Update README.md
Browse files
README.md
CHANGED
|
@@ -1,3 +1,22 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| 1 |
# Crayon 🖍️
|
| 2 |
Crayon is a high-performance, hardware-accelerated tokenizer engineered for instant vocabulary swapping and maximum throughput.
|
| 3 |
Designed to eliminate the bottleneck of data preprocessing in LLM pipelines, Crayon operates using a unique **cartridge system**—pre-built vocabulary profiles that can be loaded and swapped instantly. This allows developers to seamlessly switch between 50k (lite) and 206k (standard) vocabularies without rebuilding the tokenizer state.
|
|
|
|
| 1 |
+
---
|
| 2 |
+
language:
|
| 3 |
+
- en
|
| 4 |
+
license: mit
|
| 5 |
+
tags:
|
| 6 |
+
- tokenizers
|
| 7 |
+
- text-generation
|
| 8 |
+
- hardware-accelerated
|
| 9 |
+
- avx2
|
| 10 |
+
- cuda
|
| 11 |
+
- rocm
|
| 12 |
+
- double-array-trie
|
| 13 |
+
- crayon
|
| 14 |
+
metrics:
|
| 15 |
+
- throughput
|
| 16 |
+
pipeline_tag: text-generation
|
| 17 |
+
library_name: generic
|
| 18 |
+
---
|
| 19 |
+
|
| 20 |
# Crayon 🖍️
|
| 21 |
Crayon is a high-performance, hardware-accelerated tokenizer engineered for instant vocabulary swapping and maximum throughput.
|
| 22 |
Designed to eliminate the bottleneck of data preprocessing in LLM pipelines, Crayon operates using a unique **cartridge system**—pre-built vocabulary profiles that can be loaded and swapped instantly. This allows developers to seamlessly switch between 50k (lite) and 206k (standard) vocabularies without rebuilding the tokenizer state.
|