-
Notifications
You must be signed in to change notification settings - Fork 2
/
data_retention.py
217 lines (171 loc) · 6.22 KB
/
data_retention.py
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
175
176
177
178
179
180
181
182
183
184
185
186
187
188
189
190
191
192
193
194
195
196
197
198
199
200
201
202
203
204
205
206
207
208
209
210
211
212
213
214
215
216
217
import argparse
import json
import os
import pandas as pd
from src.common.plotting import plot_rouge_retention, plot_increase, plot_bleuvar_retention
def read_args():
parser = argparse.ArgumentParser()
parser.add_argument("--data", nargs='+', help="")
parser.add_argument("--root_path", type=str, help="")
parser.add_argument("--models", nargs='+', help="")
parser.add_argument("--n_list", nargs='+', type=int, help="")
parser.add_argument("--bases", nargs='+', help="")
args, unknown = parser.parse_known_args()
return args, unknown
def metrics_change(df):
""""""
num_samples = len(df)
percent_increase = int(num_samples / 1000)
metrics_change_df = df.expanding().mean()
metrics_change_df = metrics_change_df[1:].iloc[::percent_increase, :].reset_index(drop=True)
return metrics_change_df
def load_runs(run_list, n_list):
""""""
run_dfs = []
for run_path, n in zip(run_list, n_list):
try:
df = pd.read_csv(run_path, sep="\t")
sorted_df = df.sort_values(by=['bleuvar'], ignore_index=True)
except FileNotFoundError:
sorted_df = None
run_dfs.append(sorted_df)
return run_dfs
def load_bases(base_list):
""""""
agg_base_dfs = []
base_dfs = []
for base_path in base_list:
try:
base_df = pd.read_csv(base_path, sep="\t")
agg_base_df = base_df.agg(['mean', 'std'])
except FileNotFoundError:
base_df = None
agg_base_df = base_df
base_dfs.append(base_df)
agg_base_dfs.append(agg_base_df)
return base_dfs, agg_base_dfs
def norm_bleuvar(runs, n_list):
""""""
scaled_metrics_dfs = []
for metrics_change_df, n in zip(runs, n_list):
if metrics_change_df is not None:
metrics_change_df["bleuvar_scaled"] = metrics_change_df["bleuvar"].apply(lambda x: x / (n * (n - 1)))
scaled_metrics_dfs.append(metrics_change_df)
return scaled_metrics_dfs
def join_runs(runs, run_models, base_runs, base_models):
""""""
joined_dfs = []
for model, metrics_df in zip(run_models, runs):
base_df = base_runs[base_models.index(model)]
if base_df is not None:
join_df = metrics_df.merge(base_df, on="article_id", suffixes=["_var", "_std"])
else:
join_df = base_df
joined_dfs.append(join_df)
return joined_dfs
def compute_diffs(joined_dfs, metrics):
""""""
diff_dfs = []
for join_df in joined_dfs:
if join_df is not None:
for metric in metrics:
join_df[f"{metric}_diff"] = join_df[[f"{metric}_var", f"{metric}_std"]] \
.apply(lambda x: x[0] - x[1], axis=1) \
.clip(-1, 1)
diff_dfs.append(join_df)
return diff_dfs
def aggregate_runs(dfs):
""""""
agg_dfs = []
for df in dfs:
if df is not None:
metrics_df = metrics_change(df)
else:
metrics_df = df
agg_dfs.append(metrics_df)
return agg_dfs
def increase_metrics(runs, run_models, n_list, fractions, metrics):
""""""
all_metrics = {}
for run_df, model, n in zip(runs, run_models, n_list):
if run_df is None:
continue
model_metrics = {}
for i, frac in enumerate(fractions):
discr_df = run_df.describe()
frac_metrics_df = run_df[run_df["bleuvar"] >= discr_df["bleuvar"][frac]][
["rouge1_std", "rouge2_std", "rougeL_std"]] \
.agg(['mean', 'std'])
frac_metrics = {}
for metric in metrics:
frac_increase = (
(discr_df[f"{metric}_std"]["mean"] - frac_metrics_df[f"{metric}_std"]["mean"])
/ discr_df[f"{metric}_var"]["mean"]) * 100
frac_metrics[metric] = frac_increase
model_metrics[frac] = frac_metrics
all_metrics[f"Var{model}-{n}"] = model_metrics
return all_metrics
def main():
args, unknown = read_args()
data_list = args.data
root_data_path = args.root_path
model_list = args.models
base_model_list = args.bases
metrics_list = ["rouge1", "rouge2", "rougeL"]
n_list = args.n_list
fraction_list = ["75%", "50%", "25%"]
data_runs = {}
bases = {}
diffs = {}
for data in data_list:
print(f"Running dataset {data}...")
run_list = [
f"{root_data_path}/var{model.lower()}{n}_{data}/generated_sums.csv" for model, n in zip(model_list, n_list)
]
base_list = [
f"{root_data_path}/{base_model.lower()}_{data}/generated_sums.csv" for base_model in base_model_list
]
run_dfs = load_runs(run_list, n_list)
base_dfs, agg_base_dfs = load_bases(base_list)
bases[data] = agg_base_dfs
join_dfs = join_runs(runs=run_dfs, run_models=model_list, base_runs=base_dfs, base_models=base_model_list)
incr_metrics = increase_metrics(
runs=join_dfs,
run_models=model_list,
n_list=n_list,
fractions=fraction_list,
metrics=metrics_list)
with open(os.path.join(root_data_path, f"{data}_increase_metrics.json"), "w") as mf:
json.dump(incr_metrics, mf)
print(f"Exported {data}_increase_metrics.json")
agg_run_dfs = aggregate_runs(join_dfs)
norm_run_dfs = norm_bleuvar(runs=agg_run_dfs, n_list=n_list)
data_runs[data] = norm_run_dfs
diff_dfs = compute_diffs(joined_dfs=agg_run_dfs, metrics=metrics_list)
diffs[data] = diff_dfs
plot_rouge_retention(
metrics=metrics_list,
dataset_runs=data_runs,
run_models=model_list,
base_runs=bases,
base_models=base_model_list,
n_list=n_list,
save_path=root_data_path,
ascending=False,
)
plot_bleuvar_retention(
dataset_runs=data_runs,
run_models=model_list,
n_list=n_list,
save_path=root_data_path,
ascending=True, )
plot_increase(
dataset_runs=data_runs,
metrics=metrics_list,
# diff_runs=diffs,
run_models=model_list,
n_list=n_list,
save_path=root_data_path,
ascending=False, )
if __name__ == "__main__":
main()