File size: 6,706 Bytes
4265aea |
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 99 100 101 102 103 104 105 106 107 108 109 110 111 112 113 114 115 116 117 118 119 120 121 122 123 124 125 126 127 128 129 130 131 132 133 134 135 136 137 138 139 140 141 142 143 144 145 146 147 148 149 150 151 152 153 154 155 156 157 158 159 160 161 162 163 164 165 166 167 168 169 170 171 172 173 174 175 176 177 178 179 180 181 182 183 184 185 186 187 188 189 190 191 192 193 194 195 196 197 198 199 200 201 202 203 204 205 206 207 208 |
"""
Advanced usage example for NexForge Tokenizer Builder.
This example demonstrates:
- Custom special tokens
- Batch processing with progress tracking
- Vocabulary inspection and analysis
- Error handling and recovery
- Performance optimization
"""
import os
import json
import time
from pathlib import Path
from typing import Dict, List, Optional
from tqdm import tqdm
# Import the tokenizer components
from nexforgetokenizer import (
build_tokenizer,
SystemResources,
log_memory_usage,
TokenizerError
)
def create_large_sample_dataset(num_files: int = 50, base_dir: str = "sample_data") -> Path:
"""Create a larger sample dataset with different file types."""
base_path = Path(base_dir)
# Clean up if exists
if base_path.exists():
import shutil
shutil.rmtree(base_path)
# Create directories
base_path.mkdir(exist_ok=True)
# Create Python files
for i in range(num_files // 2):
module_content = f"""
# Sample Python module {i}
def process_data(data):
'''Process sample data.'''
result = []
for item in data:
if item % 2 == 0:
result.append(item * 2)
return result
"""
(base_path / f"module_{i}.py").write_text(module_content)
# Create text files
for i in range(num_files // 2):
doc_content = f"""
This is sample text document {i}.
It contains multiple lines of text with various tokens.
The quick brown fox jumps over the lazy dog.
Special characters: !@#$%^&*()_+-=[]{{}}|;':\",./<>?
"""
(base_path / f"document_{i}.txt").write_text(doc_content)
print(f"Created {num_files} sample files in {base_path}")
return base_path
class DataProcessor:
"""Example data processor class for demonstration."""
def __init__(self, config: dict):
self.config = config
def run(self):
"""Run the processor with the current config."""
print(f"Processing with config: {self.config}")
class TokenizerAnalyzer:
"""Helper class for analyzing tokenizer performance and vocabulary."""
def __init__(self, tokenizer_path: str):
self.tokenizer_path = tokenizer_path
self.tokenizer = None
self.vocab = None
def load(self):
"""Load the tokenizer."""
from tokenizers import Tokenizer
self.tokenizer = Tokenizer.from_file(self.tokenizer_path)
self.vocab = {
idx: self.tokenizer.id_to_token(idx)
for idx in range(self.tokenizer.get_vocab_size())
}
def analyze_vocab(self, top_n: int = 20):
"""Analyze and print vocabulary statistics."""
if not self.tokenizer:
self.load()
vocab_size = len(self.vocab)
special_tokens = [
token for token in self.vocab.values()
if token.startswith("[") and token.endswith("]")
]
print(f"\n=== Vocabulary Analysis ===")
print(f"Total vocabulary size: {vocab_size}")
print(f"Special tokens ({len(special_tokens)}): {', '.join(special_tokens[:10])}" +
("..." if len(special_tokens) > 10 else ""))
# Show sample of vocabulary
print(f"\nSample vocabulary items:")
for idx in range(min(top_n, vocab_size)):
print(f" {idx}: {self.vocab.get(idx, 'N/A')}")
if vocab_size > top_n:
print(f" ... and {vocab_size - top_n} more")
def main():
"""Run the advanced example."""
print("NexForge Tokenizer Builder - Advanced Example")
print("=========================================\n")
# 1. Setup
output_dir = Path("advanced_output")
output_dir.mkdir(exist_ok=True)
tokenizer_path = output_dir / "advanced_tokenizer.json"
# 2. Check system resources
resources = SystemResources()
print(f"\n=== System Resources ===")
print(f"CPU Cores: {resources.cpu_cores}")
print(f"Available RAM: {resources.available_ram_gb:.2f} GB")
if resources.has_cuda:
print(f"GPU: {resources.cuda_device} with {resources.cuda_mem_gb:.2f} GB")
else:
print("No CUDA GPU detected")
# 3. Create sample dataset
print("\n=== Creating Sample Dataset ===")
dataset_path = create_large_sample_dataset(num_files=50)
# 4. Custom special tokens
special_tokens = [
"[PAD]", "[UNK]", "[CLS]", "[SEP]", "[MASK]",
"[PYTHON]", "[TEXT]", "[CODE]"
]
# 5. Build the tokenizer with advanced options
print("\n=== Building Tokenizer ===")
print(f"Input directory: {dataset_path}")
print(f"Output path: {tokenizer_path}")
start_time = time.time()
try:
success = build_tokenizer(
input_dir=str(dataset_path),
output_path=str(tokenizer_path),
vocab_size=5000, # Larger vocabulary for better coverage
min_frequency=2, # Only include tokens that appear at least twice
special_tokens=special_tokens,
resources=resources,
max_files=50, # Process all files
chunk_size=100000, # Process in 100KB chunks
n_threads=max(1, resources.cpu_cores - 1) # Use all but one CPU core
)
if success:
duration = time.time() - start_time
print(f"\nTokenizer created successfully in {duration:.2f} seconds")
print(f"Tokenizer saved to: {tokenizer_path}")
# 6. Analyze the created tokenizer
print("\n=== Tokenizer Analysis ===")
analyzer = TokenizerAnalyzer(str(tokenizer_path))
analyzer.load()
analyzer.analyze_vocab()
# 7. Show example encoding/decoding
print("\n=== Example Encoding/Decoding ===")
sample_text = "def hello_world():\n print('Hello, world!') # Sample Python code"
encoded = analyzer.tokenizer.encode(sample_text)
decoded = analyzer.tokenizer.decode(encoded.ids)
print(f"Original: {sample_text}")
print(f"Encoded: {encoded.ids}")
print(f"Tokens: {encoded.tokens}")
print(f"Decoded: {decoded}")
else:
print("\nFailed to create tokenizer")
except TokenizerError as e:
print(f"\nError creating tokenizer: {e}")
except Exception as e:
print(f"\nUnexpected error: {e}")
finally:
# 8. Cleanup (optional)
# import shutil
# shutil.rmtree(dataset_path, ignore_errors=True)
pass
print("\nExample completed!")
if __name__ == "__main__":
main()
|