[pre-commit.ci] auto fixes from pre-commit.com hooks

for more information, see https://pre-commit.ci

studio/backend/core/training/diffusion_dit_trainer.py

@@ -325,6 +325,7 @@ def _apply_mxfp8_training(transformer, on_event) -> bool:
     try:
         from torchao.prototype.mx_formats import MXLinearConfig
         from torchao.quantization import quantize_
+
         quantize_(
             transformer,
             MXLinearConfig.from_recipe_name("mxfp8_cublas"),
@@ -332,9 +333,7 @@ def _apply_mxfp8_training(transformer, on_event) -> bool:
         )
         return True
     except Exception as exc:  # noqa: BLE001 -- mxfp8 is an optimisation, never fatal
-        _emit(
-            on_event, "warning", message = f"mxfp8 training unavailable, using bf16 compute: {exc}"
-        )
+        _emit(on_event, "warning", message = f"mxfp8 training unavailable, using bf16 compute: {exc}")
         return False
 
 

studio/backend/core/training/diffusion_lora_trainer.py

@@ -360,6 +360,7 @@ def run_diffusion_lora_training(
         # with a warning event). The U-Net is a dense bf16 base here, the combination that
         # wrapper compiles under "auto".
         from core.training.diffusion_dit_trainer import _maybe_compile_transformer
+
         compiled = _maybe_compile_transformer(
             unet, cfg, False, device, on_event, base_precision = "bf16"
         )
@@ -550,9 +551,7 @@ def run_diffusion_lora_training(
                     peak_gb = round(torch.cuda.max_memory_allocated() / 1e9, 2)
                 per_step = cfg.train_batch_size * cfg.gradient_accumulation_steps
                 if t_steady is not None and done > 1:
-                    samples_per_second = round(
-                        (done - 1) * per_step / max(now - t_steady, 1e-6), 3
-                    )
+                    samples_per_second = round((done - 1) * per_step / max(now - t_steady, 1e-6), 3)
                 else:
                     samples_per_second = round(done * per_step / max(now - t_start, 1e-6), 3)
                 _emit(

studio/backend/core/training/diffusion_train_common.py

@@ -335,9 +335,7 @@ class DiffusionLoraConfig:
             raise ValueError("compile_transformer must be one of off / on / auto")
         base_precision = str(self.base_precision or "nf4").strip().lower()
         if base_precision not in ("nf4", "bf16", "int8", "fp8", "mxfp8", "auto"):
-            raise ValueError(
-                "base_precision must be one of nf4 / bf16 / int8 / fp8 / mxfp8 / auto"
-            )
+            raise ValueError("base_precision must be one of nf4 / bf16 / int8 / fp8 / mxfp8 / auto")
         if base_precision in ("bf16", "int8", "fp8", "mxfp8"):
             if repo_is_prequantized(self.base_model):
                 raise ValueError(

studio/backend/tests/test_diffusion_dit_trainer.py

@@ -114,7 +114,6 @@ def test_family_train_infos_sdxl_supports_compile_without_precision_modes(monkey
 # ── mxfp8 base precision (DiT dense speed mode) ───────────────────────────────
 def _linear(in_features, out_features):
     import torch.nn as nn
-
     return nn.Linear(in_features, out_features)
 
 
@@ -172,7 +171,6 @@ def _patch_capability(monkeypatch, capability):
     # Drive train_precision_modes' GPU probe: pretend CUDA is present at the given tensor
     # core capability (fp8 needs sm89+, mxfp8 needs sm100+).
     import torch
-
     monkeypatch.setattr(torch.cuda, "is_available", lambda: True)
     monkeypatch.setattr(torch.cuda, "get_device_capability", lambda *a, **k: capability)