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

Fix gemma model conversion

Browse source
This commit is contained in:
Galunid 2024-06-18 17:24:05 +02:00
parent bc6c457fa3
commit 07f4b706e6
1 changed files with 5 additions and 4 deletions
Download patch file Download diff file Expand all files Collapse all files

9
convert-hf-to-gguf.py
View file

@ -569,7 +569,7 @@ class Model:
special_vocab._set_special_token("unk", tokenizer.special_tokens["<|endoftext|>"])
special_vocab.add_to_gguf(self.gguf_writer)
def _set_vocab_sentencepiece(self):
def _set_vocab_sentencepiece(self, special_vocab=None):
from sentencepiece import SentencePieceProcessor
tokenizer_path = self.dir_model / 'tokenizer.model'
@ -637,7 +637,8 @@ class Model:
self.gguf_writer.add_token_scores(scores)
self.gguf_writer.add_token_types(toktypes)
special_vocab = gguf.SpecialVocab(self.dir_model, n_vocab=len(tokens))
if special_vocab is None:
special_vocab = gguf.SpecialVocab(self.dir_model, n_vocab=len(tokens))
special_vocab.add_to_gguf(self.gguf_writer)
def _set_vocab_llama_hf(self):
@ -2237,7 +2238,6 @@ class GemmaModel(Model):
model_arch = gguf.MODEL_ARCH.GEMMA
def set_vocab(self):
self._set_vocab_sentencepiece()
# TODO: these special tokens should be exported only for the CodeGemma family
special_vocab = gguf.SpecialVocab(self.dir_model, load_merges=False,
@ -2247,7 +2247,8 @@ class GemmaModel(Model):
special_vocab._set_special_token("middle", 68)
special_vocab._set_special_token("fsep", 70)
special_vocab._set_special_token("eot", 107)
special_vocab.add_to_gguf(self.gguf_writer)
self._set_vocab_sentencepiece(special_vocab)
def set_gguf_parameters(self):
hparams = self.hparams