pymllm.quantize.solver

Classes

QuantizeSolver

Module Contents

class pymllm.quantize.solver.QuantizeSolver

passes: List[pymllm.quantize.quantize_pass.QuantizeBasePass] = []

register_pass(pass_)

Parameters:

pass_ (pymllm.quantize.quantize_pass.QuantizeBasePass)

stream_quantize_params_size(quant_cfg, tensor_dict, **kwargs)

Parameters:

tensor_dict (Dict)

Return type:

int

stream_quantize(quant_cfg, tensor_dict, writer, **kwargs)

Parameters:

• tensor_dict (Dict)

• writer (pymllm.convertor.ModelFileV2)

Return type:

bool