Spaces:
Sleeping
Sleeping
comi
Browse files- mlflow_config.yaml +12 -0
- mlflow_tracker.py +66 -0
mlflow_config.yaml
ADDED
@@ -0,0 +1,12 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
1 |
+
# mlflow_config.yaml
|
2 |
+
|
3 |
+
experiment_name: "billion-row-analysis"
|
4 |
+
run_name: "benchmarking"
|
5 |
+
tracking_uri: "http://localhost:5000" # Optional: set if you have a remote tracking server
|
6 |
+
metrics:
|
7 |
+
- Library
|
8 |
+
- Load Time (s)
|
9 |
+
- CPU Load (%)
|
10 |
+
- Memory Load (%)
|
11 |
+
- Peak Memory (%)
|
12 |
+
|
mlflow_tracker.py
ADDED
@@ -0,0 +1,66 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
1 |
+
import mlflow
|
2 |
+
import yaml
|
3 |
+
import os
|
4 |
+
|
5 |
+
class MLflowTracker:
|
6 |
+
"""
|
7 |
+
A reusable MLflow tracking class that reads configuration from a YAML file.
|
8 |
+
This class sets up the MLflow experiment and run, and exposes methods to log parameters,
|
9 |
+
metrics, and artifacts.
|
10 |
+
"""
|
11 |
+
def __init__(self, config_file="mlflow_config.yaml"):
|
12 |
+
# Load configuration from the YAML file.
|
13 |
+
if not os.path.exists(config_file):
|
14 |
+
raise FileNotFoundError(f"Config file '{config_file}' not found.")
|
15 |
+
with open(config_file, "r") as f:
|
16 |
+
self.config = yaml.safe_load(f)
|
17 |
+
|
18 |
+
# Set up configuration parameters
|
19 |
+
self.experiment_name = self.config.get("experiment_name", "Default_Experiment")
|
20 |
+
self.run_name = self.config.get("run_name", "Default_Run")
|
21 |
+
self.tracking_uri = self.config.get("tracking_uri", None)
|
22 |
+
self.metrics_to_track = self.config.get("metrics", [])
|
23 |
+
|
24 |
+
# Set tracking URI if provided
|
25 |
+
if self.tracking_uri:
|
26 |
+
mlflow.set_tracking_uri(self.tracking_uri)
|
27 |
+
|
28 |
+
# Set the experiment
|
29 |
+
mlflow.set_experiment(self.experiment_name)
|
30 |
+
|
31 |
+
# Start the run
|
32 |
+
self.run = mlflow.start_run(run_name=self.run_name)
|
33 |
+
print(f"MLflow run started: Experiment='{self.experiment_name}', Run='{self.run_name}'")
|
34 |
+
|
35 |
+
def log_param(self, key, value):
|
36 |
+
"""Log a single parameter."""
|
37 |
+
mlflow.log_param(key, value)
|
38 |
+
|
39 |
+
def log_params(self, params: dict):
|
40 |
+
"""Log multiple parameters from a dictionary."""
|
41 |
+
mlflow.log_params(params)
|
42 |
+
|
43 |
+
def log_metric(self, key, value, step=None):
|
44 |
+
"""Log a single metric. Optionally include a step value."""
|
45 |
+
mlflow.log_metric(key, value, step=step)
|
46 |
+
|
47 |
+
def log_metrics(self, metrics: dict, step=None):
|
48 |
+
"""Log multiple metrics from a dictionary."""
|
49 |
+
for key, value in metrics.items():
|
50 |
+
self.log_metric(key, value, step=step)
|
51 |
+
|
52 |
+
def log_artifact(self, file_path, artifact_path=None):
|
53 |
+
"""Log an artifact (file) to MLflow."""
|
54 |
+
mlflow.log_artifact(file_path, artifact_path=artifact_path)
|
55 |
+
|
56 |
+
def end_run(self):
|
57 |
+
"""End the current MLflow run."""
|
58 |
+
mlflow.end_run()
|
59 |
+
print("MLflow run ended.")
|
60 |
+
|
61 |
+
# Example usage (can be removed or placed in a separate test script):
|
62 |
+
if __name__ == "__main__":
|
63 |
+
tracker = MLflowTracker("mlflow_config.yaml")
|
64 |
+
tracker.log_param("example_param", 123)
|
65 |
+
tracker.log_metric("example_metric", 0.95)
|
66 |
+
tracker.end_run()
|