Model card updates and fixes
Browse files
README.md
CHANGED
|
@@ -8,11 +8,13 @@ license_link: LICENSE
|
|
| 8 |
datasets:
|
| 9 |
- imagenet-1k
|
| 10 |
---
|
|
|
|
| 11 |
# FastViT: A Fast Hybrid Vision Transformer using Structural Reparameterization
|
| 12 |
|
| 13 |
Please observe [original license](https://github.com/apple/ml-fastvit/blob/8af5928238cab99c45f64fc3e4e7b1516b8224ba/LICENSE).
|
| 14 |
|
| 15 |
## Model Details
|
|
|
|
| 16 |
- **Model Type:** Image classification / feature backbone
|
| 17 |
- **Model Stats:**
|
| 18 |
- Params (M): 4.0
|
|
@@ -28,10 +30,8 @@ Please observe [original license](https://github.com/apple/ml-fastvit/blob/8af59
|
|
| 28 |
|
| 29 |
| Variant | Parameters | Size (MB) | Weight precision | Act. precision | Δ Pytorch acc |
|
| 30 |
| ------------------------------------------------------- | ---------: | --------: | ---------------- | -------------- | ------------- |
|
| 31 |
-
| T8
|
| 32 |
-
| MA36
|
| 33 |
-
|
| 34 |
-
|
| 35 |
|
| 36 |
## Evaluation - Inference time
|
| 37 |
|
|
@@ -42,7 +42,28 @@ Please observe [original license](https://github.com/apple/ml-fastvit/blob/8af59
|
|
| 42 |
| MA36 | iPhone 12 Pro Max | 18.0 | 4.50 | Neural Engine |
|
| 43 |
| MA36 | M3 Max | 15.0 | 2.99 | Neural Engine |
|
| 44 |
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| 45 |
## Citation
|
|
|
|
| 46 |
```bibtex
|
| 47 |
@inproceedings{vasufastvit2023,
|
| 48 |
author = {Pavan Kumar Anasosalu Vasu and James Gabriel and Jeff Zhu and Oncel Tuzel and Anurag Ranjan},
|
|
|
|
| 8 |
datasets:
|
| 9 |
- imagenet-1k
|
| 10 |
---
|
| 11 |
+
|
| 12 |
# FastViT: A Fast Hybrid Vision Transformer using Structural Reparameterization
|
| 13 |
|
| 14 |
Please observe [original license](https://github.com/apple/ml-fastvit/blob/8af5928238cab99c45f64fc3e4e7b1516b8224ba/LICENSE).
|
| 15 |
|
| 16 |
## Model Details
|
| 17 |
+
|
| 18 |
- **Model Type:** Image classification / feature backbone
|
| 19 |
- **Model Stats:**
|
| 20 |
- Params (M): 4.0
|
|
|
|
| 30 |
|
| 31 |
| Variant | Parameters | Size (MB) | Weight precision | Act. precision | Δ Pytorch acc |
|
| 32 |
| ------------------------------------------------------- | ---------: | --------: | ---------------- | -------------- | ------------- |
|
| 33 |
+
| [T8](https://huggingface.co/apple/FastViTT8F16.mlpackage) | 3.6M | 7.8 | Float16 | Float16 | -0.9% |
|
| 34 |
+
| [MA36](https://huggingface.co/apple/FastViTMA36F16.mlpackage) | 42.7M | 84 | Float16 | Float16 | -0.06% |
|
|
|
|
|
|
|
| 35 |
|
| 36 |
## Evaluation - Inference time
|
| 37 |
|
|
|
|
| 42 |
| MA36 | iPhone 12 Pro Max | 18.0 | 4.50 | Neural Engine |
|
| 43 |
| MA36 | M3 Max | 15.0 | 2.99 | Neural Engine |
|
| 44 |
|
| 45 |
+
## Download
|
| 46 |
+
|
| 47 |
+
Install `huggingface-cli`
|
| 48 |
+
|
| 49 |
+
```bash
|
| 50 |
+
brew install huggingface-cli
|
| 51 |
+
```
|
| 52 |
+
|
| 53 |
+
To download one of the `.mlpackage` folders to the `models` directory:
|
| 54 |
+
|
| 55 |
+
```bash
|
| 56 |
+
huggingface-cli download \
|
| 57 |
+
--local-dir models --local-dir-use-symlinks False \
|
| 58 |
+
apple/coreml-FastViT-T8
|
| 59 |
+
```
|
| 60 |
+
|
| 61 |
+
## Integrate in Swift apps
|
| 62 |
+
|
| 63 |
+
The [`huggingface/coreml-examples`](https://github.com/huggingface/coreml-examples/blob/main/depth-anything-example/README.md) repository contains sample Swift code for `coreml-FastViT-T8` and other models. See [the instructions there](https://github.com/huggingface/coreml-examples/tree/main/FastViTSample) to build the demo app, which shows how to use the model in your own Swift apps.
|
| 64 |
+
|
| 65 |
## Citation
|
| 66 |
+
|
| 67 |
```bibtex
|
| 68 |
@inproceedings{vasufastvit2023,
|
| 69 |
author = {Pavan Kumar Anasosalu Vasu and James Gabriel and Jeff Zhu and Oncel Tuzel and Anurag Ranjan},
|