AIMET API

ONNX

aimet_onnx

AIMET quantization for ONNX models provides the following functionality.

• aimet_onnx.quantsim

• aimet_onnx.apply_adaround

• aimet_onnx.apply_seq_mse

• aimet_onnx.quantsim.set_grouped_blockwise_quantization_for_weights

• aimet_onnx.batch_norm_fold

• aimet_onnx.cross_layer_equalization

• aimet_onnx.lite_mp

• aimet_onnx.mixed_precision

• aimet_onnx.quant_analyzer

• aimet_onnx.layer_output_utils

PyTorch

aimet_torch

Important

aimet_torch package has been upgraded to aimet_torch.v2.

aimet_torch 2 is fully backward-compatible with all the public APIs of aimet_torch 1.x. See the Migration guide.

AIMET quantization for PyTorch models provides the following functionality.

• aimet_torch.quantsim

• aimet_torch.nn

• aimet_torch.quantization

• aimet_torch.onnx

• aimet_torch.utils

• aimet_torch.adaround

• aimet_torch.seq_mse

• aimet_torch.experimental.adascale

• aimet_torch.quantsim.config_utils

• aimet_torch.batch_norm_fold

• aimet_torch.cross_layer_equalization

• aimet_torch.model_preparer

• aimet_torch.model_validator

• aimet_torch.mixed_precision

• aimet_torch.quant_analyzer

• aimet_torch.autoquant

• aimet_torch.bn_reestimation

• aimet_torch.visualization_tools

• aimet_torch.layer_output_utils

• aimet_torch.peft

• aimet_torch.compress

• aimet_torch.experimental.omniquant

• aimet_torch.experimental.spinquant

aimet_torch.v1

• aimet_torch.v1.quantsim

• aimet_torch.v1.adaround

• aimet_torch.v1.seq_mse

• aimet_torch.v1.quant_analyzer

• aimet_torch.v1.autoquant

• aimet_torch.v1.amp