Updated with correct pre-tokenizer

In preparation for full requantization with correct pre-tokenization and gpt2 tokenizer, see the following PRs:
[lama : improve BPE pre-processing + LLaMA 3 and Deepseek support](https://github.com/ggerganov/llama.cpp/pull/6920) - Sets correct pre-tokenizer, applied to new fp16.gguf
[lama3 custom regex split](https://github.com/ggerganov/llama.cpp/pull/6965) - Fixes gpt2 tokenization

gorilla-openfunctions-v2.fp16.gguf

@@ -1,3 +1,3 @@
 version https://git-lfs.github.com/spec/v1
-oid sha256:e082a0ef78b7fe5c27a1fc47eea4a521e6e99175743fe1b346648d7948165d6d
-size 13825630784
+oid sha256:83605e12a91965ee3deae19b9cbe96bf227baf4d3ac3d5d86acf505fb9ec1c41
+size 13825218848