Closed KonstantineGoudz closed 10 months ago
Resolved an issue when attempting to use cuda to generate text audio data sets
this will not cause any issues now
from audiolm_pytorch import HubertWithKmeans, data from spear_tts_pytorch import ( TextToSemantic, SemanticToTextDatasetGenerator, ) ds = data.SoundDataset( folder="path-to-audio-files", target_sample_hz=44100 ) output_folder = "generated_audio_text_dataset" wav2vec = HubertWithKmeans( checkpoint_path = './hubert_base_ls960.pt', kmeans_path = './hubert_base_ls960_L9_km500.bin' ) model = TextToSemantic( wav2vec = wav2vec, dim = 512, heads = 8, use_openai_tokenizer=True, target_kv_heads = 2, # grouped query attention, for memory efficient decoding source_depth = 1, target_depth = 1, ) model.cuda() ds_generator = SemanticToTextDatasetGenerator( model=model, dataset=ds, folder=output_folder, ) ds_generator.forward()
@KonstantineGoudz thank you!
Resolved an issue when attempting to use cuda to generate text audio data sets
this will not cause any issues now