summaryrefslogtreecommitdiff
path: root/tokenizer/bpe/tokenizer.go
diff options
context:
space:
mode:
authorJonas Knobloch <jonas.knobloch@t-online.de>2026-03-24 13:02:49 +0100
committerJonas Knobloch <jonas.knobloch@t-online.de>2026-03-24 15:40:19 +0100
commitac670c698540488c818a17c3749abf3ed4cb764a (patch)
treebeaa4102bdbd24d2568431b0791dc6b780c1dd22 /tokenizer/bpe/tokenizer.go
parentfc9be8f81f673c0b04a0381a544b78e5a2a0d0bb (diff)
Add tokenizer module
* Wrap MBPE tokenizer
Diffstat (limited to 'tokenizer/bpe/tokenizer.go')
-rw-r--r--tokenizer/bpe/tokenizer.go25
1 files changed, 25 insertions, 0 deletions
diff --git a/tokenizer/bpe/tokenizer.go b/tokenizer/bpe/tokenizer.go
new file mode 100644
index 0000000..1b65de2
--- /dev/null
+++ b/tokenizer/bpe/tokenizer.go
@@ -0,0 +1,25 @@
+package bpe
+
+import "github.com/jonasknobloch/mbpe"
+
+type Tokenizer struct {
+ mbpe *mbpe.Tokenizer
+}
+
+func NewTokenizer(mbpe *mbpe.Tokenizer) *Tokenizer {
+ return &Tokenizer{
+ mbpe: mbpe,
+ }
+}
+
+func (t *Tokenizer) Encode(s string) []int {
+ return t.mbpe.Tokenize(s)
+}
+
+func (t *Tokenizer) Decode(ids []int) string {
+ panic("unimplemented") // TODO implement
+}
+
+func (t *Tokenizer) Tokenize(s string) []int {
+ return t.Encode(s)
+}