Update README.md

README.md

@@ -4,6 +4,30 @@ tags:
 - alpaca
 ---
 
+### Stanford Alpaca-7B-Merged
+
+This repo hosts the merged weight for [Stanford Alpaca-7B](https://github.com/tatsu-lab/stanford_alpaca/) that can be used directly.
+Below is the original model card information.
+
+-----------------------
 ### Stanford Alpaca-7B
 
-This repo hosts the merged weight for [Stanford Alpaca-7B](https://github.com/tatsu-lab/stanford_alpaca/) that can be used directly.
+This repo hosts the weight diff for [Stanford Alpaca-7B](https://github.com/tatsu-lab/stanford_alpaca/) that can be used to reconstruct the original model weights when applied to Meta's LLaMA weights. 
+
+To recover the original Alpaca-7B weights, follow these steps:
+```text
+1. Convert Meta's released weights into huggingface format. Follow this guide:
+    https://huggingface.co/docs/transformers/main/model_doc/llama
+2. Make sure you cloned the released weight diff into your local machine. The weight diff is located at:
+    https://huggingface.co/tatsu-lab/alpaca-7b/tree/main
+3. Run this function with the correct paths. E.g.,
+    python weight_diff.py recover --path_raw <path_to_step_1_dir> --path_diff <path_to_step_2_dir> --path_tuned <path_to_store_recovered_weights>
+```
+
+Once step 3 completes, you should have a directory with the recovered weights, from which you can load the model like the following
+
+```python
+import transformers
+alpaca_model = transformers.AutoModelForCausalLM.from_pretrained("<path_to_store_recovered_weights>")
+alpaca_tokenizer = transformers.AutoTokenizer.from_pretrained("<path_to_store_recovered_weights>")
+```