src/distiller/__init__.py

"""
Distiller package for code-specialized embedding model distillation and evaluation.

This package provides a complete pipeline for:
1. Distilling code-specialized embedding models using Model2Vec
2. Comprehensive evaluation including CodeSearchNet and performance benchmarks
3. Analysis and reporting of model performance

Main modules:
- distill: Model2Vec distillation with optional advanced training
- evaluate: Comprehensive evaluation (CodeSearchNet + performance benchmarks)
- analyze: Analysis and reporting tools
- config: Centralized configuration management
- beam_utils: Beam cloud utilities for distributed processing

Usage:
    from distiller import distill, evaluate, analyze
"""

from . import analyze, config, distill, evaluate
from .analyze import CodeSearchNetAnalyzer
from .config import (
	BEAM_ENV_SETTINGS,
	DEFAULT_EVALUATION_MODELS,
	IMAGE,
	codesearchnet_config,
	directories,
	distillation_config,
	get_volume_config,
	languages_config,
)
from .distill import (
	run_beam_distillation,
	run_local_distillation,
)
from .evaluate import (
	CodeSearchNetEvaluator,
	ComprehensiveModelEvaluator,
	run_evaluation,
)

__all__ = [
	# Configuration
	"BEAM_ENV_SETTINGS",
	"DEFAULT_EVALUATION_MODELS",
	"IMAGE",
	# Main classes
	"CodeSearchNetAnalyzer",
	"CodeSearchNetEvaluator",
	"ComprehensiveModelEvaluator",
	# Modules
	"analyze",
	"codesearchnet_config",
	"config",
	"directories",
	"distill",
	"distillation_config",
	"evaluate",
	"get_volume_config",
	"languages_config",
	# Main functions
	"run_beam_distillation",
	"run_evaluation",
	"run_local_distillation",
]