summaryrefslogtreecommitdiff
path: root/llmc/cmd/data/fineweb.go
diff options
context:
space:
mode:
authorJonas Knobloch <jonas.knobloch@t-online.de>2026-04-24 11:49:26 +0200
committerJonas Knobloch <jonas.knobloch@t-online.de>2026-04-24 11:50:02 +0200
commit12f15fdd8c048e5958e294f12a9f163f6f950b3e (patch)
tree16bb23c134f1fe476d977dd846e6dda6c7667c48 /llmc/cmd/data/fineweb.go
parent408536516bdfedc93d93226dc08280ed224361fc (diff)
Recover tokenization based on config
Diffstat (limited to 'llmc/cmd/data/fineweb.go')
-rw-r--r--llmc/cmd/data/fineweb.go5
1 files changed, 4 insertions, 1 deletions
diff --git a/llmc/cmd/data/fineweb.go b/llmc/cmd/data/fineweb.go
index 3d3d51b..a1bf7df 100644
--- a/llmc/cmd/data/fineweb.go
+++ b/llmc/cmd/data/fineweb.go
@@ -26,7 +26,10 @@ func fineWeb() {
var tokenizer llm.Tokenizer
- if t, err := bpe.NewTokenizerFromFiles(shelf.Abs("models/gpt2/vocab.json"), shelf.Abs("models/gpt2/merges.txt")); err != nil {
+ v := shelf.Abs("models/gpt2/vocab.json")
+ m := shelf.Abs("models/gpt2/merges.txt")
+
+ if t, err := bpe.NewTokenizerFromFiles(v, m, bpe.DefaultConfig()); err != nil {
log.Fatal(err)
} else {
tokenizer = t