若API未标明支持情况则代表该API暂无验证结果，待验证后更新表格。
API名称 |
是否支持 |
限制与说明 |
|---|---|---|
torch.ao.quantization.quantize |
||
torch.ao.quantization.quantize_dynamic |
||
torch.ao.quantization.quantize_qat |
||
torch.ao.quantization.prepare |
||
torch.ao.quantization.prepare_qat |
是 |
|
torch.ao.quantization.convert |
是 |
|
torch.ao.quantization.fuse_modules |
||
torch.ao.quantization.QuantStub |
是 |
|
torch.ao.quantization.DeQuantStub |
是 |
|
torch.ao.quantization.QuantWrapper |
是 |
|
torch.ao.quantization.add_quant_dequant |
||
torch.ao.quantization.swap_module |
||
torch.ao.quantization.propagate_qconfig_ |
||
torch.ao.quantization.default_eval_fn |
||
torch.ao.quantization.quantize_fx.prepare_fx |
||
torch.ao.quantization.quantize_fx.prepare_qat_fx |
||
torch.ao.quantization.quantize_fx.convert_fx |
||
torch.ao.quantization.quantize_fx.fuse_fx |
||
torch.ao.quantization.qconfig_mapping.QConfigMapping |
是 |
|
torch.ao.quantization.qconfig_mapping.QConfigMapping.from_dict |
是 |
|
torch.ao.quantization.qconfig_mapping.QConfigMapping.set_global |
是 |
|
torch.ao.quantization.qconfig_mapping.QConfigMapping.set_module_name |
是 |
|
torch.ao.quantization.qconfig_mapping.QConfigMapping.set_module_name_object_type_order |
是 |
|
torch.ao.quantization.qconfig_mapping.QConfigMapping.set_module_name_regex |
是 |
|
torch.ao.quantization.qconfig_mapping.QConfigMapping.set_object_type |
是 |
|
torch.ao.quantization.qconfig_mapping.QConfigMapping.to_dict |
是 |
|
torch.ao.quantization.qconfig_mapping.get_default_qconfig_mapping |
是 |
|
torch.ao.quantization.qconfig_mapping.get_default_qat_qconfig_mapping |
是 |
|
torch.ao.quantization.backend_config.BackendConfig |
是 |
|
torch.ao.quantization.backend_config.BackendConfig.configs |
是 |
|
torch.ao.quantization.backend_config.BackendConfig.from_dict |
是 |
|
torch.ao.quantization.backend_config.BackendConfig.set_backend_pattern_config |
是 |
|
torch.ao.quantization.backend_config.BackendConfig.set_backend_pattern_configs |
是 |
|
torch.ao.quantization.backend_config.BackendConfig.set_name |
是 |
|
torch.ao.quantization.backend_config.BackendConfig.to_dict |
是 |
|
torch.ao.quantization.backend_config.BackendPatternConfig |
是 |
|
torch.ao.quantization.backend_config.BackendPatternConfig.add_dtype_config |
是 |
|
torch.ao.quantization.backend_config.BackendPatternConfig.from_dict |
是 |
|
torch.ao.quantization.backend_config.BackendPatternConfig.set_dtype_configs |
是 |
|
torch.ao.quantization.backend_config.BackendPatternConfig.set_fused_module |
是 |
|
torch.ao.quantization.backend_config.BackendPatternConfig.set_fuser_method |
是 |
|
torch.ao.quantization.backend_config.BackendPatternConfig.set_observation_type |
是 |
|
torch.ao.quantization.backend_config.BackendPatternConfig.set_pattern |
是 |
|
torch.ao.quantization.backend_config.BackendPatternConfig.set_qat_module |
是 |
|
torch.ao.quantization.backend_config.BackendPatternConfig.set_reference_quantized_module |
是 |
|
torch.ao.quantization.backend_config.BackendPatternConfig.set_root_module |
是 |
|
torch.ao.quantization.backend_config.BackendPatternConfig.to_dict |
是 |
|
torch.ao.quantization.backend_config.DTypeConfig |
是 |
|
torch.ao.quantization.backend_config.DTypeConfig.from_dict |
是 |
|
torch.ao.quantization.backend_config.DTypeConfig.to_dict |
是 |
|
torch.ao.quantization.backend_config.DTypeWithConstraints |
是 |
|
torch.ao.quantization.backend_config.ObservationType |
是 |
|
torch.ao.quantization.backend_config.ObservationType.INPUT_OUTPUT_NOT_OBSERVED |
是 |
|
torch.ao.quantization.backend_config.ObservationType.OUTPUT_SHARE_OBSERVER_WITH_INPUT |
是 |
|
torch.ao.quantization.backend_config.ObservationType.OUTPUT_USE_DIFFERENT_OBSERVER_AS_INPUT |
是 |
|
torch.ao.quantization.fx.custom_config.FuseCustomConfig |
是 |
|
torch.ao.quantization.fx.custom_config.FuseCustomConfig.from_dict |
是 |
|
torch.ao.quantization.fx.custom_config.FuseCustomConfig.set_preserved_attributes |
是 |
|
torch.ao.quantization.fx.custom_config.FuseCustomConfig.to_dict |
是 |
|
torch.ao.quantization.fx.custom_config.PrepareCustomConfig |
是 |
|
torch.ao.quantization.fx.custom_config.PrepareCustomConfig.from_dict |
是 |
|
torch.ao.quantization.fx.custom_config.PrepareCustomConfig.set_float_to_observed_mapping |
是 |
|
torch.ao.quantization.fx.custom_config.PrepareCustomConfig.set_input_quantized_indexes |
是 |
|
torch.ao.quantization.fx.custom_config.PrepareCustomConfig.set_non_traceable_module_classes |
是 |
|
torch.ao.quantization.fx.custom_config.PrepareCustomConfig.set_non_traceable_module_names |
是 |
|
torch.ao.quantization.fx.custom_config.PrepareCustomConfig.set_output_quantized_indexes |
是 |
|
torch.ao.quantization.fx.custom_config.PrepareCustomConfig.set_preserved_attributes |
是 |
|
torch.ao.quantization.fx.custom_config.PrepareCustomConfig.set_standalone_module_class |
是 |
|
torch.ao.quantization.fx.custom_config.PrepareCustomConfig.set_standalone_module_name |
是 |
|
torch.ao.quantization.fx.custom_config.PrepareCustomConfig.to_dict |
是 |
|
torch.ao.quantization.fx.custom_config.ConvertCustomConfig |
是 |
|
torch.ao.quantization.fx.custom_config.ConvertCustomConfig.from_dict |
是 |
|
torch.ao.quantization.fx.custom_config.ConvertCustomConfig.set_observed_to_quantized_mapping |
是 |
|
torch.ao.quantization.fx.custom_config.ConvertCustomConfig.set_preserved_attributes |
是 |
|
torch.ao.quantization.fx.custom_config.ConvertCustomConfig.to_dict |
是 |
|
torch.ao.quantization.fx.custom_config.StandaloneModuleConfigEntry |
是 |
|
torch.ao.quantization.observer.ObserverBase |
是 |
|
torch.ao.quantization.observer.ObserverBase.with_args |
是 |
|
torch.ao.quantization.observer.ObserverBase.with_callable_args |
是 |
|
torch.ao.quantization.observer.MinMaxObserver |
是 |
|
torch.ao.quantization.observer.MinMaxObserver.calculate_qparams |
是 |
|
torch.ao.quantization.observer.MinMaxObserver.forward |
是 |
|
torch.ao.quantization.observer.MinMaxObserver.reset_min_max_vals |
是 |
|
torch.ao.quantization.observer.MovingAverageMinMaxObserver |
是 |
|
torch.ao.quantization.observer.PerChannelMinMaxObserver |
是 |
|
torch.ao.quantization.observer.PerChannelMinMaxObserver.reset_min_max_vals |
是 |
|
torch.ao.quantization.observer.MovingAveragePerChannelMinMaxObserver |
是 |
|
torch.ao.quantization.observer.HistogramObserver |
是 |
|
torch.ao.quantization.observer.PlaceholderObserver |
是 |
|
torch.ao.quantization.observer.RecordingObserver |
是 |
|
torch.ao.quantization.observer.NoopObserver |
是 |
|
torch.ao.quantization.observer.get_observer_state_dict |
是 |
|
torch.ao.quantization.observer.load_observer_state_dict |
是 |
|
torch.ao.quantization.observer.default_observer |
是 |
|
torch.ao.quantization.observer.default_placeholder_observer |
是 |
|
torch.ao.quantization.observer.default_debug_observer |
是 |
|
torch.ao.quantization.observer.default_weight_observer |
是 |
|
torch.ao.quantization.observer.default_histogram_observer |
是 |
|
torch.ao.quantization.observer.default_per_channel_weight_observer |
是 |
|
torch.ao.quantization.observer.default_dynamic_quant_observer |
是 |
|
torch.ao.quantization.observer.default_float_qparams_observer |
是 |
|
torch.ao.quantization.fake_quantize.FakeQuantizeBase |
||
torch.ao.quantization.fake_quantize.FakeQuantize |
是 |
|
torch.ao.quantization.fake_quantize.FixedQParamsFakeQuantize |
是 |
|
torch.ao.quantization.fake_quantize.FusedMovingAvgObsFakeQuantize |
是 |
|
torch.ao.quantization.fake_quantize.default_fake_quant |
||
torch.ao.quantization.fake_quantize.default_weight_fake_quant |
||
torch.ao.quantization.fake_quantize.default_per_channel_weight_fake_quant |
||
torch.ao.quantization.fake_quantize.default_histogram_fake_quant |
||
torch.ao.quantization.fake_quantize.default_fused_act_fake_quant |
||
torch.ao.quantization.fake_quantize.default_fused_wt_fake_quant |
||
torch.ao.quantization.fake_quantize.default_fused_per_channel_wt_fake_quant |
||
torch.ao.quantization.fake_quantize.disable_fake_quant |
是 |
|
torch.ao.quantization.fake_quantize.enable_fake_quant |
是 |
|
torch.ao.quantization.fake_quantize.disable_observer |
是 |
|
torch.ao.quantization.fake_quantize.enable_observer |
是 |
|
torch.ao.quantization.qconfig.QConfig |
是 |
|
torch.ao.quantization.qconfig.default_qconfig |
是 |
|
torch.ao.quantization.qconfig.default_debug_qconfig |
是 |
|
torch.ao.quantization.qconfig.default_per_channel_qconfig |
是 |
|
torch.ao.quantization.qconfig.default_dynamic_qconfig |
是 |
|
torch.ao.quantization.qconfig.float16_dynamic_qconfig |
是 |
|
torch.ao.quantization.qconfig.float16_static_qconfig |
是 |
|
torch.ao.quantization.qconfig.per_channel_dynamic_qconfig |
是 |
|
torch.ao.quantization.qconfig.float_qparams_weight_only_qconfig |
是 |
|
torch.ao.quantization.qconfig.default_qat_qconfig |
是 |
|
torch.ao.quantization.qconfig.default_weight_only_qconfig |
是 |
|
torch.ao.quantization.qconfig.default_activation_only_qconfig |
是 |
|
torch.ao.quantization.qconfig.default_qat_qconfig_v2 |
是 |
|
torch.ao.nn.intrinsic.ConvReLU1d |
||
torch.ao.nn.intrinsic.ConvReLU2d |
||
torch.ao.nn.intrinsic.ConvReLU3d |
||
torch.ao.nn.intrinsic.LinearReLU |
否 |
|
torch.ao.nn.intrinsic.ConvBn1d |
||
torch.ao.nn.intrinsic.ConvBn2d |
||
torch.ao.nn.intrinsic.ConvBn3d |
||
torch.ao.nn.intrinsic.ConvBnReLU1d |
||
torch.ao.nn.intrinsic.ConvBnReLU2d |
||
torch.ao.nn.intrinsic.ConvBnReLU3d |
||
torch.ao.nn.intrinsic.BNReLU2d |
||
torch.ao.nn.intrinsic.BNReLU3d |
||
torch.ao.nn.intrinsic.qat.LinearReLU |
否 |
|
torch.ao.nn.intrinsic.qat.ConvBn1d |
否 |
|
torch.ao.nn.intrinsic.qat.ConvBnReLU1d |
否 |
|
torch.ao.nn.intrinsic.qat.ConvBn2d |
||
torch.ao.nn.intrinsic.qat.ConvBnReLU2d |
否 |
|
torch.ao.nn.intrinsic.qat.ConvReLU2d |
||
torch.ao.nn.intrinsic.qat.ConvBn3d |
||
torch.ao.nn.intrinsic.qat.ConvBnReLU3d |
||
torch.ao.nn.intrinsic.qat.ConvReLU3d |
||
torch.ao.nn.intrinsic.qat.update_bn_stats |
是 |
|
torch.ao.nn.intrinsic.qat.freeze_bn_stats |
是 |
|
torch.ao.nn.intrinsic.quantized.BNReLU2d |
||
torch.ao.nn.intrinsic.quantized.BNReLU3d |
||
torch.ao.nn.intrinsic.quantized.ConvReLU1d |
||
torch.ao.nn.intrinsic.quantized.ConvReLU2d |
||
torch.ao.nn.intrinsic.quantized.ConvReLU3d |
||
torch.ao.nn.intrinsic.quantized.LinearReLU |
||
torch.ao.nn.intrinsic.quantized.dynamic.LinearReLU |
||
torch.ao.nn.qat.Conv2d |
||
torch.ao.nn.qat.Conv3d |
||
torch.ao.nn.qat.Linear |
否 |
|
torch.ao.nn.qat.Linear.from_float |
||
torch.ao.nn.qat.dynamic.Linear |
||
torch.ao.nn.quantized.ReLU6 |
||
torch.ao.nn.quantized.Hardswish |
||
torch.ao.nn.quantized.ELU |
||
torch.ao.nn.quantized.LeakyReLU |
||
torch.ao.nn.quantized.Sigmoid |
||
torch.ao.nn.quantized.BatchNorm2d |
||
torch.ao.nn.quantized.BatchNorm3d |
||
torch.ao.nn.quantized.Conv1d |
||
torch.ao.nn.quantized.Conv1d.from_float |
||
torch.ao.nn.quantized.Conv2d |
||
torch.ao.nn.quantized.Conv2d.from_float |
||
torch.ao.nn.quantized.Conv3d |
||
torch.ao.nn.quantized.Conv3d.from_float |
||
torch.ao.nn.quantized.ConvTranspose1d |
||
torch.ao.nn.quantized.ConvTranspose2d |
||
torch.ao.nn.quantized.ConvTranspose3d |
||
torch.ao.nn.quantized.Embedding |
||
torch.ao.nn.quantized.Embedding.from_float |
||
torch.ao.nn.quantized.EmbeddingBag |
||
torch.ao.nn.quantized.EmbeddingBag.from_float |
||
torch.ao.nn.quantized.FloatFunctional |
||
torch.ao.nn.quantized.FXFloatFunctional |
||
torch.ao.nn.quantized.QFunctional |
||
torch.ao.nn.quantized.Linear |
||
torch.ao.nn.quantized.Linear.from_float |
||
torch.ao.nn.quantized.Linear.from_reference |
||
torch.ao.nn.quantized.LayerNorm |
||
torch.ao.nn.quantized.GroupNorm |
||
torch.ao.nn.quantized.InstanceNorm1d |
||
torch.ao.nn.quantized.InstanceNorm2d |
||
torch.ao.nn.quantized.InstanceNorm3d |
||
torch.ao.nn.quantized.functional.avg_pool2d |
||
torch.ao.nn.quantized.functional.avg_pool3d |
||
torch.ao.nn.quantized.functional.adaptive_avg_pool2d |
||
torch.ao.nn.quantized.functional.adaptive_avg_pool3d |
||
torch.ao.nn.quantized.functional.conv1d |
||
torch.ao.nn.quantized.functional.conv2d |
||
torch.ao.nn.quantized.functional.conv3d |
||
torch.ao.nn.quantized.functional.interpolate |
||
torch.ao.nn.quantized.functional.linear |
||
torch.ao.nn.quantized.functional.max_pool1d |
||
torch.ao.nn.quantized.functional.max_pool2d |
||
torch.ao.nn.quantized.functional.celu |
||
torch.ao.nn.quantized.functional.leaky_relu |
||
torch.ao.nn.quantized.functional.hardtanh |
||
torch.ao.nn.quantized.functional.hardswish |
||
torch.ao.nn.quantized.functional.threshold |
||
torch.ao.nn.quantized.functional.elu |
||
torch.ao.nn.quantized.functional.hardsigmoid |
||
torch.ao.nn.quantized.functional.clamp |
||
torch.ao.nn.quantized.functional.upsample |
||
torch.ao.nn.quantized.functional.upsample_bilinear |
||
torch.ao.nn.quantized.functional.upsample_nearest |
||
torch.ao.nn.quantizable.LSTM |
否 |
|
torch.ao.nn.quantizable.MultiheadAttention |
||
torch.ao.nn.quantizable.MultiheadAttention.dequantize |
||
torch.ao.nn.quantizable.MultiheadAttention.forward |
||
torch.ao.nn.quantized.dynamic.Linear |
否 |
|
torch.ao.nn.quantized.dynamic.Linear.from_float |
||
torch.ao.nn.quantized.dynamic.Linear.from_reference |
||
torch.ao.nn.quantized.dynamic.LSTM |
是 |
支持fp32 |
torch.ao.nn.quantized.dynamic.GRU |
是 |
支持fp32 |
torch.ao.nn.quantized.dynamic.RNNCell |
是 |
支持fp32 |
torch.ao.nn.quantized.dynamic.LSTMCell |
是 |
支持fp32 |
torch.ao.nn.quantized.dynamic.GRUCell |
是 |
支持fp32 |
torch.ao.ns._numeric_suite.compare_weights |
是 |
|
torch.ao.ns._numeric_suite.get_logger_dict |
是 |
|
torch.ao.ns._numeric_suite.Logger |
是 |
|
torch.ao.ns._numeric_suite.Logger.forward |
是 |
|
torch.ao.ns._numeric_suite.ShadowLogger |
是 |
|
torch.ao.ns._numeric_suite.ShadowLogger.forward |
是 |
|
torch.ao.ns._numeric_suite.OutputLogger |
是 |
|
torch.ao.ns._numeric_suite.OutputLogger.forward |
是 |
|
torch.ao.ns._numeric_suite.Shadow |
是 |
|
torch.ao.ns._numeric_suite.Shadow.forward |
是 |
|
torch.ao.ns._numeric_suite.Shadow.add |
是 |
支持bf16，fp16，fp32，fp64，int8，uint8，int16，int32，int64，bool |
torch.ao.ns._numeric_suite.Shadow.add_scalar |
是 |
支持bf16，fp16，fp32，fp64，int8，uint8，int16，int32，int64，bool |
torch.ao.ns._numeric_suite.Shadow.mul |
是 |
支持bf16，fp16，fp32，fp64，int8，uint8，int16，int32，int64，bool |
torch.ao.ns._numeric_suite.Shadow.mul_scalar |
是 |
支持bf16，fp16，fp32，fp64，int8，uint8，int16，int32，int64，bool |
torch.ao.ns._numeric_suite.Shadow.cat |
是 |
支持bf16，fp16，fp32，fp64，int8，uint8，int16，int32，int64，bool |
torch.ao.ns._numeric_suite.Shadow.add_relu |
是 |
支持bf16，fp16，fp32，int8，uint8，int32，int64 |
torch.ao.ns._numeric_suite.prepare_model_with_stubs |
是 |
|
torch.ao.ns._numeric_suite.compare_model_stub |
是 |
|
torch.ao.ns._numeric_suite.get_matching_activations |
是 |
|
torch.ao.ns._numeric_suite.prepare_model_outputs |
是 |
|
torch.ao.ns._numeric_suite.compare_model_outputs |
是 |
|
torch.ao.ns._numeric_suite_fx.OutputLogger |
是 |
|
torch.ao.ns._numeric_suite_fx.OutputLogger.forward |
是 |
|
torch.ao.ns._numeric_suite_fx.OutputComparisonLogger |
是 |
|
torch.ao.ns._numeric_suite_fx.OutputComparisonLogger.forward |
是 |
|
torch.ao.ns._numeric_suite_fx.NSTracer |
是 |
|
torch.ao.ns._numeric_suite_fx.NSTracer.is_leaf_module |
是 |
|
torch.ao.ns._numeric_suite_fx.extract_weights |
是 |
|
torch.ao.ns._numeric_suite_fx.add_loggers |
是 |
|
torch.ao.ns._numeric_suite_fx.extract_logger_info |
是 |
|
torch.ao.ns._numeric_suite_fx.add_shadow_loggers |
是 |
|
torch.ao.ns._numeric_suite_fx.extract_shadow_logger_info |
是 |
|
torch.ao.ns._numeric_suite_fx.extend_logger_results_with_comparison |
是 |
|
torch.ao.ns._numeric_suite_fx.prepare_n_shadows_model |
是 |
|
torch.ao.ns._numeric_suite_fx.loggers_set_enabled |
是 |
|
torch.ao.ns._numeric_suite_fx.loggers_set_save_activations |
是 |
|
torch.ao.ns._numeric_suite_fx.convert_n_shadows_model |
是 |
|
torch.ao.ns._numeric_suite_fx.extract_results_n_shadows_model |
是 |
|
torch.ao.ns._numeric_suite_fx.print_comparisons_n_shadows_model |
是 |
|
torch.ao.ns.fx.utils.compute_sqnr |
否 |
|
torch.ao.ns.fx.utils.compute_normalized_l2_error |
||
torch.ao.ns.fx.utils.compute_cosine_similarity |
||
torch.ao.quantization.get_default_qat_qconfig 说明：
该API为Quantization Aware Training依赖接口，未在PyTorch对外文档中呈现，具体使用请参考LINK。 |
是 |
在aarch64系统中只能使用qnnpack backend |