o ϯi @s|ddlZddlZddlmZddlmZddlmZmZddlm Z e dZ ddZ d d Z d d Zed krsztokenizer..)tokenizeitems)textresultrrr tokenizer src Cstjrdnd}d}d}d}d}d}d}t|||||||d \}}d d g} t| } ||| } |  } t | t | j dS) Ncuda:0cpufp32 HTSAT-tinyrobertaTaff_2d4/home/la/kechen/Research/KE_CLAP/ckpt/fusion_best.pt precisiondevice enable_fusion fusion_typezI love the contrastive learningzI love the pretrain model) torchcuda is_availablerrevalget_text_embeddingdetachrnumpyprintshape) r"r!amodeltmodelr#r$ pretrainedmodel model_cfg text_data text_embedrrr infer_texts.   r5c Cstjrdnd}d}d}d}d}d}d}t|||||||d \}}tjd d d \} } tt| } t|  } i} t | | d dd|dd} | | | g} | } t| t| jdS)NrrrrrTrrr z9/home/la/kechen/Research/KE_CLAP/ckpt/test_clap_short.wavi)sriSfusion repeatpad audio_cfg)data_truncating data_fillingr9)r%r&r'rlibrosaloadrr from_numpyfloatrr(get_audio_embeddingr*rr+r,r-) r"r!r.r/r#r$r0r1r2audio_waveformr6 audio_dict audio_embedrrr infer_audio/s>   rD__main__)r%r< clap_moduler training.datarrr transformersrfrom_pretrainedrrr5rD__name__rrrrs     +