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

Update convert-hf-to-gguf.py

Browse source
This commit is contained in:
0xspringtime 2024-06-19 14:53:32 -04:00 • committed by GitHub
parent 620633e9d8
commit a05faca0d0
No known key found for this signature in database
GPG key ID: B5690EEEBB952194
1 changed files with 5 additions and 1 deletions
Download patch file Download diff file Expand all files Collapse all files

6
convert-hf-to-gguf.py
View file

@ -967,10 +967,14 @@ class XverseModel(Model):
from transformers import AutoTokenizer from transformers import AutoTokenizer
tokenizer = AutoTokenizer.from_pretrained(dir_model) tokenizer = AutoTokenizer.from_pretrained(dir_model)
vocab_size = hparams.get("vocab_size", len(tokenizer.vocab)) vocab_size = hparams.get("vocab_size", len(tokenizer.vocab))
if max(tokenizer.get_vocab().values()) >= vocab_size: # Since we are checking the maximum index, we need to ensure it's strictly less than vocab_size,
# because vocab_size is the count of items, and indexes start at 0.
max_vocab_index = max(tokenizer.get_vocab().values())
if max_vocab_index >= vocab_size:
raise ValueError("Vocabulary size exceeds expected maximum size.") raise ValueError("Vocabulary size exceeds expected maximum size.")
reverse_vocab: dict[int, str] = {id_: encoded_tok for encoded_tok, id_ in tokenizer.vocab.items()} reverse_vocab: dict[int, str] = {id_: encoded_tok for encoded_tok, id_ in tokenizer.vocab.items()}
added_vocab = tokenizer.get_added_vocab() added_vocab = tokenizer.get_added_vocab()