CS348Project/llama.cpp
5
0
Fork 0
mirror of https://github.com/ggml-org/llama.cpp.git synced 2025-11-06 09:46:50 +00:00
Code Issues Packages Projects Releases Wiki Activity

fix vocab size

Browse Source
This commit is contained in:
ibrahimkhadraoui
2025-07-04 16:25:27 +04:00
parent 1fd0574adc
commit 3ee7983961
1 changed files with 4 additions and 4 deletions
Download Patch File Download Diff File Expand all files Collapse all files

8
convert_hf_to_gguf.py
View File

@@ -607,10 +607,7 @@ class TextModel(ModelBase):
from transformers import AutoTokenizer
tokenizer = AutoTokenizer.from_pretrained(self.dir_model)
vocab_size = max(
self.hparams.get("vocab_size", len(tokenizer.vocab)),
len(tokenizer.vocab)
)
vocab_size = self.hparams.get("vocab_size", len(tokenizer.vocab))
assert max(tokenizer.vocab.values()) < vocab_size
tokpre = self.get_vocab_base_pre(tokenizer)
@@ -4885,6 +4882,9 @@ class Mamba2Model(TextModel):
pad_vocab = self.hparams.get("pad_vocab_size_multiple", 16)
# pad using ceiling division
# ref: https://stackoverflow.com/a/17511341/22827863
# if architecture is FalconH1, don't pad vocab size
if self.hparams.get("architectures", [None])[0] == "FalconH1ForCausalLM":
pad_vocab = 1
vocab_size = -(vocab_size // -pad_vocab) * pad_vocab
self.hparams["vocab_size"] = vocab_size