### LIBRARIES ### # # Data from matplotlib.pyplot import legend import numpy as np import pandas as pd import torch import json from tqdm import tqdm from math import floor from datasets import load_dataset from collections import defaultdict from transformers import AutoTokenizer # Analysis # from gensim.models.doc2vec import Doc2Vec # from sklearn.metrics import accuracy_score, f1_score, precision_score, recall_score # import nltk # nltk.download('punkt') #make sure that punkt is downloaded # App & Visualization import streamlit as st import st_aggrid import altair as alt import plotly.graph_objects as go from streamlit_vega_lite import altair_component # utils from random import sample # from PIL import Image def down_samp(embedding): """Down sample a data frame for altiar visualization """ # total number of positive and negative sentiments in the class #embedding = embedding.groupby('slice').apply(lambda x: x.sample(frac=0.3)) total_size = embedding.groupby(['slice','label'], as_index=False).count() user_data = 0 # if 'Your Sentences' in str(total_size['slice']): # tmp = embedding.groupby(['slice'], as_index=False).count() # val = int(tmp[tmp['slice'] == "Your Sentences"]['source']) # user_data = val max_sample = total_size.groupby('slice').max()['content'] # # down sample to meeting altair's max values # # but keep the proportional representation of groups down_samp = 1/(sum(max_sample.astype(float))/(1000-user_data)) max_samp = max_sample.apply(lambda x: floor(x*down_samp)).astype(int).to_dict() max_samp['Your Sentences'] = user_data # # sample down for each group in the data frame embedding = embedding.groupby('slice').apply(lambda x: x.sample(n=max_samp.get(x.name))).reset_index(drop=True) # # order the embedding return(embedding) def data_comparison(df): # set up a dropdown select bindinf # input_dropdown = alt.binding_select(options=['Negative Sentiment','Positive Sentiment']) selection = alt.selection_multi(fields=['slice','label']) color = alt.condition(alt.datum.slice == 'high-loss', alt.value("orange"), alt.value("steelblue")) # color = alt.condition(selection, # alt.Color('slice:Q', legend=None), # # scale = alt.Scale(domain = pop_domain,range=color_range)), # alt.value('lightgray')) opacity = alt.condition(selection, alt.value(0.7), alt.value(0.25)) # basic chart scatter = alt.Chart(df).mark_point(size=100, filled=True).encode( x=alt.X('x', axis=None), y=alt.Y('y', axis=None), color=color, shape=alt.Shape('label', scale=alt.Scale(range=['circle', 'diamond'])), tooltip=['slice','content','label','pred'], opacity=opacity ).properties( width=1500, height=1000 ).interactive() legend = alt.Chart(df).mark_point().encode( y=alt.Y('slice:N', axis=alt.Axis(orient='right'), title=""), x=alt.X("label"), shape=alt.Shape('label', scale=alt.Scale( range=['circle', 'diamond']), legend=None), color=color ).add_selection( selection ) layered = scatter |legend layered = layered.configure_axis( grid=False ).configure_view( strokeOpacity=0 ) return layered def quant_panel(embedding_df): """ Quantitative Panel Layout""" all_metrics = {} # st.warning("**Data Comparison**") # with st.expander("how to read this chart:"): # st.markdown("* each **point** is a single sentence") # st.markdown("* the **position** of each dot is determined mathematically based upon an analysis of the words in a sentence. The **closer** two points on the visualization the **more similar** the sentences are. The **further apart ** two points on the visualization the **more different** the sentences are") # st.markdown( # " * the **shape** of each point reflects whether it a positive (diamond) or negative sentiment (circle)") # st.markdown("* the **color** of each point is the ") st.altair_chart(data_comparison(down_samp(embedding_df))) def frequent_tokens(data, tokenizer, loss_quantile=0.95, top_k=200, smoothing=0.005): unique_tokens = [] tokens = [] for row in tqdm(data['content']): tokenized = tokenizer(row,padding=True, return_tensors='pt') tokens.append(tokenized['input_ids'].flatten()) unique_tokens.append(torch.unique(tokenized['input_ids'])) losses = data['loss'].astype(float) high_loss = losses.quantile(loss_quantile) loss_weights = (losses > high_loss) loss_weights = loss_weights / loss_weights.sum() token_frequencies = defaultdict(float) token_frequencies_error = defaultdict(float) weights_uniform = np.full_like(loss_weights, 1 / len(loss_weights)) num_examples = len(data) for i in tqdm(range(num_examples)): for token in unique_tokens[i]: token_frequencies[token.item()] += weights_uniform[i] token_frequencies_error[token.item()] += loss_weights[i] token_lrs = {k: (smoothing+token_frequencies_error[k]) / (smoothing+token_frequencies[k]) for k in token_frequencies} tokens_sorted = list(map(lambda x: x[0], sorted(token_lrs.items(), key=lambda x: x[1])[::-1])) top_tokens = [] for i, (token) in enumerate(tokens_sorted[:top_k]): top_tokens.append(['%10s' % (tokenizer.decode(token)), '%.4f' % (token_frequencies[token]), '%.4f' % ( token_frequencies_error[token]), '%4.2f' % (token_lrs[token])]) return pd.DataFrame(top_tokens, columns=['Token', 'Freq', 'Freq error slice', 'lrs']) @st.cache(ttl=600) def get_data(spotlight, emb): preds = spotlight.outputs.numpy() losses = spotlight.losses.numpy() embeddings = pd.DataFrame(emb, columns=['x', 'y']) num_examples = len(losses) # dataset_labels = [dataset[i]['label'] for i in range(num_examples)] return pd.concat([pd.DataFrame(np.transpose(np.vstack([dataset[:num_examples]['content'], dataset[:num_examples]['label'], preds, losses])), columns=['content', 'label', 'pred', 'loss']), embeddings], axis=1) def topic_distribution(weights, smoothing=0.01): topic_frequencies = defaultdict(float) topic_frequencies_spotlight = defaultdict(float) weights_uniform = np.full_like(weights, 1 / len(weights)) num_examples = len(weights) for i in range(num_examples): example = dataset[i] category = example['title'] topic_frequencies[category] += weights_uniform[i] topic_frequencies_spotlight[category] += weights[i] topic_ratios = {c: (smoothing + topic_frequencies_spotlight[c]) / ( smoothing + topic_frequencies[c]) for c in topic_frequencies} categories_sorted = map(lambda x: x[0], sorted( topic_ratios.items(), key=lambda x: x[1], reverse=True)) topic_distr = [] for category in categories_sorted: topic_distr.append(['%.3f' % topic_frequencies[category], '%.3f' % topic_frequencies_spotlight[category], '%.2f' % topic_ratios[category], '%s' % category]) return pd.DataFrame(topic_distr, columns=['Overall frequency', 'Error frequency', 'Ratio', 'Category']) # for category in categories_sorted: # return(topic_frequencies[category], topic_frequencies_spotlight[category], topic_ratios[category], category) if __name__ == "__main__": ### STREAMLIT APP CONGFIG ### st.set_page_config(layout="wide", page_title="Error Slice Analysis") lcol, rcol = st.columns([3, 2]) # ******* loading the mode and the data dataset = st.sidebar.selectbox( "Dataset", ["amazon_polarity", "squad", "movielens", "waterbirds"], index=0 ) tokenizer = AutoTokenizer.from_pretrained( "distilbert-base-uncased-finetuned-sst-2-english") model = st.sidebar.selectbox( "Model", ["distilbert-base-uncased-finetuned-sst-2-english", "distilbert-base-uncased-finetuned-sst-2-english"], index=0 ) loss_quantile = st.sidebar.selectbox( "Loss Quantile", [0.98, 0.95, 0.9, 0.8, 0.75], index = 1 ) ### LOAD DATA AND SESSION VARIABLES ### data_df = pd.read_parquet('amazon_polarity.test.parquet') embedding_umap = data_df[['x','y']] if "user_data" not in st.session_state: st.session_state["user_data"] = data_df if "selected_slice" not in st.session_state: st.session_state["selected_slice"] = None if "embedding" not in st.session_state: st.session_state["embedding"] = embedding_umap with lcol: st.title('Error Slices') dataframe = data_df[['content', 'label', 'pred', 'loss']].sort_values( by=['loss'], ascending=False) table_html = dataframe.to_html( columns=['content', 'label', 'pred', 'loss'], max_rows=100) # table_html = table_html.replace("", '') # left-align the headers st.write(dataframe) # st_aggrid.AgGrid(dataframe) # table_html = dataframe.to_html(columns=['content', 'label', 'pred', 'loss'], max_rows=100) # table_html = table_html.replace("", '') # left-align the headers # st.write(table_html) with rcol: st.title('Word Distribution in Error Slice') commontokens = frequent_tokens(data_df, tokenizer, loss_quantile=loss_quantile) st.write(commontokens) data_df['loss'] = data_df['loss'].astype(float) losses = data_df['loss'] high_loss = losses.quantile(loss_quantile) data_df['slice'] = 'high-loss' data_df['slice'] = data_df['slice'].where(data_df['loss'] > high_loss, 'low-loss') quant_panel(data_df)