Spaces:
Runtime error
Runtime error
lalanikarim
commited on
Commit
•
1e62eb5
1
Parent(s):
c7dfe89
added hugging face hub for model download and caching
Browse files- main.py +6 -16
- requirements.txt +1 -0
main.py
CHANGED
@@ -1,6 +1,4 @@
|
|
1 |
import streamlit as st
|
2 |
-
import os
|
3 |
-
from urllib.request import urlretrieve
|
4 |
from operator import itemgetter
|
5 |
from langchain.llms import LlamaCpp
|
6 |
from langchain.prompts import ChatPromptTemplate, MessagesPlaceholder
|
@@ -8,6 +6,7 @@ from langchain.memory import ConversationBufferMemory
|
|
8 |
# from langchain.callbacks.manager import CallbackManager
|
9 |
from langchain.callbacks.base import BaseCallbackHandler
|
10 |
from langchain.schema.runnable import RunnablePassthrough, RunnableLambda
|
|
|
11 |
|
12 |
|
13 |
# StreamHandler to intercept streaming output from the LLM.
|
@@ -38,21 +37,12 @@ def create_chain(system_prompt):
|
|
38 |
# responses in real time.
|
39 |
# callback_manager = CallbackManager([stream_handler])
|
40 |
|
41 |
-
|
42 |
-
|
43 |
-
"https://huggingface.co/TheBloke/Mistral-7B-Instruct-v0.1-GGUF/resolve/main/mistral-7b-instruct-v0.1.Q4_0.gguf",
|
44 |
-
"mistral-7b-instruct-v0.1.Q4_0.gguf")
|
45 |
|
46 |
-
|
47 |
-
|
48 |
-
|
49 |
-
|
50 |
-
# set model_path before providing to LlamaCpp
|
51 |
-
model_path = "models/" + model_name
|
52 |
-
|
53 |
-
# download model to model_path if missing
|
54 |
-
if not os.path.exists(model_path):
|
55 |
-
urlretrieve(url, model_path)
|
56 |
|
57 |
# initialize LlamaCpp llm model
|
58 |
llm = LlamaCpp(
|
|
|
1 |
import streamlit as st
|
|
|
|
|
2 |
from operator import itemgetter
|
3 |
from langchain.llms import LlamaCpp
|
4 |
from langchain.prompts import ChatPromptTemplate, MessagesPlaceholder
|
|
|
6 |
# from langchain.callbacks.manager import CallbackManager
|
7 |
from langchain.callbacks.base import BaseCallbackHandler
|
8 |
from langchain.schema.runnable import RunnablePassthrough, RunnableLambda
|
9 |
+
from huggingface_hub import hf_hub_download
|
10 |
|
11 |
|
12 |
# StreamHandler to intercept streaming output from the LLM.
|
|
|
37 |
# responses in real time.
|
38 |
# callback_manager = CallbackManager([stream_handler])
|
39 |
|
40 |
+
(repo_id, model_file_name) = ("TheBloke/Mistral-7B-Instruct-v0.1-GGUF",
|
41 |
+
"mistral-7b-instruct-v0.1.Q5_0.gguf")
|
|
|
|
|
42 |
|
43 |
+
model_path = hf_hub_download(repo_id=repo_id,
|
44 |
+
filename=model_file_name,
|
45 |
+
repo_type="model")
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
46 |
|
47 |
# initialize LlamaCpp llm model
|
48 |
llm = LlamaCpp(
|
requirements.txt
CHANGED
@@ -1,3 +1,4 @@
|
|
1 |
langchain==0.0.321
|
2 |
llama_cpp_python==0.2.11
|
|
|
3 |
streamlit==1.27.2
|
|
|
1 |
langchain==0.0.321
|
2 |
llama_cpp_python==0.2.11
|
3 |
+
huggingface_hub==0.18.0
|
4 |
streamlit==1.27.2
|