Add tokenizer config
Browse files- tokenizer_config.json +28 -0
tokenizer_config.json
ADDED
|
@@ -0,0 +1,28 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| 1 |
+
{
|
| 2 |
+
"tokenizer_class": "PreTrainedTokenizerFast",
|
| 3 |
+
"vocab_size": 65536,
|
| 4 |
+
"model_type": "glaurung-binary-tokenizer",
|
| 5 |
+
"version": "001",
|
| 6 |
+
"encoding": "latin-1",
|
| 7 |
+
"description": "BPE tokenizer for binary executables and malware (x86-64, ARM64, Windows PE, Linux ELF)",
|
| 8 |
+
"compression_ratio": 2.849,
|
| 9 |
+
"training_data": {
|
| 10 |
+
"size_gb": 13,
|
| 11 |
+
"files": 30738,
|
| 12 |
+
"platforms": [
|
| 13 |
+
"Linux (Alpine, Debian, Ubuntu)",
|
| 14 |
+
"Windows (8, 10, 11)"
|
| 15 |
+
],
|
| 16 |
+
"architectures": [
|
| 17 |
+
"x86-64",
|
| 18 |
+
"x86-32",
|
| 19 |
+
"ARM64"
|
| 20 |
+
]
|
| 21 |
+
},
|
| 22 |
+
"performance": {
|
| 23 |
+
"bytes_per_token": 2.849,
|
| 24 |
+
"theoretical_efficiency": 0.86,
|
| 25 |
+
"improvement_over_32k": 0.099
|
| 26 |
+
},
|
| 27 |
+
"predecessor": "mjbommar/binary-tokenizer-005"
|
| 28 |
+
}
|