Update README.md
Browse files
README.md
CHANGED
|
@@ -1,7 +1,7 @@
|
|
| 1 |
---
|
| 2 |
license: apache-2.0
|
| 3 |
library_name: pytorch
|
| 4 |
-
pipeline_tag: time-series-
|
| 5 |
tags:
|
| 6 |
- sensor-time-series
|
| 7 |
- physiological-signal
|
|
@@ -13,19 +13,16 @@ datasets:
|
|
| 13 |
- physionet/mimic-iv-ecg
|
| 14 |
---
|
| 15 |
|
| 16 |
-
# SensorGen
|
| 17 |
|
| 18 |
[](#)
|
| 19 |
[](#)
|
| 20 |
[](https://github.com/yang-ai-lab/SensorGen)
|
| 21 |
-
[](https://www.python.org/)
|
| 23 |
[](https://pytorch.org/)
|
| 24 |
|
| 25 |
|
| 26 |
-
> Scalable Interpolant Transformer (SiT-B/1d) reference models for the
|
| 27 |
-
> SensorGen benchmark on real-world sensor time series.
|
| 28 |
-
|
| 29 |
This repository hosts the pre-trained checkpoints used in the SensorGen study
|
| 30 |
("*Signal or Noise? Understanding Generative Models for Real-World Sensor
|
| 31 |
Time Series*").
|
|
@@ -36,12 +33,6 @@ Time Series*").
|
|
| 36 |
| `bp_translation.pt` | PPG and NIBP to invasive BP | VitalDB |
|
| 37 |
|
| 38 |
|
| 39 |
-
For VitalDB forecasting we observed a U-shape in validation loss past ~20K
|
| 40 |
-
steps; the **18K best-validation checkpoint** is what we use to report paper
|
| 41 |
-
numbers. The 200K extended checkpoint is provided as a reference for
|
| 42 |
-
practitioners who prefer the longer-trained generator at the cost of mild
|
| 43 |
-
overfit on the validation set.
|
| 44 |
-
|
| 45 |
|
| 46 |
## Usage
|
| 47 |
|
|
@@ -54,15 +45,15 @@ Pair this checkpoint repository with the GitHub code release at
|
|
| 54 |
from huggingface_hub import hf_hub_download
|
| 55 |
|
| 56 |
ckpt_path = hf_hub_download(
|
| 57 |
-
repo_id="yang-ai-lab/SensorGen
|
| 58 |
filename="text2ecg.pt",
|
| 59 |
)
|
| 60 |
```
|
| 61 |
|
| 62 |
-
### Download all checkpoints
|
| 63 |
|
| 64 |
```bash
|
| 65 |
-
hf download yang-ai-lab/SensorGen
|
| 66 |
```
|
| 67 |
|
| 68 |
|
|
|
|
| 1 |
---
|
| 2 |
license: apache-2.0
|
| 3 |
library_name: pytorch
|
| 4 |
+
pipeline_tag: time-series-generation
|
| 5 |
tags:
|
| 6 |
- sensor-time-series
|
| 7 |
- physiological-signal
|
|
|
|
| 13 |
- physionet/mimic-iv-ecg
|
| 14 |
---
|
| 15 |
|
| 16 |
+
# SensorGen
|
| 17 |
|
| 18 |
[](#)
|
| 19 |
[](#)
|
| 20 |
[](https://github.com/yang-ai-lab/SensorGen)
|
| 21 |
+
[](https://huggingface.co/yang-ai-lab/SensorGen-SiT)
|
| 22 |
[](https://www.python.org/)
|
| 23 |
[](https://pytorch.org/)
|
| 24 |
|
| 25 |
|
|
|
|
|
|
|
|
|
|
| 26 |
This repository hosts the pre-trained checkpoints used in the SensorGen study
|
| 27 |
("*Signal or Noise? Understanding Generative Models for Real-World Sensor
|
| 28 |
Time Series*").
|
|
|
|
| 33 |
| `bp_translation.pt` | PPG and NIBP to invasive BP | VitalDB |
|
| 34 |
|
| 35 |
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| 36 |
|
| 37 |
## Usage
|
| 38 |
|
|
|
|
| 45 |
from huggingface_hub import hf_hub_download
|
| 46 |
|
| 47 |
ckpt_path = hf_hub_download(
|
| 48 |
+
repo_id="yang-ai-lab/SensorGen",
|
| 49 |
filename="text2ecg.pt",
|
| 50 |
)
|
| 51 |
```
|
| 52 |
|
| 53 |
+
### Download all checkpoints
|
| 54 |
|
| 55 |
```bash
|
| 56 |
+
hf download yang-ai-lab/SensorGen --local-dir ./ckpts
|
| 57 |
```
|
| 58 |
|
| 59 |
|