Download RepCodec/setup.py from m-a-p/xcodec_mini_infer: direct link, hf CLI and curl.
- Browser
- Download file 883 Bytes
-
https://huggingface.co/m-a-p/xcodec_mini_infer/resolve/refs%2Fpr%2F1/RepCodec/setup.py
- Command line
-
hf download hf://m-a-p/xcodec_mini_infer@refs/pr/1/RepCodec/setup.py
-
curl -L -o setup.py https://huggingface.co/m-a-p/xcodec_mini_infer/resolve/refs%2Fpr%2F1/RepCodec/setup.py
883 Bytes
| # Copyright (c) ByteDance, Inc. and its affiliates. | |
| # Copyright (c) Chutong Meng | |
| # | |
| # This source code is licensed under the MIT license found in the | |
| # LICENSE file in the root directory of this source tree. | |
| from setuptools import setup | |
| try: | |
| with open("README.md") as fp: | |
| long_description = fp.read() | |
| except Exception: | |
| long_description = "" | |
| setup( | |
| name="RepCodec", | |
| version="v1.0.0", | |
| description="A Speech Representation Codec for Speech Tokenization", | |
| long_description=long_description, | |
| url="https://github.com/mct10/RepCodec", | |
| packages=["repcodec", "repcodec.modules", "repcodec.layers"], | |
| package_data={ | |
| "repcodec": ["configs/*.yaml"] | |
| }, | |
| install_requires=["numpy", "tqdm", "torch", "tensorboardX", "PyYAML"], | |
| entry_points={ | |
| 'console_scripts': [ | |
| "repcodec=repcodec.tokenize:cli" | |
| ] | |
| } | |
| ) | |