fakezeta commited on
Commit
88278c4
1 Parent(s): 6feb027

cleaned comments

Browse files
Files changed (3) hide show
  1. app.py +0 -1
  2. ingest_data.py +0 -5
  3. query_data.py +1 -24
app.py CHANGED
@@ -1,4 +1,3 @@
1
- from ast import Delete
2
  import streamlit as st
3
  from streamlit_chat import message
4
  from ingest_data import embed_doc
 
 
1
  import streamlit as st
2
  from streamlit_chat import message
3
  from ingest_data import embed_doc
ingest_data.py CHANGED
@@ -25,13 +25,8 @@ def embed_doc(filename):
25
  st.text("Load and split text: "+str(round(end - start,1)))
26
 
27
 
28
- # Load Data to vectorstore
29
  start = time.time()
30
- # embeddings = LlamaCppEmbeddings(model_path="ggml-model.bin")
31
- # embeddings = HuggingFaceEmbeddings(model_name="diptanuc/all-mpnet-base-v2", model_kwargs={'device': 'cpu'})
32
- # embeddings = TensorflowHubEmbeddings(model_url="https://tfhub.dev/google/universal-sentence-encoder/4")
33
  embeddings = TensorflowHubEmbeddings(model_url="https://tfhub.dev/google/universal-sentence-encoder-multilingual-qa/3")
34
- # embeddings = HuggingFaceEmbeddings(model_name="obrizum/all-MiniLM-L6-v2", model_kwargs={'device': 'cpu'})
35
  end = time.time()
36
  st.text("Embedding time: "+str(round(end - start,1)))
37
  start = time.time()
 
25
  st.text("Load and split text: "+str(round(end - start,1)))
26
 
27
 
 
28
  start = time.time()
 
 
 
29
  embeddings = TensorflowHubEmbeddings(model_url="https://tfhub.dev/google/universal-sentence-encoder-multilingual-qa/3")
 
30
  end = time.time()
31
  st.text("Embedding time: "+str(round(end - start,1)))
32
  start = time.time()
query_data.py CHANGED
@@ -1,40 +1,17 @@
1
- from langchain.prompts.prompt import PromptTemplate
2
  from langchain.llms import LlamaCpp
3
  from langchain.chains import ConversationalRetrievalChain
4
- from langchain.memory import ConversationBufferMemory
5
  from huggingface_hub import hf_hub_download
6
 
7
  import psutil
8
  import os
9
 
10
- #_template = """Given the following conversation and a follow up question, rephrase the follow up question to be a standalone question.
11
- #You can assume the question about the uploaded document.
12
 
13
- #Chat History:
14
- #{chat_history}
15
- #Follow Up Input: {question}
16
- #Standalone question:"""
17
- #CONDENSE_QUESTION_PROMPT = PromptTemplate.from_template(_template)
18
-
19
- #template = """You are an AI assistant for answering questions about the uploaded document.
20
- #You are given the following extracted parts of a long document and a question. Provide a conversational answer.
21
- #If you don't know the answer, just say "Hmm, I'm not sure." Don't try to make up an answer.
22
- #If the question is not about the uploaded document, politely inform them that you are tuned to only answer questions about the uploaded document.
23
- #Question: {question}
24
-
25
- #Answer in Markdown:"""
26
- ##QA_PROMPT = PromptTemplate(template=template, input_variables=["question", "context"])
27
- #QA_PROMPT = PromptTemplate(template=template, input_variables=["question"])
28
-
29
- #=========
30
- #{context}
31
- #=========
32
 
33
 
34
  def get_chain(vectorstore):
35
  if not os.path.exists("ggml-vic7b-q5_1.bin"):
36
  hf_hub_download(repo_id="eachadea/ggml-vicuna-7b-1.1", filename="ggml-vic7b-q5_1.bin", local_dir=".")
37
- llm = LlamaCpp(model_path="ggml-vic7b-q5_1.bin", n_ctx=2048, n_threads=psutil.cpu_count(logical=False)/2)
38
  qa_chain = ConversationalRetrievalChain.from_llm(
39
  llm,
40
  vectorstore.as_retriever(),
 
 
1
  from langchain.llms import LlamaCpp
2
  from langchain.chains import ConversationalRetrievalChain
 
3
  from huggingface_hub import hf_hub_download
4
 
5
  import psutil
6
  import os
7
 
 
 
8
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
9
 
10
 
11
  def get_chain(vectorstore):
12
  if not os.path.exists("ggml-vic7b-q5_1.bin"):
13
  hf_hub_download(repo_id="eachadea/ggml-vicuna-7b-1.1", filename="ggml-vic7b-q5_1.bin", local_dir=".")
14
+ llm = LlamaCpp(model_path="ggml-vic7b-q5_1.bin", n_ctx=2048, n_threads=psutil.cpu_count(logical=False))
15
  qa_chain = ConversationalRetrievalChain.from_llm(
16
  llm,
17
  vectorstore.as_retriever(),