Commit ·
d9b845d
0
Parent(s):
ChakraTS model card
Browse filesCo-authored-by: Shubham Agrawal <Shubham Agrawal@users.noreply.huggingface.co>
- .gitattributes +35 -0
- README.md +170 -0
.gitattributes
ADDED
|
@@ -0,0 +1,35 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| 1 |
+
*.7z filter=lfs diff=lfs merge=lfs -text
|
| 2 |
+
*.arrow filter=lfs diff=lfs merge=lfs -text
|
| 3 |
+
*.bin filter=lfs diff=lfs merge=lfs -text
|
| 4 |
+
*.bz2 filter=lfs diff=lfs merge=lfs -text
|
| 5 |
+
*.ckpt filter=lfs diff=lfs merge=lfs -text
|
| 6 |
+
*.ftz filter=lfs diff=lfs merge=lfs -text
|
| 7 |
+
*.gz filter=lfs diff=lfs merge=lfs -text
|
| 8 |
+
*.h5 filter=lfs diff=lfs merge=lfs -text
|
| 9 |
+
*.joblib filter=lfs diff=lfs merge=lfs -text
|
| 10 |
+
*.lfs.* filter=lfs diff=lfs merge=lfs -text
|
| 11 |
+
*.mlmodel filter=lfs diff=lfs merge=lfs -text
|
| 12 |
+
*.model filter=lfs diff=lfs merge=lfs -text
|
| 13 |
+
*.msgpack filter=lfs diff=lfs merge=lfs -text
|
| 14 |
+
*.npy filter=lfs diff=lfs merge=lfs -text
|
| 15 |
+
*.npz filter=lfs diff=lfs merge=lfs -text
|
| 16 |
+
*.onnx filter=lfs diff=lfs merge=lfs -text
|
| 17 |
+
*.ot filter=lfs diff=lfs merge=lfs -text
|
| 18 |
+
*.parquet filter=lfs diff=lfs merge=lfs -text
|
| 19 |
+
*.pb filter=lfs diff=lfs merge=lfs -text
|
| 20 |
+
*.pickle filter=lfs diff=lfs merge=lfs -text
|
| 21 |
+
*.pkl filter=lfs diff=lfs merge=lfs -text
|
| 22 |
+
*.pt filter=lfs diff=lfs merge=lfs -text
|
| 23 |
+
*.pth filter=lfs diff=lfs merge=lfs -text
|
| 24 |
+
*.rar filter=lfs diff=lfs merge=lfs -text
|
| 25 |
+
*.safetensors filter=lfs diff=lfs merge=lfs -text
|
| 26 |
+
saved_model/**/* filter=lfs diff=lfs merge=lfs -text
|
| 27 |
+
*.tar.* filter=lfs diff=lfs merge=lfs -text
|
| 28 |
+
*.tar filter=lfs diff=lfs merge=lfs -text
|
| 29 |
+
*.tflite filter=lfs diff=lfs merge=lfs -text
|
| 30 |
+
*.tgz filter=lfs diff=lfs merge=lfs -text
|
| 31 |
+
*.wasm filter=lfs diff=lfs merge=lfs -text
|
| 32 |
+
*.xz filter=lfs diff=lfs merge=lfs -text
|
| 33 |
+
*.zip filter=lfs diff=lfs merge=lfs -text
|
| 34 |
+
*.zst filter=lfs diff=lfs merge=lfs -text
|
| 35 |
+
*tfevents* filter=lfs diff=lfs merge=lfs -text
|
README.md
ADDED
|
@@ -0,0 +1,170 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| 1 |
+
---
|
| 2 |
+
license: other
|
| 3 |
+
license_name: yhat-labs-api-terms
|
| 4 |
+
license_link: https://yhatlabs.com/terms
|
| 5 |
+
pipeline_tag: time-series-forecasting
|
| 6 |
+
tags:
|
| 7 |
+
- time-series
|
| 8 |
+
- forecasting
|
| 9 |
+
- probabilistic-forecasting
|
| 10 |
+
- foundation-model
|
| 11 |
+
- covariates
|
| 12 |
+
- hosted-api
|
| 13 |
+
model-index:
|
| 14 |
+
- name: ChakraTS
|
| 15 |
+
results:
|
| 16 |
+
- task:
|
| 17 |
+
type: time-series-forecasting
|
| 18 |
+
dataset:
|
| 19 |
+
name: fev-bench
|
| 20 |
+
type: autogluon/fev_datasets
|
| 21 |
+
metrics:
|
| 22 |
+
- type: win rate
|
| 23 |
+
name: Win rate (all 100 tasks)
|
| 24 |
+
value: 0.893
|
| 25 |
+
- type: SQL skill score
|
| 26 |
+
name: Scaled quantile loss skill score (all 100 tasks)
|
| 27 |
+
value: 51.3
|
| 28 |
+
source:
|
| 29 |
+
name: fev-bench leaderboard
|
| 30 |
+
url: https://huggingface.co/spaces/autogluon/fev-leaderboard
|
| 31 |
+
- task:
|
| 32 |
+
type: time-series-forecasting
|
| 33 |
+
dataset:
|
| 34 |
+
name: GIFT-Eval
|
| 35 |
+
type: Salesforce/GiftEval
|
| 36 |
+
metrics:
|
| 37 |
+
- type: MASE
|
| 38 |
+
name: MASE (geometric mean, relative to seasonal naive)
|
| 39 |
+
value: 0.644
|
| 40 |
+
- type: CRPS
|
| 41 |
+
name: CRPS (geometric mean, relative to seasonal naive)
|
| 42 |
+
value: 0.438
|
| 43 |
+
source:
|
| 44 |
+
name: GIFT-Eval leaderboard
|
| 45 |
+
url: https://huggingface.co/spaces/Salesforce/GIFT-Eval
|
| 46 |
+
- task:
|
| 47 |
+
type: time-series-forecasting
|
| 48 |
+
dataset:
|
| 49 |
+
name: TIME
|
| 50 |
+
type: Real-TSF/TIME
|
| 51 |
+
metrics:
|
| 52 |
+
- type: MASE
|
| 53 |
+
name: MASE (geometric mean, relative to seasonal naive, 98 tasks)
|
| 54 |
+
value: 0.624
|
| 55 |
+
- type: CRPS
|
| 56 |
+
name: CRPS (geometric mean, relative to seasonal naive, 98 tasks)
|
| 57 |
+
value: 0.523
|
| 58 |
+
source:
|
| 59 |
+
name: TIME leaderboard
|
| 60 |
+
url: https://huggingface.co/spaces/Real-TSF/TIME-leaderboard
|
| 61 |
+
---
|
| 62 |
+
|
| 63 |
+
# ChakraTS
|
| 64 |
+
|
| 65 |
+
**ChakraTS is a probabilistic time-series forecasting model from [YHat Labs](https://yhatlabs.com), served through an API.
|
| 66 |
+
This repository holds the model card; there are no weights to download. Updated November 2026.**
|
| 67 |
+
|
| 68 |
+
Send your series and a horizon, and get quantiles for every future step. ChakraTS forecasts new series without being
|
| 69 |
+
trained on them, reads past and known-future covariates (numeric or categorical), forecasts several targets of a series
|
| 70 |
+
together, and works at any regular frequency from seconds to years.
|
| 71 |
+
|
| 72 |
+
- Website: https://yhatlabs.com
|
| 73 |
+
- API access: https://yhatlabs.com/models/access
|
| 74 |
+
- Python library: `pip install yhatlabs` ([GitHub](https://github.com/yhatlabs/yhatlabs-python))
|
| 75 |
+
- Recipes and benchmark files: [yhatlabs-cookbook](https://github.com/yhatlabs/yhatlabs-cookbook)
|
| 76 |
+
- Tabular model: [yhatlabs/ChakraTab](https://huggingface.co/yhatlabs/ChakraTab)
|
| 77 |
+
|
| 78 |
+
## Benchmark standing (as of November 2026)
|
| 79 |
+
|
| 80 |
+
ChakraTS is the best model across forecasting benchmarks.
|
| 81 |
+
|
| 82 |
+
| Benchmark | Standing | Result | Leaderboard |
|
| 83 |
+
|---|---|---|---|
|
| 84 |
+
| fev-bench (Amazon) | 1st by win rate | Win rate 89.3%, skill score 51.3 | [fev-bench](https://huggingface.co/spaces/autogluon/fev-leaderboard) |
|
| 85 |
+
| GIFT-Eval (Salesforce) | 2nd among non-agentic models | MASE 0.644 | [GIFT-Eval](https://huggingface.co/spaces/Salesforce/GIFT-Eval) |
|
| 86 |
+
| TIME | 1st by MASE rank | MASE 0.624, CRPS 0.523 | [TIME](https://huggingface.co/spaces/Real-TSF/TIME-leaderboard) |
|
| 87 |
+
|
| 88 |
+
GIFT-Eval and TIME scores are normalized by Seasonal Naive (lower is better). The results use the `pro` preset. Result
|
| 89 |
+
files and the code to reproduce them are in the [cookbook](https://github.com/yhatlabs/yhatlabs-cookbook/tree/main/benchmarks).
|
| 90 |
+
|
| 91 |
+
## Getting started
|
| 92 |
+
|
| 93 |
+
```python
|
| 94 |
+
from yhatlabs import Client
|
| 95 |
+
|
| 96 |
+
yh = Client() # reads YHAT_API_KEY
|
| 97 |
+
fc = yh.forecast(df, h=24, preset="pro") # df: one row per series and timestamp
|
| 98 |
+
print(fc.df.head())
|
| 99 |
+
```
|
| 100 |
+
|
| 101 |
+
Or call the API directly:
|
| 102 |
+
|
| 103 |
+
```bash
|
| 104 |
+
curl -X POST https://api.yhatlabs.com/v1/series/forecast \
|
| 105 |
+
-H "Authorization: Bearer $YHAT_API_KEY" \
|
| 106 |
+
-H "Content-Type: application/json" \
|
| 107 |
+
-d '{
|
| 108 |
+
"model": "chakra-ts",
|
| 109 |
+
"preset": "pro",
|
| 110 |
+
"series": [
|
| 111 |
+
{"id": "store_1", "values": [12, 15, 11, 18, 20, 17, 22, 19, 24, 21, 25, 23, 27, 26]},
|
| 112 |
+
{"id": "store_2", "values": [40, 38, 45, 47, 44, 50, 52, 49, 55, 53, 58, 57, 60, 62]}
|
| 113 |
+
],
|
| 114 |
+
"horizon": 7,
|
| 115 |
+
"freq": "D"
|
| 116 |
+
}'
|
| 117 |
+
```
|
| 118 |
+
|
| 119 |
+
Each forecast comes back as a `[horizon, 9]` array of quantiles (levels 0.1 to 0.9 by default). The median is the
|
| 120 |
+
point forecast.
|
| 121 |
+
|
| 122 |
+
## Capabilities
|
| 123 |
+
|
| 124 |
+
| Property | Detail |
|
| 125 |
+
|---|---|
|
| 126 |
+
| Training on your data | not needed; optional tuning on your own history |
|
| 127 |
+
| Presets | `fast` (default) and `pro` (most accurate, slower) |
|
| 128 |
+
| Context | up to 32,768 past points per series |
|
| 129 |
+
| Horizon | up to 2,048 steps |
|
| 130 |
+
| Targets | one or several per series, forecast together |
|
| 131 |
+
| Covariates | past and known-future, numeric or categorical |
|
| 132 |
+
| Missing values | allowed, marked as `null` |
|
| 133 |
+
| Output | quantiles at any levels (the API returns 0.1 to 0.9 by default), sample paths |
|
| 134 |
+
| Batching | up to 256 series per request; the Python library splits larger inputs |
|
| 135 |
+
|
| 136 |
+
## Intended use
|
| 137 |
+
|
| 138 |
+
- **Energy:** load, renewable generation and electricity price forecasting, with published forecasts as covariates.
|
| 139 |
+
- **Retail and supply chain:** demand per product and location, with promotions, prices and calendar effects.
|
| 140 |
+
- **Operations and infrastructure:** traffic, capacity, call volumes, cloud and sensor metrics.
|
| 141 |
+
- **Finance and planning:** revenue, cash-flow and KPI projections with prediction intervals.
|
| 142 |
+
- **Scenario analysis and backtesting:** the same series under alternative covariate paths, and rolling-origin
|
| 143 |
+
evaluation on your own history.
|
| 144 |
+
|
| 145 |
+
## Not intended use
|
| 146 |
+
|
| 147 |
+
- Series without a regular time index.
|
| 148 |
+
- Very short histories: forecasts are returned, but the intervals are wide.
|
| 149 |
+
- Classification or anomaly labelling as such; see [ChakraTab](https://huggingface.co/yhatlabs/ChakraTab) for prediction
|
| 150 |
+
on tables.
|
| 151 |
+
|
| 152 |
+
## Terms
|
| 153 |
+
|
| 154 |
+
Use of the API is governed by the [YHat Labs terms](https://yhatlabs.com/terms), which also describe how request data is
|
| 155 |
+
handled. ChakraTS is available for commercial use under those terms; enterprise and on-premises arrangements on request.
|
| 156 |
+
|
| 157 |
+
## Citation
|
| 158 |
+
|
| 159 |
+
```bibtex
|
| 160 |
+
@misc{chakrats2026,
|
| 161 |
+
title = {ChakraTS: probabilistic time-series forecasting as a service},
|
| 162 |
+
author = {YHat Labs},
|
| 163 |
+
year = {2026},
|
| 164 |
+
url = {https://huggingface.co/yhatlabs/ChakraTS}
|
| 165 |
+
}
|
| 166 |
+
```
|
| 167 |
+
|
| 168 |
+
## Contact
|
| 169 |
+
|
| 170 |
+
founders@yhatlabs.com · https://yhatlabs.com
|