File size: 7,973 Bytes
551eebe |
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 |
import os
import json
import numpy as np
import matplotlib.pyplot as plt
from typing import List, Dict, Any
from transformers import AutoTokenizer
class HFTokenizerTestSuite:
def __init__(self, model_name: str, test_data_paths: List[str]):
self.tokenizer = AutoTokenizer.from_pretrained(model_name)
self.languages = ['hindi', 'english']
self.edge_cases = {
'hindi': {
'script_test': 'नमस्ते, मैं भारत से हूँ। दिल्ली बहुत बड़ा शहर है।',
'unicode_test': 'हिन्दी १२३४५६७८९ vowels: अ आ इ ई उ ऊ',
'special_chars': 'हिन्दी! @ # $ % ^ & * ( ) _ + = [ ] { }',
},
'english': {
'script_test': 'Hello, I am from the United States. New York is a beautiful city.',
'unicode_test': 'English 0123456789 vowels: a e i o u',
'special_chars': 'English! @ # $ % ^ & * ( ) _ + = [ ] { }',
}
}
self.test_data = self._load_test_data(test_data_paths)
self.results = {
'coverage': {},
'complexity': {},
'language_analysis': {},
'edge_cases': {}
}
def _load_test_data(self, data_paths: List[str]) -> Dict[str, List[str]]:
test_data = {lang: [] for lang in self.languages}
for path in data_paths:
try:
with open(path, 'r', encoding='utf-8') as f:
texts = f.readlines()
for i, text in enumerate(texts):
lang = self.languages[i % len(self.languages)]
test_data[lang].append(text.strip())
except Exception as e:
print(f"Error loading {path}: {e}")
return test_data
def unicode_coverage_analysis(self) -> Dict[str, Any]:
unicode_results = {}
for lang, edge_cases in self.edge_cases.items():
unicode_test = edge_cases['unicode_test']
tokenizer_output = self.tokenizer(unicode_test, return_tensors="pt")
tokens = self.tokenizer.convert_ids_to_tokens(tokenizer_output['input_ids'][0])
unicode_results[lang] = {
'original_text': unicode_test,
'tokens': tokens,
'token_count': len(tokens),
'unique_tokens': len(set(tokens)),
'coverage_ratio': len(set(tokens)) / len(tokens)
}
self.results['unicode_coverage'] = unicode_results
return unicode_results
def language_specific_edge_cases(self) -> Dict[str, Any]:
edge_case_results = {}
for lang, cases in self.edge_cases.items():
language_results = {}
for case_name, text in cases.items():
try:
tokenizer_output = self.tokenizer(text, return_tensors="pt")
tokens = self.tokenizer.convert_ids_to_tokens(tokenizer_output['input_ids'][0])
language_results[case_name] = {
'tokens': tokens,
'token_count': len(tokens),
'unique_tokens': len(set(tokens))
}
except Exception as e:
language_results[case_name] = {
'error': str(e)
}
edge_case_results[lang] = language_results
self.results['edge_cases'] = edge_case_results
return edge_case_results
def script_complexity_analysis(self) -> Dict[str, Any]:
complexity_results = {}
for lang in self.languages:
text = self.edge_cases[lang]['script_test']
tokenizer_output = self.tokenizer(text, return_tensors="pt")
tokens = self.tokenizer.convert_ids_to_tokens(tokenizer_output['input_ids'][0])
# Filter out special tokens for accurate length calculation
filtered_tokens = [token for token in tokens if not token.startswith('[') or not token.endswith(']')]
complexity_results[lang] = {
'original_text_length': len(text),
'tokens': tokens,
'token_count': len(tokens),
'avg_token_length': np.mean([len(token) for token in filtered_tokens]) if filtered_tokens else 0,
'token_diversity': len(set(tokens)) / len(tokens)
}
self.results['script_complexity'] = complexity_results
return complexity_results
def generate_token_histograms(self):
plt.figure(figsize=(15, 10))
for i, lang in enumerate(self.languages):
text = self.test_data[lang][0] if self.test_data[lang] else self.edge_cases[lang]['script_test']
tokenizer_output = self.tokenizer(text, return_tensors="pt")
tokens = self.tokenizer.convert_ids_to_tokens(tokenizer_output['input_ids'][0])
# Filter out special tokens
filtered_tokens = [token for token in tokens if not token.startswith('[') or not token.endswith(']')]
token_lengths = [len(token) for token in filtered_tokens]
plt.subplot(len(self.languages), 1, i+1)
plt.hist(token_lengths, bins=range(1, max(token_lengths) + 2), alpha=0.7)
plt.title(f'Token Length Distribution for {lang.capitalize()}')
plt.xlabel('Token Length')
plt.ylabel('Frequency')
plt.grid(True, alpha=0.3)
plt.tight_layout()
plt.savefig('token_length_histograms.png')
plt.close()
def generate_unicode_visualization(self):
plt.figure(figsize=(15, 10))
unicode_results = self.results.get('unicode_coverage', {})
plt.subplot(2, 1, 1)
token_counts = [result['token_count'] for result in unicode_results.values()]
plt.bar(unicode_results.keys(), token_counts)
plt.title('Token Count in Unicode Test Texts')
plt.xlabel('Language')
plt.ylabel('Number of Tokens')
plt.xticks(rotation=45)
plt.subplot(2, 1, 2)
coverage_ratios = [result['coverage_ratio'] for result in unicode_results.values()]
plt.bar(unicode_results.keys(), coverage_ratios)
plt.title('Token Diversity Ratio')
plt.xlabel('Language')
plt.ylabel('Unique Tokens / Total Tokens')
plt.xticks(rotation=45)
plt.tight_layout()
plt.savefig('unicode_token_analysis.png')
plt.close()
def run_all_tests(self):
print("Running Tokenizer Test Suite for Hindi and English...")
print("1. Unicode Coverage Analysis...")
self.unicode_coverage_analysis()
print("2. Language-Specific Edge Cases...")
self.language_specific_edge_cases()
print("3. Script Complexity Analysis...")
self.script_complexity_analysis()
print("4. Generating Token Histograms...")
self.generate_token_histograms()
print("5. Generating Unicode Visualizations...")
self.generate_unicode_visualization()
print("Test Suite Complete!")
return self.results
if __name__ == "__main__":
MODEL_NAME = "tinycompany/ShawtyIsBad-bgem3"
TEST_DATA_PATHS = [
'./test2.txt'
]
test_suite = HFTokenizerTestSuite(MODEL_NAME, TEST_DATA_PATHS)
results = test_suite.run_all_tests()
with open('result1.json', 'w', encoding='utf-8') as f:
json.dump(results, f, ensure_ascii=False, indent=4) |