summaryrefslogtreecommitdiff
path: root/tokenizer
diff options
context:
space:
mode:
authorJonas Knobloch <jonas.knobloch@t-online.de>2026-09-11 18:36:45 +0200
committerJonas Knobloch <jonas.knobloch@t-online.de>2026-09-11 18:36:45 +0200
commit6e6a94b12d6d43c1ce4b8e587742627af6d86a43 (patch)
tree8da5c2f1abc9d46d89942890ee75cccb11a6e04b /tokenizer
parent08400210cfdfd5769ce2d48db5234fbeae2e0c6e (diff)
WIP
Diffstat (limited to 'tokenizer')
-rw-r--r--tokenizer/bpe/utility.go4
1 files changed, 4 insertions, 0 deletions
diff --git a/tokenizer/bpe/utility.go b/tokenizer/bpe/utility.go
index f8fcdbc..79e6faf 100644
--- a/tokenizer/bpe/utility.go
+++ b/tokenizer/bpe/utility.go
@@ -19,6 +19,10 @@ func NewTokenizerFromFiles(vocab, merges string, cfg Config) (*Tokenizer, error)
return NewTokenizer(tokenizer, cfg), nil
}
+func MBPE(t *Tokenizer) *mbpe.Tokenizer {
+ return t.mbpe
+}
+
func Vocab(t *Tokenizer) []string {
m, ok := t.mbpe.Model().(*mbpe.MBPE)