docs: add huggingface repo notes

notes/huggingface.md

@@ -12,3 +12,35 @@ And it has a data parallelism feature which is when you split the data across
 multiple GPUs by dividing the data into batches and then each GPU processes
 a batch. Note that this is about data and note the same as the distrubuted
 training.
+
+### Models
+To add a new model to huggingface.co you first create a new model using their
+UI and then you can use normal git command to push.
+
+### Enable files larger than 5GB
+To enable files larger than 5GB you need to use the `lfs-enable-largefiles`:
+```console
+$ huggingface-cli lfs-enable-largefiles . 
+```
+Also you need to have the `git-lfs` installed.
+```console
+$ git lfs install
+$ git lfs track "*.gguf"
+```
+
+### Configure ssh
+To configure ssh you need to create a `~/.ssh/config` file with the following:
+```console
+Host hf.co
+    HostName hf.co
+    User git
+    IdentityFile ~/.ssh/id_file
+```
+After this you should be able to authenicate to hf:
+```console
+$ ssh -T [email protected]
+```
+And we can update the git remote to use ssh:
+```console
+$ git remote add origin [email protected]:danbev/test-model
+```