fix: correct the single-token coverage claim to 98.74%

- 392 of 397 distinct characters over all 29,600 Mon validation lines
- record lines_measured, so the sampling is auditable

The card said 100% with sampling labelled "whole split, no cap", but the
figure came from the first 5,000 lines. tokenizer.json is unchanged and
still hashes to 34d181532eee7e6754bfbad693753ee4660bdda00eb36a2ccc7cd2f372c71282,
byte-identical to the packaged copy; only the card was re-measured.

README.md

@@ -60,7 +60,11 @@ Vocabulary 64,256 · Unigram · trained on the train split of an 893,936-line /
 | English | 4.112 | 34.6 | — *(n=0)* | **100%** | 0.02% |
 | mixed script | 3.804 | 24.6 | 0.81% *(n=28,133)* | **100%** | 0.19% |
 
-**100% of the characters in the Mon validation split are single tokens.**
+**98.74% of the distinct characters in the Mon validation split are single
+tokens** — 392 of 397, over all 29,600 lines. The remaining five (a combining
+diaeresis below, a Greek capital pi, two Sinhala letters and one emoji; seven
+occurrences in 2.28M characters) round-trip through byte fallback at one to four
+tokens each.
 
 Three notes on reading these:
 

model_card.json

@@ -81,8 +81,9 @@
       "unreconstructable": 11
     },
     "coverage": {
-      "distinct": 199,
-      "single_token": 1.0
+      "distinct": 397,
+      "single_token": 0.9874,
+      "lines_measured": 29600
     }
   },
   "notes": {
@@ -93,6 +94,7 @@
   },
   "eval": {
     "split": "val",
-    "sampling": "whole split, no cap"
+    "sampling": "whole split, no cap",
+    "coverage_lines_measured": 29600
   }
 }