Added Llama 2 model

langchain_model/main.py

@@ -0,0 +1,111 @@
+from langchain.text_splitter import RecursiveCharacterTextSplitter
+from langchain.embeddings import HuggingFaceEmbeddings
+from langchain.vectorstores import Pinecone
+from sentence_transformers import SentenceTransformer
+from langchain.chains.question_answering import load_qa_chain
+from pinecone import Pinecone, ServerlessSpec
+from langchain.vectorstores import Pinecone as pcone
+import os
+from langchain.llms import LlamaCpp
+from langchain.callbacks.manager import CallbackManager
+from langchain.callbacks.streaming_stdout import StreamingStdOutCallbackHandler
+from huggingface_hub import hf_hub_download
+from langchain.chains.question_answering import load_qa_chain
+from langchain.document_loaders import PyPDFLoader, OnlinePDFLoader
+import warnings
+import time
+from fastapi import FastAPI, HTTPException
+from pydantic import BaseModel
+
+
+app = FastAPI()
+
+warnings.filterwarnings("ignore")
+model_name_or_path = "TheBloke/Llama-2-13B-chat-GGUF"
+model_basename = "llama-2-13b-chat.Q5_0.gguf"
+PINECONE_API_KEY = os.environ.get('PINECONE_API_KEY', '7b51ee47-cecb-4276-9a91-f225348f4f73')
+PINECONE_API_ENV = os.environ.get('PINECONE_API_ENV', 'asia-southeast1-gcp-free')
+
+os.environ["HUGGINGFACEHUB_API_TOKEN"] = "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"
+os.environ["PINECONE_API_KEY"] = PINECONE_API_KEY
+
+# initialize pinecone
+pc = Pinecone(
+    api_key=PINECONE_API_KEY,  # find at app.pinecone.io
+    environment=PINECONE_API_ENV  # next to api key in console
+)
+callback_manager = CallbackManager([StreamingStdOutCallbackHandler()])     
+model_path = hf_hub_download(repo_id=model_name_or_path, filename=model_basename)
+
+# Loading model,
+llm = LlamaCpp(
+    model_path=model_path,
+    max_tokens=256,
+    n_batch=8,
+    callback_manager=callback_manager,
+    n_ctx=1024,
+    verbose=False,
+    device="cpu"
+)
+
+class Text(BaseModel):
+    query: str
+
+def infer_pdf( src ): 
+    loader = PyPDFLoader(src)
+    data = loader.load()
+
+    chain=load_qa_chain(llm, chain_type="stuff")
+
+    index_name = "langchainpinecone" # put in the name of your pinecone index here
+    text_splitter=RecursiveCharacterTextSplitter(chunk_size=500, chunk_overlap=0)
+    docs=text_splitter.split_documents(data)
+    embeddings=HuggingFaceEmbeddings(model_name='sentence-transformers/all-MiniLM-L6-v2')    
+    docsearch= pcone.from_texts([t.page_content for t in docs], embeddings, index_name=index_name)
+
+    query="Generate 3 MCQ questions with answers based on the content given"
+    docs=docsearch.similarity_search(query)
+    t = chain.run(input_documents=docs, question=query)
+    print( "ans", t );
+
+def infer_text( text):
+    # Loading model,
+    llm = LlamaCpp(
+        model_path=model_path,
+        max_tokens=256,
+        n_batch=32,
+        callback_manager=callback_manager,
+        n_ctx=1024,
+        verbose=False,
+        device="cpu"
+    )
+
+    chain=load_qa_chain(llm, chain_type="stuff")
+
+    query = "Give me a single question based on the following string " + text
+    t = chain.run(input_documents=[], question=query )
+    return {
+        "ans" : t
+    }
+
+@app.post("/text-reframe")
+async def reframe_text(item: Text):
+    try:
+        start = time.time() 
+        infer_text(item.query)
+        print( "text inference - ", time.time() - start )
+
+    except Exception as e:
+        raise HTTPException(status_code=500, detail=str(e))
+
+
+@app.post( "/generate-mcq" )
+async def generate_mcq( item: Text ):
+    try:
+        start = time.time() 
+        infer_pdf(item.query)
+        print( "pdf inference - ",  time.time() - start ) 
+
+    except Exception as e:
+        raise HTTPException(status_code=500, detail=str(e))
+