btx-e001 T-0: flan-t5-small + tinybert tables, gated, coverage disclosed (newline is the mutual floor; case one-way lossy into wordpiece)
Browse files- t0_extraction_ledger.json +24 -0
t0_extraction_ledger.json
ADDED
|
@@ -0,0 +1,24 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| 1 |
+
{
|
| 2 |
+
"flan-t5-small": {
|
| 3 |
+
"family": "sentencepiece",
|
| 4 |
+
"n_tokens": 32100,
|
| 5 |
+
"n_special": 103,
|
| 6 |
+
"lossy": true,
|
| 7 |
+
"gate_mode": "normalized",
|
| 8 |
+
"path": "E:\\mirel\\data\\bytelex\\vocab_flan-t5-small.jsonl",
|
| 9 |
+
"coverage_raw": 0.991032,
|
| 10 |
+
"coverage_folded": 0.991032,
|
| 11 |
+
"codex_bytes": 256016
|
| 12 |
+
},
|
| 13 |
+
"tinybert-4l-312d": {
|
| 14 |
+
"family": "wordpiece",
|
| 15 |
+
"n_tokens": 30522,
|
| 16 |
+
"n_special": 999,
|
| 17 |
+
"lossy": true,
|
| 18 |
+
"gate_mode": "normalized",
|
| 19 |
+
"path": "E:\\mirel\\data\\bytelex\\vocab_tinybert-4l-312d.jsonl",
|
| 20 |
+
"coverage_raw": 0.916341,
|
| 21 |
+
"coverage_folded": 0.991032,
|
| 22 |
+
"codex_bytes": 256016
|
| 23 |
+
}
|
| 24 |
+
}
|