runner
======

.. py:module:: runner


Classes
-------

.. autoapisummary::

   runner.Qwen3Quantizer


Functions
---------

.. autoapisummary::

   runner.freeze_qwen3_rmsnorm_weight
   runner.freeze_qwen3_linear_weight
   runner.disable_qdq_observer
   runner.enable_qdq_observer
   runner.convert_weight


Module Contents
---------------

.. py:function:: freeze_qwen3_rmsnorm_weight(m)

.. py:function:: freeze_qwen3_linear_weight(m)

.. py:function:: disable_qdq_observer(m)

.. py:function:: enable_qdq_observer(m)

.. py:function:: convert_weight(m)

.. py:class:: Qwen3Quantizer(model_path, mllm_qualcomm_max_length=2048)

   .. py:attribute:: tokenizer


   .. py:attribute:: model


   .. py:attribute:: mllm_qualcomm_max_length
      :value: 2048



   .. py:method:: freeze_activation()


   .. py:method:: enable_activation_update()


   .. py:method:: compile()


   .. py:method:: infer(prompt)


   .. py:method:: calibrate(num_samples=64, max_seq_length=512)

      Perform calibration using Wikipedia dataset (PTQ)
      :param num_samples: Number of samples for calibration
      :param max_seq_length: Maximum length for each sample (not exceeding mllm_qualcomm_max_length)



   .. py:method:: convert()