hf llm

backend.py

@@ -17,7 +17,7 @@ from IPython.display import Markdown, display
 from langchain_huggingface import HuggingFaceEmbeddings
 
 #from llama_index import LangchainEmbedding, ServiceContext
-from llama_index.embeddings.huggingface import HuggingFaceEmbedding
+#from llama_index.embeddings.huggingface import HuggingFaceEmbedding
 from llama_index.llms.huggingface import HuggingFaceInferenceAPI, HuggingFaceLLM
 from dotenv import load_dotenv
 
@@ -71,13 +71,18 @@ llm = HuggingFaceLLM(
 
 embed_model= HuggingFaceEmbeddings(model_name="sentence-transformers/all-mpnet-base-v2")
 
-service_context = ServiceContext.from_defaults(
-    chunk_size=1024,
-    llm=llm,
-    embed_model=embed_model
-)
+Settings.llm = llm
+Settings.embed_model = embed_model
+#Settings.node_parser = SentenceSplitter(chunk_size=512, chunk_overlap=20, paragraph_separator="\n\n")
+Settings.num_output = 512
+Settings.context_window = 3900
+
+
 documents = SimpleDirectoryReader('data/blockchainprova.txt').load_data()
-index = VectorStoreIndex.from_documents(documents, service_context=service_context)
+
+nodes = SentenceSplitter(chunk_size=512, chunk_overlap=20, paragraph_separator="\n\n").get_nodes_from_documents(documents)
+# Build the vector store index from the nodes
+index = VectorStoreIndex(nodes)
 
 
 

requirements.txt

@@ -9,6 +9,7 @@ sentence_transformers
 llama-index-readers-web
 llama-index-readers-file
 llama-index-embeddings-langchain
+
 pypdf
 langchain>=0.2.3
 langchain-community