Fix modelscope bug

README.md

@@ -1,21 +1,15 @@
----
-license: apache-2.0
-language:
-- en
-pipeline_tag: audio-to-audio
----
 # UniSE: A Unified Framework for Decoder-only Autoregressive LM-based Speech Enhancement
 
 <p align="center">
   <a href="https://arxiv.org/abs/2510.20441">
     <img src="https://img.shields.io/badge/Paper-ArXiv-red.svg" alt="Paper">
   </a>
-  <a href="https://hyyan2k.github.io/UniSE/">
-    <img src="https://img.shields.io/badge/Demo-Page-blue.svg" alt="Demo">
-  </a>
-  <a href="https://huggingface.co/spaces/QuarkAudio/">
+  <a href="https://huggingface.co/QuarkAudio/QuarkAudio-UniSE/">
     <img src="https://img.shields.io/badge/Model-Hugging%20Face-yellow.svg" alt="Hugging Face">
   </a>
+  <a href="https://www.modelscope.cn/models/QuarkAudio/QuarkAudio-UniSE/">
+    <img src="https://img.shields.io/badge/Model-%20%E9%AD%94%E6%90%AD-orange.svg" alt="ModelScope">
+  </a>
 </p>
 
 <p align="center">
@@ -29,7 +23,7 @@ pipeline_tag: audio-to-audio
 - 🔄 **End-to-End Compatible**: Integrates WavLM (feature extractor), BiCodec (discrete codec), and LM into one pipeline.
 - 🌍 **Multitask Support**: SE, SR, TSE, SS, and more — all in a single model.
 
-📄 **Paper**: [arXiv:2510.20441](https://arxiv.org/abs/2510.20441) | 🎤 **Listen**: [Demo Page](https://hyyan2k.github.io/UniSE/) | 🤗 **Model**: [Hugging Face Spaces](https://huggingface.co/spaces/QuarkAudio/)
+📄 **Paper**: [arXiv:2510.20441](https://arxiv.org/abs/2510.20441)  | 🤗 **Model**: [Hugging Face Spaces]https://huggingface.co/QuarkAudio/QuarkAudio-UniSE/)
 
 ---
 
@@ -71,7 +65,7 @@ QuarkAudio-UniSE requires three additional **WavLM** and **BiCodec** pre-trained
 cd checkpoints
 bash download.sh
 ```
-Additionally, download WavLM-base-plus.pt from this [URL](https://huggingface.co/microsoft/wavlm-base-plus) and put it at `./ckpt/WavLM-base-plus.pt` .
+Additionally, download WavLM-Large.pt from this [URL](https://huggingface.co/microsoft/wavlm-base-plus) and put it at `./ckpt/WavLM-Large.pt` .
 
 Alternatively, you can download them manually and place them in the `./model/bicodec/` directory.
 
@@ -91,14 +85,14 @@ python ./train.py --config conf/config.yaml
 | `speech_scp_path`        | SCP of clean audio files                                                       |
 | `noise_scp_path`        | SCP of noise audio files                                                                   
  | `rir_scp_path`        | SCP of rir audio files                                                                       |
-| `mode`           | Task type: `SE` (Noise Suppression,Speech Restoration,Packet Loss Concealment), `TSE` (Target Speaker Extraction), `SS` (Speech Separation). |
+| `mode`           | Task type: `se` (Noise Suppression,Speech Restoration,Packet Loss Concealment), `tse` (Target Speaker Extraction), `SS` (Speech Separation). |
 
 
 ## Inference
 + Quick start
 The main inference script is **`test.py`**. The inference process consists of two stages:
 
-1. Extract hidden states from all **WavLM** layers and obtain a single representation by averaging them across layers.
+1. Extract hidden states from all WavLM layers and obtain a single representation by averaging them across layers.
 2. Use the language model (LM) to predict speech tokens, and then decode them into audio using **BiCodec**.
 
 ### Running Inference
@@ -119,13 +113,10 @@ Command to run inference:
 python test.py
 ```
 
-## Results
-
-Samples processed by UniSE can be found on our [Demo Page](https://github.com/hyyan2k/UniSE/).
 
 ## Model Checkpoints
 
-Our pretrained model is available on [Hugging Face](https://huggingface.co/spaces/QuarkAudio/).
+Our pretrained model is available on [Hugging Face](https://huggingface.co/QuarkAudio/QuarkAudio-UniSE/).
 
 ## Hints
 
@@ -144,7 +135,8 @@ Our approach focuses on leveraging the LLM's comprehension capabilities to enabl
       url={https://arxiv.org/abs/2510.20441}, 
 }
 ```
-     
+
 
 ## Contact
-For any questions, please contact: `yanhaoyin.yhy@alibaba-inc.com`
+For any questions, please contact: `yanhaoyin.yhy@alibaba-inc.com`
+