Instructions to use KexuanShi/Megatron-LM with libraries, inference providers, notebooks, and local apps. Follow these links to get started.
- Libraries
- NeMo
How to use KexuanShi/Megatron-LM with NeMo:
# tag did not correspond to a valid NeMo domain.
- Notebooks
- Google Colab
- Kaggle
| # fusions package | |
| This package provides modules that provide commonly fused | |
| operations. Fusing operations improves compute efficiency by | |
| increasing the amount of work done each time a tensor is read from | |
| memory. To perform the fusion, modules in this either rely on PyTorch | |
| functionality for doing just-in-time compilation | |
| (i.e. `torch.jit.script` in older PyTorch versions of `torch.compile` | |
| in recent versions), or call into custom kernels in external libraries | |
| such as Apex or TransformerEngine. | |