Update README.md
Browse files
README.md
CHANGED
|
@@ -1,7 +1,51 @@
|
|
| 1 |
---
|
| 2 |
library_name: transformers
|
| 3 |
-
|
| 4 |
---
|
| 5 |
|
| 6 |
|
| 7 |
-
**WE are COOKED**
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| 1 |
---
|
| 2 |
library_name: transformers
|
| 3 |
+
license: apache-2.0
|
| 4 |
---
|
| 5 |
|
| 6 |
|
| 7 |
+
**WE are COOKED**
|
| 8 |
+
|
| 9 |
+
# Test Log 08 March 2025
|
| 10 |
+
|
| 11 |
+
### First Test:
|
| 12 |
+
Mean Perplexity : tested on `wikitext-2-raw-v1`, ~2k English samples was `1420.7414870547489`
|
| 13 |
+
|
| 14 |
+
### Second Test
|
| 15 |
+
Evaluated the tokenizer's performance on:
|
| 16 |
+
- Unicode coverage.
|
| 17 |
+
- Token distribution.
|
| 18 |
+
- Tokenization complexity across different scripts.
|
| 19 |
+
- Encoding and decoding capabilities &
|
| 20 |
+
- Edge cases e.g., special characters, numbers, etc.
|
| 21 |
+
- 1k samples: 500 Hindi, 500 English
|
| 22 |
+
|
| 23 |
+
### 1. Edge Case Handling
|
| 24 |
+
|
| 25 |
+
| **Language** | **Test Type** | **Token Count** | **Unique Tokens** |
|
| 26 |
+
|--------------|--------------------|-----------------|-------------------|
|
| 27 |
+
| **Hindi** | Script Test | 14 | 13 |
|
| 28 |
+
| | Unicode Test | 21 | 21 |
|
| 29 |
+
| | Special Characters | 19 | 19 |
|
| 30 |
+
| **English** | Script Test | 16 | 15 |
|
| 31 |
+
| | Unicode Test | 14 | 14 |
|
| 32 |
+
| | Special Characters | 18 | 18 |
|
| 33 |
+
|
| 34 |
+
### 2. Unicode Coverage
|
| 35 |
+
|
| 36 |
+
| **Language** | **Coverage Ratio** | **Token Count** | **Unique Tokens** |
|
| 37 |
+
|--------------|--------------------|-----------------|-------------------|
|
| 38 |
+
| **Hindi** | 100% | 21 | 21 |
|
| 39 |
+
| **English** | 100% | 14 | 14 |
|
| 40 |
+
|
| 41 |
+
### 3. Complexity
|
| 42 |
+
|
| 43 |
+
| **Language** | **Original Length** | **Token Count** | **Avg Token Length** | **Token Diversity** |
|
| 44 |
+
|--------------|---------------------|-----------------|----------------------|---------------------|
|
| 45 |
+
| **Hindi** | 49 | 14 | 9.07 | 0.928 |
|
| 46 |
+
| **English** | 65 | 16 | 4.06 | 0.937 |
|
| 47 |
+
|
| 48 |
+
|
| 49 |
+
|
| 50 |
+
|
| 51 |
+
|