bitorch_engine.utils.quant_operators

Functions

bit_set	Sets a specific bit in an integer variable to a given value.
get_binary_col	Binarizes an array column-wise, transforming each element into a binary representation.
get_binary_row	Binarizes an input NDArray (nd_row) into a binary representation (binary_row) based on the specified number of bits per binary word (bits_per_binary_word).
gptq_style_unpacking	Reconstructs the fp16 weight tensor from the input quantized weight parameter in GPTQ style.
gptq_style_zeros_packing	Packs the zeros tensor in GPTQ style for efficient storage and computation.
nv_tensor_quant	Quantizes the given tensor using specified quantization parameters.
q4_quantization	Quantizes an input tensor to 4-bit integers using uniform quantization.
q8_quantization	Quantizes an input tensor to 8-bit integers using uniform quantization.