Merge pull request #198 from kohya-ss/dev

Dev

README.md

@@ -124,6 +124,11 @@ The majority of scripts is licensed under ASL 2.0 (including codes from Diffuser
 
 ## Change History
 
+- 16 Feb. 2023, 2023/2/16:
+  - Noise offset is recorded to the metadata. Thanks to space-nuko!
+  - Show the moving average loss to prevent loss jumping in ``train_network.py`` and ``train_db.py``. Thanks to shirayu!
+  - Noise offsetがメタデータに記録されるようになりました。space-nuko氏に感謝します。
+  - ``train_network.py``と``train_db.py``で学習中に表示されるlossの値が移動平均になりました。epochの先頭で表示されるlossが大きく変動する事象を解決します。shirayu氏に感謝します。
 - 14 Feb. 2023, 2023/2/14:
   - Add support with multi-gpu trainining for ``train_network.py``. Thanks to Isotr0py!
   - Add ``--verbose`` option for ``resize_lora.py``. For details, see [this PR](https://github.com/kohya-ss/sd-scripts/pull/179). Thanks to mgz-dev!

train_db.py

@@ -206,6 +206,8 @@ def train(args):
   if accelerator.is_main_process:
     accelerator.init_trackers("dreambooth")
 
+  loss_list = []
+  loss_total = 0.0
   for epoch in range(num_train_epochs):
     print(f"epoch {epoch+1}/{num_train_epochs}")
     train_dataset.set_current_epoch(epoch + 1)
@@ -216,7 +218,6 @@ def train(args):
     if args.gradient_checkpointing or global_step < args.stop_text_encoder_training:
       text_encoder.train()
 
-    loss_total = 0
     for step, batch in enumerate(train_dataloader):
       # 指定したステップ数でText Encoderの学習を止める
       if global_step == args.stop_text_encoder_training:
@@ -294,8 +295,13 @@ def train(args):
         logs = {"loss": current_loss, "lr": lr_scheduler.get_last_lr()[0]}
         accelerator.log(logs, step=global_step)
 
+      if epoch == 0:
+        loss_list.append(current_loss)
+      else:
+        loss_total -= loss_list[step]
+        loss_list[step] = current_loss
       loss_total += current_loss
-      avr_loss = loss_total / (step+1)
+      avr_loss = loss_total / len(loss_list)
       logs = {"loss": avr_loss}  # , "lr": lr_scheduler.get_last_lr()[0]}
       progress_bar.set_postfix(**logs)
 
@@ -303,7 +309,7 @@ def train(args):
         break
 
     if args.logging_dir is not None:
-      logs = {"epoch_loss": loss_total / len(train_dataloader)}
+      logs = {"loss/epoch": loss_total / len(loss_list)}
       accelerator.log(logs, step=epoch+1)
 
     accelerator.wait_for_everyone()

train_network.py

@@ -353,6 +353,7 @@ def train(args):
       "ss_max_bucket_reso": train_dataset.max_bucket_reso,
       "ss_seed": args.seed,
       "ss_keep_tokens": args.keep_tokens,
+      "ss_noise_offset": args.noise_offset,
       "ss_dataset_dirs": json.dumps(train_dataset.dataset_dirs_info),
       "ss_reg_dataset_dirs": json.dumps(train_dataset.reg_dataset_dirs_info),
       "ss_tag_frequency": json.dumps(train_dataset.tag_frequency),
@@ -392,6 +393,8 @@ def train(args):
   if accelerator.is_main_process:
     accelerator.init_trackers("network_train")
 
+  loss_list = []
+  loss_total = 0.0
   for epoch in range(num_train_epochs):
     print(f"epoch {epoch+1}/{num_train_epochs}")
     train_dataset.set_current_epoch(epoch + 1)
@@ -400,7 +403,6 @@ def train(args):
 
     network.on_epoch_start(text_encoder, unet)
 
-    loss_total = 0
     for step, batch in enumerate(train_dataloader):
       with accelerator.accumulate(network):
         with torch.no_grad():
@@ -464,8 +466,13 @@ def train(args):
         global_step += 1
 
       current_loss = loss.detach().item()
+      if epoch == 0:
+        loss_list.append(current_loss)
+      else:
+        loss_total -= loss_list[step]
+        loss_list[step] = current_loss
       loss_total += current_loss
-      avr_loss = loss_total / (step+1)
+      avr_loss = loss_total / len(loss_list)
       logs = {"loss": avr_loss}  # , "lr": lr_scheduler.get_last_lr()[0]}
       progress_bar.set_postfix(**logs)
 
@@ -477,7 +484,7 @@ def train(args):
         break
 
     if args.logging_dir is not None:
-      logs = {"loss/epoch": loss_total / len(train_dataloader)}
+      logs = {"loss/epoch": loss_total / len(loss_list)}
       accelerator.log(logs, step=epoch+1)
 
     accelerator.wait_for_everyone()