[CPU] Enabled dynamic quantization be default

Author: dmitry-gorokhov

src/plugins/intel_cpu/src/config.h

@@ -55,7 +55,7 @@ struct Config {
     std::string dumpToDot = {};
     std::string device_id = {};
     float fcSparseWeiDecompressionRate = 1.0f;
-    uint64_t fcDynamicQuantizationGroupSize = 0;
+    uint64_t fcDynamicQuantizationGroupSize = 32;
     ov::element::Type kvCachePrecision = ov::element::f16;
 #if defined(OPENVINO_ARCH_X86_64)
     size_t rtCacheCapacity = 5000ul;

src/plugins/intel_cpu/src/transformations/transformation_pipeline.cpp

@@ -289,7 +289,7 @@ void Transformations::PreLpt(const std::vector<ov::element::Type>& defaultPrecis
                                                      ov::element::i4,
                                                      ov::element::nf4};
     CPU_REGISTER_PASS_X64(decompression_handling_manager, ov::pass::MarkDequantizationSubgraph, decompression_precisions,
-        config.fcDynamicQuantizationGroupSize != 0);
+        config.fcDynamicQuantizationGroupSize == 0);
     CPU_SET_CALLBACK_X64(decompression_handling_manager, [&](const_node_ptr &node) -> bool {
         return !is_decompression_multiply(node);
     }, ov::pass::MarkDequantizationSubgraph);