Skip to content

Fix setting fp16 dtype in AnimateDiff convert script. #7127

New issue

Have a question about this project? Sign up for a free GitHub account to open an issue and contact its maintainers and the community.

Sign up for GitHub

By clicking “Sign up for GitHub”, you agree to our terms of service and privacy statement. We’ll occasionally send you account related emails.

Already on GitHub? Sign in to your account

Merged
merged 2 commits into from
Feb 29, 2024
Merged

Fix setting fp16 dtype in AnimateDiff convert script. #7127

Changes from all commits
Commits
Show all changes
2 commits
Select commit Hold shift + click to select a range
8ba623d
update
DN6 Feb 28, 2024
2edf45a
update
DN6 Feb 28, 2024
File filter

Filter by extension

Filter by extension
Conversations
Failed to load comments.
Loading
Jump to
Jump to file
Failed to load files.
Loading
Diff view
Diff view
5 changes: 4 additions & 1 deletion scripts/convert_animatediff_motion_module_to_diffusers.py
View file
Open in desktop
Original file line number Diff line number Diff line change
Expand Up @@ -30,6 +30,7 @@ def get_args():
parser.add_argument("--output_path", type=str, required=True)
parser.add_argument("--use_motion_mid_block", action="store_true")
parser.add_argument("--motion_max_seq_length", type=int, default=32)
parser.add_argument("--save_fp16", action="store_true")

return parser.parse_args()

Expand All @@ -48,4 +49,6 @@ def get_args():
# skip loading position embeddings
adapter.load_state_dict(conv_state_dict, strict=False)
adapter.save_pretrained(args.output_path)
adapter.save_pretrained(args.output_path, variant="fp16", torch_dtype=torch.float16)

if args.save_fp16:
Copy link
Collaborator

@yiyixuxu yiyixuxu Mar 5, 2024

Choose a reason for hiding this comment

The reason will be displayed to describe this comment to others. Learn more.

are we missing an else statement here?

Copy link
Member

@sayakpaul sayakpaul Mar 5, 2024

Choose a reason for hiding this comment

The reason will be displayed to describe this comment to others. Learn more.

IIUC the scrtipt always serializes in the full precision. I think the script additionally serializes in FP16 when args.save_fp16 is True.

adapter.to(torch.float16).save_pretrained(args.output_path, variant="fp16")