MCT Quantizers Nightly

1.6.0.20250912.post11759 verified Sat May 09 auth: no python

Nightly (pre-release) builds of MCT Quantizers, providing infrastructure for neural network compression via quantization. Current version is 1.6.0.20250912.post11759. Released nightly, not for production use.

pip install mct-quantizers-nightly

Common errors

error ModuleNotFoundError: No module named 'mct_quantizers' ↓

cause Nightly package is named 'mct-quantizers-nightly' on PyPI, but import uses 'mct_quantizers'.

fix

Install the package with 'pip install mct-quantizers-nightly' and import as 'mct_quantizers'.

error AttributeError: module 'mct_quantizers' has no attribute 'QuantizationConfig' ↓

cause Incorrect import from a submodule that doesn't exist in nightly version.

fix

Use 'from mct_quantizers import QuantizationConfig' instead of from a submodule.

Warnings

deprecated Nightly builds may have APIs that are removed in the future. Do not rely on stable interfaces. ↓

fix Use the stable release mct-quantizers for production.

gotcha Import paths differ between nightly and stable releases. Check documentation before upgrading. ↓

fix Always import from the top-level 'mct_quantizers' package.

Imports

QuantizationConfig

wrong

from mct_quantizers.quantization import QuantizationConfig

correct

from mct_quantizers import QuantizationConfig

Wrong submodule path; top-level import is correct.

QuantizedLayer

wrong

from mct_quantizers.layers import QuantizedLayer

correct

from mct_quantizers import QuantizedLayer

QuantizedLayer is at package root, not in layers submodule.

Quickstart

Import the library and configure basic quantization parameters.

import mct_quantizers
from mct_quantizers import QuantizationConfig

# Example: set quantization configuration
config = QuantizationConfig(n_weights_bits=8, n_activation_bits=8)
print(config)