From 4fbd8098e63670c6ae11a8adc350f5ba191cfda3 Mon Sep 17 00:00:00 2001 From: Daniel Bevenius Date: Tue, 16 Apr 2024 08:13:13 +0200 Subject: gguf : add special tokens metadata for FIM/Infill (#6689) This commit adds special token metadata for Fill-In-the-Middle (FIM)/Infill to the GGUF model. The motivation for this is that currently there is support for CodeLlama but other models exist now like CodeGemma, but the different models use different token ids for the special tokens and this commit allows for supporting multiple models. Signed-off-by: Daniel Bevenius --- convert-hf-to-gguf.py | 15 +++++++++++++++ 1 file changed, 15 insertions(+) (limited to 'convert-hf-to-gguf.py') diff --git a/convert-hf-to-gguf.py b/convert-hf-to-gguf.py index b51d6830..6d28ab5e 100755 --- a/convert-hf-to-gguf.py +++ b/convert-hf-to-gguf.py @@ -1221,6 +1221,14 @@ class LlamaModel(Model): except FileNotFoundError: self._set_vocab_llama_hf() + special_vocab = gguf.SpecialVocab(self.dir_model, load_merges=False, + special_token_types = ['prefix', 'suffix', 'middle', 'eot']) + special_vocab._set_special_token("prefix", 32007) + special_vocab._set_special_token("suffix", 32008) + special_vocab._set_special_token("middle", 32009) + special_vocab._set_special_token("eot", 32010) + special_vocab.add_to_gguf(self.gguf_writer) + def set_gguf_parameters(self): super().set_gguf_parameters() hparams = self.hparams @@ -2240,6 +2248,13 @@ class GemmaModel(Model): def set_vocab(self): self._set_vocab_sentencepiece() + special_vocab = gguf.SpecialVocab(self.dir_model, load_merges=False, + special_token_types = ['prefix', 'suffix', 'middle', 'eot']) + special_vocab._set_special_token("prefix", 67) + special_vocab._set_special_token("suffix", 69) + special_vocab._set_special_token("middle", 68) + special_vocab._set_special_token("eot", 70) + special_vocab.add_to_gguf(self.gguf_writer) def set_gguf_parameters(self): hparams = self.hparams -- cgit v1.2.3