No project description provided
Project description
MetricVisualizer - for easy managing performance metric
Automated metric visualization for comparison experiments
- Box plot
- Trajectory plot
- Scatter plot
- Bar plot
- Violin plot
- Rank test
- On the way
Install
If you want to make tikz(latex) plots, you need to install texlive (other latex release version are not tested).
pip install metric_visualizer
用法说明 Usage
假设存在多组对比实验(或者一组参数设置),则称之为trial,每组实验存在多个metric(例如AUC,Accuracy,F1,Loss等), 每组参照实验重复n词,则使用以下方法监听实验结果: Assume that there exist multiple sets of comparison experiments (or a set of parameter settings), called trials, with multiple metrics (e.g., AUC, accuracy, F1, loss, etc.) for each set of experiments. Repeat n words for each set of reference experiments, and then listen to the results of the experiments using the following method.
import numpy as np
from metric_visualizer import MetricVisualizer
MV = MetricVisualizer()
trial_num = 5 # number of different trials,
repeat = 10 # number of repeats
metric_num = 3 # number of metrics
for trial in range(trial_num):
for r in range(repeat): # repeat the experiments to plot violin or box figure
metrics = [(np.random.random() + n) for n in range(metric_num)] # n is metric scale factor
for i, m in enumerate(metrics):
MV.add_metric('Metric-{}'.format(i + 1), round(m, 2)) # Add metric by metric name
MV.next_trial() # move to next trial
画图代码如下:
save_prefix = None
MV.summary(save_path=save_prefix, no_print=True) # save fig into .tex and .pdf format
MV.traj_plot_by_trial(save_name=save_prefix, xlabel='', xrotation=30, minorticks_on=True) # save fig into .tex and .pdf format
MV.violin_plot_by_trial(save_name=save_prefix) # save fig into .tex and .pdf format
MV.box_plot_by_trial(save_name=save_prefix) # save fig into .tex and .pdf format
MV.avg_bar_plot_by_trial(save_name=save_prefix) # save fig into .tex and .pdf format
MV.sum_bar_plot_by_trial(save_name=save_prefix) # save fig into .tex and .pdf format
# 此函数适合对比不同模型性能,每个模型代表一个trial,综合多个metric进行Scott-Knott Rank Test,并绘制箱型图
MV.scott_knott_plot(save_name=save_prefix, minorticks_on=False)
print(MV.rank_test_by_trail('trial0')) # save fig into .tex and .pdf format
print(MV.rank_test_by_metric('metric1')) # save fig into .tex and .pdf format
# save_path = None
# MV.summary(save_path=save_path) # save fig into .tex and .pdf format
# MV.traj_plot_by_metric(save_path=save_path, xlabel='', xrotation=30) # save fig into .tex and .pdf format
# MV.violin_plot_by_metric(save_path=save_path) # save fig into .tex and .pdf format
# MV.box_plot_by_metric(save_path=save_path) # save fig into .tex and .pdf format
# MV.avg_bar_plot_by_metric(save_path=save_path) # save fig into .tex and .pdf format
# MV.sum_bar_plot_by_metric(save_path=save_path) # save fig into .tex and .pdf format
-------------------- Metric Summary --------------------
╒══════════╤═════════╤══════════════════════════════════════════════════════════════╤═════════════════════════════════════════════════════════════╕
│ Metric │ Trial │ Values │ Summary │
╞══════════╪═════════╪══════════════════════════════════════════════════════════════╪═════════════════════════════════════════════════════════════╡
│ Metric-1 │ trial-0 │ [0.35, 0.65, 0.67, 0.51, 0.04, 0.43, 0.46, 0.58, 0.11, 0.66] │ ['Avg:0.45, Median: 0.48, IQR: 0.22, Max: 0.67, Min: 0.04'] │
├──────────┼─────────┼──────────────────────────────────────────────────────────────┼─────────────────────────────────────────────────────────────┤
│ Metric-1 │ trial-1 │ [0.52, 0.1, 0.11, 0.86, 0.49, 0.7, 0.77, 0.96, 0.16, 0.65] │ ['Avg:0.53, Median: 0.58, IQR: 0.41, Max: 0.96, Min: 0.1'] │
├──────────┼─────────┼──────────────────────────────────────────────────────────────┼─────────────────────────────────────────────────────────────┤
│ Metric-1 │ trial-2 │ [0.73, 0.99, 0.13, 0.72, 0.63, 0.61, 0.14, 0.85, 0.71, 0.86] │ ['Avg:0.64, Median: 0.72, IQR: 0.17, Max: 0.99, Min: 0.13'] │
├──────────┼─────────┼──────────────────────────────────────────────────────────────┼─────────────────────────────────────────────────────────────┤
│ Metric-1 │ trial-3 │ [0.99, 0.69, 0.86, 0.2, 0.4, 0.1, 0.05, 0.07, 0.95, 0.31] │ ['Avg:0.46, Median: 0.36, IQR: 0.62, Max: 0.99, Min: 0.05'] │
├──────────┼─────────┼──────────────────────────────────────────────────────────────┼─────────────────────────────────────────────────────────────┤
│ Metric-1 │ trial-4 │ [0.58, 0.95, 0.73, 0.63, 0.04, 0.19, 0.5, 0.9, 0.64, 0.89] │ ['Avg:0.6, Median: 0.64, IQR: 0.27, Max: 0.95, Min: 0.04'] │
├──────────┼─────────┼──────────────────────────────────────────────────────────────┼─────────────────────────────────────────────────────────────┤
│ Metric-2 │ trial-0 │ [1.58, 1.32, 1.98, 1.76, 1.31, 1.6, 1.6, 1.22, 1.3, 1.19] │ ['Avg:1.49, Median: 1.45, IQR: 0.29, Max: 1.98, Min: 1.19'] │
├──────────┼─────────┼──────────────────────────────────────────────────────────────┼─────────────────────────────────────────────────────────────┤
│ Metric-2 │ trial-1 │ [1.88, 1.67, 1.77, 1.94, 1.01, 1.6, 1.25, 1.63, 1.62, 1.91] │ ['Avg:1.63, Median: 1.65, IQR: 0.21, Max: 1.94, Min: 1.01'] │
├──────────┼─────────┼──────────────────────────────────────────────────────────────┼─────────────────────────────────────────────────────────────┤
│ Metric-2 │ trial-2 │ [1.4, 1.94, 1.28, 1.78, 1.01, 1.08, 1.21, 1.82, 1.78, 1.18] │ ['Avg:1.45, Median: 1.34, IQR: 0.59, Max: 1.94, Min: 1.01'] │
├──────────┼─────────┼──────────────────────────────────────────────────────────────┼─────────────────────────────────────────────────────────────┤
│ Metric-2 │ trial-3 │ [1.79, 1.35, 1.14, 1.5, 1.73, 1.06, 1.98, 1.75, 1.07, 1.49] │ ['Avg:1.49, Median: 1.5, IQR: 0.49, Max: 1.98, Min: 1.06'] │
├──────────┼─────────┼──────────────────────────────────────────────────────────────┼─────────────────────────────────────────────────────────────┤
│ Metric-2 │ trial-4 │ [1.93, 1.81, 1.18, 1.08, 1.57, 1.85, 1.95, 1.94, 1.58, 1.35] │ ['Avg:1.62, Median: 1.7, IQR: 0.43, Max: 1.95, Min: 1.08'] │
├──────────┼─────────┼──────────────────────────────────────────────────────────────┼─────────────────────────────────────────────────────────────┤
│ Metric-3 │ trial-0 │ [2.85, 2.87, 2.3, 2.05, 2.86, 2.34, 2.85, 2.3, 2.95, 2.53] │ ['Avg:2.59, Median: 2.69, IQR: 0.54, Max: 2.95, Min: 2.05'] │
├──────────┼─────────┼──────────────────────────────────────────────────────────────┼─────────────────────────────────────────────────────────────┤
│ Metric-3 │ trial-1 │ [2.31, 2.41, 2.34, 2.96, 2.48, 2.68, 2.99, 2.94, 2.01, 2.46] │ ['Avg:2.56, Median: 2.47, IQR: 0.44, Max: 2.99, Min: 2.01'] │
├──────────┼─────────┼──────────────────────────────────────────────────────────────┼─────────────────────────────────────────────────────────────┤
│ Metric-3 │ trial-2 │ [2.65, 2.5, 2.68, 2.34, 2.32, 2.61, 2.61, 2.88, 2.86, 2.36] │ ['Avg:2.58, Median: 2.61, IQR: 0.24, Max: 2.88, Min: 2.32'] │
├──────────┼─────────┼──────────────────────────────────────────────────────────────┼─────────────────────────────────────────────────────────────┤
│ Metric-3 │ trial-3 │ [2.29, 2.12, 2.4, 2.81, 2.5, 2.54, 2.82, 2.61, 2.45, 2.44] │ ['Avg:2.5, Median: 2.48, IQR: 0.16, Max: 2.82, Min: 2.12'] │
├──────────┼─────────┼──────────────────────────────────────────────────────────────┼─────────────────────────────────────────────────────────────┤
│ Metric-3 │ trial-4 │ [2.41, 2.12, 2.31, 2.29, 2.46, 2.95, 2.74, 2.66, 2.34, 2.65] │ ['Avg:2.49, Median: 2.44, IQR: 0.33, Max: 2.95, Min: 2.12'] │
╘══════════╧═════════╧══════════════════════════════════════════════════════════════╧═════════════════════════════════════════════════════════════╛
-------------------- Metric Summary --------------------
Plot via Matplotlib (or Tikz)
Traj Plot tikz version
Box Plot tikz version
Violin Plot tikz version
Average Bar Plot tikz version
Sum Bar Plot tikz version
Real Usage Example in PyABSA
To analyze the impact of max_seq_len, we can use MetricVisualizer as following:
pip install pyabsa # install pyabsa
import random
import os
from metric_visualizer import MetricVisualizer
from pyabsa.functional import Trainer
from pyabsa.functional import APCConfigManager
from pyabsa.functional import ABSADatasetList
from pyabsa.functional import APCModelList
config = APCConfigManager.get_config()
config.model = APCModelList.FAST_LCF_BERT
config.lcf = 'cdw'
config.seed = [random.randint(0, 10000) for _ in range(3)] # each trial repeats with different seed
MV = MetricVisualizer()
config.MV = MV
max_seq_lens = [60, 70, 80, 90, 100]
for max_seq_len in max_seq_lens:
config.max_seq_len = max_seq_len
dataset = ABSADatasetList.Laptop14
Trainer(config=config,
dataset=dataset, # train set and test set will be automatically detected
auto_device=True # automatic choose CUDA or CPU
)
config.MV.next_trial()
save_prefix = os.getcwd()
MV.summary(save_path=save_prefix, no_print=True) # save fig into .tex and .pdf format
# save fig into .tex and .pdf format
MV.traj_plot_by_trial(save_path=save_prefix, xticks=max_seq_lens)
MV.violin_plot_by_trial(save_path=save_prefix, xticks=max_seq_lens)
MV.box_plot_by_trial(save_path=save_prefix, xticks=max_seq_lens)
MV.avg_bar_plot_by_trial(save_path=save_prefix, xticks=max_seq_lens)
MV.sum_bar_plot_by_trial(save_path=save_prefix, xticks=max_seq_lens)
MV.scott_knott_plot(save_path=save_prefix, xticks=max_seq_lens, minorticks_on=False)
Project details
Release history Release notifications | RSS feed
Download files
Download the file for your platform. If you're not sure which to choose, learn more about installing packages.
Source Distributions
Built Distribution
Hashes for metric_visualizer-0.4.32-py3-none-any.whl
Algorithm | Hash digest | |
---|---|---|
SHA256 | 5914cd4b1a51d8964151d9225c33a0f76623dab1916cd2b44a76ac602a2381ed |
|
MD5 | 0086e49548684c02c51de2e804eb535d |
|
BLAKE2b-256 | a819311cf5ee4c84af1b9de62bc0cfdd4c038bae8511f16b32944a8a26abff86 |