Merge 1dc45f481a into fa53f71ec0

* fix: improve numerical stability by conditionally using float32 in block computations

* doc: update README for improvement stability for fp16 training on Anima in version 0.10.3

README-ja.md

@@ -50,6 +50,9 @@ 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
 
+- **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。
+
 - **Version 0.10.2 (2026-03-30):**
     - SD/SDXLのLECO学習に対応しました。[PR #2285](https://github.com/kohya-ss/sd-scripts/pull/2285) および [PR #2294](https://github.com/kohya-ss/sd-scripts/pull/2294) umisetokikaze氏に深く感謝します。
         - 詳細は[ドキュメント](./docs/train_leco.md)をご覧ください。

README.md

@@ -47,6 +47,9 @@ If you find this project helpful, please consider supporting its development via
 
 ### Change History
 
+- **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.
+
 - **Version 0.10.2 (2026-03-30):**
     - LECO training for SD/SDXL is now supported. Many thanks to umisetokikaze for [PR #2285](https://github.com/kohya-ss/sd-scripts/pull/2285) and [PR #2294](https://github.com/kohya-ss/sd-scripts/pull/2294).
         - Please refer to the [documentation](./docs/train_leco.md) for details.

library/anima_models.py

@@ -738,9 +738,9 @@ class FinalLayer(nn.Module):
         x_B_T_H_W_D: torch.Tensor,
         emb_B_T_D: torch.Tensor,
         adaln_lora_B_T_3D: Optional[torch.Tensor] = None,
+        use_fp32: bool = False,
     ):
         # Compute AdaLN modulation parameters (in float32 when fp16 to avoid overflow in Linear layers)
-        use_fp32 = x_B_T_H_W_D.dtype == torch.float16
         with torch.autocast(device_type=x_B_T_H_W_D.device.type, dtype=torch.float32, enabled=use_fp32):
             if self.use_adaln_lora:
                 assert adaln_lora_B_T_3D is not None
@@ -863,11 +863,11 @@ class Block(nn.Module):
         emb_B_T_D: torch.Tensor,
         crossattn_emb: torch.Tensor,
         attn_params: attention.AttentionParams,
+        use_fp32: bool = False,
         rope_emb_L_1_1_D: Optional[torch.Tensor] = None,
         adaln_lora_B_T_3D: Optional[torch.Tensor] = None,
         extra_per_block_pos_emb: Optional[torch.Tensor] = None,
     ) -> torch.Tensor:
-        use_fp32 = x_B_T_H_W_D.dtype == torch.float16
         if use_fp32:
             # Cast to float32 for better numerical stability in residual connections. Each module will cast back to float16 by enclosing autocast context.
             x_B_T_H_W_D = x_B_T_H_W_D.float()
@@ -959,6 +959,7 @@ class Block(nn.Module):
         emb_B_T_D: torch.Tensor,
         crossattn_emb: torch.Tensor,
         attn_params: attention.AttentionParams,
+        use_fp32: bool = False,
         rope_emb_L_1_1_D: Optional[torch.Tensor] = None,
         adaln_lora_B_T_3D: Optional[torch.Tensor] = None,
         extra_per_block_pos_emb: Optional[torch.Tensor] = None,
@@ -972,6 +973,7 @@ class Block(nn.Module):
                     emb_B_T_D,
                     crossattn_emb,
                     attn_params,
+                    use_fp32,
                     rope_emb_L_1_1_D,
                     adaln_lora_B_T_3D,
                     extra_per_block_pos_emb,
@@ -994,6 +996,7 @@ class Block(nn.Module):
                     emb_B_T_D,
                     crossattn_emb,
                     attn_params,
+                    use_fp32,
                     rope_emb_L_1_1_D,
                     adaln_lora_B_T_3D,
                     extra_per_block_pos_emb,
@@ -1007,6 +1010,7 @@ class Block(nn.Module):
                     emb_B_T_D,
                     crossattn_emb,
                     attn_params,
+                    use_fp32,
                     rope_emb_L_1_1_D,
                     adaln_lora_B_T_3D,
                     extra_per_block_pos_emb,
@@ -1018,6 +1022,7 @@ class Block(nn.Module):
                 emb_B_T_D,
                 crossattn_emb,
                 attn_params,
+                use_fp32,
                 rope_emb_L_1_1_D,
                 adaln_lora_B_T_3D,
                 extra_per_block_pos_emb,
@@ -1338,16 +1343,19 @@ class Anima(nn.Module):
 
         attn_params = attention.AttentionParams.create_attention_params(self.attn_mode, self.split_attn)
 
+        # Determine whether to use float32 for block computations based on input dtype (use float32 for better stability when input is float16)
+        use_fp32 = x_B_T_H_W_D.dtype == torch.float16
+
         for block_idx, block in enumerate(self.blocks):
             if self.blocks_to_swap:
                 self.offloader.wait_for_block(block_idx)
 
-            x_B_T_H_W_D = block(x_B_T_H_W_D, t_embedding_B_T_D, crossattn_emb, attn_params, **block_kwargs)
+            x_B_T_H_W_D = block(x_B_T_H_W_D, t_embedding_B_T_D, crossattn_emb, attn_params, use_fp32, **block_kwargs)
 
             if self.blocks_to_swap:
                 self.offloader.submit_move_blocks(self.blocks, block_idx)
 
-        x_B_T_H_W_O = self.final_layer(x_B_T_H_W_D, t_embedding_B_T_D, adaln_lora_B_T_3D=adaln_lora_B_T_3D)
+        x_B_T_H_W_O = self.final_layer(x_B_T_H_W_D, t_embedding_B_T_D, adaln_lora_B_T_3D=adaln_lora_B_T_3D, use_fp32=use_fp32)
         x_B_C_Tt_Hp_Wp = self.unpatchify(x_B_T_H_W_O)
         return x_B_C_Tt_Hp_Wp
 

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"
     )