From 7fcefe55b4ce5e333f44c1b88fc8f7599b3ebcb4 Mon Sep 17 00:00:00 2001
From: Brian <23239305+b-chu@users.noreply.github.com>
Date: Fri, 26 Apr 2024 09:49:03 -0400
Subject: [PATCH] Add new FT instructions

---
 scripts/train/yamls/finetune/README.md | 6 ++++++
 1 file changed, 6 insertions(+)
 create mode 100644 scripts/train/yamls/finetune/README.md

diff --git a/scripts/train/yamls/finetune/README.md b/scripts/train/yamls/finetune/README.md
new file mode 100644
index 0000000000..3ed238a4ca
--- /dev/null
+++ b/scripts/train/yamls/finetune/README.md
@@ -0,0 +1,6 @@
+# How to finetune a new Hugging Face model
+Using the [DBRX yaml](https://github.com/mosaicml/llm-foundry/blob/main/scripts/train/yamls/finetune/dbrx-full-ft.yaml) as a template, change the following fields for the new model:
+- `max_seq_len`
+- `model/pretrained_model_name_or_path`
+- `tokenizer/name`
+Open a GitHub issue if these steps don't work. Most new causal LMs should work out of the box.