Skip to content
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

Jump to bottom

Brandon/cast unquantized flux to bfloat16 #6815

Merged
merged 5 commits into from
Sep 5, 2024
Merged

Brandon/cast unquantized flux to bfloat16 #6815

Changes from all commits
Commits
Show all changes
5 commits
Select commit Hold shift + click to select a range
991d264
Cast tensors in unquantized flux models to bfloat16 during loading
brandonrising Sep 4, 2024
8150a58
Update flux transformer loader to more efficiently use and release me…
brandonrising Sep 5, 2024
289ee12
Add comment explaining the cache make room call
brandonrising Sep 5, 2024
816aac8
Remove dependency of asizeof
brandonrising Sep 5, 2024
f08e942
Simplify flux model dtype conversion in model loader
brandonrising Sep 5, 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
9 changes: 8 additions & 1 deletion invokeai/backend/model_manager/load/model_loaders/flux.py
View file
Open in desktop
Original file line number Diff line number Diff line change
Expand Up @@ -32,7 +32,9 @@
)
from invokeai.backend.model_manager.load.load_default import ModelLoader
from invokeai.backend.model_manager.load.model_loader_registry import ModelLoaderRegistry
from invokeai.backend.model_manager.util.model_util import convert_bundle_to_flux_transformer_checkpoint
from invokeai.backend.model_manager.util.model_util import (
convert_bundle_to_flux_transformer_checkpoint,
)
from invokeai.backend.util.silence_warnings import SilenceWarnings

try:
Expand Down Expand Up @@ -193,6 +195,11 @@ def _load_from_singlefile(
sd = load_file(model_path)
if "model.diffusion_model.double_blocks.0.img_attn.norm.key_norm.scale" in sd:
sd = convert_bundle_to_flux_transformer_checkpoint(sd)
new_sd_size = sum([ten.nelement() * torch.bfloat16.itemsize for ten in sd.values()])
self._ram_cache.make_room(new_sd_size)
for k in sd.keys():
# We need to cast to bfloat16 due to it being the only currently supported dtype for inference
sd[k] = sd[k].to(torch.bfloat16)
model.load_state_dict(sd, assign=True)
return model

Expand Down
Loading