auto: pre-test commit

tests/fmha_v3_stage_c_example9.py

@@ -243,7 +243,7 @@ class FmhaV3StageCMulti:
             kvc.reset()
             acc_st = pipeline.make_pipeline_state(pipeline.PipelineUserType.Producer, self.num_acc_stage)
             acc_pipe.producer_acquire(acc_st)
-            for kt in range(n_kv_tiles):
+            for kt in range(self.n_kv_tiles):
                 kvh = kvc.wait_and_advance()
                 sh = s_prod.acquire_and_advance()
                 qk_mma.set(tcgen05.Field.ACCUMULATE, False)
@@ -321,7 +321,9 @@ class FmhaV3StageCMulti:
             scale_log2 = Float32(self.scale_softmax_log2)
 
             # Per-tile softmax loop with online rescale.
-            for kt in range(n_kv_tiles):
+            # Use self.n_kv_tiles (Python int) not the CuTeDSL symbolic n_kv_tiles
+            # — Python range() needs a Python int.
+            for kt in range(self.n_kv_tiles):
                 si_handle = s_cons.wait_and_advance()
                 cute.printf("SOFTMAX kt=%d row_max_before=%f row_sum_before=%f\n", kt, row_max, row_sum)