From 280dd2dcb7434ad084578b60457854fd768ebf5f Mon Sep 17 00:00:00 2001 From: ibrahimkhadraoui Date: Mon, 7 Jul 2025 10:25:57 +0400 Subject: [PATCH] falcon-h1 specefic vocab resolved --- convert_hf_to_gguf.py | 6 +++--- 1 file changed, 3 insertions(+), 3 deletions(-) diff --git a/convert_hf_to_gguf.py b/convert_hf_to_gguf.py index 17feb9430c..96cc8d28c7 100755 --- a/convert_hf_to_gguf.py +++ b/convert_hf_to_gguf.py @@ -4882,9 +4882,6 @@ 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 @@ -6590,6 +6587,9 @@ class FalconH1Model(Mamba2Model): keys = list(keys) + prefixed return super().find_hparam(keys, *args, **kwargs) + def set_vocab(self): + self._set_vocab_gpt2() + def _generate_mup_vector(self, block_id: int) -> torch.Tensor: zxbcdt_multipliers = self.hparams["ssm_multipliers"] intermediate_size = self.hparams["mamba_d_ssm"]