amphion

/

MaskGCT

like 238

Follow

Amphion 192

I mean load model on different cards,
e.g.
semantic_model, semantic_mean, semantic_std = build_semantic_model(torch.device("cuda:0"))
# 2. build semantic codec
semantic_codec = build_semantic_codec(cfg.model.semantic_codec, torch.device("cuda:1"))
# 3. build acoustic codec
codec_encoder, codec_decoder = build_acoustic_codec(cfg.model.acoustic_codec, torch.device("cuda:2"))
# 4. build t2s model
t2s_model = build_t2s_model(cfg.model.t2s_model, torch.device("cuda:3"))
# 5. build s2a model