[AINode] Integrate PatchTST-FM-R1 as builtin forecasting model

[karthikreddy-02]

Search before asking

• I searched in the issues and found nothing similar.

Motivation

Description:

Summary: This issue tracks the integration of IBM Research's PatchTST-FM-R1 into IoTDB AINode's built-in model registry, following the contribution guide documented here.

Motivation: PatchTST-FM-R1 is a SOTA patch-based Transformer foundation model for time series forecasting. Adding it expands the range of zero-shot forecasting options available to IoTDB users without requiring model training.

Implementation Plan:

Create iotdb-core/ainode/iotdb/ainode/core/model/patchtst/ directory with a PatchTSTFMPipeline extending

ForecastPipeline
Register the model in BUILTIN_HF_TRANSFORMERS_MODEL_MAP in

model_info.py
Add integration test
Reference: PR #16903 (Chronos2 integration) as the implementation pattern.

Solution

No response

Alternatives

No response

Are you willing to submit a PR?

• I'm willing to submit a PR!

[shubhika123]

Hi @karthikreddy-02 , is this issue still open?
I’d like to take this up. I’ve been going through the Chronos2 integration PR to understand how models are added and registered in AINode.
I’m planning to use a similar structure to integrate PatchTST-FM-R1, starting with the pipeline setup and model registration, then adding basic tests.
Let me know if I can work on this.

[karthikreddy-02]

Hi @shubhika123. I'm waiting for my PR to be merged wrt this issue. Thanks!

[shubhika123]

Okay, sure.