LATS_powered_by_SambaNova

Runtime error

Etash Guha commited on Jul 10

Commit

5731c25

•

1 Parent(s): d4db51c

changed name:

Files changed (1) hide show

app.py CHANGED Viewed

@@ -15,7 +15,7 @@ if 'response_content' not in st.session_state:
 # Creating main columns for the chat and runtime notifications
 chat_col = st.container()
-chat_col.title("SambaLATS")
 description = """This demo is an implementation of Language Agent Tree Search (LATS) (https://arxiv.org/abs/2310.04406) with Samba-1 in the backend. Thank you to the original authors of demo on which this is based from [Lapis Labs](https://lapis.rocks/)!
 Given Samba-1's lightning quick inference, not only can we accelerate our system's speeds but also improve our system's accuracy. Using many inference calls in this LATS style, we can solve programming questions with higher accuracy. In fact, this system reaches **GPT-3.5 accuracy on HumanEval Python**, 74% accuracy, with LLaMa 3 8B, taking 8 seconds on average. This is a 15.5% boost on LLaMa 3 8B alone.

 # Creating main columns for the chat and runtime notifications
 chat_col = st.container()
+chat_col.title("LATS powered by SambaNova")
 description = """This demo is an implementation of Language Agent Tree Search (LATS) (https://arxiv.org/abs/2310.04406) with Samba-1 in the backend. Thank you to the original authors of demo on which this is based from [Lapis Labs](https://lapis.rocks/)!
 Given Samba-1's lightning quick inference, not only can we accelerate our system's speeds but also improve our system's accuracy. Using many inference calls in this LATS style, we can solve programming questions with higher accuracy. In fact, this system reaches **GPT-3.5 accuracy on HumanEval Python**, 74% accuracy, with LLaMa 3 8B, taking 8 seconds on average. This is a 15.5% boost on LLaMa 3 8B alone.