return to bge

app.py

@@ -15,8 +15,7 @@ PERSIST_DIR = './storage'
 # Configure the settings
 DEVICE = torch.device("cuda" if torch.cuda.is_available() else "cpu")
 
-Settings.embed_model = OpenAIEmbedding()
-# Settings.embed_model = HuggingFaceEmbedding(model_name="BAAI/bge-base-en-v1.5", device="cpu")
+Settings.embed_model = HuggingFaceEmbedding(model_name="BAAI/bge-base-en-v1.5", device="cpu")
 
 Settings.llm = HuggingFaceLLM(
     model_name="meta-llama/Meta-Llama-3-8B-Instruct",
@@ -39,17 +38,28 @@ query_engine = index.as_query_engine(streaming=True, similarity_top_k=1, node_po
 
 
 @spaces.GPU
-def chatbot_response(message, history):
+def chatbot_response(message, context_window, max_new_tokens, temperature, top_k, top_p):
+    Settings.llm.context_window = context_window
+    Settings.llm.max_new_tokens = max_new_tokens
+    Settings.llm.generate_kwargs = {"temperature": temperature, "top_k": top_k, "top_p": top_p, "do sample": True}
     response = query_engine.query(message)
     return str(response)
 
-iface = gr.ChatInterface(
+iface = gr.Interface(
     fn=chatbot_response,
-    title="UESP Lore Chatbot: Running on top of Meta-Llama-3-8B-Instruct (currently) It works 'okay'",
+    inputs=[
+        gr.Slider(minimum=512, maximum=4096, step=256, value=2048, label="Context Window"),
+        gr.Slider(minimum=32, maximum=512, step=32, value=256, label="Max New Tokens"),
+        gr.Slider(minimum=0.1, maximum=1.0, step=0.1, value=0.7, label="Temperature: Lower number is by the book, higher number is make stuff up"),
+        gr.Slider(minimum=1, maximum=100, step=1, value=50, label="Top K"),
+        gr.Slider(minimum=0.5, maximum=1.0, step=0.05, value=0.95, label="Top P"),
+    ],
+    outputs=gr.Textbox(label="Response"),
+    title="UESP Lore Chatbot: Running on top of Meta-Llama-3-8B-Instruct + BGE_LARGE. This is inferior to the downloadable ones.",
     description="Github page for use case, general information, local installs, etc: https://github.com/emarron/UESP-lore",
-    examples=["Who is Zaraphus?", "What is the relation between dragonbreak and chim?", "What is the Lunar Lorkhan?"],
-    cache_examples=True,
 )
 
+
+
 if __name__ == "__main__":
     iface.launch()

storage/default__vector_store.json

@@ -1,3 +1,3 @@
 version https://git-lfs.github.com/spec/v1
-oid sha256:eb0befaddec7950c565951a9cc8e60acaba5d197c2d4886f419dcee6896680b2
-size 397136247
+oid sha256:9d5b72dceac75e8648eeafd9a9e1b6d62aa24ef560b3fb1116e0e43c65d5ccb4
+size 197975710

storage/docstore.json

@@ -1,3 +1,3 @@
 version https://git-lfs.github.com/spec/v1
-oid sha256:46fd63226ff854bab49a645a2194317b3883247b4ecb98f7944438044e06349f
+oid sha256:f9793071f910c4d3c66e206d6cf0930d9a869c243f1fad99e19b25d3612df161
 size 44043711

storage/index_store.json

@@ -1,3 +1,3 @@
 version https://git-lfs.github.com/spec/v1
-oid sha256:00d5d0f5b20c06f13b0dd622d898396862fd12d77a1fd5e55550704bda465c6d
+oid sha256:c31d504b2c3a15e815dcfe5af9215cb8bbd7e5d5530cf0cd7d0dd8511413477a
 size 959947