kshitijthakkar/smoltrace-metrics-20260304_071138

Tiny Agents. Total Visibility.

---

SMOLTRACE GPU & Environmental Metrics

This dataset contains time-series GPU metrics and environmental impact data from a SMOLTRACE benchmark run.

Dataset Information

Field	Value
Model	Nanbeige/Nanbeige4.1-3B
Run ID	cf816e4a-1745-4010-a7cb-2bde4dfc02d5
Total Samples	334
Generated	2026-03-04 08:08:56 UTC
GPU Metrics	Available

Schema

Column	Type	Description
run_id	string	Unique run identifier
timestamp	string	ISO timestamp of measurement
timestamp_unix_nano	string	Unix nanosecond timestamp
service_name	string	Service identifier
gpu_id	string	GPU device ID
gpu_name	string	GPU model name
gpu_utilization_percent	float	GPU compute utilization (0-100%)
gpu_memory_used_mib	float	GPU memory used (MiB)
gpu_memory_total_mib	float	Total GPU memory (MiB)
gpu_temperature_celsius	float	GPU temperature (°C)
gpu_power_watts	float	GPU power consumption (W)
co2_emissions_gco2e	float	Cumulative CO2 emissions (gCO2e)
power_cost_usd	float	Cumulative power cost (USD)

Environmental Impact

SMOLTRACE tracks environmental metrics to help you understand the carbon footprint of your AI workloads:

• CO2 Emissions: Calculated based on GPU power consumption and regional carbon intensity
• Power Cost: Estimated electricity cost based on configurable rates

Usage

from datasets import load_dataset
import pandas as pd

# Load metrics
ds = load_dataset("YOUR_USERNAME/smoltrace-metrics-TIMESTAMP")

# Convert to DataFrame for analysis
df = pd.DataFrame(ds['train'])

# Plot GPU utilization over time
import matplotlib.pyplot as plt
plt.plot(df['timestamp'], df['gpu_utilization_percent'])
plt.xlabel('Time')
plt.ylabel('GPU Utilization (%)')
plt.title('GPU Utilization During Evaluation')
plt.show()

# Get total environmental impact
total_co2 = df['co2_emissions_gco2e'].max()
total_cost = df['power_cost_usd'].max()
print(f"Total CO2: {total_co2:.4f} gCO2e")
print(f"Total Cost: ${total_cost:.6f}")

Related Datasets

This evaluation run also generated:

• Results Dataset: Pass/fail outcomes for each test case
• Traces Dataset: Detailed OpenTelemetry execution traces
• Leaderboard: Aggregated metrics for model comparison

---

About SMOLTRACE

SMOLTRACE is a comprehensive benchmarking and evaluation framework for Smolagents - HuggingFace's lightweight agent library.

Key Features

• Automated agent evaluation with customizable test cases
• OpenTelemetry-based tracing for detailed execution insights
• GPU metrics collection (utilization, memory, temperature, power)
• CO2 emissions and power cost tracking
• Leaderboard aggregation and comparison

Quick Links

• GitHub Repository
• PyPI Package
• Documentation
• Report Issues

Installation

pip install smoltrace

Citation

If you use SMOLTRACE in your research, please cite:

@software{smoltrace,
  title = {SMOLTRACE: Benchmarking Framework for Smolagents},
  author = {Thakkar, Kshitij},
  url = {https://github.com/Mandark-droid/SMOLTRACE},
  year = {2025}
}

---

_{Generated by SMOLTRACE}