CoolFace
Modelpublic

ts-arena/chronos_bolt_tiny_9m_forecasting

sourceHugging Faceapache-2.0updated 9mo agoView on Hugging Face
0likes9downloads
Model Card

chronosbolttiny9mforecasting

TS Arena wrapper for Amazon Chronos BOLT-Tiny time series forecasting model.

Model Description

Chronos-Bolt is an optimized version of the Chronos family, designed for faster inference while maintaining accuracy. It uses an optimized T5 architecture with improved tokenization and reduced computational overhead.

AttributeValue
Parameters9M
ArchitectureT5 Optimized (Bolt)
Original Repoamazon/chronos-bolt-tiny
PaperChronos: Learning the Language of Time Series
TaskTime Series Forecasting

Usage with TS Arena

python
import ts_arena

# Load model
model = ts_arena.load_model("chronos-bolt-tiny")

# Generate forecasts
import numpy as np
context = np.random.randn(96)  # 96 timesteps of history
output = model.predict(context, prediction_length=24, num_samples=20)

# Access results
print(output.predictions.shape)      # Point forecasts (median)
print(output.quantiles[0.5].shape)   # Median forecast
print(output.quantiles[0.1].shape)   # 10th percentile
print(output.quantiles[0.9].shape)   # 90th percentile

Direct Usage with Chronos

python
from chronos import ChronosPipeline
import torch

pipeline = ChronosPipeline.from_pretrained(
    "amazon/chronos-bolt-tiny",
    device_map="cuda",
    torch_dtype=torch.bfloat16,
)

context = torch.randn(1, 96)  # (batch, time)
forecast = pipeline.predict(context, prediction_length=24, num_samples=20)

Evaluation Results

ETTh1 Dataset (context=96, horizon=96)

MetricValue
MSE4.37
MAE1.66
RMSE2.09

Features

  • —Zero-shot forecasting: No training required
  • —Probabilistic forecasts: Returns samples and quantiles
  • —Variable horizons: Supports different prediction lengths
  • —Multivariate support: Processes each channel independently

Limitations

  • —Univariate model (multivariate handled channel-by-channel)
  • —No exogenous variable support
  • —Recommended max prediction length: 64

Citation

bibtex
@article{ansari2024chronos,
  title={Chronos: Learning the Language of Time Series},
  author={Ansari, Abdul Fatir and Stella, Lorenzo and Turkmen, Caner and Zhang, Xiyuan and others},
  journal={arXiv preprint arXiv:2403.07815},
  year={2024}
}

License

Apache-2.0 (following the original Chronos license)

Links