fix(moore): enable Mate flash-attn inference

Author: spike-zhu

csrc/infinicore/src/ops/mha_kvcache/mha_kvcache_infiniops.cc

@@ -25,7 +25,7 @@ std::size_t implementation_index_for_device(
         return 0;
     }
     if (device_type == infini::ops::Device::Type::kMoore) {
-        return 8;
+        return 16;
     }
     return 16;
 }

csrc/infinicore/src/ops/multi_head_attention_varlen/mha_varlen_infiniops.cc

@@ -23,7 +23,7 @@ std::size_t implementation_index_for_device(
         return 0;
     }
     if (device_type == infini::ops::Device::Type::kMoore) {
-        return 8;
+        return 16;
     }
     return 16;
 }

csrc/pybind11/engine/engine.hpp

@@ -121,7 +121,9 @@ inline void bind_infer_engine(py::module &m) {
             }
             return state_dict_tp_all;
         })
-        .def("process_weights_after_loading", &InferEngine::process_weights_after_loading, "Process the weights after loading on all workers (e.g., for quantization)")
+       .def("process_weights_after_loading", &InferEngine::process_weights_after_loading,
+             "Process the weights after loading on all workers (e.g., for quantization)",
+             py::call_guard<py::gil_scoped_release>())
         .def(
             "forward", [](InferEngine &self, const InferEngine::Input &input) -> InferEngine::Output {
                 // IMPORTANT: Release the GIL before calling forward() to allow other Python threads