Skip to content

Actions: duj12/CosyVoice

Actions

Lint

Actions

Loading...
Loading

Show workflow options

Create status badge

Loading
lint.yml
28 workflow runs
28 workflow runs

Filter by Event

Filter by Status

Filter by Branch

Filter by Actor

修改验证集Dataset，避免训练中验证集ddp stuck导致异常退出. Lint #28: Commit 2ae94d2 pushed by duj12
December 27, 2024 07:24 16s dj
dj
December 27, 2024 07:24 16s
预留s3tokenizer-v2的代码。 Lint #27: Commit 130a472 pushed by duj12
December 24, 2024 06:57 18s dj
dj
December 24, 2024 06:57 18s
添加tts数据微调训练配置。 Lint #26: Commit ed6057f pushed by duj12
December 24, 2024 04:23 17s dj
dj
December 24, 2024 04:23 17s
每此evaluate之前先保存一下模型，evaluate完之后，更新一下yaml即可。避免evaluate异常退出导致模型ckpt丢失。 Lint #25: Commit f2e4019 pushed by duj12
December 24, 2024 04:00 16s dj
dj
December 24, 2024 04:00 16s
添加大数据集hifigan训练配置。 Lint #24: Commit 690602e pushed by duj12
December 18, 2024 06:43 15s dj
dj
December 18, 2024 06:43 15s
更新lora Lint #23: Commit 9a92d12 pushed by duj12
December 17, 2024 08:46 19s dj
dj
December 17, 2024 08:46 19s
添加lora代码 Lint #22: Commit d794d04 pushed by duj12
December 17, 2024 03:52 17s dj
dj
December 17, 2024 03:52 17s
更新flow大数据集训练配置。 Lint #21: Commit 278a47d pushed by duj12
December 10, 2024 04:43 19s dj
dj
December 10, 2024 04:43 19s
更新训练代码。将codec和speaker_encoder的ckpt放到配置文件里面。 Lint #20: Commit 2f37257 pushed by duj12
December 7, 2024 03:50 17s dj
dj
December 7, 2024 03:50 17s
flow-matching步数参数开放。 Lint #19: Commit a765873 pushed by duj12
December 4, 2024 08:08 16s dj
dj
December 4, 2024 08:08 16s
去掉Matcha-TTS的引用。 Lint #18: Commit 145d850 pushed by duj12
December 4, 2024 03:28 17s dj
dj
December 4, 2024 03:28 17s
添加SpeakerEmbedding模块。 Lint #17: Commit 985c7b0 pushed by duj12
December 3, 2024 10:25 20s dj
dj
December 3, 2024 10:25 20s
fix Lint #16: Commit 5d62d37 pushed by duj12
December 3, 2024 08:46 26s dj
dj
December 3, 2024 08:46 26s
实验llm和flow是否使用prompt speech token Lint #15: Commit 1d183ba pushed by duj12
November 28, 2024 09:22 22s dj
dj
November 28, 2024 09:22 22s
更改token2wav传参，只需要输入音色向量和合成文本token。 Lint #14: Commit 9a815fd pushed by duj12
November 28, 2024 03:52 21s dj
dj
November 28, 2024 03:52 21s
更新推理代码。 Lint #13: Commit da5e0d7 pushed by duj12
November 27, 2024 07:08 15s dj
dj
November 27, 2024 07:08 15s
添加文本转写，方便测试的时候，直接从参考音频获取文本信息。 Lint #12: Commit 44e8e13 pushed by duj12
November 27, 2024 02:36 16s dj
dj
November 27, 2024 02:36 16s
更新25万小时数据训练的代码和配置 Lint #11: Commit 957eda5 pushed by duj12
November 23, 2024 03:24 17s dj
dj
November 23, 2024 03:24 17s
解决推理死循环的问题。 Lint #10: Commit b924190 pushed by duj12
November 22, 2024 09:33 20s dj
dj
November 22, 2024 09:33 20s
data pipeline 增加一些日志。 Lint #9: Commit 121a276 pushed by duj12
November 21, 2024 07:04 16s dj
dj
November 21, 2024 07:04 16s
添加phoneme_llm的推理，it works!!! Lint #8: Commit 1dac1b6 pushed by duj12
November 21, 2024 04:03 16s dj
dj
November 21, 2024 04:03 16s
添加音素llm训练。读取json格式。 Lint #7: Commit 0675830 pushed by duj12
November 20, 2024 09:56 26s dj
dj
November 20, 2024 09:56 26s
修复flow中采样率的问题，flow训练和推理都正常了。 Lint #6: Commit 64bfe29 pushed by duj12
November 20, 2024 02:04 16s dj
dj
November 20, 2024 02:04 16s
修复s3tokenizer输入音频不是16k的bug. Lint #5: Commit fc1c214 pushed by duj12
November 19, 2024 04:03 18s dj
dj
November 19, 2024 04:03 18s
添加flow和hifigan训练，以及cosyvoice的整体推理。 Lint #4: Commit edad7ef pushed by duj12
November 18, 2024 11:14 16s dj
dj
November 18, 2024 11:14 16s