File size: 2,198 Bytes
8aac646
 
 
 
 
 
b19b634
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
8aac646
 
b19b634
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
8aac646
b19b634
8aac646
b19b634
 
 
8aac646
b19b634
 
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
import streamlit as st
from datasets import load_dataset
import os 

HF_TOKEN = os.environ.get("HF_TOKEN", None)

st.set_page_config(page_title="FW Clusters inspection", layout="wide")
st.title("FW clusters inspection")

st.markdown("""
We clustered 100k FineWeb samples using [text-clustering](https://github.com/huggingface/text-clustering). 

Our approach involved prompting Mixtral to evaluate whether the topics in each cluster are educational or could be considered college material. 

Additionally, the model was tasked with assigning a category to each cluster from 23 predefined categories found in [AFAIK](https://afaik.io/). 

Sometimes, the model may define its own category. This can happen either within the context of AFAIK topics seperately. Hence the `Select Category Type` dropdown in our interface.
""")

@st.cache_data
def load_data(educational_topic):
    ds = load_dataset("HuggingFaceTB/FW_clusters_under_afaik_topics", split="train", token=HF_TOKEN, num_proc=2)
    if educational_topic in ['Yes', 'No']:
        ds = ds.filter(lambda x: x['is_topic_educational'] == educational_topic)
    return ds

@st.cache_data
def get_categories_by_type(_ds, category_type):
    filtered_ds = _ds.filter(lambda x: x['category_type'] == category_type)
    return list(set(filtered_ds['category']))


st.subheader("Cluster information")
col_1, col_2, col_3 = st.columns(3)
with col_1:
    educational_topic = st.selectbox('Are the topics deemed educational by the LLM?', ["Yes", "No"])

ds = load_data(educational_topic)

with col_2:
    category_types = ['afaik', 'defined_by_llm', 'defined_by_llm_under_afaik']
    selected_category_type = st.selectbox("Select Category Type", category_types)
with col_3:
    categories = get_categories_by_type(ds, selected_category_type)
    selected_category = st.selectbox("Select Category", categories)

selected_cluster = ds.filter(lambda x: x['category'] == selected_category)

# Select sample index
n_samples = len(selected_cluster["examples"])
index_example = st.number_input(f"Index of a sample: 0 - {n_samples}",  min_value=0, max_value=n_samples-1, value=0, step=1)

sample = selected_cluster["examples"][index_example]
st.markdown(sample)