*`i2 lddlZddlZddlmZddlmZddlmZdgZej e Z defdZ dS)N)quantize)QuantizationStatus)Modulecompress_quantized_weightsmodulect|dd}|r|jsdSt|dd}|tjurdSt|dd}t|dd}t|dd}t|dd}||tj|_dSd|j_t|||||jtj |j_ tj|_dS) z Quantizes the module weight representation to use fewer bits in memory apply to full model with `model.apply(compress_quantized_weights)` :param module: module to compress to quantized representation quantization_schemeNquantization_statusweight weight_scaleweight_zero_point weight_g_idxF)xscale zero_pointg_idxargsdtype) getattrweightsr COMPRESSEDr r requires_gradrtorchint8data)rschemestatusr rrrs /home/jaya/work/projects/VOICE-AGENT/VIET/agent-env/lib/python3.11/site-packages/compressed_tensors/quantization/lifecycle/compressed.pyrr sV2D 9 9F  V2D 9 9F #... VXt , ,F FND 1 1E!4d;;J FND 1 1E ~&8%B""'FM!  ^j FM"4!>F) loggingr1compressed_tensors.quantization.lifecycle.forwardr,compressed_tensors.quantization.quant_configrtorch.nnr__all__ getLogger__name___LOGGERrrrr)s  FFFFFFKKKKKK!  ' H % %(?v(?(?(?(?(?(?r