File size: 5,130 Bytes
f24563f |
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 99 100 101 102 103 104 105 106 107 108 109 110 111 112 113 114 115 116 117 118 119 120 121 122 123 124 125 126 127 128 129 130 131 132 133 134 135 136 137 138 139 140 141 142 143 144 145 146 147 148 149 150 151 152 153 154 155 156 157 158 159 160 161 162 163 164 165 166 167 168 169 170 171 172 173 174 175 176 177 178 179 180 181 182 183 184 185 186 187 188 189 190 191 192 193 194 195 196 197 198 199 200 201 202 203 204 205 206 207 208 209 210 211 212 213 214 |
"""
Logging utilities for LLM training.
"""
import os
import time
import json
import logging
from typing import Dict, Any, Optional, List
import numpy as np
import jax
import jax.numpy as jnp
import tensorflow as tf
def setup_logger(
name: str,
log_file: Optional[str] = None,
level: int = logging.INFO
) -> logging.Logger:
"""
Set up logger.
Args:
name: Logger name
log_file: Path to log file
level: Logging level
Returns:
Logger
"""
# Create logger
logger = logging.getLogger(name)
logger.setLevel(level)
# Create formatter
formatter = logging.Formatter(
"%(asctime)s - %(name)s - %(levelname)s - %(message)s"
)
# Create console handler
console_handler = logging.StreamHandler()
console_handler.setLevel(level)
console_handler.setFormatter(formatter)
logger.addHandler(console_handler)
# Create file handler if log file is provided
if log_file is not None:
# Create directory if it doesn't exist
os.makedirs(os.path.dirname(log_file), exist_ok=True)
# Create file handler
file_handler = logging.FileHandler(log_file)
file_handler.setLevel(level)
file_handler.setFormatter(formatter)
logger.addHandler(file_handler)
return logger
def log_metrics(
metrics: Dict[str, Any],
step: int,
logger: Optional[logging.Logger] = None,
prefix: str = "",
log_to_console: bool = True
) -> None:
"""
Log metrics.
Args:
metrics: Dictionary of metrics
step: Training step
logger: Logger
prefix: Prefix for metric names
log_to_console: Whether to log to console
"""
# Add prefix to metric names
if prefix:
metrics = {f"{prefix}/{k}": v for k, v in metrics.items()}
# Convert metrics to Python types
metrics = {
k: float(v) if isinstance(v, (np.ndarray, jnp.ndarray)) else v
for k, v in metrics.items()
}
# Log to console
if log_to_console:
print(f"Step {step}:")
for k, v in metrics.items():
if isinstance(v, float):
print(f" {k}: {v:.4f}")
else:
print(f" {k}: {v}")
# Log to logger
if logger is not None:
logger.info(f"Step {step}: {metrics}")
def create_summary_writer(log_dir: str) -> tf.summary.SummaryWriter:
"""
Create TensorBoard summary writer.
Args:
log_dir: Directory for TensorBoard logs
Returns:
TensorBoard summary writer
"""
# Create directory if it doesn't exist
os.makedirs(log_dir, exist_ok=True)
# Create summary writer
return tf.summary.create_file_writer(log_dir)
def log_metrics_to_tensorboard(
metrics: Dict[str, Any],
step: int,
writer: tf.summary.SummaryWriter,
prefix: str = ""
) -> None:
"""
Log metrics to TensorBoard.
Args:
metrics: Dictionary of metrics
step: Training step
writer: TensorBoard summary writer
prefix: Prefix for metric names
"""
# Add prefix to metric names
if prefix:
metrics = {f"{prefix}/{k}": v for k, v in metrics.items()}
# Convert metrics to Python types
metrics = {
k: float(v) if isinstance(v, (np.ndarray, jnp.ndarray)) else v
for k, v in metrics.items()
}
# Log metrics to TensorBoard
with writer.as_default():
for k, v in metrics.items():
if isinstance(v, float):
tf.summary.scalar(k, v, step=step)
elif isinstance(v, (list, tuple)) and all(isinstance(x, float) for x in v):
tf.summary.histogram(k, v, step=step)
# Flush writer
writer.flush()
def log_text_to_tensorboard(
text: str,
tag: str,
step: int,
writer: tf.summary.SummaryWriter
) -> None:
"""
Log text to TensorBoard.
Args:
text: Text to log
tag: Tag for text
step: Training step
writer: TensorBoard summary writer
"""
# Log text to TensorBoard
with writer.as_default():
tf.summary.text(tag, text, step=step)
# Flush writer
writer.flush()
def log_model_summary(
model: Any,
input_shape: tuple,
logger: Optional[logging.Logger] = None
) -> None:
"""
Log model summary.
Args:
model: Model
input_shape: Input shape
logger: Logger
"""
# Create dummy input
dummy_input = jnp.ones(input_shape, dtype=jnp.int32)
# Initialize model
params = model.init(jax.random.PRNGKey(0), dummy_input)
# Count parameters
param_count = sum(
np.prod(p.shape) for p in jax.tree_util.tree_leaves(params)
)
# Log model summary
summary = f"Model summary:\n"
summary += f" Input shape: {input_shape}\n"
summary += f" Parameter count: {param_count:,}\n"
# Log to console
print(summary)
# Log to logger
if logger is not None:
logger.info(summary)
|