LLM inference in C/C++
C++ ggml
convert : respect Mistral format for DeepseekV2Model (#30669)
aiqubits2a667032dc78
diff --git a/conversion/deepseek.py b/conversion/deepseek.pyindex 817eb76128..946de53476 100644--- a/conversion/deepseek.py+++ b/conversion/deepseek.py@@ -243,7 +243,7 @@ class DeepseekV2Model(TextModel):def __init__(self, *args, **kwargs):super().__init__(*args, **kwargs)- hparams: dict = ModelBase.load_hparams(self.dir_model, is_mistral_format=False)+ hparams: dict = ModelBase.load_hparams(self.dir_model, is_mistral_format=self.is_mistral_format)self.origin_hf_arch = hparams.get('architectures', [None])[0]# special handling for Deepseek OCR