IBM logo

PatchTST (ETTh1)

online
ibm-granite/granite-timeseries-patchtst

616K params | 512 context | $0.5000 input | $1.50 output

This checkpoint is IBM's hosted PatchTST reference model for ETTh1 rather than a broad multi-domain TSFM family card. It uses patch tokenization and Transformer blocks to model long-horizon time-series behavior efficiently, and the published checkpoint is tuned around the ETTh1 workload. It is most useful as a strong baseline and interpretable point of comparison against the larger foundation models in the catalog.

Model Classification

Family

PatchTST

Type

pretrained forecasting model

Pretrained forecasting checkpoint hosted on TSFM.ai as a reference model, but not positioned upstream as a general-purpose TSFM family.

Training Data

ETTh1 train split covering all seven ETTh1 channels, per the official model card.

Recommended For

  • Reference long-horizon forecasting baselines
  • Interpretable comparisons against stronger hosted TSFMs

Strengths

  • Strong and widely cited long-horizon patch-transformer baseline
  • Easy comparison point for benchmark readers

Limitations

  • This hosted checkpoint is dataset-specific rather than a universal foundation model
  • Narrower operational scope than the broader TSFM families in the catalog

Capabilities

forecastingmultivariatelong-horizon

Tags

ibmpatchtstbaselinelong-horizon

Specifications

Parameters
616K
Architecture
PatchTST transformer with patch tokenization
Context length
512
Max output
1,024
Avg latency
n/a
Uptime
n/a
Rate limit
n/a
Accelerator
NVIDIA GPU
Regions
Virginia, US
License
n/a

Pricing

Input / 1M tokens
$0.5000
Output / 1M tokens
$1.50

Performance

Average latency
n/a
Availability
n/a
Rate limit
n/a