Skip to content

[task] fix: fix bug for args.train.accelerator.fsdp_config.mixed_precision.enable #638

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
FoolPlayer merged 4 commits into from
Apr 18, 2026
+11 −7
Merged

[task] fix: fix bug for args.train.accelerator.fsdp_config.mixed_precision.enable #638

Show file tree
Hide file tree
Changes from all commits
Commits
Show all changes
4 commits
Select commit Hold shift + click to select a range
6010234
fix bug for args.train.accelerator.fsdp_config.mixed_precision.enable
UserChen666 Apr 9, 2026
a1e0f45
fix bug for args.train.accelerator.fsdp_config.mixed_precision.enable
UserChen666 Apr 10, 2026
79d6657
fix bug for args.train.accelerator.fsdp_config.mixed_precision.enable
UserChen666 Apr 17, 2026
07cc6e4
code format
FoolPlayer Apr 18, 2026
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
8 changes: 6 additions & 2 deletions tasks/deprecated_task/train_flux.py
View file
Open in desktop
Original file line number Diff line number Diff line change
Expand Up @@ -308,7 +308,7 @@ def main():
weights_path=args.model.model_path,
enable_full_shard=args.train.accelerator.fsdp_config.full_shard,
enable_reshard_after_forward=args.train.accelerator.fsdp_config.reshard_after_forward,
enable_mixed_precision=args.train.enable_mixed_precision,
mixed_precision=args.train.accelerator.fsdp_config.mixed_precision,
enable_gradient_checkpointing=args.train.gradient_checkpointing.enable,
init_device=args.train.init_device,
enable_fsdp_offload=args.train.accelerator.fsdp_config.offload,
Expand Down Expand Up @@ -473,7 +473,11 @@ def main():
extra_input = model.prepare_extra_input(latents)
# noise and target
noisy_latents = flow_scheduler.add_noise(
latents, noise, timestep, args.train.micro_batch_size, args.train.enable_mixed_precision
latents,
noise,
timestep,
args.train.micro_batch_size,
args.train.accelerator.fsdp_config.mixed_precision.enable,
)
training_target = flow_scheduler.training_target(latents, noise, timestep)
# predict noise
Expand Down
2 changes: 1 addition & 1 deletion tasks/deprecated_task/train_qwen_vl.py
View file
Open in desktop
Original file line number Diff line number Diff line change
Expand Up @@ -227,7 +227,7 @@ def main():
weights_path=args.model.model_path,
enable_full_shard=args.train.accelerator.fsdp_config.full_shard,
enable_reshard_after_forward=args.train.accelerator.fsdp_config.reshard_after_forward,
enable_mixed_precision=args.train.enable_mixed_precision,
mixed_precision=args.train.accelerator.fsdp_config.mixed_precision,
enable_gradient_checkpointing=args.train.gradient_checkpointing.enable,
enable_fsdp_offload=args.train.accelerator.fsdp_config.offload,
basic_modules=model._no_split_modules + args.model.basic_modules,
Expand Down
4 changes: 2 additions & 2 deletions tasks/deprecated_task/train_torch.py
View file
Open in desktop
Original file line number Diff line number Diff line change
Expand Up @@ -136,7 +136,7 @@ def main():
model = build_foundation_model(
config_path=args.model.config_path,
weights_path=args.model.model_path,
torch_dtype="float32" if args.train.enable_mixed_precision else "bfloat16",
torch_dtype="float32" if args.train.accelerator.fsdp_config.mixed_precision.enable else "bfloat16",
attn_implementation=args.model.ops_implementation.attn_implementation,
moe_implementation=args.model.ops_implementation.moe_implementation,
init_device=args.train.init_device,
Expand All @@ -151,7 +151,7 @@ def main():
weights_path=args.model.model_path,
enable_full_shard=args.train.accelerator.fsdp_config.full_shard,
enable_reshard_after_forward=args.train.accelerator.fsdp_config.reshard_after_forward,
enable_mixed_precision=args.train.enable_mixed_precision,
mixed_precision=args.train.accelerator.fsdp_config.mixed_precision,
enable_gradient_checkpointing=args.train.gradient_checkpointing.enable,
enable_fsdp_offload=args.train.accelerator.fsdp_config.offload,
basic_modules=list(set(getattr(model, "_no_split_modules", None) or []) | set(args.model.basic_modules)),
Expand Down
4 changes: 2 additions & 2 deletions tasks/deprecated_task/train_wan.py
View file
Open in desktop
Original file line number Diff line number Diff line change
Expand Up @@ -197,7 +197,7 @@ def main():
weights_path=args.model.model_path,
enable_full_shard=args.train.accelerator.fsdp_config.full_shard,
enable_reshard_after_forward=args.train.accelerator.fsdp_config.reshard_after_forward,
enable_mixed_precision=args.train.enable_mixed_precision,
mixed_precision=args.train.accelerator.fsdp_config.mixed_precision,
enable_gradient_checkpointing=args.train.gradient_checkpointing.enable,
init_device=args.train.init_device,
enable_fsdp_offload=args.train.accelerator.fsdp_config.offload,
Expand Down Expand Up @@ -366,7 +366,7 @@ def main():
noise,
timestep,
args.train.micro_batch_size,
args.train.enable_mixed_precision,
args.train.accelerator.fsdp_config.mixed_precision.enable,
)
training_target = flow_scheduler.training_target(latents, noise, timestep)
# predict noise
Expand Down
Loading