yujunjun/LLMs-from-scratch
1
0
Fork 0
mirror of https://github.com/rasbt/LLMs-from-scratch.git synced 2025-09-16 11:44:29 +00:00
Code Issues Packages Projects Releases Wiki Activity

Update README.md

Browse Source
This commit is contained in:
Sebastian Raschka 2024-08-10 07:54:51 -05:00 committed by GitHub
parent 13dbc548f8
commit 7feb8cad86
No known key found for this signature in database
GPG Key ID: B5690EEEBB952194
1 changed files with 2 additions and 2 deletions
Show Stats Download Patch File Download Diff File Expand all files Collapse all files

4
ch05/03_bonus_pretraining_on_gutenberg/README.md
View File

@ -15,7 +15,7 @@ Please read the [Project Gutenberg Permissions, Licensing and other Common Reque
In this section, we download books from Project Gutenberg using code from the [`pgcorpus/gutenberg`](https://github.com/pgcorpus/gutenberg) GitHub repository. In this section, we download books from Project Gutenberg using code from the [`pgcorpus/gutenberg`](https://github.com/pgcorpus/gutenberg) GitHub repository.
As of this writing, this will require approximately 50 GB of disk space, but it may be more depending on how much Project Gutenberg grew since then. As of this writing, this will require approximately 50 GB of disk space and take about 10-15 hours, but it may be more depending on how much Project Gutenberg grew since then.
   
#### Download instructions for Linux and macOS users #### Download instructions for Linux and macOS users
@ -171,4 +171,4 @@ Note that this code focuses on keeping things simple and minimal for educational
6. Add distributed data parallelism (DDP) and train the model on multiple GPUs (see section *A.9.3 Training with multiple GPUs* in appendix A; [DDP-script.py](../../appendix-A/01_main-chapter-code/DDP-script.py)). 6. Add distributed data parallelism (DDP) and train the model on multiple GPUs (see section *A.9.3 Training with multiple GPUs* in appendix A; [DDP-script.py](../../appendix-A/01_main-chapter-code/DDP-script.py)).
7. Swap the from scratch `MultiheadAttention` class in the `previous_chapter.py` script with the efficient `MHAPyTorchScaledDotProduct` class implemented in the [Efficient Multi-Head Attention Implementations](../../ch03/02_bonus_efficient-multihead-attention/mha-implementations.ipynb) bonus section, which uses Flash Attention via PyTorch's `nn.functional.scaled_dot_product_attention` function. 7. Swap the from scratch `MultiheadAttention` class in the `previous_chapter.py` script with the efficient `MHAPyTorchScaledDotProduct` class implemented in the [Efficient Multi-Head Attention Implementations](../../ch03/02_bonus_efficient-multihead-attention/mha-implementations.ipynb) bonus section, which uses Flash Attention via PyTorch's `nn.functional.scaled_dot_product_attention` function.
8. Speeding up the training by optimizing the model via [torch.compile](https://pytorch.org/tutorials/intermediate/torch_compile_tutorial.html) (`model = torch.compile`) or [thunder](https://github.com/Lightning-AI/lightning-thunder) (`model = thunder.jit(model)`). 8. Speeding up the training by optimizing the model via [torch.compile](https://pytorch.org/tutorials/intermediate/torch_compile_tutorial.html) (`model = torch.compile`) or [thunder](https://github.com/Lightning-AI/lightning-thunder) (`model = thunder.jit(model)`).
9. Implement Gradient Low-Rank Projection (GaLore) to further speed up the pretraining process. This can be achieved by just replacing the `AdamW` optimizer with the provided `GaLoreAdamW` provided in the [GaLore Python library](https://github.com/jiaweizzhao/GaLore). 9. Implement Gradient Low-Rank Projection (GaLore) to further speed up the pretraining process. This can be achieved by just replacing the `AdamW` optimizer with the provided `GaLoreAdamW` provided in the [GaLore Python library](https://github.com/jiaweizzhao/GaLore).