[https://nvbugs/5550283][fix] update to the latest MoE API (#8169)

Signed-off-by: xxi <[email protected]>

Author: xxi-nv

tests/unittest/_torch/modules/test_fused_moe.py

@@ -771,13 +771,11 @@ def per_rank_test_fused_moe_alltoall_fp8_blockwise(job_id):
                     x,
                     router_logits,
                     all_rank_num_tokens=all_rank_num_tokens,
-                    all_rank_max_num_tokens=m,
                     use_dp_padding=False)
                 ref_output = ref_model.forward(
                     x,
                     router_logits,
                     all_rank_num_tokens=all_rank_num_tokens,
-                    all_rank_max_num_tokens=m,
                     use_dp_padding=False)
 
             # Evaluate outputs with relaxed tolerance for FP8