ros/nanochat
1
0
Fork 0
Code Issues Pull Requests Actions Packages Projects Releases Wiki Activity

update readme with a link to the CPU|MPS branch

Browse Source
This commit is contained in:
Andrej Karpathy
2025-10-16 22:03:39 +00:00
parent ccfe7915ac
commit d6d86cbf4c
1 changed files with 4 additions and 0 deletions
Download Patch File Download Diff File Expand all files Collapse all files

4
README.md
View File

@@ -93,6 +93,10 @@ And a bit more about computing environments that will run nanochat:
- If your GPU(s) have less than 80GB, you'll have to tune some of the hyperparameters or you will OOM / run out of VRAM. Look for `--device_batch_size` in the scripts and reduce it until things fit. E.g. from 32 (default) to 16, 8, 4, 2, or even 1. Less than that you'll have to know a bit more what you're doing and get more creative. - If your GPU(s) have less than 80GB, you'll have to tune some of the hyperparameters or you will OOM / run out of VRAM. Look for `--device_batch_size` in the scripts and reduce it until things fit. E.g. from 32 (default) to 16, 8, 4, 2, or even 1. Less than that you'll have to know a bit more what you're doing and get more creative.
- Most of the code is fairly vanilla PyTorch so it should run on anything that supports that - xpu, mps, or etc, but I haven't implemented this out of the box so it might take a bit of tinkering. - Most of the code is fairly vanilla PyTorch so it should run on anything that supports that - xpu, mps, or etc, but I haven't implemented this out of the box so it might take a bit of tinkering.
## Running on CPU / MPS
If you'd like to tinker with nanochat on your Macbook or a CPU machine, there is a work in progress [CPU|MPS PR](https://github.com/karpathy/nanochat/pull/88) up here. If you're on Macbook, use `--device_type=mps` when running `base_train.py`. See the PR and its diff for more. You're not going to get too far without GPU nodes, but at least you'll be able to run the code and maybe train a very tiny LLM with some patience.
## Questions ## Questions
nanochat is designed to be short and sweet. One big advantage of this is that we can package up all of the files together and copy paste them to your favorite LLM to ask arbitrary questions. As an example, I like to package up the repo using the [files-to-prompt](https://github.com/simonw/files-to-prompt) utility like so: nanochat is designed to be short and sweet. One big advantage of this is that we can package up all of the files together and copy paste them to your favorite LLM to ask arbitrary questions. As an example, I like to package up the repo using the [files-to-prompt](https://github.com/simonw/files-to-prompt) utility like so: