DataComp-LM: In search of the next generation of training sets for language models
Fuente:
arXiv
Enregistré dans:
| Auteurs principaux: | , , , , , , , , , , , , , , , , , , , , , , , , , , , , , , , , , , , , , , , , , , , , , , , , , , , , , , , , , , |
|---|---|
| Format: | Preprint |
| Publié: |
2024
|
| Sujets: | |
| Accès en ligne: | |
| Tags: |
Ajouter un tag
Pas de tags, Soyez le premier à ajouter un tag!
|
| _version_ | 1866913800115978240 |
|---|---|
| author | Li, Jeffrey Fang, Alex Smyrnis, Georgios Ivgi, Maor Jordan, Matt Gadre, Samir Bansal, Hritik Guha, Etash Keh, Sedrick Arora, Kushal Garg, Saurabh Xin, Rui Muennighoff, Niklas Heckel, Reinhard Mercat, Jean Chen, Mayee Gururangan, Suchin Wortsman, Mitchell Albalak, Alon Bitton, Yonatan Nezhurina, Marianna Abbas, Amro Hsieh, Cheng-Yu Ghosh, Dhruba Gardner, Josh Kilian, Maciej Zhang, Hanlin Shao, Rulin Pratt, Sarah Sanyal, Sunny Ilharco, Gabriel Daras, Giannis Marathe, Kalyani Gokaslan, Aaron Zhang, Jieyu Chandu, Khyathi Nguyen, Thao Vasiljevic, Igor Kakade, Sham Song, Shuran Sanghavi, Sujay Faghri, Fartash Oh, Sewoong Zettlemoyer, Luke Lo, Kyle El-Nouby, Alaaeldin Pouransari, Hadi Toshev, Alexander Wang, Stephanie Groeneveld, Dirk Soldaini, Luca Koh, Pang Wei Jitsev, Jenia Kollar, Thomas Dimakis, Alexandros G. Carmon, Yair Dave, Achal Schmidt, Ludwig Shankar, Vaishaal |
| author_facet | Li, Jeffrey Fang, Alex Smyrnis, Georgios Ivgi, Maor Jordan, Matt Gadre, Samir Bansal, Hritik Guha, Etash Keh, Sedrick Arora, Kushal Garg, Saurabh Xin, Rui Muennighoff, Niklas Heckel, Reinhard Mercat, Jean Chen, Mayee Gururangan, Suchin Wortsman, Mitchell Albalak, Alon Bitton, Yonatan Nezhurina, Marianna Abbas, Amro Hsieh, Cheng-Yu Ghosh, Dhruba Gardner, Josh Kilian, Maciej Zhang, Hanlin Shao, Rulin Pratt, Sarah Sanyal, Sunny Ilharco, Gabriel Daras, Giannis Marathe, Kalyani Gokaslan, Aaron Zhang, Jieyu Chandu, Khyathi Nguyen, Thao Vasiljevic, Igor Kakade, Sham Song, Shuran Sanghavi, Sujay Faghri, Fartash Oh, Sewoong Zettlemoyer, Luke Lo, Kyle El-Nouby, Alaaeldin Pouransari, Hadi Toshev, Alexander Wang, Stephanie Groeneveld, Dirk Soldaini, Luca Koh, Pang Wei Jitsev, Jenia Kollar, Thomas Dimakis, Alexandros G. Carmon, Yair Dave, Achal Schmidt, Ludwig Shankar, Vaishaal |
| contents | We introduce DataComp for Language Models (DCLM), a testbed for controlled dataset experiments with the goal of improving language models. As part of DCLM, we provide a standardized corpus of 240T tokens extracted from Common Crawl, effective pretraining recipes based on the OpenLM framework, and a broad suite of 53 downstream evaluations. Participants in the DCLM benchmark can experiment with data curation strategies such as deduplication, filtering, and data mixing at model scales ranging from 412M to 7B parameters. As a baseline for DCLM, we conduct extensive experiments and find that model-based filtering is key to assembling a high-quality training set. The resulting dataset, DCLM-Baseline enables training a 7B parameter language model from scratch to 64% 5-shot accuracy on MMLU with 2.6T training tokens. Compared to MAP-Neo, the previous state-of-the-art in open-data language models, DCLM-Baseline represents a 6.6 percentage point improvement on MMLU while being trained with 40% less compute. Our baseline model is also comparable to Mistral-7B-v0.3 and Llama 3 8B on MMLU (63% & 66%), and performs similarly on an average of 53 natural language understanding tasks while being trained with 6.6x less compute than Llama 3 8B. Our results highlight the importance of dataset design for training language models and offer a starting point for further research on data curation. |
| format | Preprint |
| id |
arxiv_https___arxiv_org_abs_2406_11794 |
| institution | arXiv |
| publishDate | 2024 |
| record_format | arxiv |
| spellingShingle | DataComp-LM: In search of the next generation of training sets for language models Li, Jeffrey Fang, Alex Smyrnis, Georgios Ivgi, Maor Jordan, Matt Gadre, Samir Bansal, Hritik Guha, Etash Keh, Sedrick Arora, Kushal Garg, Saurabh Xin, Rui Muennighoff, Niklas Heckel, Reinhard Mercat, Jean Chen, Mayee Gururangan, Suchin Wortsman, Mitchell Albalak, Alon Bitton, Yonatan Nezhurina, Marianna Abbas, Amro Hsieh, Cheng-Yu Ghosh, Dhruba Gardner, Josh Kilian, Maciej Zhang, Hanlin Shao, Rulin Pratt, Sarah Sanyal, Sunny Ilharco, Gabriel Daras, Giannis Marathe, Kalyani Gokaslan, Aaron Zhang, Jieyu Chandu, Khyathi Nguyen, Thao Vasiljevic, Igor Kakade, Sham Song, Shuran Sanghavi, Sujay Faghri, Fartash Oh, Sewoong Zettlemoyer, Luke Lo, Kyle El-Nouby, Alaaeldin Pouransari, Hadi Toshev, Alexander Wang, Stephanie Groeneveld, Dirk Soldaini, Luca Koh, Pang Wei Jitsev, Jenia Kollar, Thomas Dimakis, Alexandros G. Carmon, Yair Dave, Achal Schmidt, Ludwig Shankar, Vaishaal Machine Learning Computation and Language We introduce DataComp for Language Models (DCLM), a testbed for controlled dataset experiments with the goal of improving language models. As part of DCLM, we provide a standardized corpus of 240T tokens extracted from Common Crawl, effective pretraining recipes based on the OpenLM framework, and a broad suite of 53 downstream evaluations. Participants in the DCLM benchmark can experiment with data curation strategies such as deduplication, filtering, and data mixing at model scales ranging from 412M to 7B parameters. As a baseline for DCLM, we conduct extensive experiments and find that model-based filtering is key to assembling a high-quality training set. The resulting dataset, DCLM-Baseline enables training a 7B parameter language model from scratch to 64% 5-shot accuracy on MMLU with 2.6T training tokens. Compared to MAP-Neo, the previous state-of-the-art in open-data language models, DCLM-Baseline represents a 6.6 percentage point improvement on MMLU while being trained with 40% less compute. Our baseline model is also comparable to Mistral-7B-v0.3 and Llama 3 8B on MMLU (63% & 66%), and performs similarly on an average of 53 natural language understanding tasks while being trained with 6.6x less compute than Llama 3 8B. Our results highlight the importance of dataset design for training language models and offer a starting point for further research on data curation. |
| title | DataComp-LM: In search of the next generation of training sets for language models |
| topic | Machine Learning Computation and Language |
| url | https://arxiv.org/abs/2406.11794 |