Index

_

__getattr__() (bitorch_engine.utils.safe_import.ExtensionModulePlaceholder method) __init__() (bitorch_engine.layers.qconv.binary.cpp.layer.BinaryConv2dCPP method) (bitorch_engine.layers.qconv.binary.cutlass.layer.BinaryConv2dCutlass method) (bitorch_engine.layers.qconv.binary.layer.BinaryConv2dBase method) (bitorch_engine.layers.qconv.nbit.cutlass.layer.Q4Conv2dCutlass method) (bitorch_engine.layers.qconv.nbit.layer.nBitConv2dBase method) (bitorch_engine.layers.qembedding.binary.layer.BinaryEmbeddingBag method) (bitorch_engine.layers.qembedding.binary.layer.BinaryEmbeddingCuda method) (bitorch_engine.layers.qlinear.binary.cpp.layer.BinaryLinearCPP method) (bitorch_engine.layers.qlinear.binary.cuda.layer.BinaryLinearCuda method) (bitorch_engine.layers.qlinear.binary.cutlass.layer.BinaryLinearCutlass method) (bitorch_engine.layers.qlinear.binary.cutlass.layer.BinaryMatMul method) (bitorch_engine.layers.qlinear.binary.layer.BinaryLinearBase method) (bitorch_engine.layers.qlinear.layer.QLinearInf method) (bitorch_engine.layers.qlinear.nbit.cuda.mbwq_layer.MBWQLinearCuda method) (bitorch_engine.layers.qlinear.nbit.cuda.mpq_layer.MPQLinearCuda method) (bitorch_engine.layers.qlinear.nbit.cutlass.q4_layer.Q4LinearCutlass method) (bitorch_engine.layers.qlinear.nbit.cutlass.q4_layer.Q4MatMul method) (bitorch_engine.layers.qlinear.nbit.cutlass.q8_layer.Q8LinearCutlass method) (bitorch_engine.layers.qlinear.nbit.layer.MPQLinearBase method) (bitorch_engine.layers.qlinear.nbit.layer.MPQWeightParameter method) (bitorch_engine.layers.qlinear.nbit.layer.nBitLinearBase method) (bitorch_engine.layers.qlinear.nbit.mps.mpq_layer.MPQLinearMlx method) (bitorch_engine.layers.qmha.binary.layer.BMHA method) (bitorch_engine.layers.qmha.binary.layer.LearnableBias method) (bitorch_engine.optim.diode_beta.DiodeMix method), [1] (bitorch_engine.optim.galore_projector.GaLoreProjector method) (bitorch_engine.utils.safe_import.ExtensionModulePlaceholder method), [1]

__setattr__() (bitorch_engine.utils.safe_import.ExtensionModulePlaceholder method) _check_forward() (bitorch_engine.layers.qlinear.binary.layer.BinaryLinearBase method) (bitorch_engine.layers.qlinear.nbit.cutlass.q4_layer.Q4LinearCutlass method) (bitorch_engine.layers.qlinear.nbit.cutlass.q8_layer.Q8LinearCutlass method) _name (bitorch_engine.utils.safe_import.ExtensionModulePlaceholder attribute)

A

a_bit (bitorch_engine.layers.qlinear.nbit.cuda.mpq_layer.MPQLinearCuda attribute) (bitorch_engine.layers.qlinear.nbit.layer.MPQLinearBase attribute) (bitorch_engine.layers.qlinear.nbit.layer.nBitLinearBase attribute) (bitorch_engine.layers.qlinear.nbit.mps.mpq_layer.MPQLinearMlx attribute)

active_indices (bitorch_engine.layers.qembedding.binary.layer.BinaryEmbeddingParameter attribute) ADAPTIVE (bitorch_engine.layers.qlinear.binary.cuda.bmm.BMM attribute) ARCH_CPU (class in bitorch_engine.utils.arch_helper) asym (bitorch_engine.layers.qlinear.nbit.layer.MPQLinearBase attribute) (bitorch_engine.layers.qlinear.nbit.layer.MPQWeightParameter attribute)

B

backward (bitorch_engine.layers.qconv.nbit.cutlass.layer.Q4Conv2dCutlassForward attribute) backward() (bitorch_engine.layers.qconv.binary.cutlass.layer.BinaryConv2dForward static method) (bitorch_engine.layers.qconv.nbit.cutlass.layer.Q4Conv2dCutlassForward static method) (bitorch_engine.layers.qembedding.binary.layer.BinaryEmbeddingBagForward static method) (bitorch_engine.layers.qembedding.binary.layer.BinaryEmbeddingForward static method) (bitorch_engine.layers.qlinear.binary.cuda.layer.BinaryLinearForward static method) (bitorch_engine.layers.qlinear.binary.cutlass.layer.BinaryLinearForward static method) (bitorch_engine.layers.qlinear.binary.cutlass.layer.BinaryMatMulFunction static method) (bitorch_engine.layers.qlinear.nbit.cuda.mbwq_layer.MBWQLinearCudaFunction static method) (bitorch_engine.layers.qlinear.nbit.cuda.mpq_layer.MPQLinearCudaFunction static method) (bitorch_engine.layers.qlinear.nbit.cutlass.q4_layer.Q4LinearFunction static method) (bitorch_engine.layers.qlinear.nbit.cutlass.q4_layer.Q4MatMulFunction static method) (bitorch_engine.layers.qlinear.nbit.cutlass.q8_layer.Q8LinearFunction static method) (bitorch_engine.layers.qlinear.nbit.mps.mpq_layer.MPQLinearMlxFunction static method) betas (bitorch_engine.optim.diode_beta.DiodeMix attribute) bias (bitorch_engine.layers.qmha.binary.layer.LearnableBias attribute) bias_a (bitorch_engine.layers.qconv.binary.cutlass.layer.BinaryConv2dCutlass attribute) (bitorch_engine.layers.qconv.nbit.cutlass.layer.Q4Conv2dCutlass attribute) (bitorch_engine.layers.qlinear.binary.cuda.layer.BinaryLinearCuda attribute) (bitorch_engine.layers.qlinear.binary.cutlass.layer.BinaryLinearCutlass attribute) (bitorch_engine.layers.qlinear.nbit.cutlass.q4_layer.Q4LinearCutlass attribute) (bitorch_engine.layers.qlinear.nbit.cutlass.q8_layer.Q8LinearCutlass attribute) binary_matmul_forward_post_processing() (in module bitorch_engine.utils.model_helper) BinaryConv2dBase (class in bitorch_engine.layers.qconv.binary.layer) BinaryConv2dCPP (class in bitorch_engine.layers.qconv.binary.cpp.layer) BinaryConv2dCutlass (class in bitorch_engine.layers.qconv.binary.cutlass.layer) BinaryConv2dForward (class in bitorch_engine.layers.qconv.binary.cpp.layer) (class in bitorch_engine.layers.qconv.binary.cutlass.layer) BinaryConvParameter (class in bitorch_engine.layers.qconv.binary.layer) BinaryEmbeddingBag (class in bitorch_engine.layers.qembedding.binary.layer) BinaryEmbeddingBagForward (class in bitorch_engine.layers.qembedding.binary.layer) BinaryEmbeddingCuda (class in bitorch_engine.layers.qembedding.binary.layer) BinaryEmbeddingForward (class in bitorch_engine.layers.qembedding.binary.layer) BinaryEmbeddingParameter (class in bitorch_engine.layers.qembedding.binary.layer) BinaryLinearBase (class in bitorch_engine.layers.qlinear.binary.layer) BinaryLinearCPP (class in bitorch_engine.layers.qlinear.binary.cpp.layer) BinaryLinearCuda (class in bitorch_engine.layers.qlinear.binary.cuda.layer) BinaryLinearCutlass (class in bitorch_engine.layers.qlinear.binary.cutlass.layer) BinaryLinearForward (class in bitorch_engine.layers.qlinear.binary.cpp.layer) (class in bitorch_engine.layers.qlinear.binary.cuda.layer) (class in bitorch_engine.layers.qlinear.binary.cutlass.layer) BinaryLinearImplementationMixin (class in bitorch_engine.layers.qlinear.binary.binary_implementation) BinaryLinearParameter (class in bitorch_engine.layers.qlinear.binary.layer) BinaryMatMul (class in bitorch_engine.layers.qlinear.binary.cutlass.layer) BinaryMatMulFunction (class in bitorch_engine.layers.qlinear.binary.cutlass.layer) bit_set() (in module bitorch_engine.utils.quant_operators) bitorch_engine module bitorch_engine.functions module bitorch_engine.functions.cuda module bitorch_engine.functions.cuda.extension module bitorch_engine.functions.cuda.functions module bitorch_engine.layers module bitorch_engine.layers.qconv module bitorch_engine.layers.qconv.binary module bitorch_engine.layers.qconv.binary.cpp module bitorch_engine.layers.qconv.binary.cpp.extension module bitorch_engine.layers.qconv.binary.cpp.layer module bitorch_engine.layers.qconv.binary.cutlass module bitorch_engine.layers.qconv.binary.cutlass.extension module bitorch_engine.layers.qconv.binary.cutlass.layer module bitorch_engine.layers.qconv.binary.layer module bitorch_engine.layers.qconv.nbit module bitorch_engine.layers.qconv.nbit.cutlass module bitorch_engine.layers.qconv.nbit.cutlass.extension module bitorch_engine.layers.qconv.nbit.cutlass.layer module bitorch_engine.layers.qconv.nbit.layer module bitorch_engine.layers.qembedding module bitorch_engine.layers.qembedding.binary module bitorch_engine.layers.qembedding.binary.layer module bitorch_engine.layers.qlinear module bitorch_engine.layers.qlinear.binary module bitorch_engine.layers.qlinear.binary.binary_implementation module

bitorch_engine.layers.qlinear.binary.cpp module bitorch_engine.layers.qlinear.binary.cpp.extension module bitorch_engine.layers.qlinear.binary.cpp.layer module bitorch_engine.layers.qlinear.binary.cuda module bitorch_engine.layers.qlinear.binary.cuda.bmm module bitorch_engine.layers.qlinear.binary.cuda.extension module bitorch_engine.layers.qlinear.binary.cuda.layer module bitorch_engine.layers.qlinear.binary.cutlass module bitorch_engine.layers.qlinear.binary.cutlass.extension module bitorch_engine.layers.qlinear.binary.cutlass.layer module bitorch_engine.layers.qlinear.binary.layer module bitorch_engine.layers.qlinear.layer module bitorch_engine.layers.qlinear.nbit module bitorch_engine.layers.qlinear.nbit.cuda module bitorch_engine.layers.qlinear.nbit.cuda.extension module bitorch_engine.layers.qlinear.nbit.cuda.mbwq_layer module bitorch_engine.layers.qlinear.nbit.cuda.mpq_layer module bitorch_engine.layers.qlinear.nbit.cuda.utils module bitorch_engine.layers.qlinear.nbit.cutlass module bitorch_engine.layers.qlinear.nbit.cutlass.extension module bitorch_engine.layers.qlinear.nbit.cutlass.q4_layer module bitorch_engine.layers.qlinear.nbit.cutlass.q8_layer module bitorch_engine.layers.qlinear.nbit.layer module bitorch_engine.layers.qlinear.nbit.mps module bitorch_engine.layers.qlinear.nbit.mps.extension module bitorch_engine.layers.qlinear.nbit.mps.mpq_layer module bitorch_engine.layers.qlinear.qlinear_implementation module bitorch_engine.layers.qmha module bitorch_engine.layers.qmha.binary module bitorch_engine.layers.qmha.binary.layer module bitorch_engine.optim module bitorch_engine.optim.diode_beta module bitorch_engine.optim.galore_projector module bitorch_engine.utils module bitorch_engine.utils.arch_helper module bitorch_engine.utils.convert module bitorch_engine.utils.cpp_extension module bitorch_engine.utils.cuda_extension module bitorch_engine.utils.cutlass_path module bitorch_engine.utils.mlx_extension module bitorch_engine.utils.mlx_path module bitorch_engine.utils.model_helper module bitorch_engine.utils.quant_operators module bitorch_engine.utils.safe_import module bits_binary_word (bitorch_engine.layers.qconv.binary.cpp.layer.BinaryConv2dCPP attribute) (bitorch_engine.layers.qconv.binary.cutlass.layer.BinaryConv2dCutlass attribute) (bitorch_engine.layers.qlinear.binary.cuda.layer.BinaryLinearCuda attribute) (bitorch_engine.layers.qlinear.binary.cutlass.layer.BinaryLinearCutlass attribute) (bitorch_engine.layers.qlinear.binary.layer.BinaryLinearBase attribute) BMHA (class in bitorch_engine.layers.qmha.binary.layer) BMM (class in bitorch_engine.layers.qlinear.binary.cuda.bmm) bmm_type (bitorch_engine.layers.qlinear.binary.cuda.layer.BinaryLinearCuda attribute) (bitorch_engine.layers.qlinear.binary.cuda.layer.BinaryLinearForward attribute) BSTC32 (bitorch_engine.layers.qlinear.binary.cuda.bmm.BMM attribute) BTC32 (bitorch_engine.layers.qlinear.binary.cuda.bmm.BMM attribute)

C

can_clone() (bitorch_engine.layers.qlinear.binary.binary_implementation.BinaryLinearImplementationMixin class method) (bitorch_engine.layers.qlinear.binary.binary_implementation.BinaryLinearImplementationMixin method) (bitorch_engine.layers.qlinear.qlinear_implementation.QLinearImplementationMixin class method) check_cpu_instruction_support() (in module bitorch_engine.utils.arch_helper) check_parameters() (bitorch_engine.layers.qlinear.nbit.cuda.mbwq_layer.MBWQLinearCuda method), [1] (bitorch_engine.layers.qlinear.nbit.cuda.mpq_layer.MPQLinearCuda method), [1] (bitorch_engine.layers.qlinear.nbit.layer.MPQLinearBase method), [1] (bitorch_engine.layers.qlinear.nbit.mps.mpq_layer.MPQLinearMlx method), [1]

check_path() (in module bitorch_engine.utils.cutlass_path) check_pytorch_version() (in module bitorch_engine.optim.diode_beta) collect_layers() (in module bitorch_engine.utils.convert) correct_bias (bitorch_engine.optim.diode_beta.DiodeMix attribute) create_clone_from() (bitorch_engine.layers.qlinear.binary.cpp.layer.BinaryLinearCPP class method) (bitorch_engine.layers.qlinear.binary.cuda.layer.BinaryLinearCuda class method) (bitorch_engine.layers.qlinear.layer.QLinearInf class method) ctx (bitorch_engine.layers.qlinear.binary.cuda.layer.BinaryLinearForward attribute)

D

device (bitorch_engine.layers.qlinear.binary.layer.BinaryLinearBase attribute) (bitorch_engine.layers.qlinear.nbit.cutlass.q4_layer.Q4MatMul attribute) (bitorch_engine.layers.qlinear.nbit.layer.nBitLinearBase attribute) device_id (bitorch_engine.layers.qlinear.binary.cuda.layer.BinaryLinearCuda property) DiodeMix (class in bitorch_engine.optim.diode_beta) disable_bias (bitorch_engine.layers.qlinear.nbit.layer.MPQLinearBase attribute) dq_group_size (bitorch_engine.layers.qlinear.nbit.layer.MPQLinearBase attribute) dq_mode (bitorch_engine.layers.qlinear.nbit.layer.MPQLinearBase attribute)

dropout (bitorch_engine.layers.qmha.binary.layer.BMHA attribute) dtype (bitorch_engine.layers.qlinear.binary.cutlass.layer.BinaryMatMul attribute) (bitorch_engine.layers.qlinear.binary.layer.BinaryLinearBase attribute) (bitorch_engine.layers.qlinear.nbit.cutlass.q4_layer.Q4MatMul attribute) (bitorch_engine.layers.qlinear.nbit.layer.MPQLinearBase attribute) (bitorch_engine.layers.qlinear.nbit.layer.nBitLinearBase attribute) (bitorch_engine.layers.qmha.binary.layer.BMHA attribute) (bitorch_engine.optim.diode_beta.DiodeMix attribute)

E

eps (bitorch_engine.layers.qconv.nbit.cutlass.layer.Q4Conv2dCutlass attribute) (bitorch_engine.layers.qlinear.nbit.cutlass.q4_layer.Q4LinearCutlass attribute) (bitorch_engine.layers.qlinear.nbit.cutlass.q4_layer.Q4MatMul attribute) (bitorch_engine.layers.qlinear.nbit.cutlass.q8_layer.Q8LinearCutlass attribute) (bitorch_engine.optim.diode_beta.DiodeMix attribute)

exl2fp_weight() (bitorch_engine.layers.qlinear.nbit.cuda.mbwq_layer.MBWQLinearCuda method) (bitorch_engine.layers.qlinear.nbit.cuda.mbwq_layer.MBWQLinearCuda static method) ExtensionModulePlaceholder (class in bitorch_engine.utils.safe_import)

F

find_cutlass() (in module bitorch_engine.utils.cutlass_path) flatten_x() (in module bitorch_engine.utils.model_helper) forward (bitorch_engine.layers.qconv.nbit.cutlass.layer.Q4Conv2dCutlassForward attribute) forward() (bitorch_engine.layers.qconv.binary.cpp.layer.BinaryConv2dCPP method) (bitorch_engine.layers.qconv.binary.cpp.layer.BinaryConv2dForward static method) (bitorch_engine.layers.qconv.binary.cutlass.layer.BinaryConv2dCutlass method) (bitorch_engine.layers.qconv.binary.cutlass.layer.BinaryConv2dForward static method) (bitorch_engine.layers.qconv.nbit.cutlass.layer.Q4Conv2dCutlass method) (bitorch_engine.layers.qconv.nbit.cutlass.layer.Q4Conv2dCutlassForward static method) (bitorch_engine.layers.qembedding.binary.layer.BinaryEmbeddingBag method) (bitorch_engine.layers.qembedding.binary.layer.BinaryEmbeddingBagForward static method) (bitorch_engine.layers.qembedding.binary.layer.BinaryEmbeddingCuda method) (bitorch_engine.layers.qembedding.binary.layer.BinaryEmbeddingForward static method) (bitorch_engine.layers.qlinear.binary.cpp.layer.BinaryLinearCPP method) (bitorch_engine.layers.qlinear.binary.cpp.layer.BinaryLinearForward static method) (bitorch_engine.layers.qlinear.binary.cuda.layer.BinaryLinearCuda method) (bitorch_engine.layers.qlinear.binary.cuda.layer.BinaryLinearForward static method) (bitorch_engine.layers.qlinear.binary.cutlass.layer.BinaryLinearCutlass method), [1] (bitorch_engine.layers.qlinear.binary.cutlass.layer.BinaryLinearForward static method) (bitorch_engine.layers.qlinear.binary.cutlass.layer.BinaryMatMul method) (bitorch_engine.layers.qlinear.binary.cutlass.layer.BinaryMatMulFunction static method) (bitorch_engine.layers.qlinear.layer.QLinearInf method) (bitorch_engine.layers.qlinear.nbit.cuda.mbwq_layer.MBWQLinearCuda method), [1] (bitorch_engine.layers.qlinear.nbit.cuda.mbwq_layer.MBWQLinearCudaFunction static method) (bitorch_engine.layers.qlinear.nbit.cuda.mpq_layer.MPQLinearCuda method), [1] (bitorch_engine.layers.qlinear.nbit.cuda.mpq_layer.MPQLinearCudaFunction static method) (bitorch_engine.layers.qlinear.nbit.cutlass.q4_layer.Q4LinearCutlass method), [1] (bitorch_engine.layers.qlinear.nbit.cutlass.q4_layer.Q4LinearFunction static method) (bitorch_engine.layers.qlinear.nbit.cutlass.q4_layer.Q4MatMul method) (bitorch_engine.layers.qlinear.nbit.cutlass.q4_layer.Q4MatMulFunction static method) (bitorch_engine.layers.qlinear.nbit.cutlass.q8_layer.Q8LinearCutlass method), [1] (bitorch_engine.layers.qlinear.nbit.cutlass.q8_layer.Q8LinearFunction static method) (bitorch_engine.layers.qlinear.nbit.mps.mpq_layer.MPQLinearMlx method), [1] (bitorch_engine.layers.qlinear.nbit.mps.mpq_layer.MPQLinearMlxFunction static method) (bitorch_engine.layers.qmha.binary.layer.BMHA method) (bitorch_engine.layers.qmha.binary.layer.LearnableBias method)

fp32toint4() (in module bitorch_engine.functions.cuda.functions)

G

g_idx (bitorch_engine.layers.qlinear.nbit.layer.MPQWeightParameter attribute) GaLoreProjector (class in bitorch_engine.optim.galore_projector) gcc_version() (in module bitorch_engine.utils.cuda_extension) gemm_kernel_id (bitorch_engine.layers.qlinear.binary.cutlass.layer.BinaryLinearCutlass attribute) generate_quantized_weight() (bitorch_engine.layers.qconv.binary.cpp.layer.BinaryConv2dCPP method) (bitorch_engine.layers.qconv.binary.cutlass.layer.BinaryConv2dCutlass method) (bitorch_engine.layers.qconv.binary.layer.BinaryConv2dBase method) (bitorch_engine.layers.qconv.nbit.cutlass.layer.Q4Conv2dCutlass method) (bitorch_engine.layers.qconv.nbit.layer.nBitConv2dBase method) (bitorch_engine.layers.qlinear.binary.cpp.layer.BinaryLinearCPP method) (bitorch_engine.layers.qlinear.binary.cuda.layer.BinaryLinearCuda method) (bitorch_engine.layers.qlinear.binary.cutlass.layer.BinaryLinearCutlass method), [1] (bitorch_engine.layers.qlinear.binary.layer.BinaryLinearBase method), [1] (bitorch_engine.layers.qlinear.layer.QLinearInf method) (bitorch_engine.layers.qlinear.nbit.cutlass.q4_layer.Q4LinearCutlass method), [1] (bitorch_engine.layers.qlinear.nbit.cutlass.q8_layer.Q8LinearCutlass method), [1] (bitorch_engine.layers.qlinear.nbit.layer.MPQLinearBase method), [1] (bitorch_engine.layers.qlinear.nbit.layer.nBitLinearBase method) get_arm_model() (bitorch_engine.utils.arch_helper.linux_arch_ident static method) get_best_binary_implementation() (in module bitorch_engine.layers.qlinear.binary) get_binary_col() (in module bitorch_engine.utils.quant_operators) get_binary_row() (in module bitorch_engine.utils.quant_operators) get_cpp_extension() (in module bitorch_engine.utils.cpp_extension)

get_cuda_arch() (in module bitorch_engine.utils.cuda_extension) get_cuda_extension() (in module bitorch_engine.utils.cuda_extension) get_cutlass_include_path() (in module bitorch_engine.utils.cutlass_path) get_ext() (in module bitorch_engine.functions.cuda.extension) (in module bitorch_engine.layers.qconv.binary.cpp.extension) (in module bitorch_engine.layers.qconv.binary.cutlass.extension) (in module bitorch_engine.layers.qconv.nbit.cutlass.extension) (in module bitorch_engine.layers.qlinear.binary.cpp.extension) (in module bitorch_engine.layers.qlinear.binary.cuda.extension) (in module bitorch_engine.layers.qlinear.binary.cutlass.extension) (in module bitorch_engine.layers.qlinear.nbit.cuda.extension) (in module bitorch_engine.layers.qlinear.nbit.cutlass.extension) (in module bitorch_engine.layers.qlinear.nbit.mps.extension) get_kwargs() (in module bitorch_engine.utils.cpp_extension) (in module bitorch_engine.utils.cuda_extension) get_mlx_extension() (in module bitorch_engine.utils.mlx_extension) get_mlx_include_path() (in module bitorch_engine.utils.mlx_path) get_mlx_lib_path() (in module bitorch_engine.utils.mlx_path) get_mpq_config() (in module bitorch_engine.utils.convert) gptq_style_unpacking() (in module bitorch_engine.utils.quant_operators) gptq_style_zeros_packing() (in module bitorch_engine.utils.quant_operators) group_size (bitorch_engine.layers.qlinear.nbit.layer.MPQLinearBase attribute) (bitorch_engine.layers.qlinear.nbit.layer.MPQWeightParameter attribute)

H

head_dim (bitorch_engine.layers.qmha.binary.layer.BMHA attribute)

hidden_dim (bitorch_engine.layers.qmha.binary.layer.BMHA attribute)

I

import_extension() (in module bitorch_engine.utils.safe_import) in_channels (bitorch_engine.layers.qlinear.nbit.layer.MPQLinearBase attribute) (bitorch_engine.layers.qlinear.nbit.layer.nBitLinearBase attribute) init_gba() (bitorch_engine.layers.qlinear.nbit.layer.MPQLinearBase method), [1] init_gptq() (bitorch_engine.layers.qlinear.nbit.layer.MPQLinearBase method), [1] init_weight() (bitorch_engine.layers.qembedding.binary.layer.BinaryEmbeddingCuda method) (in module bitorch_engine.utils.model_helper)

initialize() (bitorch_engine.layers.qlinear.nbit.layer.MPQLinearBase method), [1] input (bitorch_engine.layers.qlinear.binary.cuda.layer.BinaryLinearForward attribute) input_dim (bitorch_engine.layers.qmha.binary.layer.BMHA attribute) input_features (bitorch_engine.layers.qlinear.binary.layer.BinaryLinearBase attribute) is_arm() (bitorch_engine.utils.arch_helper.linux_arch_ident static method) is_cutlass_available() (in module bitorch_engine.utils.cutlass_path) is_mlx_available() (in module bitorch_engine.utils.mlx_path) is_train (bitorch_engine.layers.qlinear.binary.cuda.layer.BinaryLinearForward attribute)

K

k_linear (bitorch_engine.layers.qmha.binary.layer.BMHA attribute)

L

layer_type (bitorch_engine.layers.qlinear.nbit.layer.MPQWeightParameter attribute) LearnableBias (class in bitorch_engine.layers.qmha.binary.layer) linux_arch_ident (class in bitorch_engine.utils.arch_helper)

load_checkpoint() (in module bitorch_engine.utils.model_helper) load_state_dict() (bitorch_engine.layers.qlinear.nbit.cuda.mbwq_layer.MBWQLinearCuda method), [1] lr (bitorch_engine.optim.diode_beta.DiodeMix attribute)

M

make_group_map() (in module bitorch_engine.layers.qlinear.nbit.cuda.utils) MBWQLinearCuda (class in bitorch_engine.layers.qlinear.nbit.cuda.mbwq_layer) MBWQLinearCudaFunction (class in bitorch_engine.layers.qlinear.nbit.cuda.mbwq_layer) module bitorch_engine bitorch_engine.functions bitorch_engine.functions.cuda bitorch_engine.functions.cuda.extension bitorch_engine.functions.cuda.functions bitorch_engine.layers bitorch_engine.layers.qconv bitorch_engine.layers.qconv.binary bitorch_engine.layers.qconv.binary.cpp bitorch_engine.layers.qconv.binary.cpp.extension bitorch_engine.layers.qconv.binary.cpp.layer bitorch_engine.layers.qconv.binary.cutlass bitorch_engine.layers.qconv.binary.cutlass.extension bitorch_engine.layers.qconv.binary.cutlass.layer bitorch_engine.layers.qconv.binary.layer bitorch_engine.layers.qconv.nbit bitorch_engine.layers.qconv.nbit.cutlass bitorch_engine.layers.qconv.nbit.cutlass.extension bitorch_engine.layers.qconv.nbit.cutlass.layer bitorch_engine.layers.qconv.nbit.layer bitorch_engine.layers.qembedding bitorch_engine.layers.qembedding.binary bitorch_engine.layers.qembedding.binary.layer bitorch_engine.layers.qlinear bitorch_engine.layers.qlinear.binary bitorch_engine.layers.qlinear.binary.binary_implementation bitorch_engine.layers.qlinear.binary.cpp bitorch_engine.layers.qlinear.binary.cpp.extension bitorch_engine.layers.qlinear.binary.cpp.layer bitorch_engine.layers.qlinear.binary.cuda bitorch_engine.layers.qlinear.binary.cuda.bmm bitorch_engine.layers.qlinear.binary.cuda.extension bitorch_engine.layers.qlinear.binary.cuda.layer bitorch_engine.layers.qlinear.binary.cutlass bitorch_engine.layers.qlinear.binary.cutlass.extension bitorch_engine.layers.qlinear.binary.cutlass.layer bitorch_engine.layers.qlinear.binary.layer bitorch_engine.layers.qlinear.layer bitorch_engine.layers.qlinear.nbit bitorch_engine.layers.qlinear.nbit.cuda bitorch_engine.layers.qlinear.nbit.cuda.extension bitorch_engine.layers.qlinear.nbit.cuda.mbwq_layer bitorch_engine.layers.qlinear.nbit.cuda.mpq_layer bitorch_engine.layers.qlinear.nbit.cuda.utils bitorch_engine.layers.qlinear.nbit.cutlass bitorch_engine.layers.qlinear.nbit.cutlass.extension bitorch_engine.layers.qlinear.nbit.cutlass.q4_layer bitorch_engine.layers.qlinear.nbit.cutlass.q8_layer bitorch_engine.layers.qlinear.nbit.layer bitorch_engine.layers.qlinear.nbit.mps bitorch_engine.layers.qlinear.nbit.mps.extension bitorch_engine.layers.qlinear.nbit.mps.mpq_layer bitorch_engine.layers.qlinear.qlinear_implementation bitorch_engine.layers.qmha bitorch_engine.layers.qmha.binary bitorch_engine.layers.qmha.binary.layer bitorch_engine.optim bitorch_engine.optim.diode_beta bitorch_engine.optim.galore_projector bitorch_engine.utils bitorch_engine.utils.arch_helper bitorch_engine.utils.convert bitorch_engine.utils.cpp_extension bitorch_engine.utils.cuda_extension bitorch_engine.utils.cutlass_path bitorch_engine.utils.mlx_extension bitorch_engine.utils.mlx_path bitorch_engine.utils.model_helper bitorch_engine.utils.quant_operators bitorch_engine.utils.safe_import

MPQLinearBase (class in bitorch_engine.layers.qlinear.nbit.layer) MPQLinearCuda (class in bitorch_engine.layers.qlinear.nbit.cuda.mpq_layer) MPQLinearCudaFunction (class in bitorch_engine.layers.qlinear.nbit.cuda.mpq_layer) MPQLinearMlx (class in bitorch_engine.layers.qlinear.nbit.mps.mpq_layer) MPQLinearMlxFunction (class in bitorch_engine.layers.qlinear.nbit.mps.mpq_layer) MPQWeightParameter (class in bitorch_engine.layers.qlinear.nbit.layer)

N

nBitConv2dBase (class in bitorch_engine.layers.qconv.nbit.layer) nBitConvParameter (class in bitorch_engine.layers.qconv.nbit.layer) nBitLinearBase (class in bitorch_engine.layers.qlinear.nbit.layer)

nBitLinearParameter (class in bitorch_engine.layers.qlinear.nbit.layer) num_heads (bitorch_engine.layers.qmha.binary.layer.BMHA attribute) nv_tensor_quant() (in module bitorch_engine.utils.quant_operators)

O

opt_weight (bitorch_engine.layers.qconv.binary.layer.BinaryConv2dBase property) (bitorch_engine.layers.qconv.nbit.layer.nBitConv2dBase property) (bitorch_engine.layers.qlinear.binary.layer.BinaryLinearBase property) (bitorch_engine.layers.qlinear.layer.QLinearInf property) (bitorch_engine.layers.qlinear.nbit.layer.nBitLinearBase property)

opt_weight() (bitorch_engine.layers.qlinear.binary.layer.BinaryLinearBase method) out (bitorch_engine.layers.qmha.binary.layer.BMHA attribute) out_channels (bitorch_engine.layers.qlinear.nbit.layer.MPQLinearBase attribute) (bitorch_engine.layers.qlinear.nbit.layer.nBitLinearBase attribute) output_features (bitorch_engine.layers.qlinear.binary.layer.BinaryLinearBase attribute)

P

pack_bie_layers() (in module bitorch_engine.utils.model_helper) pack_fp_weight() (in module bitorch_engine.layers.qlinear.nbit.cuda.utils) pad_embedding_dim() (in module bitorch_engine.utils.model_helper) pad_last_2_dims_to_multiple_of_128() (in module bitorch_engine.utils.model_helper) params (bitorch_engine.optim.diode_beta.DiodeMix attribute) prepare_bie_layers() (in module bitorch_engine.utils.model_helper) prepare_params() (bitorch_engine.layers.qconv.binary.cpp.layer.BinaryConv2dCPP method) (bitorch_engine.layers.qconv.binary.cutlass.layer.BinaryConv2dCutlass method) (bitorch_engine.layers.qconv.binary.layer.BinaryConv2dBase method) (bitorch_engine.layers.qconv.nbit.cutlass.layer.Q4Conv2dCutlass method) (bitorch_engine.layers.qconv.nbit.layer.nBitConv2dBase method) (bitorch_engine.layers.qembedding.binary.layer.BinaryEmbeddingCuda method) (bitorch_engine.layers.qlinear.binary.cpp.layer.BinaryLinearCPP method) (bitorch_engine.layers.qlinear.binary.cuda.layer.BinaryLinearCuda method) (bitorch_engine.layers.qlinear.binary.cutlass.layer.BinaryLinearCutlass method), [1] (bitorch_engine.layers.qlinear.binary.layer.BinaryLinearBase method) (bitorch_engine.layers.qlinear.layer.QLinearInf method) (bitorch_engine.layers.qlinear.nbit.cuda.mbwq_layer.MBWQLinearCuda method), [1] (bitorch_engine.layers.qlinear.nbit.cuda.mpq_layer.MPQLinearCuda method), [1] (bitorch_engine.layers.qlinear.nbit.cutlass.q4_layer.Q4LinearCutlass method), [1] (bitorch_engine.layers.qlinear.nbit.cutlass.q8_layer.Q8LinearCutlass method), [1] (bitorch_engine.layers.qlinear.nbit.layer.MPQLinearBase method), [1] (bitorch_engine.layers.qlinear.nbit.layer.nBitLinearBase method) (bitorch_engine.layers.qlinear.nbit.mps.mpq_layer.MPQLinearMlx method), [1]

privileged_grad (bitorch_engine.layers.qlinear.nbit.layer.MPQWeightParameter attribute)

Q

q42fp_weight() (bitorch_engine.layers.qlinear.nbit.cuda.mbwq_layer.MBWQLinearCuda method) (bitorch_engine.layers.qlinear.nbit.cuda.mbwq_layer.MBWQLinearCuda static method) q4_pack_tensor() (in module bitorch_engine.functions.cuda.functions) q4_quantization() (in module bitorch_engine.utils.quant_operators) q4_unpack_and_scaling_tensor() (in module bitorch_engine.functions.cuda.functions) q4_unpack_tensor() (in module bitorch_engine.functions.cuda.functions) Q4Conv2dCutlass (class in bitorch_engine.layers.qconv.nbit.cutlass.layer) Q4Conv2dCutlassForward (class in bitorch_engine.layers.qconv.nbit.cutlass.layer) Q4LinearCutlass (class in bitorch_engine.layers.qlinear.nbit.cutlass.q4_layer) Q4LinearFunction (class in bitorch_engine.layers.qlinear.nbit.cutlass.q4_layer) Q4MatMul (class in bitorch_engine.layers.qlinear.nbit.cutlass.q4_layer) Q4MatMulFunction (class in bitorch_engine.layers.qlinear.nbit.cutlass.q4_layer) q8_quantization() (in module bitorch_engine.utils.quant_operators) Q8LinearCutlass (class in bitorch_engine.layers.qlinear.nbit.cutlass.q8_layer)

Q8LinearFunction (class in bitorch_engine.layers.qlinear.nbit.cutlass.q8_layer) q_group_map (bitorch_engine.layers.qlinear.nbit.layer.MPQWeightParameter attribute) q_linear (bitorch_engine.layers.qmha.binary.layer.BMHA attribute) q_perm (bitorch_engine.layers.qlinear.nbit.layer.MPQWeightParameter attribute) QLinearImplementationMixin (class in bitorch_engine.layers.qlinear.qlinear_implementation) QLinearInf (class in bitorch_engine.layers.qlinear.layer) quantize_linear_with_binary_linear_cuda() (in module bitorch_engine.utils.convert) quantize_linear_with_mpq_linear_cuda() (in module bitorch_engine.utils.convert) quantize_linear_with_q4_linear_cutlass() (in module bitorch_engine.utils.convert) qweight (bitorch_engine.layers.qembedding.binary.layer.BinaryEmbeddingCuda attribute) (bitorch_engine.layers.qlinear.binary.layer.BinaryLinearBase attribute) (bitorch_engine.layers.qlinear.nbit.cuda.mbwq_layer.MBWQLinearCuda attribute) (bitorch_engine.layers.qlinear.nbit.cuda.mpq_layer.MPQLinearCuda attribute) (bitorch_engine.layers.qlinear.nbit.mps.mpq_layer.MPQLinearMlx attribute) qweight_update_fn() (in module bitorch_engine.utils.model_helper)

R

replace_layers() (in module bitorch_engine.utils.convert) reset_parameters() (bitorch_engine.layers.qconv.binary.layer.BinaryConv2dBase method) (bitorch_engine.layers.qconv.nbit.layer.nBitConv2dBase method) (bitorch_engine.layers.qembedding.binary.layer.BinaryEmbeddingBag method) (bitorch_engine.layers.qembedding.binary.layer.BinaryEmbeddingCuda method) (bitorch_engine.layers.qlinear.binary.layer.BinaryLinearBase method), [1] (bitorch_engine.layers.qlinear.nbit.layer.nBitLinearBase method)

rows (bitorch_engine.layers.qlinear.nbit.cuda.mbwq_layer.MBWQLinearCuda attribute) (bitorch_engine.layers.qlinear.nbit.layer.MPQWeightParameter attribute)

S

save_checkpoint() (in module bitorch_engine.utils.model_helper) scale_a (bitorch_engine.layers.qconv.binary.cutlass.layer.BinaryConv2dCutlass attribute) (bitorch_engine.layers.qconv.nbit.cutlass.layer.Q4Conv2dCutlass attribute) (bitorch_engine.layers.qlinear.binary.cuda.layer.BinaryLinearCuda attribute) (bitorch_engine.layers.qlinear.binary.cuda.layer.BinaryLinearForward attribute) (bitorch_engine.layers.qlinear.binary.cutlass.layer.BinaryLinearCutlass attribute) (bitorch_engine.layers.qlinear.nbit.cutlass.q4_layer.Q4LinearCutlass attribute) (bitorch_engine.layers.qlinear.nbit.cutlass.q8_layer.Q8LinearCutlass attribute) scale_w (bitorch_engine.layers.qconv.binary.cutlass.layer.BinaryConv2dCutlass attribute) (bitorch_engine.layers.qconv.nbit.cutlass.layer.Q4Conv2dCutlass attribute) (bitorch_engine.layers.qembedding.binary.layer.BinaryEmbeddingCuda attribute) (bitorch_engine.layers.qlinear.binary.cuda.layer.BinaryLinearCuda attribute) (bitorch_engine.layers.qlinear.binary.cuda.layer.BinaryLinearForward attribute) (bitorch_engine.layers.qlinear.binary.cutlass.layer.BinaryLinearCutlass attribute) (bitorch_engine.layers.qlinear.nbit.cutlass.q4_layer.Q4LinearCutlass attribute) (bitorch_engine.layers.qlinear.nbit.cutlass.q8_layer.Q8LinearCutlass attribute) scales (bitorch_engine.layers.qlinear.nbit.cuda.mbwq_layer.MBWQLinearCuda attribute) (bitorch_engine.layers.qlinear.nbit.cuda.mpq_layer.MPQLinearCuda attribute) (bitorch_engine.layers.qlinear.nbit.mps.mpq_layer.MPQLinearMlx attribute) select_gemm_kernel() (bitorch_engine.layers.qlinear.binary.cutlass.layer.BinaryLinearCutlass method), [1] set_activation() (bitorch_engine.layers.qconv.binary.cutlass.layer.BinaryConv2dCutlass method) (bitorch_engine.layers.qconv.nbit.cutlass.layer.Q4Conv2dCutlass method) (bitorch_engine.layers.qlinear.binary.cuda.layer.BinaryLinearCuda method) (bitorch_engine.layers.qlinear.binary.cutlass.layer.BinaryLinearCutlass method), [1] (bitorch_engine.layers.qlinear.nbit.cutlass.q4_layer.Q4LinearCutlass method), [1] (bitorch_engine.layers.qlinear.nbit.cutlass.q8_layer.Q8LinearCutlass method), [1]

set_activation_scale() (bitorch_engine.layers.qlinear.binary.cutlass.layer.BinaryMatMul method) (bitorch_engine.layers.qlinear.nbit.cutlass.q4_layer.Q4MatMul method) set_bits_binary_word() (bitorch_engine.layers.qconv.binary.layer.BinaryConv2dBase method) (bitorch_engine.layers.qlinear.binary.layer.BinaryLinearBase method), [1] set_quantized_weight_data() (bitorch_engine.layers.qconv.binary.layer.BinaryConv2dBase method) (bitorch_engine.layers.qconv.nbit.layer.nBitConv2dBase method) (bitorch_engine.layers.qlinear.binary.layer.BinaryLinearBase method), [1] (bitorch_engine.layers.qlinear.layer.QLinearInf method) (bitorch_engine.layers.qlinear.nbit.layer.nBitLinearBase method) set_qweight_data() (bitorch_engine.layers.qlinear.nbit.layer.MPQLinearBase method), [1] set_scales() (bitorch_engine.layers.qlinear.nbit.cuda.mbwq_layer.MBWQLinearCuda method), [1] set_weight_data() (bitorch_engine.layers.qconv.binary.cutlass.layer.BinaryConv2dCutlass method) (bitorch_engine.layers.qconv.binary.layer.BinaryConv2dBase method) (bitorch_engine.layers.qconv.nbit.layer.nBitConv2dBase method) (bitorch_engine.layers.qlinear.binary.cuda.layer.BinaryLinearCuda method) (bitorch_engine.layers.qlinear.binary.cutlass.layer.BinaryLinearCutlass method), [1] (bitorch_engine.layers.qlinear.binary.layer.BinaryLinearBase method), [1] (bitorch_engine.layers.qlinear.layer.QLinearInf method) (bitorch_engine.layers.qlinear.nbit.layer.nBitLinearBase method) set_zeros() (bitorch_engine.layers.qlinear.nbit.cuda.mbwq_layer.MBWQLinearCuda method), [1] step() (bitorch_engine.optim.diode_beta.DiodeMix method), [1] symmetric (bitorch_engine.layers.qlinear.binary.layer.BinaryLinearBase attribute)

T

tensor_to_packed_uint8() (in module bitorch_engine.functions.cuda.functions)

U

unflatten_x() (in module bitorch_engine.utils.model_helper) unpack_qweight() (in module bitorch_engine.layers.qlinear.nbit.cuda.utils) unpack_uint8_tensor() (in module bitorch_engine.functions.cuda.functions) update() (bitorch_engine.layers.qconv.binary.layer.BinaryConvParameter static method) (bitorch_engine.layers.qconv.nbit.layer.nBitConvParameter static method) (bitorch_engine.layers.qembedding.binary.layer.BinaryEmbeddingParameter static method) (bitorch_engine.layers.qlinear.binary.layer.BinaryLinearParameter static method) (bitorch_engine.layers.qlinear.nbit.layer.MPQWeightParameter static method) (bitorch_engine.layers.qlinear.nbit.layer.nBitLinearParameter static method)

update_zeros() (in module bitorch_engine.utils.model_helper) use_gba_quant (bitorch_engine.layers.qlinear.nbit.layer.MPQLinearBase attribute) use_mbw (bitorch_engine.layers.qlinear.nbit.cuda.mbwq_layer.MBWQLinearCuda attribute)

V

v_linear (bitorch_engine.layers.qmha.binary.layer.BMHA attribute)

W

w_bit (bitorch_engine.layers.qlinear.nbit.cuda.mpq_layer.MPQLinearCuda attribute) (bitorch_engine.layers.qlinear.nbit.layer.MPQLinearBase attribute) (bitorch_engine.layers.qlinear.nbit.layer.MPQWeightParameter attribute) (bitorch_engine.layers.qlinear.nbit.layer.nBitLinearBase attribute) (bitorch_engine.layers.qlinear.nbit.mps.mpq_layer.MPQLinearMlx attribute)

w_pack() (bitorch_engine.layers.qlinear.binary.cuda.layer.BinaryLinearCuda static method) weight (bitorch_engine.layers.qembedding.binary.layer.BinaryEmbeddingCuda attribute) (bitorch_engine.layers.qlinear.binary.cuda.layer.BinaryLinearForward attribute) (bitorch_engine.layers.qlinear.binary.layer.BinaryLinearBase attribute) (bitorch_engine.layers.qlinear.layer.QLinearInf property) weight_decay (bitorch_engine.optim.diode_beta.DiodeMix attribute)

X

x_clip (bitorch_engine.layers.qlinear.binary.cutlass.layer.BinaryMatMul attribute) (bitorch_engine.layers.qlinear.nbit.cutlass.q4_layer.Q4MatMul attribute)

Y

y_clip (bitorch_engine.layers.qlinear.binary.cutlass.layer.BinaryMatMul attribute) (bitorch_engine.layers.qlinear.nbit.cutlass.q4_layer.Q4MatMul attribute)

Z

zeros (bitorch_engine.layers.qlinear.nbit.cuda.mbwq_layer.MBWQLinearCuda attribute) (bitorch_engine.layers.qlinear.nbit.cuda.mpq_layer.MPQLinearCuda attribute) (bitorch_engine.layers.qlinear.nbit.mps.mpq_layer.MPQLinearMlx attribute)