quantization: make x86 as default backend (#88799)

Pull Request resolved: #88799
Approved by: https://github.com/kit1980

aten/src/ATen/Context.cpp

@@ -332,8 +332,8 @@ const std::vector<at::QEngine>& Context::supportedQEngines() {
 
 #ifdef USE_FBGEMM
     if (fbgemm::fbgemmSupportedCPU()) {
-      // The X86 qengine is available if and only if FBGEMM is available
       engines.push_back(at::kX86);
+      // The X86 qengine is available if and only if FBGEMM is available
       engines.push_back(at::kFBGEMM);
     }
 #endif

torch/ao/quantization/qconfig.py

@@ -218,13 +218,13 @@ def __new__(cls, activation=torch.nn.Identity, weight=torch.nn.Identity):
 Default qconfig for operators that reuse the observers from input Tensor, e.g. reshape
 """
 
-def get_default_qconfig(backend='fbgemm', version=0):
+def get_default_qconfig(backend='x86', version=0):
     """
     Returns the default PTQ qconfig for the specified backend.
 
     Args:
       * `backend` (str): a string representing the target backend. Currently supports
-        `x86`, `fbgemm` (default), `qnnpack` and `onednn`.
+        `x86` (default), `fbgemm`, `qnnpack` and `onednn`.
 
     Return:
         qconfig
@@ -301,13 +301,13 @@ def get_default_qconfig(backend='fbgemm', version=0):
 default_embedding_qat_qconfig_4bit = QConfig(activation=NoopObserver.with_args(dtype=torch.float32),
                                              weight=default_embedding_fake_quant_4bit)
 
-def get_default_qat_qconfig(backend='fbgemm', version=1):
+def get_default_qat_qconfig(backend='x86', version=1):
     """
     Returns the default QAT qconfig for the specified backend.
 
     Args:
       * `backend` (str): a string representing the target backend. Currently supports
-        `x86`, `fbgemm` (default), `qnnpack` and `onednn`.
+        `x86` (default), `fbgemm`, `qnnpack` and `onednn`.
       * `version`: version, for backwards compatibility. Can be `None` or `1`.
 
     Return:
@@ -402,13 +402,13 @@ def get_default_qat_qconfig(backend='fbgemm', version=1):
                                                        eps=2 ** -12),
     weight=fused_per_channel_wt_fake_quant_range_neg_127_to_127)
 
-def get_default_qconfig_dict(backend='fbgemm', version=0):
+def get_default_qconfig_dict(backend='x86', version=0):
     warnings.warn(
         "torch.ao.quantization.get_default_qconfig_dict is deprecated and will be removed in "
         "a future version. Please use torch.ao.quantization.get_default_qconfig_mapping instead.")
     return torch.ao.quantization.get_default_qconfig_mapping(backend, version).to_dict()
 
-def get_default_qat_qconfig_dict(backend='fbgemm', version=1):
+def get_default_qat_qconfig_dict(backend='x86', version=1):
     warnings.warn(
         "torch.ao.quantization.get_default_qat_qconfig_dict is deprecated and will be removed in "
         "a future version. Please use torch.ao.quantization.get_default_qat_qconfig_mapping instead.")

torch/ao/quantization/qconfig_mapping.py

@@ -122,25 +122,25 @@ def _get_default_qconfig_mapping(is_qat: bool, backend: str, version: int) -> QC
 
     return qconfig_mapping
 
-def get_default_qconfig_mapping(backend="fbgemm", version=0) -> QConfigMapping:
+def get_default_qconfig_mapping(backend="x86", version=0) -> QConfigMapping:
     """
     Return the default QConfigMapping for post training quantization.
 
     Args:
       * ``backend`` (str) : the quantization backend for the default qconfig mapping, should be
-         one of ["x86", "fbgemm" (default), "qnnpack", "onednn"]
+         one of ["x86" (default), "fbgemm", "qnnpack", "onednn"]
       * ``version`` (int) : the version for the default qconfig mapping
     """
     # TODO: add assert for backend choices
     return _get_default_qconfig_mapping(False, backend, version)
 
-def get_default_qat_qconfig_mapping(backend="fbgemm", version=1) -> QConfigMapping:
+def get_default_qat_qconfig_mapping(backend="x86", version=1) -> QConfigMapping:
     """
     Return the default QConfigMapping for quantization aware training.
 
     Args:
       * ``backend`` (str) : the quantization backend for the default qconfig mapping, should be
-         one of ["x86", "fbgemm" (default), "qnnpack", "onednn"]
+         one of ["x86" (default), "fbgemm", "qnnpack", "onednn"]
       * ``version`` (int) : the version for the default qconfig mapping
     """
     return _get_default_qconfig_mapping(True, backend, version)