diff options
| author | Jonas Knobloch <jonas.knobloch@t-online.de> | 2026-04-14 19:44:06 +0200 |
|---|---|---|
| committer | Jonas Knobloch <jonas.knobloch@t-online.de> | 2026-04-14 19:44:06 +0200 |
| commit | d9bed216c45aa1e0b817f4bccc7969bdac1adbff (patch) | |
| tree | 2d865fecae1e59d16626893a9a8f1404fff89f81 /tokenizer | |
| parent | 0c97f22069d77cf3cac92d68173e38f83b258a9a (diff) | |
Fix reachable merges validator
* Previous implementation ignored byte replacements
Diffstat (limited to 'tokenizer')
| -rw-r--r-- | tokenizer/bpe/validate.go | 33 |
1 files changed, 29 insertions, 4 deletions
diff --git a/tokenizer/bpe/validate.go b/tokenizer/bpe/validate.go index 84910cf..ffa8e37 100644 --- a/tokenizer/bpe/validate.go +++ b/tokenizer/bpe/validate.go @@ -108,14 +108,39 @@ func ByteCoverage(t *Tokenizer) bool { } func ReachableMerges(t *Tokenizer, merges [][2]string) []bool { + atoi := make(map[string]int) + + vocab := Vocab(t) + + for i, token := range vocab { + if _, ok := atoi[token]; ok { + continue + } + + atoi[token] = i + } + + reachable := make(map[string]struct{}) + + for _, a := range InitialAlphabet() { + if _, ok := atoi[string(a)]; ok { + reachable[string(a)] = struct{}{} + } + } + mask := make([]bool, len(merges)) for i, merge := range merges { - a := t.Tokenize(merge[0]) - b := t.Tokenize(merge[1]) - c := t.Tokenize(merge[0] + merge[1]) + _, a := reachable[merge[0]] + _, b := reachable[merge[1]] + + if !a || !b { + continue + } + + reachable[merge[0]+merge[1]] = struct{}{} - mask[i] = len(a) == 1 && len(b) == 1 && len(c) == 1 + mask[i] = true } return mask |
