- Enable FC int8 (#37078)

paddle/fluid/inference/tests/api/analyzer_int8_image_classification_tester.cc

@@ -47,6 +47,10 @@ TEST(Analyzer_int8_image_classification, quantization) {
     std::shared_ptr<std::vector<PaddleTensor>> warmup_data =
         paddle::inference::GetWarmupData(input_slots_all);
 
+    // INT8 implies FC oneDNN passes to be used
+    q_cfg.pass_builder()->AppendPass("fc_mkldnn_pass");
+    q_cfg.pass_builder()->AppendPass("fc_act_mkldnn_fuse_pass");
+
     // configure quantizer
     q_cfg.EnableMkldnnQuantizer();
     q_cfg.mkldnn_quantizer_config()->SetWarmupData(warmup_data);