Add Requirements section to model card
Browse files
README.md
CHANGED
|
@@ -54,6 +54,31 @@ cheap to self-host and fast enough for live, word-by-word typing.
|
|
| 54 |
| Languages | English, Tamil, Hindi, Telugu, Malayalam, Kannada |
|
| 55 |
| Dataset | [AI4Bharat Aksharantar](https://huggingface.co/datasets/ai4bharat/Aksharantar) |
|
| 56 |
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| 57 |
## Quickstart
|
| 58 |
|
| 59 |
### 1. Install
|
|
|
|
| 54 |
| Languages | English, Tamil, Hindi, Telugu, Malayalam, Kannada |
|
| 55 |
| Dataset | [AI4Bharat Aksharantar](https://huggingface.co/datasets/ai4bharat/Aksharantar) |
|
| 56 |
|
| 57 |
+
## Requirements
|
| 58 |
+
|
| 59 |
+
- **Python 3.10+**
|
| 60 |
+
- **ONNX inference (CPU only):** `huggingface-hub`, `numpy`, `onnxruntime`,
|
| 61 |
+
`tokenizers`
|
| 62 |
+
- **Web app:** `gradio`
|
| 63 |
+
- **Training / PyTorch inference:** `torch` (CUDA recommended) and
|
| 64 |
+
`liger-kernel`; training additionally needs `datasets`
|
| 65 |
+
|
| 66 |
+
Versions used by the Space (`requirements.txt`):
|
| 67 |
+
|
| 68 |
+
```text
|
| 69 |
+
gradio==6.32.0
|
| 70 |
+
huggingface-hub>=1.16.0,<3.0.0
|
| 71 |
+
numpy>=1.24,<3.0
|
| 72 |
+
onnxruntime>=1.17.0
|
| 73 |
+
tokenizers>=0.15.0
|
| 74 |
+
```
|
| 75 |
+
|
| 76 |
+
Minimal install for ONNX-only inference:
|
| 77 |
+
|
| 78 |
+
```bash
|
| 79 |
+
pip install huggingface-hub numpy onnxruntime tokenizers
|
| 80 |
+
```
|
| 81 |
+
|
| 82 |
## Quickstart
|
| 83 |
|
| 84 |
### 1. Install
|