Squish42 commited on
Commit
255ac07
1 Parent(s): e6c69a2

Add some plots

Browse files
.gitignore CHANGED
@@ -4,4 +4,3 @@ original
4
  # Generated data
5
  resources/gen
6
  *.pyc
7
- scripts
 
4
  # Generated data
5
  resources/gen
6
  *.pyc
 
assets/full-test.png ADDED

Git LFS Details

  • SHA256: 5144083a67fe051ae4906459fdbf3340445ec824c3df0c0a00ad9519c129f631
  • Pointer size: 130 Bytes
  • Size of remote file: 14.5 kB
assets/full-train.png ADDED

Git LFS Details

  • SHA256: 75570638ce5521008cb5ccdee0647a539f2f2dbc292aa02682875e7e601ffca5
  • Pointer size: 130 Bytes
  • Size of remote file: 16 kB
scripts/dsbuild.py ADDED
@@ -0,0 +1,56 @@
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
1
+ import gzip
2
+ import json
3
+ import os
4
+
5
+ from matplotlib import pyplot as plt
6
+
7
+
8
+ def load_stats(path: str) -> dict:
9
+ stats: dict = {}
10
+ for filename in os.listdir(path):
11
+ file_path: str = os.path.join(path, filename)
12
+ if os.path.isdir(filename):
13
+ continue
14
+ data: dict
15
+ if filename.endswith('.gz'):
16
+ with gzip.open(file_path, mode='rt') as file:
17
+ data = json.loads(file.read())
18
+ else:
19
+ with open(file_path, mode='rt') as file:
20
+ data = json.loads(file.read())
21
+ print(f'Loaded stats from {file_path}')
22
+ stats.update(**data)
23
+ return stats
24
+
25
+
26
+ def stat_filter(stats: dict, deviation_cutoff=(1.0, 0.0), clamp=(200.0, 2048.0), min_messages=4) -> list[dict]:
27
+ cutoff_threshold: (float, float) = (
28
+ stats['wordsStdDev'] * deviation_cutoff[0], stats['wordsStdDev'] * deviation_cutoff[1])
29
+ if cutoff_threshold[1] <= 0:
30
+ cutoff_threshold = (cutoff_threshold[0], stats['wordsMax'])
31
+ cutoff_min: float = max(max(clamp[0], cutoff_threshold[0]), stats['wordsMean'] - cutoff_threshold[0])
32
+ cutoff_max: float = stats['wordsMean'] + cutoff_threshold[1]
33
+ if clamp[1] > 0:
34
+ cutoff_max = min(clamp[1], cutoff_max)
35
+
36
+ conversations: list[dict] = [v for k, v in stats['conversations'].items() if
37
+ v['wordsMax'] <= cutoff_max and v['wordsMin'] >= cutoff_min and v[
38
+ 'messagesCount'] >= min_messages]
39
+ print(
40
+ f'Min: {cutoff_min:0.0f}\tMax: {cutoff_max:0.0f}\n'
41
+ f'Clamped from {cutoff_threshold[0]:0.0f}, {cutoff_threshold[1]:0.0f}')
42
+ print(f'{len(conversations)} conversations')
43
+
44
+ return conversations
45
+
46
+
47
+ def build_mean_word_plot(conv_stats: list[float], title: str = 'Conversation Message Mean Words', xlabel: str = 'Mean',
48
+ ylabel: str = 'Conversations', text: str = '',
49
+ **kwargs):
50
+ fig, ax = plt.subplots(figsize=(5, 2.7), layout='constrained')
51
+ n, bins, patches = ax.hist(conv_stats, density=True,
52
+ facecolor='C0', alpha=0.75, **kwargs)
53
+ ax.set_title(title)
54
+ ax.set_xlabel(xlabel)
55
+ ax.set_ylabel(ylabel)
56
+ plt.figtext(0, 0.95, text)
scripts/generate_plot.py ADDED
@@ -0,0 +1,31 @@
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
1
+ import os
2
+
3
+ import matplotlib.pyplot as plt
4
+
5
+ import dsbuild
6
+
7
+ stats_path: list[str] = ['../stats/full/test/', '../stats/full/train']
8
+
9
+
10
+ def __plot_mean_word_counts(path: list[str], show: bool = False, save: bool = False):
11
+ for p in path:
12
+ p: str = os.path.normpath(p)
13
+ split_name: str = p.split(os.sep)[-1]
14
+ config_name: str = p.split(os.sep)[-2]
15
+
16
+ stats: dict = dsbuild.load_stats(p)
17
+ conversations: list[float] = [conversation['wordsMean'] for _, conversation in stats['conversations'].items()]
18
+ dsbuild.build_mean_word_plot(conversations, f'Mean Words ({config_name}-{split_name})',
19
+ text=f'{len(conversations)} total')
20
+ if save:
21
+ plt.savefig(f'../assets/{config_name}-{split_name}.png')
22
+ if show:
23
+ plt.show()
24
+
25
+
26
+ def main():
27
+ __plot_mean_word_counts(stats_path, show=True, save=True)
28
+
29
+
30
+ if __name__ == '__main__':
31
+ main()