Spaces:
Sleeping
Sleeping
Upload 2 files
Browse files- app_anno.py +146 -0
- requirements.txt +5 -0
app_anno.py
ADDED
@@ -0,0 +1,146 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
1 |
+
import pandas as pd
|
2 |
+
import streamlit as st
|
3 |
+
from langchain import PromptTemplate, HuggingFaceHub, LLMChain
|
4 |
+
from langchain.llms import OpenAI
|
5 |
+
from transformers import AutoTokenizer, AutoModelForSequenceClassification
|
6 |
+
import os
|
7 |
+
import re
|
8 |
+
|
9 |
+
|
10 |
+
def extract_positive_negative(text):
|
11 |
+
pattern = r'\b(?:positive|negative)\b'
|
12 |
+
result = re.findall(pattern, text)
|
13 |
+
return result
|
14 |
+
|
15 |
+
def classify_text(text, llm_chain, api):
|
16 |
+
if api == "HuggingFace":
|
17 |
+
classification = llm_chain.run(str(text))
|
18 |
+
elif api == "OpenAI":
|
19 |
+
classification = llm_chain.run(str(text))
|
20 |
+
classification = re.sub(r'\s', '', classification)
|
21 |
+
return classification.lower()
|
22 |
+
|
23 |
+
def classify_csv(df, llm_chain, api):
|
24 |
+
df["label_gold"] = df["label"]
|
25 |
+
del df["label"]
|
26 |
+
df["label_pred"] = df["text"].apply(classify_text, llm_chain=llm_chain, api=api)
|
27 |
+
return df
|
28 |
+
|
29 |
+
def classify_csv_zero(zero_file, llm_chain, api):
|
30 |
+
df = pd.read_csv(zero_file, sep=';')
|
31 |
+
df["label"] = df["text"].apply(classify_text, llm_chain=llm_chain, api=api)
|
32 |
+
return df
|
33 |
+
|
34 |
+
def evaluate_performance(df):
|
35 |
+
merged_df = df
|
36 |
+
correct_preds = sum(merged_df["label_gold"] == merged_df["label_pred"])
|
37 |
+
total_preds = len(merged_df)
|
38 |
+
percentage_overlap = correct_preds / total_preds * 100
|
39 |
+
|
40 |
+
return percentage_overlap
|
41 |
+
|
42 |
+
def display_home():
|
43 |
+
st.write("Please select an API and a model to classify the text. We currently support HuggingFace and OpenAI.")
|
44 |
+
api = st.selectbox("Select an API", ["HuggingFace", "OpenAI"])
|
45 |
+
|
46 |
+
if api == "HuggingFace":
|
47 |
+
model = st.selectbox("Select a model", ["google/flan-t5-xl", "databricks/dolly-v1-6b"])
|
48 |
+
api_key_hug = st.text_input("HuggingFace API Key")
|
49 |
+
elif api == "OpenAI":
|
50 |
+
model = None
|
51 |
+
api_key_openai = st.text_input("OpenAI API Key")
|
52 |
+
|
53 |
+
st.write("Please select a temperature for the model. The higher the temperature, the more creative the model will be.")
|
54 |
+
temperature = st.slider("Set the temperature", min_value=0.0, max_value=1.0, value=0.0, step=0.01)
|
55 |
+
|
56 |
+
st.write("We provide two different setups for the annotation task. In the first setup (**Test**), you can upload a CSV file with gold labels and evaluate the performance of the model. In the second setup (**Zero-Shot**), you can upload a CSV file without gold labels and use the model to classify the text.")
|
57 |
+
setup = st.selectbox("Setup", ["Test", "Zero-Shot"])
|
58 |
+
|
59 |
+
if setup == "Test":
|
60 |
+
gold_file = st.file_uploader("Upload Gold Labels CSV file with a text and a label column", type=["csv"])
|
61 |
+
elif setup == "Zero-Shot":
|
62 |
+
gold_file = None
|
63 |
+
zero_file = st.file_uploader("Upload CSV file with a text column", type=["csv"])
|
64 |
+
|
65 |
+
st.write("Please enter the prompt template below. You can use the following variables: {text} (text to classify).")
|
66 |
+
prompt_template = st.text_area("Enter your task description", """Instruction: Identify the sentiment of a text. Please read the text and provide one of these responses: "positive" or "negative".\nText to classify in "positive" or "negative": {text}\nAnswer:""", height=200)
|
67 |
+
|
68 |
+
classify_button = st.button("Run Classification/ Annotation")
|
69 |
+
|
70 |
+
if classify_button:
|
71 |
+
if prompt_template:
|
72 |
+
prompt = PromptTemplate(
|
73 |
+
template=prompt_template,
|
74 |
+
input_variables=["text"]
|
75 |
+
)
|
76 |
+
|
77 |
+
if api == "HuggingFace":
|
78 |
+
if api_key_hug:
|
79 |
+
os.environ["HUGGINGFACEHUB_API_TOKEN"] = api_key_hug
|
80 |
+
llm_chain = LLMChain(prompt=prompt, llm=HuggingFaceHub(repo_id=model, model_kwargs={"temperature": temperature, "max_length": 128}))
|
81 |
+
elif not api_key_hug:
|
82 |
+
st.warning("Please enter your HuggingFace API key to classify the text.")
|
83 |
+
elif api == "OpenAI":
|
84 |
+
if api_key_openai:
|
85 |
+
os.environ["OPENAI_API_KEY"] = api_key_openai
|
86 |
+
llm_chain = LLMChain(prompt=prompt, llm=OpenAI(temperature=temperature))
|
87 |
+
elif not api_key_openai:
|
88 |
+
st.warning("Please enter your OpenAI API key to classify the text.")
|
89 |
+
|
90 |
+
if setup == "Zero-Shot":
|
91 |
+
if zero_file is not None:
|
92 |
+
df_predicted = classify_csv_zero(zero_file, llm_chain, api)
|
93 |
+
st.write(df_predicted)
|
94 |
+
st.download_button(
|
95 |
+
label="Download CSV",
|
96 |
+
data=df_predicted.to_csv(index=False),
|
97 |
+
file_name="classified_zero-shot_data.csv",
|
98 |
+
mime="text/csv"
|
99 |
+
)
|
100 |
+
elif setup == "Test":
|
101 |
+
if gold_file is not None:
|
102 |
+
df = pd.read_csv(gold_file, sep=';')
|
103 |
+
if "label" not in df.columns:
|
104 |
+
st.warning("Please make sure that the gold labels CSV file contains a column named 'label'.")
|
105 |
+
else:
|
106 |
+
df = classify_csv(df, llm_chain, api)
|
107 |
+
st.write(df)
|
108 |
+
st.download_button(
|
109 |
+
label="Download CSV",
|
110 |
+
data=df.to_csv(index=False),
|
111 |
+
file_name="classified_test_data.csv",
|
112 |
+
mime="text/csv"
|
113 |
+
)
|
114 |
+
percentage_overlap = evaluate_performance(df)
|
115 |
+
st.write("**Performance Evaluation**")
|
116 |
+
st.write(f"Percentage overlap between gold labels and predicted labels: {percentage_overlap:.2f}%")
|
117 |
+
elif gold_file is None:
|
118 |
+
st.warning("Please upload a gold labels CSV file to evaluate the performance of the model.")
|
119 |
+
elif not prompt:
|
120 |
+
st.warning("Please enter a prompt question to classify the text.")
|
121 |
+
|
122 |
+
def main():
|
123 |
+
st.set_page_config(page_title="PromptCards Playground", page_icon=":pencil2:")
|
124 |
+
st.title("AInnotator")
|
125 |
+
|
126 |
+
# add a menu to the sidebar
|
127 |
+
if "current_page" not in st.session_state:
|
128 |
+
st.session_state.current_page = "homepage"
|
129 |
+
|
130 |
+
# Initialize selected_prompt in session_state if not set
|
131 |
+
if "selected_prompt" not in st.session_state:
|
132 |
+
st.session_state.selected_prompt = ""
|
133 |
+
|
134 |
+
# Add a menu
|
135 |
+
menu = ["Homepage", "Playground", "Prompt Archive", "Annotator", "About"]
|
136 |
+
st.sidebar.title("About")
|
137 |
+
st.sidebar.write("AInnotator 🤖🏷️ is a tool for creating artificial labels/ annotations. It is based on the concept of PromptCards, which are small, self-contained descriptions of a task that can be used to generate labels for a wide range of NLP tasks. Check out the GitHub repository and the PromptCards Archive for more information.")
|
138 |
+
st.sidebar.write("---")
|
139 |
+
st.sidebar.write("Check out the [PromptCards archive]() to find a wide range of prompts for different NLP tasks.")
|
140 |
+
st.sidebar.write("---")
|
141 |
+
st.sidebar.write("Made with ❤️ and 🤖.")
|
142 |
+
|
143 |
+
display_home()
|
144 |
+
|
145 |
+
if __name__ == "__main__":
|
146 |
+
main()
|
requirements.txt
ADDED
@@ -0,0 +1,5 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
1 |
+
langchain
|
2 |
+
pandas
|
3 |
+
streamlit
|
4 |
+
transformers
|
5 |
+
sklearn
|