Spaces:
Paused
Paused
Updated app.py
Browse files
app.py
CHANGED
@@ -1,8 +1,3 @@
|
|
1 |
-
import os
|
2 |
-
os.system("langchain upgrade-imports")
|
3 |
-
os.system("pip install accelerate")
|
4 |
-
os.system("pip install -i https://pypi.org/simple/ bitsandbytes")
|
5 |
-
|
6 |
import gradio as gr
|
7 |
# import fitz # PyMuPDF for extracting text from PDFs
|
8 |
from langchain.embeddings import HuggingFaceEmbeddings
|
@@ -61,12 +56,14 @@ query_pipeline = transformers.pipeline(
|
|
61 |
return_full_text=True,
|
62 |
torch_dtype=torch.float16,
|
63 |
device_map=device,
|
64 |
-
|
|
|
65 |
top_p=0.9,
|
66 |
top_k=50,
|
67 |
max_new_tokens=256
|
68 |
)
|
69 |
|
|
|
70 |
llm = HuggingFacePipeline(pipeline=query_pipeline)
|
71 |
|
72 |
books_db_client_retriever = RetrievalQA.from_chain_type(
|
@@ -91,7 +88,9 @@ def test_rag(query):
|
|
91 |
return corrected_text_books
|
92 |
|
93 |
# Define the Gradio interface
|
94 |
-
def chat(query, history=
|
|
|
|
|
95 |
answer = test_rag(query)
|
96 |
history.append((query, answer))
|
97 |
return history, history
|
|
|
|
|
|
|
|
|
|
|
|
|
1 |
import gradio as gr
|
2 |
# import fitz # PyMuPDF for extracting text from PDFs
|
3 |
from langchain.embeddings import HuggingFaceEmbeddings
|
|
|
56 |
return_full_text=True,
|
57 |
torch_dtype=torch.float16,
|
58 |
device_map=device,
|
59 |
+
do_sample=True, # Enable sampling
|
60 |
+
temperature=0.7, # Keep if sampling is used
|
61 |
top_p=0.9,
|
62 |
top_k=50,
|
63 |
max_new_tokens=256
|
64 |
)
|
65 |
|
66 |
+
|
67 |
llm = HuggingFacePipeline(pipeline=query_pipeline)
|
68 |
|
69 |
books_db_client_retriever = RetrievalQA.from_chain_type(
|
|
|
88 |
return corrected_text_books
|
89 |
|
90 |
# Define the Gradio interface
|
91 |
+
def chat(query, history=None):
|
92 |
+
if history is None:
|
93 |
+
history = []
|
94 |
answer = test_rag(query)
|
95 |
history.append((query, answer))
|
96 |
return history, history
|