theluyuan/DiffSynth-Studio
1
0
Fork 0
mirror of https://github.com/modelscope/DiffSynth-Studio.git synced 2026-03-23 09:28:12 +00:00
Code Issues Packages Projects Releases Wiki Activity

del swanlab

Browse Source
This commit is contained in:
Artiprocher
2025-02-28 16:13:06 +08:00
parent 6fa8dbe077
commit a466fdca8f
2 changed files with 0 additions and 55 deletions
Download Patch File Download Diff File Expand all files Collapse all files

29
diffsynth/trainers/text_to_image.py
View File

@@ -250,18 +250,6 @@ def add_general_parsers(parser):
default=None, default=None,
help="Pretrained LoRA path. Required if the training is resumed.", help="Pretrained LoRA path. Required if the training is resumed.",
) )
parser.add_argument(
"--use_swanlab",
default=False,
action="store_true",
help="Whether to use SwanLab logger.",
)
parser.add_argument(
"--swanlab_mode",
default=None,
help="SwanLab mode (cloud or local).",
)
return parser return parser
@@ -282,22 +270,6 @@ def launch_training_task(model, args):
num_workers=args.dataloader_num_workers num_workers=args.dataloader_num_workers
) )
# set swanlab logger
if args.use_swanlab:
from swanlab.integration.pytorch_lightning import SwanLabLogger
swanlab_config = {"UPPERFRAMEWORK": "DiffSynth-Studio"}
swanlab_config.update(vars(args))
swanlab_logger = SwanLabLogger(
project="diffsynth_studio",
name="diffsynth_studio",
config=swanlab_config,
mode=args.swanlab_mode,
logdir=args.output_path,
)
logger = [swanlab_config]
else:
logger = []
# train # train
trainer = pl.Trainer( trainer = pl.Trainer(
max_epochs=args.max_epochs, max_epochs=args.max_epochs,
@@ -308,7 +280,6 @@ def launch_training_task(model, args):
default_root_dir=args.output_path, default_root_dir=args.output_path,
accumulate_grad_batches=args.accumulate_grad_batches, accumulate_grad_batches=args.accumulate_grad_batches,
callbacks=[pl.pytorch.callbacks.ModelCheckpoint(save_top_k=-1)], callbacks=[pl.pytorch.callbacks.ModelCheckpoint(save_top_k=-1)],
logger=logger,
) )
trainer.fit(model=model, train_dataloaders=train_loader) trainer.fit(model=model, train_dataloaders=train_loader)

26
examples/wanvideo/train_wan_t2v.py
View File

@@ -405,17 +405,6 @@ def parse_args():
choices=["lora", "full"], choices=["lora", "full"],
help="Model structure to train. LoRA training or full training.", help="Model structure to train. LoRA training or full training.",
) )
parser.add_argument(
"--use_swanlab",
default=False,
action="store_true",
help="Whether to use SwanLab logger.",
)
parser.add_argument(
"--swanlab_mode",
default=None,
help="SwanLab mode (cloud or local).",
)
args = parser.parse_args() args = parser.parse_args()
return args return args
@@ -473,20 +462,6 @@ def train(args):
init_lora_weights=args.init_lora_weights, init_lora_weights=args.init_lora_weights,
use_gradient_checkpointing=args.use_gradient_checkpointing use_gradient_checkpointing=args.use_gradient_checkpointing
) )
if args.use_swanlab:
from swanlab.integration.pytorch_lightning import SwanLabLogger
swanlab_config = {"UPPERFRAMEWORK": "DiffSynth-Studio"}
swanlab_config.update(vars(args))
swanlab_logger = SwanLabLogger(
project="wan",
name="wan",
config=swanlab_config,
mode=args.swanlab_mode,
logdir=args.output_path,
)
logger = [swanlab_logger]
else:
logger = []
trainer = pl.Trainer( trainer = pl.Trainer(
max_epochs=args.max_epochs, max_epochs=args.max_epochs,
accelerator="gpu", accelerator="gpu",
@@ -495,7 +470,6 @@ def train(args):
default_root_dir=args.output_path, default_root_dir=args.output_path,
accumulate_grad_batches=args.accumulate_grad_batches, accumulate_grad_batches=args.accumulate_grad_batches,
callbacks=[pl.pytorch.callbacks.ModelCheckpoint(save_top_k=-1)], callbacks=[pl.pytorch.callbacks.ModelCheckpoint(save_top_k=-1)],
logger=logger,
) )
trainer.fit(model, dataloader) trainer.fit(model, dataloader)