Datasets:
Add Hart-36M evaluation results
#49
by pass-lin - opened
Model Information
- Model: Hart-36M
- Paper: HART: Hybrid Attention-RWKV Time series Foundation Model
- Code: https://github.com/pass-lin/HART
- Weights: https://www.modelscope.cn/models/q935499957/HART-36M
Description
Hart-36M is a ~36M parameter time series foundation model that interleaves RWKV-7 blocks with causal Transformer blocks and outputs quantile forecasts. This PR submits its zero-shot evaluation results on the GIFT-Eval benchmark.
zqiao11 changed pull request status to merged