diff --git a/README.md b/README.md
index 51a7586..cb40411 100644
--- a/README.md
+++ b/README.md
@@ -1 +1,6 @@
-[Project doc](https://docs.google.com/document/d/1RCQDzlVU7YGoTwR84gLQfhxTv0RFnW6srQlqfCC2bvQ/edit?usp=sharing)
\ No newline at end of file
+[Project doc](https://docs.google.com/document/d/1RCQDzlVU7YGoTwR84gLQfhxTv0RFnW6srQlqfCC2bvQ/edit?usp=sharing)
+
+### Finetuning the base model with LoRA adaptor
+```bash
+python hyperlora/intx_sft.py configs/default.yaml --model_name_or_path=meta-llama/Llama-3.2-1B-Instruct --num_train_epochs=100 --per_device_train_batch_size=8 --per_device_eval_batch_size=32 --overwrite_output_dir=True
+```
diff --git a/hyperlora/data_utils.py b/hyperlora/data_utils.py
index 6dd84a7..7c81b47 100644
--- a/hyperlora/data_utils.py
+++ b/hyperlora/data_utils.py
@@ -314,7 +314,7 @@ def tokenize_chat_messages(
         text,
         return_offsets_mapping=mask_assistant_inputs,
         add_special_tokens=False,
-        **tokenizer_kwargs if tokenizer_kwargs is not None else {},
+        **(tokenizer_kwargs or {}),
     )
     if mask_assistant_inputs:
         assistant_ranges = get_assistant_start_end_indices(messages, text)
diff --git a/hyperlora/model_loading.py b/hyperlora/model_loading.py
index b15249b..8835211 100644
--- a/hyperlora/model_loading.py
+++ b/hyperlora/model_loading.py
@@ -101,7 +101,6 @@ def get_model(
     device="cuda:0",
     dtype=torch.bfloat16,
 ):
-    # TODO: we could use lower precision during finetuning
     model_init_kwargs = dict(
         pretrained_model_name_or_path=model_name_or_path,
         device_map=device,