llama : update vocab type descriptions to reflect actual meaning

This commit is contained in:
Jared Van Bortel 2024-03-27 17:04:53 -04:00
parent ebad773e9d
commit 80e9fc7c4d

View file

@ -60,9 +60,9 @@ extern "C" {
enum llama_vocab_type { enum llama_vocab_type {
LLAMA_VOCAB_TYPE_NONE = 0, // For models without vocab LLAMA_VOCAB_TYPE_NONE = 0, // For models without vocab
LLAMA_VOCAB_TYPE_SPM = 1, // SentencePiece LLAMA_VOCAB_TYPE_SPM = 1, // LLaMA tokenizer based on byte-level BPE with byte fallback
LLAMA_VOCAB_TYPE_BPE = 2, // Byte Pair Encoding LLAMA_VOCAB_TYPE_BPE = 2, // GPT-2 tokenizer based on byte-level BPE
LLAMA_VOCAB_TYPE_WPM = 3, // WordPiece LLAMA_VOCAB_TYPE_WPM = 3, // BERT tokenizer based on WordPiece
}; };
// note: these values should be synchronized with ggml_rope // note: these values should be synchronized with ggml_rope