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"""
里程碑4 CLI - 概率编码处理
"""
import argparse
from pathlib import Path
import json
import numpy as np
from .probability_encoder import ProbabilityEncoder
def encode_command(args):
"""编码训练样本为概率分布"""
encoder = ProbabilityEncoder()
print(f"Loading samples from {args.input}")
encoder.load_samples(args.input)
print(f"Encoding probabilities with smoothing={args.smoothing}")
encoder.encode_probabilities(smoothing=args.smoothing)
if args.clean:
print(f"Cleaning samples (min_count={args.min_count}, max_length={args.max_length})")
encoder.clean_samples(min_count=args.min_count, max_context_length=args.max_length)
stats = encoder.get_statistics()
print("\nStatistics:")
for key, value in stats.items():
if isinstance(value, float):
print(f" {key}: {value:.2f}")
else:
print(f" {key}: {value}")
if args.output:
encoder.save_encoded(args.output)
def analyze_command(args):
"""分析概率编码的样本"""
encoder = ProbabilityEncoder()
print(f"Loading encoded samples from {args.input}")
encoder.load_encoded(args.input)
stats = encoder.get_statistics()
print("\nDataset Statistics:")
for key, value in stats.items():
if isinstance(value, float):
print(f" {key}: {value:.2f}")
else:
print(f" {key}: {value}")
print(f"\nBranching Analysis:")
branching = encoder.analyze_branching_patterns()
# 显示按长度的平均分叉
print("\nAverage branching by context length:")
for length in sorted(branching['avg_branching_by_length'].keys())[:10]:
avg = branching['avg_branching_by_length'][length]
print(f" Length {length}: {avg:.2f} average targets")
# 显示高熵上下文
if branching['high_entropy_contexts']:
print("\nHigh entropy contexts (top 5):")
for ctx_info in branching['high_entropy_contexts'][:5]:
print(f" Context length {len(ctx_info['context'])}: "
f"entropy={ctx_info['entropy']:.2f}, "
f"targets={ctx_info['targets']}")
if args.output:
report = {
'statistics': stats,
'branching_analysis': {
'avg_branching_by_length': branching['avg_branching_by_length'],
'num_deterministic': len(branching['deterministic_contexts']),
'num_high_entropy': len(branching['high_entropy_contexts'])
}
}
with open(args.output, 'w') as f:
json.dump(report, f, indent=2)
print(f"\nReport saved to {args.output}")
def export_command(args):
"""导出为训练格式"""
encoder = ProbabilityEncoder()
print(f"Loading encoded samples from {args.input}")
encoder.load_encoded(args.input)
print(f"Exporting to {args.format} format...")
encoder.export_for_training(args.output, format=args.format)
def inspect_command(args):
"""检查特定样本"""
encoder = ProbabilityEncoder()
print(f"Loading encoded samples from {args.input}")
encoder.load_encoded(args.input)
# 显示前N个样本
for i, sample in enumerate(encoder.probabilistic_samples[:args.num_samples]):
print(f"\nSample {i + 1}:")
print(f" Context: {sample.context}")
print(f" Count: {sample.count}")
print(f" Target distribution:")
# 排序显示概率分布
sorted_dist = sorted(sample.target_distribution.items(), key=lambda x: x[1], reverse=True)
for target_id, prob in sorted_dist[:5]: # 显示前5个
print(f" Node {target_id}: {prob:.4f}")
if len(sorted_dist) > 5:
print(f" ... and {len(sorted_dist) - 5} more targets")
def compare_command(args):
"""比较编码前后的差异"""
# 加载原始样本统计
with open(args.original, 'r') as f:
original_data = json.load(f)
encoder = ProbabilityEncoder()
encoder.load_encoded(args.encoded)
original_samples = len(original_data['samples'])
encoded_contexts = len(encoder.probabilistic_samples)
print("\nComparison Report:")
print(f" Original samples: {original_samples}")
print(f" Unique contexts after encoding: {encoded_contexts}")
print(f" Compression ratio: {original_samples / encoded_contexts:.2f}x")
# 分析分叉情况
branching = encoder.analyze_branching_patterns()
deterministic = len(branching['deterministic_contexts'])
multi_target = encoded_contexts - deterministic
print(f"\nBranching Analysis:")
print(f" Deterministic contexts: {deterministic} ({deterministic/encoded_contexts*100:.1f}%)")
print(f" Multi-target contexts: {multi_target} ({multi_target/encoded_contexts*100:.1f}%)")
def main():
parser = argparse.ArgumentParser(description="Milestone 4: Probability Encoding")
subparsers = parser.add_subparsers(dest='command', help='Available commands')
# Encode command
encode_parser = subparsers.add_parser('encode', help='Encode samples with probabilities')
encode_parser.add_argument('input', help='Path to training samples (from milestone3)')
encode_parser.add_argument('--output', '-o', help='Path to save encoded samples')
encode_parser.add_argument('--smoothing', type=float, default=0.01,
help='Laplace smoothing parameter (default: 0.01)')
encode_parser.add_argument('--clean', action='store_true',
help='Clean samples after encoding')
encode_parser.add_argument('--min-count', type=int, default=2,
help='Minimum count for cleaning (default: 2)')
encode_parser.add_argument('--max-length', type=int,
help='Maximum context length for cleaning')
# Analyze command
analyze_parser = subparsers.add_parser('analyze', help='Analyze probabilistic samples')
analyze_parser.add_argument('input', help='Path to encoded samples')
analyze_parser.add_argument('--output', '-o', help='Save analysis report')
# Export command
export_parser = subparsers.add_parser('export', help='Export for training')
export_parser.add_argument('input', help='Path to encoded samples')
export_parser.add_argument('output', help='Output path')
export_parser.add_argument('--format', choices=['numpy', 'sparse'], default='numpy',
help='Export format (default: numpy)')
# Inspect command
inspect_parser = subparsers.add_parser('inspect', help='Inspect individual samples')
inspect_parser.add_argument('input', help='Path to encoded samples')
inspect_parser.add_argument('--num-samples', '-n', type=int, default=5,
help='Number of samples to show (default: 5)')
# Compare command
compare_parser = subparsers.add_parser('compare', help='Compare before/after encoding')
compare_parser.add_argument('original', help='Path to original samples')
compare_parser.add_argument('encoded', help='Path to encoded samples')
args = parser.parse_args()
if args.command == 'encode':
encode_command(args)
elif args.command == 'analyze':
analyze_command(args)
elif args.command == 'export':
export_command(args)
elif args.command == 'inspect':
inspect_command(args)
elif args.command == 'compare':
compare_command(args)
else:
parser.print_help()
if __name__ == '__main__':
main()