LoQT: Low-Rank Adapters for Quantized Pretraining
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_ | 1866910681086820352 |
|---|---|
| author | Loeschcke, Sebastian Toftrup, Mads Kastoryano, Michael J. Belongie, Serge Snæbjarnarson, Vésteinn |
| author_facet | Loeschcke, Sebastian Toftrup, Mads Kastoryano, Michael J. Belongie, Serge Snæbjarnarson, Vésteinn |
| contents | Despite advances using low-rank adapters and quantization, pretraining of large models on consumer hardware has not been possible without model sharding, offloading during training, or per-layer gradient updates. To address these limitations, we propose Low-Rank Adapters for Quantized Training (LoQT), a method for efficiently training quantized models. LoQT uses gradient-based tensor factorization to initialize low-rank trainable weight matrices that are periodically merged into quantized full-rank weight matrices. Our approach is suitable for both pretraining and fine-tuning models. We demonstrate this for language modeling and downstream task adaptation, finding that LoQT enables efficient training of models up to 7B parameters on a 24GB GPU. We also demonstrate the feasibility of training a 13B model using per-layer gradient updates on the same hardware. |
| format | Preprint |
| id |
arxiv_https___arxiv_org_abs_2405_16528 |
| institution | arXiv |
| publishDate | 2024 |
| record_format | arxiv |
| spellingShingle | LoQT: Low-Rank Adapters for Quantized Pretraining Loeschcke, Sebastian Toftrup, Mads Kastoryano, Michael J. Belongie, Serge Snæbjarnarson, Vésteinn Machine Learning Computation and Language Despite advances using low-rank adapters and quantization, pretraining of large models on consumer hardware has not been possible without model sharding, offloading during training, or per-layer gradient updates. To address these limitations, we propose Low-Rank Adapters for Quantized Training (LoQT), a method for efficiently training quantized models. LoQT uses gradient-based tensor factorization to initialize low-rank trainable weight matrices that are periodically merged into quantized full-rank weight matrices. Our approach is suitable for both pretraining and fine-tuning models. We demonstrate this for language modeling and downstream task adaptation, finding that LoQT enables efficient training of models up to 7B parameters on a 24GB GPU. We also demonstrate the feasibility of training a 13B model using per-layer gradient updates on the same hardware. |
| title | LoQT: Low-Rank Adapters for Quantized Pretraining |
| topic | Machine Learning Computation and Language |
| url | https://arxiv.org/abs/2405.16528 |