Outlier_Identification/visual.py at main · HPCI-Lab/Outlier_Identification · GitHub

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
from prov4ml.utils.prov_getters import *
import matplotlib.pyplot as plt
import seaborn as sns
import json
import pandas as pd
from configs.paths import CACHE_PATH


file_path = f"{CACHE_PATH}/prov/IBM_outliers_13/provgraph_IBM_outliers.json"
start_at = 0


data = json.load(open(file_path))
print(get_metrics(data))
indices = get_metric(data, metric="Indices_Context.TRAINING", sort_by="time")["value"]
df1 = get_metric(data, "Loss_Context.TRAINING", start_at=start_at, sort_by="time")
is_outlier = get_metric(data, "Outlier_Context.TRAINING", start_at=start_at, sort_by="time")["value"]
is_outlier = [any(eval(o)) for o in is_outlier]
print(pd.Series(is_outlier).value_counts())

# epochs = df1["epoch"]
# df1.to_csv("loss_noshuffle.csv")


sns.set_theme(rc={'figure.figsize': (19.7, 8.27)})
color = 'tab:blue'
sns.lineplot(data=df1, x="time", y="value", color=color)
plt.title("Loss at each step of the training (red lines are outliers)")
plt.ylabel("Loss (MSE)")
plt.xlabel("Batch")
# plt.xticks(df1["time"], ["No" if not o else "Yes" for o in is_outlier], rotation=90)
# for position in df1[is_outlier]["time"]:
#     plt.axvline(x=position, color='red', linestyle='--')  # Customize the line color and style

plt.tight_layout()
plt.savefig(f"loss_noshuffle.png")
plt.close()


# df1 = get_metric(data, "Backward_time_Context.TRAINING", start_at=start_at, sort_by="time")
# # df1 = df1[df1.index % 2 == 1]
# sns.set_theme(rc={'figure.figsize': (19.7, 8.27)})
# sns.lineplot(data=df1, x="time", y="value", color=color, ci=None)
# plt.title("Backward_time")
# # plt.xticks(df1["time"], ["No" if not o else "Yes" for o in is_outlier], rotation=90)
# print(len(df1), len(is_outlier))
# for position in df1[is_outlier]["time"]:
#     plt.axvline(x=position, color='red', linestyle='--')  # Customize the line color and style

# plt.tight_layout()
# plt.savefig(f"Backward_time_Context.TRAINING.png")
# plt.close()

# df1 = get_metric(data, "Step_time_Context.TRAINING", start_at=start_at, sort_by="time")
# df1["epoch"] = epochs
# sns.set_theme(rc={'figure.figsize': (19.7, 8.27)})
# sns.lineplot(data=df1, x="time", y="value", color=color)
# for position in df1[is_outlier]["time"]:
#     plt.axvline(x=position, color='red', linestyle='--')  # Customize the line color and style

# # plt.xticks(df1["time"], ["No" if not o else "Yes" for o in is_outlier], rotation=90)
# plt.title("Step_time_Context")
# plt.tight_layout()
# plt.savefig(f"Step_time_Context.TRAINING.png")
# plt.close()


# df1 = get_metric(data, "gpu_power_usage_Context.TRAINING", start_at=start_at, sort_by="time")
# df1["epoch"] = epochs
# sns.set_theme(rc={'figure.figsize': (19.7, 8.27)})
# sns.lineplot(data=df1, x="time", y="value", color=color)
# plt.title("gpu_power_usage_Context.TRAINING")
# # plt.xticks(df1["time"], ["No" if not o else "Yes" for o in is_outlier], rotation=90)
# for position in df1[is_outlier]["time"]:
#     plt.axvline(x=position, color='red', linestyle='--')  # Customize the line color and style


# plt.tight_layout()
# plt.savefig(f"gpu_power_usage_Context.TRAINING.png")
# plt.close()

# df1 = get_metric(data, "gpu_usage_Context.TRAINING", start_at=start_at, sort_by="time")
# df1["epoch"] = epochs
# sns.set_theme(rc={'figure.figsize': (19.7, 8.27)})
# sns.lineplot(data=df1, x="time", y="value", color=color)
# # plt.xticks(df1["time"], ["No" if not o else "Yes" for o in is_outlier], rotation=90)
# plt.title("gpu_usage_Context.TRAINING")
# plt.tight_layout()
# plt.savefig(f"gpu_usage_Context.TRAINING.png")
# plt.close()

# df1 = get_metric(data, "Gpu_power_nvidia_Context.TRAINING", start_at=start_at, sort_by="time")
# df1["epoch"] = epochs
# sns.set_theme(rc={'figure.figsize': (19.7, 8.27)})
# sns.lineplot(data=df1, x="time", y="value", color=color)
# # plt.xticks(df1["time"], ["No" if not o else "Yes" for o in is_outlier], rotation=90)
# plt.title("Gpu_power_nvidia_Context.TRAINING")
# for position in df1[is_outlier]["time"]:
#     plt.axvline(x=position, color='red', linestyle='--')  # Customize the line color and style


# plt.tight_layout()
# plt.savefig(f"Gpu_power_nvidia_Context.TRAINING.png")
# plt.close()


# df1 = get_metric(data, "Gpu_energy_nvidia_Context.TRAINING", start_at=start_at, sort_by="time")
# df1["epoch"] = epochs
# sns.set_theme(rc={'figure.figsize': (19.7, 8.27)})
# sns.lineplot(data=df1, x="time", y="value", color=color)
# # plt.xticks(df1["time"], ["No" if not o else "Yes" for o in is_outlier], rotation=90)
# plt.title("Gpu_energy_nvidia_Context.TRAINING")
# for position in df1[is_outlier]["time"]:
#     plt.axvline(x=position, color='red', linestyle='--')  # Customize the line color and style


# plt.tight_layout()
# plt.savefig(f"Gpu_energy_nvidia_Context.TRAINING.png")
# plt.close()


# df1 = get_metric(data, "Gpu_usage_nvidia_Context.TRAINING", start_at=start_at, sort_by="time")
# df1["epoch"] = epochs
# df1["value"] = df1["value"].diff().fillna(0.0)
# sns.set_theme(rc={'figure.figsize': (19.7, 8.27)})
# sns.lineplot(data=df1, x="time", y="value", color=color)
# # plt.xticks(df1["time"], ["No" if not o else "Yes" for o in is_outlier], rotation=90)
# plt.title("Gpu_usage_nvidia_Context.TRAINING")
# for position in df1[is_outlier]["time"]:
#     plt.axvline(x=position, color='red', linestyle='--')  # Customize the line color and style


# plt.tight_layout()
# plt.savefig(f"Gpu_usage_nvidia_Context.TRAINING.png")
# plt.close()