vbatts/llama.cpp
1
0
Fork 0
Code Issues Pull requests Projects Releases Packages Wiki Activity Actions

Add special token type

Browse source
This commit is contained in:
wonjun Jang 2023-11-05 23:18:26 +09:00 • committed by GitHub
parent 4adb8b9862
commit 13f07013ee
No known key found for this signature in database
GPG key ID: 4AEE18F83AFDEB23
1 changed files with 4 additions and 3 deletions
Download patch file Download diff file Expand all files Collapse all files

7
convert.py
View file

@ -311,17 +311,18 @@ class VocabLoader:
vocab_set = {encoded_tok for encoded_tok, id in self.tokenizer.vocab.items()} vocab_set = {encoded_tok for encoded_tok, id in self.tokenizer.vocab.items()}
self.added_tokens_list = [] self.added_tokens_list = []
self.vocab_size_base: int = len(self.tokenizer.vocab) self.vocab_size_base: int = len(vocab_set)
self.vocab_size: int = len(self.tokenizer.vocab) self.vocab_size: int = len(vocab_set)
self.fname_tokenizer = fname_tokenizer self.fname_tokenizer = fname_tokenizer
def hf_tokens(self) -> Iterable[tuple[bytes, float, gguf.TokenType]]: def hf_tokens(self) -> Iterable[tuple[bytes, float, gguf.TokenType]]:
tokenizer = self.tokenizer tokenizer = self.tokenizer
reverse_vocab = {id: encoded_tok for encoded_tok, id in tokenizer.vocab.items()} reverse_vocab = {id: encoded_tok for encoded_tok, id in tokenizer.vocab.items()}
special_ids = set(tokenizer.all_special_ids)
for i in range(self.vocab_size_base): for i in range(self.vocab_size_base):
text = reverse_vocab[i].encode("utf-8") text = reverse_vocab[i].encode("utf-8")
yield text, 0.0, gguf.TokenType.NORMAL yield text, 0.0, gguf.TokenType.NORMAL if i not in special_ids else gguf.TokenType.CONTROL
def added_tokens(self) -> Iterable[tuple[bytes, float, gguf.TokenType]]: def added_tokens(self) -> Iterable[tuple[bytes, float, gguf.TokenType]]:
for text in self.added_tokens_list: for text in self.added_tokens_list: