Add library_name, pipeline_tag and license

README.md

@@ -3,7 +3,11 @@ datasets:
 - HuggingFaceTB/smollm-corpus
 language:
 - en
+license: apache-2.0
+library_name: transformers
+pipeline_tag: text-generation
 ---
+
 # Outlier-Safe Pre-Training
 
 [![arXiv](https://img.shields.io/badge/arXiv-2506.19697-b31b1b?style=flat-square)](https://arxiv.org/abs/2506.19697)
@@ -25,17 +29,14 @@ A method that prevents outliers but significantly reduces efficiency is unlikely
 3. 🧩**Ensuring full compatibility with existing inference pipelines**<br/>
 We prioritize compatibility with widely adopted inference frameworks such as vLLM and SGLang. Rather than introducing architectural changes that break compatibility, OSP preserves computational invariance, allowing models to be directly integrated into existing pipelines without additional effort.
 
-
-
 ## Model Checkpoints
 
 ### Final Models
 
 The models were trained on 1 trillion tokens, following the pre-training recipe of [SmolLM](https://huggingface.co/blog/smollm). Specifically, training was conducted using the [smollm-corpus](https://huggingface.co/datasets/HuggingFaceTB/smollm-corpus), a mixture of FineWeb-Edu, Cosmopedia, and Python-Edu.
 
-- [🤗 OSP-1.4B-1T-Adam](https://huggingface.co/dmis-lab/OSP-1.4B-1T-Adam): Trained on the standard Adam optimizer, without any modifications.
-- [🤗 OSP-1.4B-1T-Muon-SSNorm-EmbProj](https://huggingface.co/dmis-lab/OSP-1.4B-1T-Muon-SSNorm-EmbProj): Trained on the OSP framework. This is our final model.
-
+- 🤗 OSP-1.4B-1T-Adam](https://huggingface.co/dmis-lab/OSP-1.4B-1T-Adam): Trained on the standard Adam optimizer, without any modifications.
+- 🤗 OSP-1.4B-1T-Muon-SSNorm-EmbProj](https://huggingface.co/dmis-lab/OSP-1.4B-1T-Muon-SSNorm-EmbProj): Trained on the OSP framework. This is our final model.
 
 ### Ablation Models
 
@@ -175,8 +176,7 @@ The models were trained on 1 trillion tokens, following the pre-training recipe
         </tr>
     </tbody>
 </table>
-&dagger;Model configuration that disables decoupled embedding optimization by training with Muon optimizer without Adam optimization on embedding layers 
-
+&dagger;Model configuration that disables decoupled embedding optimization by training with Muon optimizer without Adam optimization on embedding layers
 
 ## Training
 
@@ -185,7 +185,7 @@ The models were trained on 1 trillion tokens, following the pre-training recipe
 - Architecture: Llama
 - Pretraining tokens: 100 billion tokens
 - Precision: bfloat16
-  
+
 ### Hardware
 
 - TPUs: TPU-v4-512 Pod Slice (supported by [TRC Program](https://sites.research.google/trc/about/))
@@ -202,12 +202,12 @@ This model family was trained to demonstrate the effectiveness of eliminating ou
 
 ```bibtex
 @article{park2025osp,
-      title={Outlier-Safe Pre-Training for Robust 4-Bit Quantization of Large Language Models}, 
+      title={Outlier-Safe Pre-Training for Robust 4-Bit Quantization of Large Language Models},
       author={Jungwoo Park and Taewhoo Lee and Chanwoong Yoon and Hyeon Hwang and Jaewoo Kang},
       year={2025},
       eprint={2506.19697},
       archivePrefix={arXiv},
       primaryClass={cs.LG},
-      url={https://arxiv.org/abs/2506.19697}, 
+      url={https://arxiv.org/abs/2506.19697},
 }
 ```