Upload README.md with huggingface_hub
Browse files
README.md
ADDED
|
@@ -0,0 +1,52 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| 1 |
+
\---
|
| 2 |
+
|
| 3 |
+
license: apache-2.0
|
| 4 |
+
|
| 5 |
+
language:
|
| 6 |
+
|
| 7 |
+
  - ur
|
| 8 |
+
|
| 9 |
+
  - pa
|
| 10 |
+
|
| 11 |
+
  - sd
|
| 12 |
+
|
| 13 |
+
  - ps
|
| 14 |
+
|
| 15 |
+
  - fa
|
| 16 |
+
|
| 17 |
+
  - ku
|
| 18 |
+
|
| 19 |
+
tags:
|
| 20 |
+
|
| 21 |
+
  - tokenizer
|
| 22 |
+
|
| 23 |
+
  - multilingual
|
| 24 |
+
|
| 25 |
+
  - pakistan
|
| 26 |
+
|
| 27 |
+
  - ir
|
| 28 |
+
|
| 29 |
+
\---
|
| 30 |
+
|
| 31 |
+
|
| 32 |
+
|
| 33 |
+
\# Strategic Multilingual Tokenizer
|
| 34 |
+
|
| 35 |
+
|
| 36 |
+
|
| 37 |
+
Optimized for Pakistan, Iran, and South Asian languages. Provides up to 3X better compression on Urdu, Punjabi, Sindhi, and Persian compared to Mistral and Qwen2.
|
| 38 |
+
|
| 39 |
+
|
| 40 |
+
|
| 41 |
+
\## Usage
|
| 42 |
+
|
| 43 |
+
|
| 44 |
+
|
| 45 |
+
```python
|
| 46 |
+
|
| 47 |
+
from tokenizers import Tokenizer
|
| 48 |
+
|
| 49 |
+
tok = Tokenizer.from\_file("strategic\_tokenizer\_gpt4.json")
|
| 50 |
+
|
| 51 |
+
enc = tok.encode("مصنوعی ذہانت دنیا کو بدل رہی ہے")
|
| 52 |
+
|