root/llama.cpp
1
0
Fork 0
mirror of https://github.com/ggerganov/llama.cpp.git synced 2024-11-14 06:49:54 +00:00
Code Issues Actions 14 Packages Projects Releases Wiki Activity

convert-hf : fix exception in sentencepiece with added tokens (#6320)

Browse Source
This commit is contained in:
Pedro Cuenca 2024-03-26 13:32:19 +01:00 committed by GitHub
parent d25b1c31b0
commit e097633f63
No known key found for this signature in database
GPG Key ID: B5690EEEBB952194
1 changed files with 8 additions and 4 deletions
Show Stats Download Patch File Download Diff File Expand all files Collapse all files

8
convert-hf-to-gguf.py
View File

@ -331,7 +331,7 @@ class Model(ABC):
tokenizer = SentencePieceProcessor(str(tokenizer_path)) tokenizer = SentencePieceProcessor(str(tokenizer_path))
vocab_size = self.hparams.get('vocab_size', tokenizer.vocab_size()) vocab_size = self.hparams.get('vocab_size', tokenizer.vocab_size())
for token_id in range(vocab_size): for token_id in range(tokenizer.vocab_size()):
piece = tokenizer.id_to_piece(token_id) piece = tokenizer.id_to_piece(token_id)
text = piece.encode("utf-8") text = piece.encode("utf-8")
score = tokenizer.get_score(token_id) score = tokenizer.get_score(token_id)
@ -356,10 +356,14 @@ class Model(ABC):
added_tokens_json = json.load(f) added_tokens_json = json.load(f)
for key in added_tokens_json: for key in added_tokens_json:
tokens.append(key.encode("utf-8")) key = key.encode("utf-8")
if key not in tokens:
tokens.append(key)
scores.append(-1000.0) scores.append(-1000.0)
toktypes.append(SentencePieceTokenTypes.USER_DEFINED) toktypes.append(SentencePieceTokenTypes.USER_DEFINED)
assert len(tokens) == vocab_size
self.gguf_writer.add_tokenizer_model("llama") self.gguf_writer.add_tokenizer_model("llama")
self.gguf_writer.add_token_list(tokens) self.gguf_writer.add_token_list(tokens)
self.gguf_writer.add_token_scores(scores) self.gguf_writer.add_token_scores(scores)