Merge 1dc45f481a into fa53f71ec0

README-ja.md

@@ -50,9 +50,6 @@ Stable Diffusion等の画像生成モデルの学習、モデルによる画像
 
 ### 更新履歴
 
-- 次のリリースに含まれる予定の主な変更点は以下の通りです。リリース前の変更点は予告なく変更される可能性があります。
-    - Intel GPUの互換性を向上しました。[PR #2307](https://github.com/kohya-ss/sd-scripts/pull/2307) WhitePr氏に感謝します。
-
 - **Version 0.10.3 (2026-04-02):**
     - Animaでfp16で学習する際の安定性をさらに改善しました。[PR #2302](https://github.com/kohya-ss/sd-scripts/pull/2302) 問題をご報告いただいた方々に深く感謝します。
 

README.md

@@ -47,9 +47,6 @@ If you find this project helpful, please consider supporting its development via
 
 ### Change History
 
-- The following are the main changes planned for the next release. Please note that these changes may be subject to change without notice before the release.
-    - Improved compatibility with Intel GPUs. Thanks to WhitePr for [PR #2307](https://github.com/kohya-ss/sd-scripts/pull/2307).
-
 - **Version 0.10.3 (2026-04-02):**
     - Stability when training with fp16 on Anima has been further improved. See [PR #2302](https://github.com/kohya-ss/sd-scripts/pull/2302) for details. We deeply appreciate those who reported the issue.
 

library/ipex/__init__.py

@@ -1,7 +1,6 @@
 import os
 import sys
 import torch
-from packaging import version
 try:
     import intel_extension_for_pytorch as ipex # pylint: disable=import-error, unused-import
     has_ipex = True
@@ -9,7 +8,7 @@ except Exception:
     has_ipex = False
 from .hijacks import ipex_hijacks
 
-torch_version = version.parse(torch.__version__)
+torch_version = float(torch.__version__[:3])
 
 # pylint: disable=protected-access, missing-function-docstring, line-too-long
 
@@ -57,6 +56,7 @@ def ipex_init(): # pylint: disable=too-many-statements
             torch.cuda.__path__ = torch.xpu.__path__
             torch.cuda.set_stream = torch.xpu.set_stream
             torch.cuda.torch = torch.xpu.torch
+            torch.cuda.Union = torch.xpu.Union
             torch.cuda.__annotations__ = torch.xpu.__annotations__
             torch.cuda.__package__ = torch.xpu.__package__
             torch.cuda.__builtins__ = torch.xpu.__builtins__
@@ -64,12 +64,14 @@ def ipex_init(): # pylint: disable=too-many-statements
             torch.cuda.StreamContext = torch.xpu.StreamContext
             torch.cuda._lazy_call = torch.xpu._lazy_call
             torch.cuda.random = torch.xpu.random
+            torch.cuda._device = torch.xpu._device
             torch.cuda.__name__ = torch.xpu.__name__
+            torch.cuda._device_t = torch.xpu._device_t
             torch.cuda.__spec__ = torch.xpu.__spec__
             torch.cuda.__file__ = torch.xpu.__file__
             # torch.cuda.is_current_stream_capturing = torch.xpu.is_current_stream_capturing
 
-            if torch_version < version.parse("2.3"):
+            if torch_version < 2.3:
                 torch.cuda._initialization_lock = torch.xpu.lazy_init._initialization_lock
                 torch.cuda._initialized = torch.xpu.lazy_init._initialized
                 torch.cuda._is_in_bad_fork = torch.xpu.lazy_init._is_in_bad_fork
@@ -112,22 +114,17 @@ def ipex_init(): # pylint: disable=too-many-statements
                 torch.cuda.threading = torch.xpu.threading
                 torch.cuda.traceback = torch.xpu.traceback
 
-            if torch_version < version.parse("2.5"):
+            if torch_version < 2.5:
                 torch.cuda.os = torch.xpu.os
                 torch.cuda.Device = torch.xpu.Device
                 torch.cuda.warnings = torch.xpu.warnings
                 torch.cuda.classproperty = torch.xpu.classproperty
                 torch.UntypedStorage.cuda = torch.UntypedStorage.xpu
 
-            if torch_version < version.parse("2.7"):
+            if torch_version < 2.7:
                 torch.cuda.Tuple = torch.xpu.Tuple
                 torch.cuda.List = torch.xpu.List
 
-            if torch_version < version.parse("2.11"):
-                torch.cuda._device_t = torch.xpu._device_t
-                torch.cuda._device = torch.xpu._device
-                torch.cuda.Union = torch.xpu.Union
-
 
             # Memory:
             if 'linux' in sys.platform and "WSL2" in os.popen("uname -a").read():
@@ -163,7 +160,7 @@ def ipex_init(): # pylint: disable=too-many-statements
             torch.cuda.initial_seed = torch.xpu.initial_seed
 
             # C
-            if torch_version < version.parse("2.3"):
+            if torch_version < 2.3:
                 torch._C._cuda_getCurrentRawStream = ipex._C._getCurrentRawStream
                 ipex._C._DeviceProperties.multi_processor_count = ipex._C._DeviceProperties.gpu_subslice_count
                 ipex._C._DeviceProperties.major = 12

train_network.py

@@ -1439,6 +1439,24 @@ class NetworkTrainer:
                         if hasattr(network, "update_norms"):
                             network.update_norms()
 
+                    if (args.debug_info_steps > 0) and (step % args.debug_info_steps == 0):
+                        params = accelerator.unwrap_model(network).get_trainable_params()
+                        grads, weights, numels = [], [], []
+                        for p in params:
+                            if p.requires_grad:
+                                p_detached = p.detach()
+                                weights.append(p_detached.norm(p=1).item())
+                                if p.grad is not None:
+                                    grads.append(p.grad.detach().norm(p=1).item())
+                                else:
+                                    grads.append(0.0)
+                                numels.append(p_detached.numel())
+                        total_grad = sum(grads) / sum(numels)
+                        total_weight = sum(weights) / sum(numels)
+                        accelerator.print(
+                            f"\n[Step {step}] avr_grad={total_grad:.4E}, avr_weights={total_weight:.4E}"
+                        )
+
                     optimizer.step()
                     lr_scheduler.step()
                     optimizer.zero_grad(set_to_none=True)
@@ -1734,6 +1752,9 @@ def setup_parser() -> argparse.ArgumentParser:
         help="[EXPERIMENTAL] enable offloading of tensors to CPU during checkpointing for U-Net or DiT, if supported"
         " / 勾配チェックポイント時にテンソルをCPUにオフロードする（U-NetまたはDiTのみ、サポートされている場合）",
     )
+    parser.add_argument(
+        "--debug_info_steps", type=int, default=0, help="Log gradient/weight norms every N steps"
+    )
     parser.add_argument(
         "--no_metadata", action="store_true", help="do not save metadata in output model / メタデータを出力先モデルに保存しない"
     )