Hugging Face's logo

Spaces:
jordyvl
/
ece
Configuration error

App Files Community
1
ece
File size: 4,919 Bytes 
b81628e
3f722df
 
 
 
 
b81628e
3f722df
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
b81628e
0c94397
543b514
a3659c1
 
 
543b514
 
 
 
3f722df
 
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
56
57
58
59
60
61
62
63
64
65
66
67
68
69
70
71
72
73
74
75
76
77
78
79
80
81
82
83
84
85
86
87
88
89
90
91
92
93
94
95
96
97
98
99
100
101
102
103
104
105
106
107
108
109
110
111
112
113
114
115
116
117
118
119
120
121
122
123
124
125
126
127
128
129
130
131
132
133
134
135
136
137
138
139
140
141
142
143
144
145
146
147
148
149
150
151
152
153
154
155
156
157
158
159
160
161
162
163
164
165
166
167
168
169
170
171
172
173
174
 
import evaluate
import numpy as np
import pandas as pd
import ast
import json
import gradio as gr
from evaluate.utils import launch_gradio_widget
from ece import ECE


sliders = [
    gr.Slider(0, 100, value=10, label="n_bins"),
    gr.Slider(0, 100, value=None, label="bin_range", visible=False), #DEV: need to have a double slider
    gr.Dropdown(choices=["equal-range", "equal-mass"], value="equal-range", label="scheme"),
    gr.Dropdown(choices=["upper-edge", "center"], value="upper-edge", label="proxy"),
    gr.Dropdown(choices=[1, 2, np.inf], value=1, label="p"),
]

slider_defaults = [slider.value for slider in sliders]

# example data
df = dict()
df["predictions"] = [[0.6, 0.2, 0.2], [0, 0.95, 0.05], [0.7, 0.1, 0.2]]
df["references"] = [0, 1, 2]

component = gr.inputs.Dataframe(
    headers=["predictions", "references"], col_count=2, datatype="number", type="pandas"
)

component.value = [
    [[0.6, 0.2, 0.2], 0],
    [[0.7, 0.1, 0.2], 2],
    [[0, 0.95, 0.05], 1],
]
sample_data = [[component] + slider_defaults]  ##json.dumps(df)


metric = ECE()
# module = evaluate.load("jordyvl/ece")
# launch_gradio_widget(module)

"""
Switch inputs and compute_fn
"""

def reliability_plot(results):
    #CE, calibrated_acc, empirical_acc, weights_ece
    #{"ECE": ECE[0], "y_bar": ECE[1], "p_bar": ECE[2], "bin_freq": ECE[3]}
    import matplotlib.pyplot as plt
    import seaborn as sns
    sns.set_style('white')
    sns.set_context("paper", font_scale=1)  # 2
    # plt.rcParams['figure.figsize'] = [10, 7]
    plt.rcParams['figure.dpi'] = 300


    fig = plt.figure()
    ax1 = plt.subplot2grid((3, 1), (0, 0), rowspan=2)
    ax2 = plt.subplot2grid((3, 1), (2, 0))

    n_bins = len(results["y_bar"])
    bin_range = [
        results["y_bar"][0] - results["y_bar"][0],
        results["y_bar"][-1],
    ]  # np.linspace(0, 1, n_bins)
    # if upper edge then minus binsize; same for center [but half]

    ax1.plot(
        np.linspace(bin_range[0], bin_range[1], n_bins),
        np.linspace(bin_range[0], bin_range[1], n_bins),
        color="darkgreen",
        ls="dotted",
        label="Perfect",
    )
    # ax1.plot(results["y_bar"], results["y_bar"], color="darkblue", label="Perfect")

    anindices = np.where(~np.isnan(results["p_bar"][:-1]))[0]
    bin_freqs = np.zeros(n_bins)
    bin_freqs[anindices] = results["bin_freq"]
    ax2.hist(results["y_bar"], results["y_bar"], weights=bin_freqs)

    widths = np.diff(results["y_bar"])
    for j, bin in enumerate(results["y_bar"]):
        perfect = results["y_bar"][j]
        empirical = results["p_bar"][j]

        if np.isnan(empirical):
            continue

        ax1.bar([perfect], height=[empirical], width=-widths[j], align="edge", color="lightblue")

        if perfect == empirical:
            continue

    acc_plt = ax2.axvline(
        x=results["accuracy"], ls="solid", lw=3, c="black", label="Accuracy"
    )
    conf_plt = ax2.axvline(
        x=results["p_bar_cont"], ls="dotted", lw=3, c="#444", label="Avg. confidence"
    )
    ax2.legend(handles=[acc_plt, conf_plt])

    #Bin differences
    ax1.set_ylabel("Conditional Expectation")
    ax1.set_ylim([-0.05, 1.05]) #respective to bin range
    ax1.legend(loc="lower right")
    ax1.set_title("Reliability Diagram")

    #Bin frequencies
    ax2.set_xlabel("Confidence")
    ax2.set_ylabel("Count")
    ax2.legend(loc="upper left")#, ncol=2
    plt.tight_layout()
    return fig

def compute_and_plot(data, n_bins, bin_range, scheme, proxy, p):
    # DEV: check on invalid datatypes with better warnings

    if isinstance(data, pd.DataFrame):
        data.dropna(inplace=True)

    predictions = [
        ast.literal_eval(prediction) if not isinstance(prediction, list) else prediction
        for prediction in data["predictions"]
    ]
    references = [reference for reference in data["references"]]

    results = metric._compute(
        predictions,
        references,
        n_bins=n_bins,
        # bin_range=None,#not needed
        scheme=scheme,
        proxy=proxy,
        p=p,
        detail=True,
    )

    plot = reliability_plot(results)
    return results["ECE"], plt.gcf()


outputs = [gr.outputs.Textbox(label="ECE"), gr.outputs.Plot(label="Reliability diagram")]

iface = gr.Interface(
    fn=compute_and_plot,
    inputs=[component] + sliders,
    outputs=outputs,
    description=metric.info.description,
    article=metric.info.citation,
    # examples=sample_data
)

# ValueError: Examples argument must either be a directory or a nested list, where each sublist represents a set of inputs.

iface.launch()

# dict = {"ECE": ECE[0], "y_bar": ECE[1], "p_bar": ECE[2], "bin_freq": ECE[3]}

# references=[0, 1, 2], predictions=)
# https://gradio.app/getting_started/#multiple-inputs-and-outputs
## fix with sliders for all kwargs


"""
DEV: #might be nice to also plot reliability diagram
have sliders for kwargs :)


metric = ECE()

 
"""