[CI][bugfix] Fix MiniMax H3 FP8 quality test remote-code loading (#7617)

tests/diffusion/models/minimax_h3/test_minimax_h3_quantization_quality.py

@@ -189,6 +189,8 @@ def test_minimax_h3_quantization_quality(config: QualityTestConfig):
         # The fused BF16 baseline only fits on H100-80GB with encoder TP.
         "stage_0_tensor_parallel_size": 2,
         "vae_use_tiling": True,
+        # MiniMax H3's VAE is implemented by code shipped with the checkpoint.
+        "trust_remote_code": True,
     }
 
     omni_bl = Omni(**common_kwargs)