[minimax_m3_vl] Gate the integration tests on total accelerator memory (#50025)

MiniMaxAI/MiniMax-M3-MXFP8 is 413 GiB of weights. On smaller runners
`device_map="auto"` offloads the MoE weights to disk, which fails since
they need re-saving. Skip unless 480 GiB of accelerator memory is visible.

Fixes #50023

Co-authored-by: Claude Opus 5.5 <noreply@anthropic.com>

Author: tarekziade

tests/models/minimax_m3_vl/test_modeling_minimax_m3_vl.py

@@ -32,6 +32,7 @@
 )
 from transformers.testing_utils import (
     require_torch,
+    require_torch_accelerator_memory,
     slow,
     torch_device,
 )
@@ -380,6 +381,7 @@ def test_three_axes_frequency_ladder(self, head_dim, axis_dim):
 
 @slow
 @require_torch
+@require_torch_accelerator_memory(memory=480)  # 413 GiB of MXFP8 weights, plus activations
 class MiniMaxM3VLIntegrationTest(unittest.TestCase):
     model_id = "MiniMaxAI/MiniMax-M3-MXFP8"