[Advanced dreambooth lora] adjustments to align with canonical script #8406
Conversation
|
The docs for this PR live here. All of your documentation changes will be reflected on that endpoint. The docs are available until 30 days after the last update. |
examples/advanced_diffusion_training/train_dreambooth_lora_sdxl_advanced.py
Outdated
Show resolved
Hide resolved
| @@ -571,7 +636,7 @@ def parse_args(input_args=None): | |||
| parser.add_argument( | |||
| "--optimizer", | |||
| type=str, | |||
| default="adamW", | |||
| default="AdamW", | |||
| help=('The optimizer type to use. Choose between ["AdamW", "prodigy"]'), | |||
There was a problem hiding this comment.
you can set this like:
available_optimizers = ['AdamW', 'prodigy']
parser.add_argument(
...
choices=available_optimizers,
help=f'The optimizer type to use. Choose between {available_optimizers}'
)There was a problem hiding this comment.
which ensures the user doesn't select a nonexistent option
There was a problem hiding this comment.
Yes, +1 to this suggestion. choices greatly simplifies things.
There was a problem hiding this comment.
atm - if a user specifies an optimizer other than Adam/prodigy, we log a warning default to Adam (see below)
so changing to choices would error, I'm not sure we should modify that.
# Optimizer creation
if not (args.optimizer.lower() == "prodigy" or args.optimizer.lower() == "adamw"):
logger.warning(
f"Unsupported choice of optimizer: {args.optimizer}.Supported optimizers include [adamW, prodigy]."
"Defaulting to adamW"
)
args.optimizer = "adamw"
examples/advanced_diffusion_training/train_dreambooth_lora_sdxl_advanced.py
Show resolved
Hide resolved
examples/advanced_diffusion_training/train_dreambooth_lora_sdxl_advanced.py
Show resolved
Hide resolved
examples/advanced_diffusion_training/train_dreambooth_lora_sdxl_advanced.py
Show resolved
Hide resolved
examples/advanced_diffusion_training/train_dreambooth_lora_sdxl_advanced.py
Outdated
Show resolved
Hide resolved
| if args.push_to_hub: | ||
| repo_id = create_repo(repo_id=model_id, exist_ok=True, token=args.hub_token).repo_id | ||
| repo_id = create_repo( | ||
| repo_id=args.hub_model_id or Path(args.output_dir).name, exist_ok=True, token=args.hub_token |
There was a problem hiding this comment.
If hub_token is not being used let's remove it from the CLI args, altogether.
There was a problem hiding this comment.
Hmm I'm not sure I follow? it is still in use
|
LGTM, just a small nit that I think because of pivotal tuning it may be worth to generate the README.md even without push_to_hub |
examples/advanced_diffusion_training/train_dreambooth_lora_sdxl_advanced.py
Outdated
Show resolved
Hide resolved
|
Is the failing test unrelated? |
|
100 percent not. Let’s merge once this CI run round is done. |
|
ready to go |
|
"real artists ship" |
|
Seems like the changes made in this one broke the fix issued here #6464 |
…#8406) * minor changes * minor changes * minor changes * minor changes * minor changes * minor changes * minor changes * fix * fix * aligning with blora script * aligning with blora script * aligning with blora script * aligning with blora script * aligning with blora script * remove prints * style * default val * license * move save_model_card to outside push_to_hub * Update train_dreambooth_lora_sdxl_advanced.py --------- Co-authored-by: Sayak Paul <spsayakpaul@gmail.com>
minor changes to align with canonical script - such as validation logging etc