rag_with_mircosoftphi2_and_hf_embeddings.py

from llama_index.core import VectorStoreIndex,SimpleDirectoryReader,ServiceContext
import torch

documents = SimpleDirectoryReader("SansarChat").load_data()

"""New sectiond"""

from llama_index.core.prompts.prompts import SimpleInputPrompt
from llama_index.llms.llama_cpp import LlamaCPP

system_prompt = "You are a Q&A assistant. Your goal is to answer questions as accurately as possible based on the instructions and context provided."

# This will wrap the default prompts that are internal to llama-index
query_wrapper_prompt = SimpleInputPrompt("<|USER|>{query_str}<|ASSISTANT|>")

# model_url = "https://huggingface.co/TheBloke/Llama-2-13B-chat-GGML/resolve/main/llama-2-13b-chat.ggmlv3.q4_0.bin"
model_url = "https://huggingface.co/TheBloke/Llama-2-13B-chat-GGUF/resolve/main/llama-2-13b-chat.Q4_0.gguf"

llm = LlamaCPP(
    # optionally, you can set the path to a pre-downloaded model instead of model_url
    model_path="model.gguf",
    temperature=0.1,
    max_new_tokens=256,
    # llama2 has a context window of 4096 tokens, but we set it lower to allow for some wiggle room
    context_window=4096,
    # kwargs to pass to __call__()
    generate_kwargs={},
    # kwargs to pass to __init__()
    # set to at least 1 to use GPU
    model_kwargs={"n_gpu_layers": 0},
    verbose=True
)

"""HuggingFace Embeddings"""

from llama_index.embeddings.huggingface import HuggingFaceEmbedding
# loads BAAI/bge-small-en-v1.5
embed_model = HuggingFaceEmbedding(model_name="BAAI/bge-small-en-v1.5")

service_context = ServiceContext.from_defaults(
    chunk_size=256,
    llm=llm,
    embed_model=embed_model
)

"""predict"""

index = VectorStoreIndex.from_documents(documents, service_context=service_context)

query_engine = index.as_query_engine()

def predict(input, history):
  response = query_engine.query(input)
  return str(response)

"""Gradio"""

import gradio as gr

gr.ChatInterface(predict).launch(share=True)