Source Review: SAP-RPT-1-OSS Predictor
Review date: 2026-06-18
Plugin: sap-rpt1
Evidence type: public source/model/product-page review
Live tenant/system validation: Not performed
Local model execution validation: Not performed
Sources
| Source | Role | Notes |
|---|---|---|
https://github.com/SAP-samples/sap-rpt-1-oss |
Primary SAP source repository | Use for installation, local model workflow, requirements, examples, license notes, and issue tracking. |
https://huggingface.co/SAP/sap-rpt-1-oss |
Primary model source | Use for model access, model-card context, library usage, and license/access notes. |
https://www.sap.com/products/artificial-intelligence/sap-rpt.html |
SAP product context | Use only for product positioning, hosted playground URL, and future SAP-RPT-1.5 context. |
Scope Covered
This review covers public source, public model-card, and public product-page review for a local SAP-RPT-1-OSS FI/CO prototype skill.
Scope Excluded
This review does not cover:
- Live SAP tenant validation.
- Live SAP system validation.
- Production finance workflow validation.
- SAP AI Core deployment validation.
- Hosted SAP-RPT API validation.
- Local model inference benchmark validation.
- Commercial/legal approval for model use.
- Customer-specific FI/CO table mappings.
- Customer-specific S/4HANA or ECC release behavior.
Source Findings
- SAP-samples
sap-rpt-1-ossdescribes SAP-RPT-1-OSS as the implementation of the inference pipeline from the ConTextTab paper. - The repository states Python 3.11 requirements and installation from source with
pip install git+https://github.com/SAP-samples/sap-rpt-1-oss. - The repository states model checkpoints are downloaded from Hugging Face and require Hugging Face login.
- The repository supports classification and regression through
SAP_RPT_OSS_ClassifierandSAP_RPT_OSS_Regressor. - The repository presents GPU guidance as upstream guidance: 80 GB memory, context size 8192, and bagging 8 for best performance; smaller context and bagging 1 for lighter use.
- The repository license is Apache-2.0 except as noted in the repository license files.
- The SAP product page states the SAP RPT playground is available at
https://rpt.cloud.sapand states SAP-RPT-1.5 commercial availability/productization context. Treat SAP-RPT-1.5 as future/H2 2026 context for this skill.
Product Distinction
- SAP-RPT-1-OSS: local open model workflow to document in this skill.
- Hosted SAP-RPT-1 playground: product context only; do not bundle a hosted client.
- SAP-RPT-1.5: future/H2 2026 context from this review; do not describe as currently available in this skill.
Upstream Open Issues
Reviewed from https://github.com/SAP-samples/sap-rpt-1-oss/issues on 2026-06-18.
| Issue | Title | Status as reviewed | Skill handling |
|---|---|---|---|
| #29 | predict() returns all-NaN when X is a DataFrame and y is a numpy array |
Open | Prefer consistent pandas inputs in examples without representing this workaround as a complete upstream resolution. |
| #27 | Single-sample prediction fails with ValueError: zero-dimensional arrays cannot be concatenated |
Open | The bundled wrapper blocks single-row scoring by default unless --allow-single-row is explicit. |
| #26 | Training code availability? |
Open | Describe this skill as inference workflow guidance, not model training guidance. |
| #25 | SAP_RPT_OSS_Classifier fails to recognize Pandas Int64Dtype |
Open | Normalize nullable integer columns before local inference when needed. |
| #9 | SAP BTP AI Core deployment? |
Open | Route deployment questions to sap-ai-core; do not document deployment as validated here. |
Status: upstream open as of 2026-06-18. Workarounds are unverified in this repository unless separately documented.
License Notes
- The SAP Skills plugin frontmatter must use
GPL-3.0. - SAP-samples
sap-rpt-1-osssource repository is Apache-2.0 except as noted in that repository. - This plugin does not imply SAP-RPT-1-OSS model weights are covered by the SAP Skills repository license.