, beautiful studio soft light, rim light, vibrant details, hyperrealistic" Negative: ""illustration, 3d, sepia, painting, cartoons, sketch, worst quality, low quality, normal quality"
However, ~4-8gb would be in use with a tags few pruned: Positive: "Ultra detailed, 150 mm, beautiful studio soft light, rim light, vibrant details, hyperrealistic" Negative: "illustration, sepia, painting, cartoons, sketch, worst quality, low quality, normal quality"
As an addendum, if "shuffle" is set to false, this does not happen,