Enhancing Training Efficiency Using Packing with Flash Attention
Fuente:
arXiv
Gespeichert in:
| Hauptverfasser: | , , , , |
|---|---|
| Format: | Preprint |
| Veröffentlicht: |
2024
|
| Schlagworte: | |
| Online-Zugang: | |
| Tags: |
Tag hinzufügen
Keine Tags, Fügen Sie den ersten Tag hinzu!
|
| _version_ | 1866916376766054400 |
|---|---|
| author | Kundu, Achintya Lee, Rhui Dih Wynter, Laura Ganti, Raghu Kiran Mishra, Mayank |
| author_facet | Kundu, Achintya Lee, Rhui Dih Wynter, Laura Ganti, Raghu Kiran Mishra, Mayank |
| contents | Padding is often used in tuning LLM models by adding special tokens to shorter training examples to match the length of the longest sequence in each batch. While this ensures uniformity for batch processing, it introduces inefficiencies by including irrelevant padding tokens in the computation and wastes GPU resources. Hugging Face SFT trainer has always offered the option to use packing to combine multiple training examples, allowing for maximal utilization of GPU resources. However, up till now, it did not offer proper masking of each packed training example. This capability has been added to Hugging Face Transformers 4.44. We analyse this new feature and show the benefits across different variations of packing. |
| format | Preprint |
| id |
arxiv_https___arxiv_org_abs_2407_09105 |
| institution | arXiv |
| publishDate | 2024 |
| record_format | arxiv |
| spellingShingle | Enhancing Training Efficiency Using Packing with Flash Attention Kundu, Achintya Lee, Rhui Dih Wynter, Laura Ganti, Raghu Kiran Mishra, Mayank Machine Learning Artificial Intelligence Padding is often used in tuning LLM models by adding special tokens to shorter training examples to match the length of the longest sequence in each batch. While this ensures uniformity for batch processing, it introduces inefficiencies by including irrelevant padding tokens in the computation and wastes GPU resources. Hugging Face SFT trainer has always offered the option to use packing to combine multiple training examples, allowing for maximal utilization of GPU resources. However, up till now, it did not offer proper masking of each packed training example. This capability has been added to Hugging Face Transformers 4.44. We analyse this new feature and show the benefits across different variations of packing. |
| title | Enhancing Training Efficiency Using Packing with Flash Attention |
| topic | Machine Learning Artificial Intelligence |
| url | https://arxiv.org/abs/2407.09105 |