runner¶

Classes¶

Qwen3Quantizer

Functions¶

`freeze_qwen3_rmsnorm_weight`(m)
`freeze_qwen3_linear_weight`(m)
`disable_qdq_observer`(m)
`enable_qdq_observer`(m)
`convert_weight`(m)

Module Contents¶

runner.freeze_qwen3_rmsnorm_weight(m)¶

runner.freeze_qwen3_linear_weight(m)¶

runner.disable_qdq_observer(m)¶

runner.enable_qdq_observer(m)¶

runner.convert_weight(m)¶

class runner.Qwen3Quantizer(model_path, mllm_qualcomm_max_length=2048)¶

Parameters:: model_path (str)

tokenizer¶

model¶

mllm_qualcomm_max_length = 2048¶

freeze_activation()¶

enable_activation_update()¶

compile()¶

infer(prompt)¶

Parameters:: prompt (str)

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)

convert()¶