summaryrefslogtreecommitdiff
path: root/tokenizer/bpe/utility.go
diff options
context:
space:
mode:
authorJonas Knobloch <jonas.knobloch@t-online.de>2026-04-11 01:09:16 +0200
committerJonas Knobloch <jonas.knobloch@t-online.de>2026-04-14 19:38:34 +0200
commit0c97f22069d77cf3cac92d68173e38f83b258a9a (patch)
tree127bf00ad1f69ce1368089247471fd52b8f9d305 /tokenizer/bpe/utility.go
parentaa1fdb5f3b5d6c54c3bc44269482c31f4dfafe67 (diff)
Add lesci module
Diffstat (limited to 'tokenizer/bpe/utility.go')
-rw-r--r--tokenizer/bpe/utility.go20
1 files changed, 20 insertions, 0 deletions
diff --git a/tokenizer/bpe/utility.go b/tokenizer/bpe/utility.go
index f47c073..fbe99fc 100644
--- a/tokenizer/bpe/utility.go
+++ b/tokenizer/bpe/utility.go
@@ -18,3 +18,23 @@ func NewTokenizerFromFiles(vocab, merges string) (*Tokenizer, error) {
return NewTokenizer(tokenizer), nil
}
+
+func Vocab(t *Tokenizer) []string {
+ m, ok := t.mbpe.Model().(*mbpe.MBPE)
+
+ if !ok {
+ panic("unimplemented")
+ }
+
+ return m.Vocab()
+}
+
+func Merges(t *Tokenizer) [][2]string {
+ m, ok := t.mbpe.Model().(*mbpe.MBPE)
+
+ if !ok {
+ panic("unimplemented")
+ }
+
+ return m.Merges()
+}