search

haoheliu / SemantiCodec-inference

Ultra-low bitrate neural audio codec (0.31~1.40 kbps) with a better semantic in the latent space.
MIT License
149 stars 9 forks source link

Do you provide checkpoints? #3

Open TeresaTsang opened 4 months ago

TeresaTsang commented 4 months ago

Do you provide checkpoints?

yaotingwangofficial commented 4 months ago

Try pip install git+https://github.com/haoheliu/SemantiCodec-inference.git and it will download the ckpt automatically.

haoheliu commented 4 months ago

@TeresaTsang Checkpoint will be automatically downloaded when you run the code. Thanks!

cantabile-kwok commented 1 month ago

Tip for those whose computing server does not have a huggingface network connection: checkpoints can be manually downloaded here https://huggingface.co/haoheliu/SemantiCodec/tree/main

wincing2 commented 3 weeks ago

Tip for those whose computing server does not have a huggingface network connection: checkpoints can be manually downloaded here https://huggingface.co/haoheliu/SemantiCodec/tree/main

Could you tell me the exact location to put the checkpoints after download?

cantabile-kwok commented 3 weeks ago

@wincing2 It is indeed a little bit tricky. You can refer to my location of saving:

image

Name this directory "pretrained", and you need to specify the loading path in the inference code like:

semanticodec = SemantiCodec(token_rate=token_rate, semantic_vocab_size=vocab_size, 
               checkpoint_path=f"pretrained/semanticodec_tokenrate_100")