Merge pull request #630 from ddPn08/sdxl

make tracker init_kwargs configurable

fine_tune.py

@@ -6,6 +6,7 @@ import gc
 import math
 import os
 from multiprocessing import Value
+import toml
 
 from tqdm import tqdm
 import torch
@@ -275,7 +276,10 @@ def train(args):
         custom_train_functions.fix_noise_scheduler_betas_for_zero_terminal_snr(noise_scheduler)
 
     if accelerator.is_main_process:
-        accelerator.init_trackers("finetuning" if args.log_tracker_name is None else args.log_tracker_name)
+        init_kwargs = {}
+        if args.log_tracker_config is not None:
+            init_kwargs = toml.load(args.log_tracker_config)
+        accelerator.init_trackers("finetuning" if args.log_tracker_name is None else args.log_tracker_name, init_kwargs=init_kwargs)
 
     for epoch in range(num_train_epochs):
         accelerator.print(f"\nepoch {epoch+1}/{num_train_epochs}")

library/train_util.py

@@ -2716,6 +2716,12 @@ def add_training_arguments(parser: argparse.ArgumentParser, support_dreambooth:
         default=None,
         help="name of tracker to use for logging, default is script-specific default name / ログ出力に使用するtrackerの名前、省略時はスクリプトごとのデフォルト名",
     )
+    parser.add_argument(
+        "--log_tracker_config",
+        type=str,
+        default=None,
+        help="path to tracker config file to use for logging / ログ出力に使用するtrackerの設定ファイルのパス",
+    )
     parser.add_argument(
         "--wandb_api_key",
         type=str,

sdxl_train.py

@@ -5,6 +5,7 @@ import gc
 import math
 import os
 from multiprocessing import Value
+import toml
 
 from tqdm import tqdm
 import torch
@@ -355,7 +356,10 @@ def train(args):
         custom_train_functions.fix_noise_scheduler_betas_for_zero_terminal_snr(noise_scheduler)
 
     if accelerator.is_main_process:
-        accelerator.init_trackers("finetuning" if args.log_tracker_name is None else args.log_tracker_name)
+        init_kwargs = {}
+        if args.log_tracker_config is not None:
+            init_kwargs = toml.load(args.log_tracker_config)
+        accelerator.init_trackers("finetuning" if args.log_tracker_name is None else args.log_tracker_name, init_kwargs=init_kwargs)
 
     for epoch in range(num_train_epochs):
         accelerator.print(f"\nepoch {epoch+1}/{num_train_epochs}")

train_controlnet.py

@@ -7,6 +7,7 @@ import random
 import time
 from multiprocessing import Value
 from types import SimpleNamespace
+import toml
 
 from tqdm import tqdm
 import torch
@@ -324,7 +325,10 @@ def train(args):
         clip_sample=False,
     )
     if accelerator.is_main_process:
-        accelerator.init_trackers("controlnet_train" if args.log_tracker_name is None else args.log_tracker_name)
+        init_kwargs = {}
+        if args.log_tracker_config is not None:
+            init_kwargs = toml.load(args.log_tracker_config)
+        accelerator.init_trackers("controlnet_train" if args.log_tracker_name is None else args.log_tracker_name, init_kwargs=init_kwargs)
 
     loss_list = []
     loss_total = 0.0

train_db.py

@@ -7,6 +7,7 @@ import itertools
 import math
 import os
 from multiprocessing import Value
+import toml
 
 from tqdm import tqdm
 import torch
@@ -250,7 +251,10 @@ def train(args):
         custom_train_functions.fix_noise_scheduler_betas_for_zero_terminal_snr(noise_scheduler)
 
     if accelerator.is_main_process:
-        accelerator.init_trackers("dreambooth" if args.log_tracker_name is None else args.log_tracker_name)
+        init_kwargs = {}
+        if args.log_tracker_config is not None:
+            init_kwargs = toml.load(args.log_tracker_config)
+        accelerator.init_trackers("dreambooth" if args.log_tracker_name is None else args.log_tracker_name, init_kwargs=init_kwargs)
 
     loss_list = []
     loss_total = 0.0

train_network.py

@@ -8,6 +8,7 @@ import random
 import time
 import json
 from multiprocessing import Value
+import toml
 
 from tqdm import tqdm
 import torch
@@ -682,7 +683,10 @@ class NetworkTrainer:
             custom_train_functions.fix_noise_scheduler_betas_for_zero_terminal_snr(noise_scheduler)
 
         if accelerator.is_main_process:
-            accelerator.init_trackers("network_train" if args.log_tracker_name is None else args.log_tracker_name)
+            init_kwargs = {}
+            if args.log_tracker_config is not None:
+                init_kwargs = toml.load(args.log_tracker_config)
+            accelerator.init_trackers("network_train" if args.log_tracker_name is None else args.log_tracker_name, init_kwargs=init_kwargs)
 
         loss_list = []
         loss_total = 0.0

train_textual_inversion.py

@@ -3,6 +3,7 @@ import gc
 import math
 import os
 from multiprocessing import Value
+import toml
 
 from tqdm import tqdm
 import torch
@@ -492,7 +493,10 @@ class TextualInversionTrainer:
             custom_train_functions.fix_noise_scheduler_betas_for_zero_terminal_snr(noise_scheduler)
 
         if accelerator.is_main_process:
-            accelerator.init_trackers("textual_inversion" if args.log_tracker_name is None else args.log_tracker_name)
+            init_kwargs = {}
+            if args.log_tracker_config is not None:
+                init_kwargs = toml.load(args.log_tracker_config)
+            accelerator.init_trackers("textual_inversion" if args.log_tracker_name is None else args.log_tracker_name, init_kwargs=init_kwargs)
 
         # function for saving/removing
         def save_model(ckpt_name, embs_list, steps, epoch_no, force_sync_upload=False):

train_textual_inversion_XTI.py

@@ -388,7 +388,10 @@ def train(args):
         custom_train_functions.fix_noise_scheduler_betas_for_zero_terminal_snr(noise_scheduler)
 
     if accelerator.is_main_process:
-        accelerator.init_trackers("textual_inversion" if args.log_tracker_name is None else args.log_tracker_name)
+        init_kwargs = {}
+        if args.log_tracker_config is not None:
+            init_kwargs = toml.load(args.log_tracker_config)
+        accelerator.init_trackers("textual_inversion" if args.log_tracker_name is None else args.log_tracker_name, init_kwargs=init_kwargs)
 
     # function for saving/removing
     def save_model(ckpt_name, embs, steps, epoch_no, force_sync_upload=False):