You can deploy this model and start earning money today!

LingweiMeng/
speechllm_multispk

tensorboard

Views

No views yet

代码

https://github.com/XiaoshanHsj/speechllm/tree/multispk_lingmeng

看代码的README_lingmeng.md

数据：

英文部分: ./data/Whisper-Sidecar-data-metadata/data_for_wavllm

德文相关: ./data/de-en-mix

targetASR (target-talker ASR)用到的reference audio: ./data/reference_enroll_audio/all

英文部分只有metadata，可以从librispeech生成。德文部分还备份了测试集音频。

模型：

tokenizer: ./llama_model/llama/tokenizer.model

llama-2-chat: ./llama_model/llama-2-7b-chat/consolidated.00.pth

训练好的MT-LLM模型目录:

./lingmeng_multispk_multitask_retrain_speechllm_v0.1_llama2_chat_wavlm_weighted_update_lora_32_32_prompt_build_multispk_multitask_de.yaml_16gpu_1accum

LingweiMeng/speechllm_multispk

代码

数据：

模型：

LingweiMeng/
speechllm_multispk