Index

_

__init__() (dmx.compressor.advanced_recipe.DmxApproximationFunctionTuningHyperparams method) (dmx.compressor.advanced_recipe.DmxApproximationFunctionTuningRecipe method) (dmx.compressor.advanced_recipe.DmxBaseRecipe method) (dmx.compressor.advanced_recipe.DmxGPTQRecipe method) (dmx.compressor.advanced_recipe.DmxModuleGPTQHyperparams method) (dmx.compressor.advanced_recipe.DmxModuleQuantizerCalibrationHyperparams method) (dmx.compressor.advanced_recipe.DmxModuleSmoothQuantHyperparams method) (dmx.compressor.advanced_recipe.DmxQuantizerCalibrationHyperparams method) (dmx.compressor.advanced_recipe.DmxQuantizerCalibrationRecipe method) (dmx.compressor.advanced_recipe.DmxSLaNCHyperparams method) (dmx.compressor.advanced_recipe.DmxSLaNCRecipe method) (dmx.compressor.advanced_recipe.DmxSmoothQuantRecipe method) (dmx.compressor.functional.approximate.Approximate method) (dmx.compressor.functional.approximate.ApproximationFunction method) (dmx.compressor.functional.approximate.ApproximationMixin method) (dmx.compressor.functional.approximate.Approximator method) (dmx.compressor.functional.approximate.CustomFunctionApproximation method) (dmx.compressor.functional.approximate.NoApproximation method) (dmx.compressor.functional.approximate.TorchFunctionApproximation method) (dmx.compressor.functional.functions.experimental method) (dmx.compressor.fx.tracer.DmxHFTracer method) (dmx.compressor.fx.tracer.HFQuantTracer method) (dmx.compressor.fx.tracer.QuantTracer method) (dmx.compressor.fx.tracer.Scope method) (dmx.compressor.fx.tracer.ScopeContextManager method) (dmx.compressor.fx.transformer.configuration_transformer.ConfigurationTransformer method) (dmx.compressor.fx.transformer.dmx_aware_transformer.DMXAwareTransformer method) (dmx.compressor.fx.transformer.export_transformer.ExportSubstituteTransformer method) (dmx.compressor.fx.transformer.input_output_transformer.InputOutputTransformer method) (dmx.compressor.fx.transformer.nodedict_transformer.NodeDictTransformer method) (dmx.compressor.fx.transformer.qdq_transformer.QdQTransformer method) (dmx.compressor.fx.transformer.record_input_interpreter.RecordInputInterpreter method) (dmx.compressor.fx.transformer.record_input_interpreter.RecordInputInterpreterExport method) (dmx.compressor.layer_reconstruction.ApproximationFunctionTuner method) (dmx.compressor.layer_reconstruction.LayerReconstructionMixin method) (dmx.compressor.layer_reconstruction.OptimalBrainCompressor method) (dmx.compressor.modeling.hf.DmxPreTrainedModel method) (dmx.compressor.modeling.model.DmxConfig method) (dmx.compressor.modeling.model.DmxConfigRule method) (dmx.compressor.modeling.model.DmxModel method) (dmx.compressor.modeling.model.DmxModelMixin method) (dmx.compressor.modeling.model.DmxPipelineMixin method) (dmx.compressor.modeling.model.DmxSimplePipeline method) (dmx.compressor.modeling.model.Model method) (dmx.compressor.modeling.nn.core.DmxGraph method) (dmx.compressor.modeling.nn.core.DmxModule method) (dmx.compressor.modeling.nn.core.DmxModuleConfig method) (dmx.compressor.modeling.nn.core.DmxModuleType method) (dmx.compressor.modeling.nn.custom_modules.ApplyRotaryPosEmb method) (dmx.compressor.modeling.nn.custom_modules.ApplyRotaryPosEmbBase method) (dmx.compressor.modeling.nn.custom_modules.BloomGELU method) (dmx.compressor.modeling.nn.custom_modules.ClippedGELU method) (dmx.compressor.modeling.nn.custom_modules.FastGELU method) (dmx.compressor.modeling.nn.custom_modules.GemmaRMSNorm method) (dmx.compressor.modeling.nn.custom_modules.NewGELU method) (dmx.compressor.modeling.nn.custom_modules.QuickGELU method) (dmx.compressor.modeling.nn.custom_modules.RotaryEmbedding method) (dmx.compressor.modeling.nn.experimental.Conv1dScatter method) (dmx.compressor.modeling.nn.experimental.Conv1dUnfold method) (dmx.compressor.modeling.nn.experimental.Conv2dGather method) (dmx.compressor.modeling.nn.experimental.Conv2dUnfold method) (dmx.compressor.modeling.nn.torch_modules.ActActMatMul method) (dmx.compressor.modeling.nn.torch_modules.AdaptiveAvgPool2d method) (dmx.compressor.modeling.nn.torch_modules.AvgPool2d method) (dmx.compressor.modeling.nn.torch_modules.BAddBMM method) (dmx.compressor.modeling.nn.torch_modules.BatchNorm2d method) (dmx.compressor.modeling.nn.torch_modules.Conv1d method) (dmx.compressor.modeling.nn.torch_modules.Conv2d method) (dmx.compressor.modeling.nn.torch_modules.ConvTranspose2d method) (dmx.compressor.modeling.nn.torch_modules.Dropout method) (dmx.compressor.modeling.nn.torch_modules.Embedding method) (dmx.compressor.modeling.nn.torch_modules.Exp method) (dmx.compressor.modeling.nn.torch_modules.GELU method) (dmx.compressor.modeling.nn.torch_modules.GELUBase method) (dmx.compressor.modeling.nn.torch_modules.GroupNorm method) (dmx.compressor.modeling.nn.torch_modules.LayerNorm method) (dmx.compressor.modeling.nn.torch_modules.Linear method) (dmx.compressor.modeling.nn.torch_modules.MaxPool2d method) (dmx.compressor.modeling.nn.torch_modules.Mul method) (dmx.compressor.modeling.nn.torch_modules.ReLU method) (dmx.compressor.modeling.nn.torch_modules.ReLU6 method) (dmx.compressor.modeling.nn.torch_modules.ResAdd method) (dmx.compressor.modeling.nn.torch_modules.RMSNorm method) (dmx.compressor.modeling.nn.torch_modules.ScaledDotProductAttention method) (dmx.compressor.modeling.nn.torch_modules.SiLU method) (dmx.compressor.modeling.nn.torch_modules.Softmax method) (dmx.compressor.modeling.nn.torch_modules.Tanh method) (dmx.compressor.numerical.cast.CastTo method) (dmx.compressor.numerical.cast.CastToDict method) (dmx.compressor.numerical.cast.CastToFormat method) (dmx.compressor.numerical.cast.DeQuantize method) (dmx.compressor.numerical.cast.NumericalCastMixin method) (dmx.compressor.numerical.cast.Quantize method) (dmx.compressor.numerical.format.BlockFloatingPoint method) (dmx.compressor.numerical.format.FixedPoint method) (dmx.compressor.numerical.format.FloatingPoint method) (dmx.compressor.numerical.format.Format method) (dmx.compressor.numerical.format.MXFP method) (dmx.compressor.numerical.format.MXINT method) (dmx.compressor.numerical.format.Same method) (dmx.compressor.numerical.format.ScaledBlockFloatingPoint method) (dmx.compressor.numerical.observer.DMXObserverBase method) (dmx.compressor.numerical.observer.DummyObserver method) (dmx.compressor.numerical.observer.HistogramObserver method) (dmx.compressor.numerical.observer.MinMaxObserver method) (dmx.compressor.numerical.observer.PercentileObserver method) (dmx.compressor.numerical.onnx.BFPTypeEnum method) (dmx.compressor.numerical.smoothquant.ActivationWeightSmoothQuant method) (dmx.compressor.numerical.smoothquant.SmoothQuant method) (dmx.compressor.perf_proxy.PerformanceProxyMixin method) (dmx.compressor.plugins.ActivatePlugins method) (dmx.compressor.plugins.PluginBase method) (dmx.compressor.plugins.PluginLayerData method) (dmx.compressor.sparse.Bernoulli method) (dmx.compressor.sparse.BlockTopK method) (dmx.compressor.sparse.Dense method) (dmx.compressor.sparse.LazySparsify method) (dmx.compressor.sparse.Sparseness method) (dmx.compressor.sparse.SparsificationManager method) (dmx.compressor.sparse.Sparsify method) (dmx.compressor.sparse.TopK method) (dmx.compressor.sparse.WeightSparseMixin method) (dmx.compressor.utils.benchmark.EVALUATION_MODE method)

_bias (dmx.compressor.modeling.nn.torch_modules.Conv1d attribute), [1] (dmx.compressor.modeling.nn.torch_modules.Conv2d attribute), [1] (dmx.compressor.modeling.nn.torch_modules.ConvTranspose2d attribute), [1] (dmx.compressor.modeling.nn.torch_modules.Linear attribute), [1] _weight (dmx.compressor.modeling.nn.torch_modules.Conv1d attribute), [1] (dmx.compressor.modeling.nn.torch_modules.Conv2d attribute), [1] (dmx.compressor.modeling.nn.torch_modules.ConvTranspose2d attribute), [1] (dmx.compressor.modeling.nn.torch_modules.Embedding attribute), [1] (dmx.compressor.modeling.nn.torch_modules.Linear attribute), [1]

A

a_maxabs_exists (dmx.compressor.numerical.smoothquant.SmoothQuant property) accum_format (dmx.compressor.numerical.cast.NumericalCastMixin property) ActActMatMul (class in dmx.compressor.modeling.nn.torch_modules), [1] ActivatePlugins (class in dmx.compressor.plugins), [1] ActivationWeightSmoothQuant (class in dmx.compressor.numerical.smoothquant), [1] AdaptiveAvgPool2d (class in dmx.compressor.modeling.nn.torch_modules), [1] add_dmx_aware_functional_mapping() (dmx.compressor.fx.transformer.dmx_aware_transformer.DMXAwareTransformer method) add_submod() (dmx.compressor.fx.transformer.dmx_aware_transformer.DMXAwareTransformer method) (dmx.compressor.fx.transformer.export_transformer.ExportSubstituteTransformer method) additional_dmx_aware_mappings (dmx.compressor.modeling.model.DmxModelMixin attribute) align_device() (dmx.compressor.modeling.nn.core.DmxModule method) applied_to() (dmx.compressor.advanced_recipe.DmxBaseRecipe method) (dmx.compressor.plugins.ActivatePlugins method) apply() (dmx.compressor.layer_reconstruction.OptimalBrainCompressor method) apply_rotary_embeddings() (in module dmx.compressor.modeling.nn.custom_modules), [1] apply_shaping_seq() (dmx.compressor.numerical.cast.CastTo method) apply_to() (dmx.compressor.modeling.model.DmxConfigRule method)

ApplyRotaryPosEmb (class in dmx.compressor.modeling.nn.custom_modules), [1] ApplyRotaryPosEmbBase (class in dmx.compressor.modeling.nn.custom_modules), [1] approx_forward() (dmx.compressor.functional.approximate.ApproximationMixin method) Approximate (class in dmx.compressor.functional.approximate), [1] approximation_function (dmx.compressor.functional.approximate.ApproximationMixin property) approximation_function() (dmx.compressor.functional.approximate.Approximator method) approximation_function_constructor() (in module dmx.compressor.utils.io), [1] approximation_function_representer() (in module dmx.compressor.utils.io), [1] ApproximationFunction (class in dmx.compressor.functional.approximate), [1] ApproximationFunctionTuner (class in dmx.compressor.layer_reconstruction), [1] ApproximationMixin (class in dmx.compressor.functional.approximate), [1] Approximator (class in dmx.compressor.functional.approximate), [1] approximator_wrapper() (dmx.compressor.functional.approximate.ApproximationMixin method) (dmx.compressor.modeling.nn.torch_modules.LayerNorm method) (dmx.compressor.modeling.nn.torch_modules.Softmax method) args (dmx.compressor.plugins.PluginLayerData attribute) AvgPool2d (class in dmx.compressor.modeling.nn.torch_modules), [1]

B

b_maxabs_exists (dmx.compressor.numerical.smoothquant.SmoothQuant property) backward() (dmx.compressor.numerical.cast.CastToFormat static method) (dmx.compressor.sparse.Bernoulli static method) (dmx.compressor.sparse.BlockTopK static method) (dmx.compressor.sparse.Sparseness static method) BAddBMM (class in dmx.compressor.modeling.nn.torch_modules), [1] balanced_device_map() (in module dmx.compressor.modeling.hf), [1] BASELINE (dmx.compressor.utils.benchmark.EVALUATION_MODE attribute) BASIC (dmx.compressor.utils.benchmark.EVALUATION_MODE attribute) BASIC_NOVSIMD (dmx.compressor.utils.benchmark.EVALUATION_MODE attribute) BatchNorm2d (class in dmx.compressor.modeling.nn.torch_modules), [1] Bernoulli (class in dmx.compressor.sparse), [1] bfp_id (dmx.compressor.numerical.format.BlockFloatingPoint property) (dmx.compressor.numerical.format.Format attribute) (dmx.compressor.numerical.format.ScaledBlockFloatingPoint property) BFPTypeEnum (class in dmx.compressor.numerical.onnx), [1] bias_format (dmx.compressor.numerical.cast.NumericalCastMixin property) bit_precision (dmx.compressor.numerical.format.BlockFloatingPoint property) (dmx.compressor.numerical.format.FixedPoint property) (dmx.compressor.numerical.format.FloatingPoint property) (dmx.compressor.numerical.format.Format property) (dmx.compressor.numerical.format.MXFP property) (dmx.compressor.numerical.format.Same property) (dmx.compressor.numerical.format.ScaledBlockFloatingPoint property)

block_size (dmx.compressor.advanced_recipe.DmxModuleGPTQHyperparams attribute) blocked (dmx.compressor.numerical.format.BlockFloatingPoint attribute) (dmx.compressor.numerical.format.FixedPoint attribute) (dmx.compressor.numerical.format.FloatingPoint attribute) (dmx.compressor.numerical.format.Format attribute) (dmx.compressor.numerical.format.MXFP attribute) (dmx.compressor.numerical.format.Same attribute) (dmx.compressor.numerical.format.ScaledBlockFloatingPoint attribute) (dmx.compressor.sparse.Bernoulli attribute) (dmx.compressor.sparse.BlockTopK attribute) (dmx.compressor.sparse.Dense attribute) (dmx.compressor.sparse.Sparseness attribute) (dmx.compressor.sparse.TopK attribute) BlockFloatingPoint (class in dmx.compressor.numerical.format), [1] BlockTopK (class in dmx.compressor.sparse), [1] BloomGELU (class in dmx.compressor.modeling.nn.custom_modules), [1] bops (dmx.compressor.perf_proxy.PerformanceProxyMixin property) bytes_per_elem (dmx.compressor.numerical.format.BlockFloatingPoint property) (dmx.compressor.numerical.format.FixedPoint property) (dmx.compressor.numerical.format.FloatingPoint property) (dmx.compressor.numerical.format.Format property) (dmx.compressor.numerical.format.MXFP property) (dmx.compressor.numerical.format.Same property) (dmx.compressor.numerical.format.ScaledBlockFloatingPoint property)

C

calculate_qparams() (dmx.compressor.numerical.observer.DummyObserver method) (dmx.compressor.numerical.observer.HistogramObserver method) (dmx.compressor.numerical.observer.MinMaxObserver method) (dmx.compressor.numerical.observer.PercentileObserver method) calibrating (dmx.compressor.numerical.smoothquant.SmoothQuant attribute) calibrating_quantizers() (dmx.compressor.layer_reconstruction.LayerReconstructionMixin method) calibrating_smoothquant() (dmx.compressor.layer_reconstruction.LayerReconstructionMixin method) call_function() (dmx.compressor.fx.transformer.configuration_transformer.ConfigurationTransformer method) (dmx.compressor.fx.transformer.dmx_aware_transformer.DMXAwareTransformer method) (dmx.compressor.fx.transformer.export_transformer.ExportSubstituteTransformer method) (dmx.compressor.fx.transformer.input_output_transformer.InputOutputTransformer method) (dmx.compressor.fx.transformer.nodedict_transformer.NodeDictTransformer method) (dmx.compressor.modeling.nn.core.DmxGraph method) call_method() (dmx.compressor.fx.transformer.dmx_aware_transformer.DMXAwareTransformer method) (dmx.compressor.fx.transformer.export_transformer.ExportSubstituteTransformer method) (dmx.compressor.fx.transformer.input_output_transformer.InputOutputTransformer method) (dmx.compressor.fx.transformer.nodedict_transformer.NodeDictTransformer method) (dmx.compressor.modeling.nn.core.DmxGraph method) call_module() (dmx.compressor.fx.tracer.HFQuantTracer method) (dmx.compressor.fx.tracer.QuantTracer method) (dmx.compressor.fx.transformer.dmx_aware_transformer.DMXAwareTransformer method) (dmx.compressor.fx.transformer.export_transformer.ExportSubstituteTransformer method) (dmx.compressor.fx.transformer.nodedict_transformer.NodeDictTransformer method) (dmx.compressor.fx.transformer.qdq_transformer.QdQTransformer method) cast() (dmx.compressor.numerical.format.BlockFloatingPoint method) (dmx.compressor.numerical.format.FixedPoint method) (dmx.compressor.numerical.format.FloatingPoint method) (dmx.compressor.numerical.format.Format method) (dmx.compressor.numerical.format.MXFP method) (dmx.compressor.numerical.format.Same method) (dmx.compressor.numerical.format.ScaledBlockFloatingPoint method) cast_input_output_transform() (in module dmx.compressor.fx.transform), [1] CastTo (class in dmx.compressor.numerical.cast), [1] CastToDict (class in dmx.compressor.numerical.cast), [1] CastToFormat (class in dmx.compressor.numerical.cast), [1] cfg (dmx.compressor.fx.transformer.configuration_transformer.ConfigurationTransformer attribute), [1] (dmx.compressor.fx.transformer.input_output_transformer.InputOutputTransformer attribute), [1] ch_axis (dmx.compressor.advanced_recipe.DmxQuantizerCalibrationHyperparams attribute) check_dim_consistency() (dmx.compressor.modeling.model.DmxModelMixin method)

check_format_dim_consistency() (dmx.compressor.numerical.cast.NumericalCastMixin method) check_input_format_dim_consistency() (dmx.compressor.numerical.cast.NumericalCastMixin method) check_residual_format_dim_consistency() (dmx.compressor.numerical.cast.NumericalCastMixin method) check_sparseness_dim_consistency() (dmx.compressor.sparse.WeightSparseMixin method) check_weight_format_dim_consistency() (dmx.compressor.numerical.cast.NumericalCastMixin method) ClippedGELU (class in dmx.compressor.modeling.nn.custom_modules), [1] cls_to_become (dmx.compressor.sparse.LazySparsify attribute) collect_layer_activations() (in module dmx.compressor.utils.benchmark), [1] compute_error() (in module dmx.compressor.utils.benchmark), [1] compute_maxdelta_error() (in module dmx.compressor.utils.benchmark), [1] compute_md5() (in module dmx.compressor.utils.io), [1] compute_mse_error() (in module dmx.compressor.utils.benchmark), [1] compute_scale() (dmx.compressor.numerical.smoothquant.ActivationWeightSmoothQuant method) (dmx.compressor.numerical.smoothquant.SmoothQuant method) ConfigurationTransformer (class in dmx.compressor.fx.transformer.configuration_transformer), [1] configure() (dmx.compressor.modeling.model.DmxModelMixin method) (dmx.compressor.modeling.model.DmxPipelineMixin method) (dmx.compressor.modeling.nn.core.DmxModule method) (dmx.compressor.sparse.Sparsify method) configure_transform() (in module dmx.compressor.fx.transform), [1] contains_number() (in module dmx.compressor.modeling.hf), [1] Conv1d (class in dmx.compressor.modeling.nn.torch_modules), [1] Conv1dScatter (class in dmx.compressor.modeling.nn.experimental), [1] Conv1dUnfold (class in dmx.compressor.modeling.nn.experimental), [1] Conv2d (class in dmx.compressor.modeling.nn.torch_modules), [1] Conv2dGather (class in dmx.compressor.modeling.nn.experimental), [1] Conv2dUnfold (class in dmx.compressor.modeling.nn.experimental), [1] ConvTranspose2d (class in dmx.compressor.modeling.nn.torch_modules), [1] count_flops() (dmx.compressor.perf_proxy.PerformanceProxyMixin method) counting_flops() (dmx.compressor.modeling.model.DmxModelMixin method) (dmx.compressor.modeling.model.DmxPipelineMixin method) (dmx.compressor.perf_proxy.PerformanceProxyMixin method) create_node() (dmx.compressor.fx.tracer.HFQuantTracer method) (dmx.compressor.fx.tracer.QuantTracer method) (dmx.compressor.modeling.nn.core.DmxGraph method) create_placeholders() (dmx.compressor.modeling.nn.core.DmxGraph method) create_submod_transform_forward() (dmx.compressor.modeling.model.DmxModelMixin class method) create_unique_name_in_scope() (dmx.compressor.fx.transformer.dmx_aware_transformer.DMXAwareTransformer method) CustomFunctionApproximation (class in dmx.compressor.functional.approximate), [1]

D

deepcopy_args() (dmx.compressor.modeling.model.DmxModelMixin static method) Dense (class in dmx.compressor.sparse), [1] density (dmx.compressor.sparse.Sparsify property) DeQuantize (class in dmx.compressor.numerical.cast), [1] determine_scaler_exponent_bias_from() (dmx.compressor.numerical.format.ScaledBlockFloatingPoint method) device (dmx.compressor.advanced_recipe.DmxSLaNCHyperparams attribute) disable() (dmx.compressor.numerical.smoothquant.SmoothQuant method) disable_fake_quant() (dmx.compressor.numerical.cast.CastToDict method) disable_hooked_forward() (in module dmx.compressor.fx.tracer), [1] disable_observer() (dmx.compressor.numerical.cast.CastToDict method) dmx module dmx.compressor module dmx.compressor.advanced_recipe module dmx.compressor.functional module dmx.compressor.functional.approximate module dmx.compressor.functional.functions module dmx.compressor.fx module dmx.compressor.fx.tracer module dmx.compressor.fx.transform module dmx.compressor.fx.transformer module dmx.compressor.fx.transformer.configuration_transformer module dmx.compressor.fx.transformer.custom_ops module dmx.compressor.fx.transformer.dmx_aware_transformer module dmx.compressor.fx.transformer.export_transformer module dmx.compressor.fx.transformer.input_output_transformer module dmx.compressor.fx.transformer.nodedict_transformer module dmx.compressor.fx.transformer.qdq_transformer module dmx.compressor.fx.transformer.record_input_interpreter module dmx.compressor.fx.transformer.utils module dmx.compressor.layer_reconstruction module dmx.compressor.modeling module dmx.compressor.modeling.hf module dmx.compressor.modeling.model module dmx.compressor.modeling.nn module dmx.compressor.modeling.nn.core module dmx.compressor.modeling.nn.custom_modules module dmx.compressor.modeling.nn.experimental module dmx.compressor.modeling.nn.torch_modules module dmx.compressor.numerical module dmx.compressor.numerical.cast module dmx.compressor.numerical.custom_lib module dmx.compressor.numerical.format module dmx.compressor.numerical.observer module dmx.compressor.numerical.onnx module dmx.compressor.numerical.smoothquant module dmx.compressor.perf_proxy module dmx.compressor.plugins module dmx.compressor.quant module dmx.compressor.sparse module dmx.compressor.utils module dmx.compressor.utils.benchmark module dmx.compressor.utils.io module dmx.compressor.utils.torch_helpers module

dmx.compressor.utils.visualization module DMX_BFP_12_128 (dmx.compressor.numerical.onnx.BFPTypeEnum attribute) DMX_BFP_12_16 (dmx.compressor.numerical.onnx.BFPTypeEnum attribute) DMX_BFP_12_32 (dmx.compressor.numerical.onnx.BFPTypeEnum attribute) DMX_BFP_12_64 (dmx.compressor.numerical.onnx.BFPTypeEnum attribute) DMX_BFP_12A_128 (dmx.compressor.numerical.onnx.BFPTypeEnum attribute) DMX_BFP_12A_16 (dmx.compressor.numerical.onnx.BFPTypeEnum attribute) DMX_BFP_12A_32 (dmx.compressor.numerical.onnx.BFPTypeEnum attribute) DMX_BFP_12A_64 (dmx.compressor.numerical.onnx.BFPTypeEnum attribute) DMX_BFP_14_128 (dmx.compressor.numerical.onnx.BFPTypeEnum attribute) DMX_BFP_14_16 (dmx.compressor.numerical.onnx.BFPTypeEnum attribute) DMX_BFP_14_32 (dmx.compressor.numerical.onnx.BFPTypeEnum attribute) DMX_BFP_14_64 (dmx.compressor.numerical.onnx.BFPTypeEnum attribute) DMX_BFP_14A_128 (dmx.compressor.numerical.onnx.BFPTypeEnum attribute) DMX_BFP_14A_16 (dmx.compressor.numerical.onnx.BFPTypeEnum attribute) DMX_BFP_14A_32 (dmx.compressor.numerical.onnx.BFPTypeEnum attribute) DMX_BFP_14A_64 (dmx.compressor.numerical.onnx.BFPTypeEnum attribute) DMX_BFP_16_128 (dmx.compressor.numerical.onnx.BFPTypeEnum attribute) DMX_BFP_16_16 (dmx.compressor.numerical.onnx.BFPTypeEnum attribute) DMX_BFP_16_32 (dmx.compressor.numerical.onnx.BFPTypeEnum attribute) DMX_BFP_16_64 (dmx.compressor.numerical.onnx.BFPTypeEnum attribute) DMX_BFP_24_16 (dmx.compressor.numerical.onnx.BFPTypeEnum attribute) DMX_BFP_24_32 (dmx.compressor.numerical.onnx.BFPTypeEnum attribute) DMX_BFP_24_64 (dmx.compressor.numerical.onnx.BFPTypeEnum attribute) DMX_BFP_32_1 (dmx.compressor.numerical.onnx.BFPTypeEnum attribute) dmx_config (dmx.compressor.modeling.model.DmxModelMixin property) dmx_config() (dmx.compressor.modeling.nn.core.DmxModule method) dmx_config_dict (dmx.compressor.modeling.model.DmxPipelineMixin property) dmx_module_instance_constructor() (in module dmx.compressor.utils.io), [1] dmx_module_instance_representer() (in module dmx.compressor.utils.io), [1] dmx_module_names (dmx.compressor.modeling.model.DmxModelMixin property) DMX_SBFP_12_16_10 (dmx.compressor.numerical.onnx.BFPTypeEnum attribute) DMX_SBFP_12_16_11 (dmx.compressor.numerical.onnx.BFPTypeEnum attribute) DMX_SBFP_12_16_12 (dmx.compressor.numerical.onnx.BFPTypeEnum attribute) DMX_SBFP_12_16_13 (dmx.compressor.numerical.onnx.BFPTypeEnum attribute) DMX_SBFP_12_16_14 (dmx.compressor.numerical.onnx.BFPTypeEnum attribute) DMX_SBFP_12_16_15 (dmx.compressor.numerical.onnx.BFPTypeEnum attribute) DMX_SBFP_12_16_16 (dmx.compressor.numerical.onnx.BFPTypeEnum attribute) DMX_SBFP_12_16_17 (dmx.compressor.numerical.onnx.BFPTypeEnum attribute) DMX_SBFP_12_16_18 (dmx.compressor.numerical.onnx.BFPTypeEnum attribute) DMX_SBFP_12_16_4 (dmx.compressor.numerical.onnx.BFPTypeEnum attribute) DMX_SBFP_12_16_5 (dmx.compressor.numerical.onnx.BFPTypeEnum attribute) DMX_SBFP_12_16_6 (dmx.compressor.numerical.onnx.BFPTypeEnum attribute) DMX_SBFP_12_16_7 (dmx.compressor.numerical.onnx.BFPTypeEnum attribute) DMX_SBFP_12_16_8 (dmx.compressor.numerical.onnx.BFPTypeEnum attribute) DMX_SBFP_12_16_9 (dmx.compressor.numerical.onnx.BFPTypeEnum attribute) dmx_transform() (in module dmx.compressor.modeling.hf), [1] DMX_UBFP_12_128 (dmx.compressor.numerical.onnx.BFPTypeEnum attribute) DMX_UBFP_12_16 (dmx.compressor.numerical.onnx.BFPTypeEnum attribute) DMX_UBFP_12_32 (dmx.compressor.numerical.onnx.BFPTypeEnum attribute) DMX_UBFP_12_64 (dmx.compressor.numerical.onnx.BFPTypeEnum attribute) DMX_UBFP_14_128 (dmx.compressor.numerical.onnx.BFPTypeEnum attribute) DMX_UBFP_14_16 (dmx.compressor.numerical.onnx.BFPTypeEnum attribute) DMX_UBFP_14_32 (dmx.compressor.numerical.onnx.BFPTypeEnum attribute) DMX_UBFP_14_64 (dmx.compressor.numerical.onnx.BFPTypeEnum attribute) DMX_UBFP_16_128 (dmx.compressor.numerical.onnx.BFPTypeEnum attribute) DMX_UBFP_16_16 (dmx.compressor.numerical.onnx.BFPTypeEnum attribute) DMX_UBFP_16_32 (dmx.compressor.numerical.onnx.BFPTypeEnum attribute) DMX_UBFP_16_64 (dmx.compressor.numerical.onnx.BFPTypeEnum attribute) DMX_UBFP_24_16 (dmx.compressor.numerical.onnx.BFPTypeEnum attribute) DMX_UBFP_24_32 (dmx.compressor.numerical.onnx.BFPTypeEnum attribute) DMX_UBFP_24_64 (dmx.compressor.numerical.onnx.BFPTypeEnum attribute) DMX_UBFP_32_1 (dmx.compressor.numerical.onnx.BFPTypeEnum attribute) DmxApproximationFunctionTuningHyperparams (class in dmx.compressor.advanced_recipe), [1] DmxApproximationFunctionTuningRecipe (class in dmx.compressor.advanced_recipe), [1] DMXAwareTransformer (class in dmx.compressor.fx.transformer.dmx_aware_transformer), [1] DmxBaseRecipe (class in dmx.compressor.advanced_recipe), [1] DmxConfig (class in dmx.compressor.modeling.model), [1] DmxConfigRule (class in dmx.compressor.modeling.model), [1] DmxGPTQRecipe (class in dmx.compressor.advanced_recipe), [1] DmxGraph (class in dmx.compressor.modeling.nn.core), [1] DmxHFTracer (class in dmx.compressor.fx.tracer), [1] DmxModel (class in dmx.compressor.modeling.model), [1] DmxModelMixin (class in dmx.compressor.modeling.model), [1] DmxModule (class in dmx.compressor.modeling.nn.core), [1] DmxModuleConfig (class in dmx.compressor.modeling.nn.core), [1] DmxModuleGPTQHyperparams (class in dmx.compressor.advanced_recipe), [1] DmxModuleQuantizerCalibrationHyperparams (class in dmx.compressor.advanced_recipe), [1] DmxModuleSmoothQuantHyperparams (class in dmx.compressor.advanced_recipe), [1] DmxModuleType (class in dmx.compressor.modeling.nn.core), [1] DMXObserverBase (class in dmx.compressor.numerical.observer), [1] DmxPipelineMixin (class in dmx.compressor.modeling.model), [1] DmxPreTrainedModel (class in dmx.compressor.modeling.hf), [1] DmxQuantizerCalibrationHyperparams (class in dmx.compressor.advanced_recipe), [1] DmxQuantizerCalibrationRecipe (class in dmx.compressor.advanced_recipe), [1] DmxSimplePipeline (class in dmx.compressor.modeling.model), [1] DmxSLaNCHyperparams (class in dmx.compressor.advanced_recipe), [1] DmxSLaNCRecipe (class in dmx.compressor.advanced_recipe), [1] DmxSmoothQuantRecipe (class in dmx.compressor.advanced_recipe), [1] DmxTransformation (in module dmx.compressor.modeling.model), [1] do_forward_on() (in module dmx.compressor.modeling.hf), [1] down_proj (dmx.compressor.advanced_recipe.DmxSLaNCHyperparams attribute) Dropout (class in dmx.compressor.modeling.nn.torch_modules), [1] DummyObserver (class in dmx.compressor.numerical.observer), [1] dynamic (dmx.compressor.numerical.smoothquant.ActivationWeightSmoothQuant property)

E

effective_weight (dmx.compressor.sparse.WeightSparseMixin property) Embedding (class in dmx.compressor.modeling.nn.torch_modules), [1] enable() (dmx.compressor.numerical.smoothquant.SmoothQuant method) enable_approximation_function_tuning() (dmx.compressor.layer_reconstruction.LayerReconstructionMixin method) enable_calibration() (dmx.compressor.numerical.cast.CastTo method) enable_fake_quant() (dmx.compressor.numerical.cast.CastToDict method) enable_flop_counter() (dmx.compressor.perf_proxy.PerformanceProxyMixin method) enable_observer() (dmx.compressor.numerical.cast.CastToDict method) enable_optimal_brain_compression() (dmx.compressor.layer_reconstruction.LayerReconstructionMixin method) enable_quantizer_calib() (dmx.compressor.layer_reconstruction.LayerReconstructionMixin method) enable_smoothquant_calib() (dmx.compressor.layer_reconstruction.LayerReconstructionMixin method) eps (dmx.compressor.numerical.observer.DMXObserverBase attribute) eval() (dmx.compressor.modeling.model.DmxPipelineMixin method) eval_question_answering() (in module dmx.compressor.modeling.hf), [1] eval_text_generation() (in module dmx.compressor.modeling.hf), [1] evaluate_vsimd_ops_deltas() (in module dmx.compressor.utils.benchmark), [1] EVALUATION_MODE (class in dmx.compressor.utils.benchmark), [1]

execute() (dmx.compressor.functional.approximate.ApproximationFunction method) (dmx.compressor.functional.approximate.CustomFunctionApproximation method) (dmx.compressor.functional.approximate.NoApproximation method) (dmx.compressor.functional.approximate.TorchFunctionApproximation method) Exp (class in dmx.compressor.modeling.nn.torch_modules), [1] experimental (class in dmx.compressor.functional.functions), [1] export_substitute_transform() (in module dmx.compressor.fx.transform), [1] ExportSubstituteTransformer (class in dmx.compressor.fx.transformer.export_transformer), [1] extra_repr() (dmx.compressor.functional.approximate.Approximate method) (dmx.compressor.functional.approximate.Approximator method) (dmx.compressor.numerical.cast.CastTo method) (dmx.compressor.numerical.observer.DMXObserverBase method) (dmx.compressor.numerical.observer.HistogramObserver method) (dmx.compressor.numerical.observer.MinMaxObserver method) (dmx.compressor.numerical.observer.PercentileObserver method) (dmx.compressor.numerical.smoothquant.ActivationWeightSmoothQuant method) (dmx.compressor.numerical.smoothquant.SmoothQuant method) (dmx.compressor.sparse.Sparsify method)

F

FastGELU (class in dmx.compressor.modeling.nn.custom_modules), [1] fc1 (dmx.compressor.advanced_recipe.DmxSLaNCHyperparams attribute) fc2 (dmx.compressor.advanced_recipe.DmxSLaNCHyperparams attribute) FixedPoint (class in dmx.compressor.numerical.format), [1] FloatingPoint (class in dmx.compressor.numerical.format), [1] flop_counter (dmx.compressor.perf_proxy.PerformanceProxyMixin attribute) flop_counter_enabled (dmx.compressor.perf_proxy.PerformanceProxyMixin attribute) flops (dmx.compressor.perf_proxy.PerformanceProxyMixin property) fold_weight_and_bias() (dmx.compressor.modeling.nn.core.DmxModule method) fold_weights_and_biases() (dmx.compressor.modeling.model.DmxModelMixin method) Format (class in dmx.compressor.numerical.format), [1] format_constructor() (in module dmx.compressor.utils.io), [1] format_representer() (in module dmx.compressor.utils.io), [1] forward() (dmx.compressor.functional.approximate.Approximate method) (dmx.compressor.functional.approximate.Approximator method) (dmx.compressor.modeling.nn.core.DmxModule method) (dmx.compressor.modeling.nn.custom_modules.ApplyRotaryPosEmbBase method) (dmx.compressor.modeling.nn.torch_modules.ScaledDotProductAttention method) (dmx.compressor.numerical.cast.CastTo method) (dmx.compressor.numerical.cast.CastToDict method) (dmx.compressor.numerical.cast.CastToFormat static method) (dmx.compressor.numerical.cast.DeQuantize method) (dmx.compressor.numerical.cast.Quantize method) (dmx.compressor.numerical.observer.DummyObserver method) (dmx.compressor.numerical.observer.HistogramObserver method) (dmx.compressor.numerical.observer.MinMaxObserver method) (dmx.compressor.numerical.smoothquant.ActivationWeightSmoothQuant method) (dmx.compressor.numerical.smoothquant.SmoothQuant method) (dmx.compressor.sparse.Bernoulli static method) (dmx.compressor.sparse.BlockTopK static method) (dmx.compressor.sparse.Sparsify method) (dmx.compressor.sparse.TopK static method) forward_weight_hypernets() (dmx.compressor.modeling.model.DmxModelMixin method) FP8 (dmx.compressor.utils.benchmark.EVALUATION_MODE attribute) freeze() (dmx.compressor.modeling.model.DmxModelMixin method) from_model() (dmx.compressor.modeling.model.DmxConfig class method) from_module() (dmx.compressor.modeling.nn.core.DmxModuleConfig class method) from_pretrained() (dmx.compressor.modeling.hf.DmxPreTrainedModel class method) from_raw() (dmx.compressor.modeling.nn.custom_modules.GemmaRMSNorm class method) (dmx.compressor.modeling.nn.custom_modules.RotaryEmbedding class method) (dmx.compressor.modeling.nn.torch_modules.AdaptiveAvgPool2d class method) (dmx.compressor.modeling.nn.torch_modules.AvgPool2d class method) (dmx.compressor.modeling.nn.torch_modules.BatchNorm2d class method) (dmx.compressor.modeling.nn.torch_modules.Conv1d class method) (dmx.compressor.modeling.nn.torch_modules.Conv2d class method) (dmx.compressor.modeling.nn.torch_modules.ConvTranspose2d class method) (dmx.compressor.modeling.nn.torch_modules.Dropout class method) (dmx.compressor.modeling.nn.torch_modules.Embedding class method) (dmx.compressor.modeling.nn.torch_modules.GELUBase class method) (dmx.compressor.modeling.nn.torch_modules.LayerNorm class method) (dmx.compressor.modeling.nn.torch_modules.Linear class method) (dmx.compressor.modeling.nn.torch_modules.MaxPool2d class method) (dmx.compressor.modeling.nn.torch_modules.ReLU class method) (dmx.compressor.modeling.nn.torch_modules.ReLU6 class method) (dmx.compressor.modeling.nn.torch_modules.RMSNorm class method) (dmx.compressor.modeling.nn.torch_modules.SiLU class method) (dmx.compressor.modeling.nn.torch_modules.Softmax class method) (dmx.compressor.modeling.nn.torch_modules.Tanh class method)

from_shorthand() (dmx.compressor.functional.approximate.ApproximationFunction static method) (dmx.compressor.functional.approximate.CustomFunctionApproximation class method) (dmx.compressor.functional.approximate.NoApproximation class method) (dmx.compressor.functional.approximate.TorchFunctionApproximation class method) (dmx.compressor.numerical.format.BlockFloatingPoint class method) (dmx.compressor.numerical.format.FixedPoint class method) (dmx.compressor.numerical.format.FloatingPoint class method) (dmx.compressor.numerical.format.Format static method) (dmx.compressor.numerical.format.MXFP class method) (dmx.compressor.numerical.format.MXINT class method) (dmx.compressor.numerical.format.Same class method) (dmx.compressor.numerical.format.ScaledBlockFloatingPoint class method) (dmx.compressor.sparse.Bernoulli class method) (dmx.compressor.sparse.BlockTopK class method) (dmx.compressor.sparse.Dense class method) (dmx.compressor.sparse.Sparseness class method) (dmx.compressor.sparse.TopK class method) from_torch() (dmx.compressor.modeling.model.DmxModel class method) from_yaml() (dmx.compressor.modeling.model.DmxConfig class method) functional_forward (dmx.compressor.modeling.nn.core.DmxModule attribute) functional_forward() (dmx.compressor.modeling.nn.custom_modules.BloomGELU method) (dmx.compressor.modeling.nn.custom_modules.ClippedGELU method) (dmx.compressor.modeling.nn.custom_modules.FastGELU method) (dmx.compressor.modeling.nn.custom_modules.GemmaRMSNorm method) (dmx.compressor.modeling.nn.custom_modules.NewGELU method) (dmx.compressor.modeling.nn.custom_modules.QuickGELU method) (dmx.compressor.modeling.nn.torch_modules.GELU method) fuse_to_weight (dmx.compressor.advanced_recipe.DmxModuleSmoothQuantHyperparams attribute) fuse_to_weight() (dmx.compressor.numerical.smoothquant.ActivationWeightSmoothQuant method)

G

gap_analysis() (in module dmx.compressor.fx.transformer.utils), [1] gate_proj (dmx.compressor.advanced_recipe.DmxSLaNCHyperparams attribute) gather_tensors() (in module dmx.compressor.utils.benchmark), [1] GELU (class in dmx.compressor.modeling.nn.torch_modules), [1] GELUBase (class in dmx.compressor.modeling.nn.torch_modules), [1] GemmaRMSNorm (class in dmx.compressor.modeling.nn.custom_modules), [1] get_attr() (dmx.compressor.fx.transformer.configuration_transformer.ConfigurationTransformer method) (dmx.compressor.fx.transformer.dmx_aware_transformer.DMXAwareTransformer method) (dmx.compressor.fx.transformer.export_transformer.ExportSubstituteTransformer method) (dmx.compressor.fx.transformer.input_output_transformer.InputOutputTransformer method) (dmx.compressor.fx.transformer.nodedict_transformer.NodeDictTransformer method) (dmx.compressor.modeling.nn.core.DmxGraph method) get_config_file() (in module dmx.compressor.modeling.hf), [1] get_dumper() (in module dmx.compressor.utils.io), [1] get_loader() (in module dmx.compressor.utils.io), [1]

get_mask() (dmx.compressor.sparse.Bernoulli method) (dmx.compressor.sparse.BlockTopK method) (dmx.compressor.sparse.Dense method) (dmx.compressor.sparse.Sparseness method) (dmx.compressor.sparse.TopK method) get_model_by_name() (dmx.compressor.modeling.model.DmxPipelineMixin method) get_modules() (in module dmx.compressor.modeling.hf), [1] get_monitoring_records() (dmx.compressor.modeling.model.DmxModelMixin method) get_name_for_func_nodes() (in module dmx.compressor.fx.transformer.utils), [1] get_nested_attr() (in module dmx.compressor.fx.transformer.export_transformer), [1] get_op_set_from() (in module dmx.compressor.fx.transformer.utils), [1] get_precision() (dmx.compressor.numerical.cast.CastTo method) get_runtime_records() (dmx.compressor.modeling.model.DmxModelMixin method) group_size (dmx.compressor.advanced_recipe.DmxQuantizerCalibrationHyperparams attribute) GroupNorm (class in dmx.compressor.modeling.nn.torch_modules), [1]

H

H (dmx.compressor.layer_reconstruction.OptimalBrainCompressor attribute) has_module() (dmx.compressor.fx.transformer.configuration_transformer.ConfigurationTransformer method) hf_symbolic_trace() (in module dmx.compressor.fx.tracer), [1]

HFQuantTracer (class in dmx.compressor.fx.tracer), [1] histogram (dmx.compressor.numerical.observer.HistogramObserver attribute) HistogramObserver (class in dmx.compressor.numerical.observer), [1]

I

Identity (in module dmx.compressor.functional.approximate), [1] infer_ch_axis() (dmx.compressor.numerical.cast.NumericalCastMixin method) init_casts() (dmx.compressor.numerical.cast.NumericalCastMixin method) init_smoothquant() (dmx.compressor.numerical.cast.NumericalCastMixin method) init_sparsifier() (dmx.compressor.sparse.WeightSparseMixin method) initialize_parameters() (dmx.compressor.sparse.LazySparsify method) input_after_cast (dmx.compressor.plugins.PluginLayerData attribute) input_before_cast (dmx.compressor.plugins.PluginLayerData attribute) input_formats (dmx.compressor.numerical.cast.NumericalCastMixin property) input_maxabs_exists (dmx.compressor.numerical.smoothquant.ActivationWeightSmoothQuant property)

input_precision (dmx.compressor.numerical.cast.NumericalCastMixin property) InputOutputTransformer (class in dmx.compressor.fx.transformer.input_output_transformer), [1] inputs (dmx.compressor.advanced_recipe.DmxModuleQuantizerCalibrationHyperparams attribute) is_compound (dmx.compressor.modeling.nn.core.DmxModule attribute) (dmx.compressor.modeling.nn.torch_modules.ScaledDotProductAttention attribute) is_configurable() (in module dmx.compressor.modeling.nn.core), [1] is_leaf_module() (dmx.compressor.fx.tracer.DmxHFTracer method) (dmx.compressor.fx.tracer.HFQuantTracer method) (dmx.compressor.fx.tracer.QuantTracer method) is_same_signature() (dmx.compressor.modeling.model.DmxModelMixin static method)

K

keep_dmx_config() (dmx.compressor.modeling.model.DmxModelMixin method)

kwargs (dmx.compressor.plugins.PluginLayerData attribute) kwargs_to_string() (in module dmx.compressor.utils.io), [1]

L

largest_representable_power_of_two (dmx.compressor.numerical.format.FloatingPoint property) last_input_shape (dmx.compressor.perf_proxy.PerformanceProxyMixin attribute) last_output_shape (dmx.compressor.perf_proxy.PerformanceProxyMixin attribute) LayerNorm (class in dmx.compressor.modeling.nn.torch_modules), [1]

LayerReconstructionMixin (class in dmx.compressor.layer_reconstruction), [1] LazySparsify (class in dmx.compressor.sparse), [1] Linear (class in dmx.compressor.modeling.nn.torch_modules), [1] load_config_file() (in module dmx.compressor.utils.io), [1] load_state_dict_and_register_url() (dmx.compressor.modeling.nn.core.DmxModule method)

M

make_compiler_graph() (in module dmx.compressor.fx.transform), [1] make_compiler_graphs() (dmx.compressor.modeling.model.DmxModel method) make_mantissa_asymmetric() (dmx.compressor.numerical.format.BlockFloatingPoint static method) mask2braille() (in module dmx.compressor.utils.visualization), [1] mask_str() (dmx.compressor.sparse.Sparsify method) max_val (dmx.compressor.numerical.observer.HistogramObserver attribute) (dmx.compressor.numerical.observer.MinMaxObserver attribute) MaxPool2d (class in dmx.compressor.modeling.nn.torch_modules), [1] measure_hessian() (dmx.compressor.layer_reconstruction.OptimalBrainCompressor method) measure_mode_perf() (in module dmx.compressor.utils.benchmark), [1] measure_model_accuracy() (in module dmx.compressor.utils.benchmark), [1] measure_model_error() (in module dmx.compressor.utils.benchmark), [1] measure_model_runtime() (in module dmx.compressor.utils.benchmark), [1] measure_runtimes() (dmx.compressor.modeling.model.DmxModelMixin method) measuring_runtime() (dmx.compressor.modeling.nn.core.DmxModule method) merge_graph() (dmx.compressor.fx.transformer.dmx_aware_transformer.DMXAwareTransformer method) (dmx.compressor.fx.transformer.export_transformer.ExportSubstituteTransformer method) microblock_size (dmx.compressor.advanced_recipe.DmxModuleGPTQHyperparams attribute) migration_strength (dmx.compressor.advanced_recipe.DmxModuleSmoothQuantHyperparams attribute) min_val (dmx.compressor.numerical.observer.HistogramObserver attribute) (dmx.compressor.numerical.observer.MinMaxObserver attribute) MinMaxObserver (class in dmx.compressor.numerical.observer), [1] mlp_type (dmx.compressor.advanced_recipe.DmxSLaNCHyperparams attribute) mod (dmx.compressor.plugins.PluginLayerData attribute) Model (class in dmx.compressor.modeling.model), [1] module dmx dmx.compressor dmx.compressor.advanced_recipe dmx.compressor.functional dmx.compressor.functional.approximate dmx.compressor.functional.functions dmx.compressor.fx dmx.compressor.fx.tracer dmx.compressor.fx.transform dmx.compressor.fx.transformer dmx.compressor.fx.transformer.configuration_transformer dmx.compressor.fx.transformer.custom_ops dmx.compressor.fx.transformer.dmx_aware_transformer dmx.compressor.fx.transformer.export_transformer dmx.compressor.fx.transformer.input_output_transformer dmx.compressor.fx.transformer.nodedict_transformer dmx.compressor.fx.transformer.qdq_transformer dmx.compressor.fx.transformer.record_input_interpreter dmx.compressor.fx.transformer.utils dmx.compressor.layer_reconstruction dmx.compressor.modeling dmx.compressor.modeling.hf dmx.compressor.modeling.model dmx.compressor.modeling.nn dmx.compressor.modeling.nn.core dmx.compressor.modeling.nn.custom_modules dmx.compressor.modeling.nn.experimental dmx.compressor.modeling.nn.torch_modules dmx.compressor.numerical dmx.compressor.numerical.cast dmx.compressor.numerical.custom_lib dmx.compressor.numerical.format dmx.compressor.numerical.observer dmx.compressor.numerical.onnx dmx.compressor.numerical.smoothquant dmx.compressor.perf_proxy dmx.compressor.plugins dmx.compressor.quant dmx.compressor.sparse dmx.compressor.utils dmx.compressor.utils.benchmark dmx.compressor.utils.io dmx.compressor.utils.torch_helpers dmx.compressor.utils.visualization

module (dmx.compressor.fx.transformer.configuration_transformer.ConfigurationTransformer attribute), [1] (dmx.compressor.fx.transformer.dmx_aware_transformer.DMXAwareTransformer attribute), [1] (dmx.compressor.fx.transformer.input_output_transformer.InputOutputTransformer attribute), [1] (dmx.compressor.fx.transformer.nodedict_transformer.NodeDictTransformer attribute), [1] module_config (dmx.compressor.modeling.model.DmxConfigRule attribute), [1] module_graph() (dmx.compressor.modeling.nn.torch_modules.ScaledDotProductAttention method) module_names (dmx.compressor.modeling.model.DmxConfig property) module_path (dmx.compressor.fx.tracer.Scope attribute), [1] module_type (dmx.compressor.fx.tracer.Scope attribute), [1] module_types (dmx.compressor.modeling.model.DmxConfigRule attribute), [1] monitoring() (dmx.compressor.modeling.model.DmxModelMixin method) (dmx.compressor.modeling.nn.core.DmxModule method) Mul (class in dmx.compressor.modeling.nn.torch_modules), [1] multiplier_format (dmx.compressor.numerical.cast.NumericalCastMixin property) MXFP (class in dmx.compressor.numerical.format), [1] MXINT (class in dmx.compressor.numerical.format), [1]

N

name_rule (dmx.compressor.modeling.model.DmxConfigRule attribute), [1] named_dmx_models() (dmx.compressor.modeling.model.DmxSimplePipeline method) named_dmx_modules() (dmx.compressor.modeling.model.DmxModelMixin method) (dmx.compressor.modeling.model.DmxPipelineMixin method) names_in() (dmx.compressor.modeling.model.DmxConfigRule method) NewGELU (class in dmx.compressor.modeling.nn.custom_modules), [1]

NoApproximation (class in dmx.compressor.functional.approximate), [1] node_name_to_scope (dmx.compressor.fx.tracer.HFQuantTracer attribute), [1] (dmx.compressor.fx.tracer.QuantTracer attribute), [1] (dmx.compressor.fx.transformer.dmx_aware_transformer.DMXAwareTransformer attribute), [1] nodeDict (dmx.compressor.fx.transformer.nodedict_transformer.NodeDictTransformer attribute), [1] NodeDictTransformer (class in dmx.compressor.fx.transformer.nodedict_transformer), [1] NumericalCastMixin (class in dmx.compressor.numerical.cast), [1]

O

o_proj (dmx.compressor.advanced_recipe.DmxSLaNCHyperparams attribute) observer_cls (dmx.compressor.advanced_recipe.DmxQuantizerCalibrationHyperparams attribute) op_set (dmx.compressor.modeling.model.DmxModelMixin property) (dmx.compressor.modeling.model.DmxSimplePipeline property) (dmx.compressor.modeling.model.Model property) optimal_brain_compressing() (dmx.compressor.layer_reconstruction.LayerReconstructionMixin method) OptimalBrainCompressor (class in dmx.compressor.layer_reconstruction), [1]

optimize() (dmx.compressor.layer_reconstruction.ApproximationFunctionTuner method) output() (dmx.compressor.fx.transformer.configuration_transformer.ConfigurationTransformer method) (dmx.compressor.fx.transformer.input_output_transformer.InputOutputTransformer method) (dmx.compressor.fx.transformer.nodedict_transformer.NodeDictTransformer method) output_after_cast (dmx.compressor.plugins.PluginLayerData attribute) output_before_cast (dmx.compressor.plugins.PluginLayerData attribute) output_formats (dmx.compressor.numerical.cast.NumericalCastMixin property) outputs (dmx.compressor.advanced_recipe.DmxModuleQuantizerCalibrationHyperparams attribute)

P

pack_to_dict() (dmx.compressor.numerical.cast.CastToDict method) percdamp (dmx.compressor.advanced_recipe.DmxModuleGPTQHyperparams attribute) PercentileObserver (class in dmx.compressor.numerical.observer), [1] PerformanceProxyMixin (class in dmx.compressor.perf_proxy), [1] pipe_eval() (in module dmx.compressor.modeling.hf), [1] pipeline() (in module dmx.compressor.modeling.hf), [1] placeholder() (dmx.compressor.fx.transformer.configuration_transformer.ConfigurationTransformer method) (dmx.compressor.fx.transformer.dmx_aware_transformer.DMXAwareTransformer method) (dmx.compressor.fx.transformer.export_transformer.ExportSubstituteTransformer method) (dmx.compressor.fx.transformer.input_output_transformer.InputOutputTransformer method) (dmx.compressor.fx.transformer.nodedict_transformer.NodeDictTransformer method) (dmx.compressor.modeling.nn.core.DmxGraph method) PluginBase (class in dmx.compressor.plugins), [1] PluginLayerData (class in dmx.compressor.plugins), [1] plugins (dmx.compressor.modeling.nn.core.DmxModule attribute) position (dmx.compressor.advanced_recipe.DmxSLaNCHyperparams attribute)

post_process_gm() (dmx.compressor.modeling.model.DmxModelMixin static method) prepare_dataset_and_column() (in module dmx.compressor.modeling.hf), [1] prepare_dynamic_shapes() (in module dmx.compressor.fx.transform), [1] prepare_model() (in module dmx.compressor.utils.benchmark), [1] prepare_module_call_signature() (in module dmx.compressor.fx.transform), [1] prepare_tracing_inputs() (in module dmx.compressor.fx.transform), [1] prev_ln_weight (dmx.compressor.advanced_recipe.DmxSLaNCHyperparams attribute) prev_module_path (dmx.compressor.fx.tracer.ScopeContextManager attribute), [1] prev_module_type (dmx.compressor.fx.tracer.ScopeContextManager attribute), [1] print_model_tree() (dmx.compressor.modeling.model.DmxModelMixin method) (in module dmx.compressor.utils.visualization), [1] process_args() (in module dmx.compressor.fx.transformer.utils), [1] process_inputs_for_export() (dmx.compressor.modeling.model.DmxModelMixin static method) process_kwargs() (in module dmx.compressor.fx.transformer.utils), [1] process_layer() (dmx.compressor.plugins.PluginBase method) process_model() (dmx.compressor.plugins.PluginBase method)

Q

qdq_node() (dmx.compressor.modeling.nn.core.DmxGraph method) qDq_transform() (in module dmx.compressor.fx.transform), [1] QdQTransformer (class in dmx.compressor.fx.transformer.qdq_transformer), [1]

qscheme_to_overload (dmx.compressor.advanced_recipe.DmxQuantizerCalibrationHyperparams attribute) Quantize (class in dmx.compressor.numerical.cast), [1] QuantTracer (class in dmx.compressor.fx.tracer), [1] QuickGELU (class in dmx.compressor.modeling.nn.custom_modules), [1]

R

record_stack_traces (dmx.compressor.fx.tracer.HFQuantTracer attribute), [1] (dmx.compressor.fx.tracer.QuantTracer attribute), [1] RecordInputInterpreter (class in dmx.compressor.fx.transformer.record_input_interpreter), [1] RecordInputInterpreterExport (class in dmx.compressor.fx.transformer.record_input_interpreter), [1] ReLU (class in dmx.compressor.modeling.nn.torch_modules), [1] ReLU6 (class in dmx.compressor.modeling.nn.torch_modules), [1] ResAdd (class in dmx.compressor.modeling.nn.torch_modules), [1] reset_a_maxabs() (dmx.compressor.numerical.smoothquant.SmoothQuant method) reset_b_maxabs() (dmx.compressor.numerical.smoothquant.SmoothQuant method)

reset_min_max_vals() (dmx.compressor.numerical.observer.MinMaxObserver method) reset_parameters() (dmx.compressor.sparse.LazySparsify method) reset_scale() (dmx.compressor.numerical.smoothquant.SmoothQuant method) reset_weight_maxabs() (dmx.compressor.numerical.smoothquant.ActivationWeightSmoothQuant method) residual_format (dmx.compressor.numerical.cast.NumericalCastMixin property) RMSNorm (class in dmx.compressor.modeling.nn.torch_modules), [1] RotaryEmbedding (class in dmx.compressor.modeling.nn.custom_modules), [1] rotate_half() (dmx.compressor.modeling.nn.custom_modules.ApplyRotaryPosEmbBase method) run_node() (dmx.compressor.fx.transformer.record_input_interpreter.RecordInputInterpreter method) (dmx.compressor.fx.transformer.record_input_interpreter.RecordInputInterpreterExport method)

S

Same (class in dmx.compressor.numerical.format), [1] save_config_file() (in module dmx.compressor.utils.io), [1] save_state_dict_and_register_url() (dmx.compressor.modeling.nn.core.DmxModule method) save_state_dict_and_return_url() (in module dmx.compressor.utils.io), [1] scale_a() (dmx.compressor.numerical.smoothquant.SmoothQuant method) scale_b() (dmx.compressor.numerical.smoothquant.SmoothQuant method) scale_input() (dmx.compressor.numerical.smoothquant.ActivationWeightSmoothQuant method) scale_weight() (dmx.compressor.numerical.smoothquant.ActivationWeightSmoothQuant method) ScaledBlockFloatingPoint (class in dmx.compressor.numerical.format), [1] ScaledDotProductAttention (class in dmx.compressor.modeling.nn.torch_modules), [1] Scope (class in dmx.compressor.fx.tracer), [1] scope (dmx.compressor.fx.tracer.HFQuantTracer attribute), [1] (dmx.compressor.fx.tracer.QuantTracer attribute), [1] (dmx.compressor.fx.tracer.ScopeContextManager attribute), [1] ScopeContextManager (class in dmx.compressor.fx.tracer), [1] scopeDict (dmx.compressor.fx.transformer.configuration_transformer.ConfigurationTransformer attribute), [1] (dmx.compressor.fx.transformer.input_output_transformer.InputOutputTransformer attribute), [1] score (dmx.compressor.sparse.LazySparsify attribute) search_space (dmx.compressor.advanced_recipe.DmxApproximationFunctionTuningHyperparams attribute) set_dynamic() (dmx.compressor.numerical.smoothquant.ActivationWeightSmoothQuant method) (dmx.compressor.numerical.smoothquant.SmoothQuant method) set_format() (dmx.compressor.numerical.cast.CastTo method) (dmx.compressor.numerical.cast.CastToDict method)

set_function() (dmx.compressor.functional.approximate.Approximate method) set_migration_strength() (dmx.compressor.numerical.smoothquant.SmoothQuant method) set_pre_transform() (dmx.compressor.numerical.cast.CastTo method) (dmx.compressor.numerical.cast.CastToDict method) set_scale_format() (dmx.compressor.numerical.smoothquant.SmoothQuant method) SiLU (class in dmx.compressor.modeling.nn.torch_modules), [1] silu() (dmx.compressor.functional.functions.experimental method) slanc_tuning() (dmx.compressor.layer_reconstruction.LayerReconstructionMixin method) SmoothQuant (class in dmx.compressor.numerical.smoothquant), [1] Softmax (class in dmx.compressor.modeling.nn.torch_modules), [1] Sparseness (class in dmx.compressor.sparse), [1] sparseness_constructor() (in module dmx.compressor.utils.io), [1] sparseness_representer() (in module dmx.compressor.utils.io), [1] SparsificationManager (class in dmx.compressor.sparse), [1] Sparsify (class in dmx.compressor.sparse), [1] state_dict_url (dmx.compressor.modeling.nn.core.DmxModule attribute), [1] step() (dmx.compressor.sparse.SparsificationManager method) string_to_kwargs() (in module dmx.compressor.utils.io), [1] substitute_compiler_graph() (dmx.compressor.fx.transformer.qdq_transformer.QdQTransformer static method) substitute_transform() (in module dmx.compressor.fx.transform), [1] symbolic() (dmx.compressor.numerical.cast.CastToFormat static method) symbolic_trace() (in module dmx.compressor.fx.tracer), [1]

T

Tanh (class in dmx.compressor.modeling.nn.torch_modules), [1] thaw() (dmx.compressor.modeling.model.DmxModelMixin method) to() (dmx.compressor.modeling.model.DmxSimplePipeline method) to_baseline_mode() (dmx.compressor.modeling.model.DmxModelMixin method) to_basic_mode() (dmx.compressor.modeling.model.DmxModelMixin method) to_compiler_graph() (dmx.compressor.modeling.nn.core.DmxModule method) (dmx.compressor.modeling.nn.custom_modules.ApplyRotaryPosEmb method) (dmx.compressor.modeling.nn.custom_modules.GemmaRMSNorm method) (dmx.compressor.modeling.nn.custom_modules.RotaryEmbedding method) (dmx.compressor.modeling.nn.experimental.Conv1dScatter method) (dmx.compressor.modeling.nn.experimental.Conv1dUnfold method) (dmx.compressor.modeling.nn.experimental.Conv2dGather method) (dmx.compressor.modeling.nn.experimental.Conv2dUnfold method) (dmx.compressor.modeling.nn.torch_modules.ActActMatMul method) (dmx.compressor.modeling.nn.torch_modules.BAddBMM method) (dmx.compressor.modeling.nn.torch_modules.Dropout method) (dmx.compressor.modeling.nn.torch_modules.Embedding method) (dmx.compressor.modeling.nn.torch_modules.Exp method) (dmx.compressor.modeling.nn.torch_modules.GELUBase method) (dmx.compressor.modeling.nn.torch_modules.GroupNorm method) (dmx.compressor.modeling.nn.torch_modules.LayerNorm method) (dmx.compressor.modeling.nn.torch_modules.Linear method) (dmx.compressor.modeling.nn.torch_modules.Mul method) (dmx.compressor.modeling.nn.torch_modules.ReLU method) (dmx.compressor.modeling.nn.torch_modules.ReLU6 method) (dmx.compressor.modeling.nn.torch_modules.ResAdd method) (dmx.compressor.modeling.nn.torch_modules.RMSNorm method) (dmx.compressor.modeling.nn.torch_modules.ScaledDotProductAttention method) (dmx.compressor.modeling.nn.torch_modules.SiLU method) (dmx.compressor.modeling.nn.torch_modules.Softmax method) (dmx.compressor.modeling.nn.torch_modules.Tanh method) to_fp8_mode() (dmx.compressor.modeling.model.DmxModelMixin method) to_old_forward() (dmx.compressor.modeling.model.DmxModelMixin static method) to_signature_key() (dmx.compressor.modeling.model.DmxModelMixin static method) to_transformed_forward() (dmx.compressor.modeling.model.DmxModelMixin static method) to_yaml() (dmx.compressor.modeling.model.DmxConfig method) TopK (class in dmx.compressor.sparse), [1] TorchFunctionApproximation (class in dmx.compressor.functional.approximate), [1] training (dmx.compressor.modeling.nn.custom_modules.ApplyRotaryPosEmb attribute) (dmx.compressor.modeling.nn.custom_modules.BloomGELU attribute) (dmx.compressor.modeling.nn.custom_modules.ClippedGELU attribute) (dmx.compressor.modeling.nn.custom_modules.FastGELU attribute) (dmx.compressor.modeling.nn.custom_modules.GemmaRMSNorm attribute) (dmx.compressor.modeling.nn.custom_modules.NewGELU attribute) (dmx.compressor.modeling.nn.custom_modules.QuickGELU attribute) (dmx.compressor.modeling.nn.custom_modules.RotaryEmbedding attribute) (dmx.compressor.modeling.nn.experimental.Conv1dScatter attribute) (dmx.compressor.modeling.nn.experimental.Conv1dUnfold attribute) (dmx.compressor.modeling.nn.experimental.Conv2dGather attribute) (dmx.compressor.modeling.nn.experimental.Conv2dUnfold attribute) (dmx.compressor.modeling.nn.torch_modules.ActActMatMul attribute) (dmx.compressor.modeling.nn.torch_modules.AdaptiveAvgPool2d attribute) (dmx.compressor.modeling.nn.torch_modules.AvgPool2d attribute) (dmx.compressor.modeling.nn.torch_modules.BAddBMM attribute) (dmx.compressor.modeling.nn.torch_modules.BatchNorm2d attribute) (dmx.compressor.modeling.nn.torch_modules.Conv1d attribute) (dmx.compressor.modeling.nn.torch_modules.Conv2d attribute) (dmx.compressor.modeling.nn.torch_modules.ConvTranspose2d attribute) (dmx.compressor.modeling.nn.torch_modules.Dropout attribute) (dmx.compressor.modeling.nn.torch_modules.Embedding attribute) (dmx.compressor.modeling.nn.torch_modules.Exp attribute) (dmx.compressor.modeling.nn.torch_modules.GELU attribute) (dmx.compressor.modeling.nn.torch_modules.GELUBase attribute) (dmx.compressor.modeling.nn.torch_modules.GroupNorm attribute) (dmx.compressor.modeling.nn.torch_modules.LayerNorm attribute) (dmx.compressor.modeling.nn.torch_modules.Linear attribute) (dmx.compressor.modeling.nn.torch_modules.MaxPool2d attribute) (dmx.compressor.modeling.nn.torch_modules.Mul attribute) (dmx.compressor.modeling.nn.torch_modules.ReLU attribute) (dmx.compressor.modeling.nn.torch_modules.ReLU6 attribute) (dmx.compressor.modeling.nn.torch_modules.ResAdd attribute) (dmx.compressor.modeling.nn.torch_modules.RMSNorm attribute) (dmx.compressor.modeling.nn.torch_modules.ScaledDotProductAttention attribute) (dmx.compressor.modeling.nn.torch_modules.SiLU attribute) (dmx.compressor.modeling.nn.torch_modules.Softmax attribute) (dmx.compressor.modeling.nn.torch_modules.Tanh attribute)

transform() (dmx.compressor.fx.transformer.configuration_transformer.ConfigurationTransformer method) (dmx.compressor.fx.transformer.nodedict_transformer.NodeDictTransformer method) (dmx.compressor.fx.transformer.qdq_transformer.QdQTransformer method) (dmx.compressor.modeling.model.DmxModelMixin method) (dmx.compressor.modeling.model.DmxPipelineMixin method) (dmx.compressor.modeling.nn.core.DmxModule method) transform_submodule() (in module dmx.compressor.utils.torch_helpers), [1] transformed (dmx.compressor.modeling.model.DmxModelMixin attribute) tuning_approximation_function() (dmx.compressor.layer_reconstruction.LayerReconstructionMixin method)

U

up_proj (dmx.compressor.advanced_recipe.DmxSLaNCHyperparams attribute) update_mask() (dmx.compressor.sparse.Sparsify method)

update_params_with_raw() (dmx.compressor.modeling.nn.core.DmxModule method) update_smoothquant_scale() (dmx.compressor.layer_reconstruction.LayerReconstructionMixin method)

V

v_proj (dmx.compressor.advanced_recipe.DmxSLaNCHyperparams attribute)

VANILLA (dmx.compressor.utils.benchmark.EVALUATION_MODE attribute) visualize_graph() (dmx.compressor.modeling.model.DmxModel method)

W

weight (dmx.compressor.advanced_recipe.DmxModuleQuantizerCalibrationHyperparams attribute) weight_elem_count (dmx.compressor.perf_proxy.PerformanceProxyMixin property) weight_format (dmx.compressor.numerical.cast.NumericalCastMixin property) weight_hypernet (dmx.compressor.modeling.nn.core.DmxModule property) weight_maxabs_computed (dmx.compressor.numerical.smoothquant.ActivationWeightSmoothQuant property) weight_precision (dmx.compressor.numerical.cast.NumericalCastMixin property) weight_scale (dmx.compressor.numerical.cast.NumericalCastMixin property) weight_size_in_bytes (dmx.compressor.perf_proxy.PerformanceProxyMixin property)

weight_sparseness (dmx.compressor.sparse.WeightSparseMixin property) weight_storage (dmx.compressor.advanced_recipe.DmxModuleQuantizerCalibrationHyperparams attribute) weight_storage_format (dmx.compressor.numerical.cast.NumericalCastMixin property) weight_storage_precision (dmx.compressor.numerical.cast.NumericalCastMixin property) weight_storage_scale (dmx.compressor.numerical.cast.NumericalCastMixin property) weight_storage_zero_point (dmx.compressor.numerical.cast.NumericalCastMixin property) weight_zero_point (dmx.compressor.numerical.cast.NumericalCastMixin property) WeightSparseMixin (class in dmx.compressor.sparse), [1]

Z

zero_flop_counter() (dmx.compressor.perf_proxy.PerformanceProxyMixin method)