Spaces:
Build error
Build error
File size: 2,578 Bytes
26175df |
1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 18 19 20 21 22 23 24 25 26 27 28 29 30 31 32 33 34 35 36 37 38 39 40 41 42 43 44 45 46 47 48 49 50 51 52 53 54 55 56 57 58 59 60 61 62 63 64 65 66 67 68 69 70 71 72 73 74 75 76 77 78 79 |
import os
import subprocess
from dotenv import load_dotenv
load_dotenv()
try:
os.environ["HUGGINGFACEHUB_API_TOKEN"] = os.getenv("HUGGINGFACEHUB_API_TOKEN")
PINECONE_API_KEY = os.getenv("PINECONE_API_KEY")
except:
PINECONE_API_KEY = subprocess.check_output(["bash", "-c", "echo ${{ secrets.PINECONE_API_KEY }}"]).decode("utf-8").strip()
from langchain.embeddings import HuggingFaceEmbeddings
import pinecone
import torch
from langchain import PromptTemplate, LLMChain,HuggingFacePipeline
from langchain.vectorstores import Pinecone
from langchain.chains.question_answering import load_qa_chain
from langchain.chains import RetrievalQA
from transformers import pipeline
def get_llm(model_name,pinecone_index,llm):
# model_name = "bert-large-uncased" #"t5-large"
model_kwargs = {'device': 'cuda' if torch.cuda.is_available() else 'cpu'}
embeddings = HuggingFaceEmbeddings(model_name=model_name, model_kwargs=model_kwargs)
pinecone.init(
api_key=PINECONE_API_KEY,
environment="us-east-1-aws"
)
index = pinecone.Index(pinecone_index)
print(index.describe_index_stats())
docsearch = Pinecone(index, embeddings.embed_query,"text")
# print("About to load the model")
instruct_pipeline = pipeline(model=llm, torch_dtype=torch.bfloat16, trust_remote_code=True, device_map="auto",
return_full_text=True, do_sample=False, max_new_tokens=128)
llm = HuggingFacePipeline(pipeline=instruct_pipeline)
# print("Loaded the LLM")
# print("Prompting")
template = """Context: {context}
Question: {question}
Answer: Let's go step by step."""
prompt = PromptTemplate(template=template, input_variables=["question","context"])
llm_chain = LLMChain(prompt=prompt, llm=llm)
return llm_chain, docsearch
if __name__ == "__main__":
model_name = "bert-large-uncased"
pinecone_index = "bert-large-uncased"
llm = "databricks/dolly-v2-3b"
llm_chain, docsearch = get_llm(model_name,pinecone_index,llm)
print(":"*40)
questions = ["what is the name of the first Hindi newspaper published in Bihar?",
"what is the capital of Bihar?",
"Brief about the Gupta Dynasty"]
for question in questions:
context = docsearch.similarity_search(question, k=3,metadata=False)
content = ""
for i in context:
content= content + f"{i.__dict__['page_content']}"
print(f"{question}")
response = llm_chain.predict(question=question,context=content)
print(f"{response}\n{'--'*25}")
|