From de97c33b402f27fe5f757c3b89aeb302deee4f25 Mon Sep 17 00:00:00 2001 From: firecoperana Date: Thu, 11 Sep 2025 00:59:24 -0500 Subject: [PATCH] fix convert error for ernie 4.5 (#774) --- convert_hf_to_gguf.py | 10 +++++----- 1 file changed, 5 insertions(+), 5 deletions(-) diff --git a/convert_hf_to_gguf.py b/convert_hf_to_gguf.py index 7b2aa7e6..ba8b4a80 100644 --- a/convert_hf_to_gguf.py +++ b/convert_hf_to_gguf.py @@ -320,7 +320,7 @@ class Model: gguf.LlamaFileType.MOSTLY_Q4_0, gguf.LlamaFileType.MOSTLY_Q4_1, ): - data_qtype = gguf.GGMLQuantizationType.Q5_0 + data_qtype = gguf.GGMLQuantizationType.Q5_0 elif self.ftype in ( gguf.LlamaFileType.MOSTLY_Q5_0, gguf.LlamaFileType.MOSTLY_Q5_1, @@ -418,7 +418,7 @@ class Model: logger.info("Set model quantization version") self.gguf_writer.add_quantization_version(gguf.GGML_QUANT_VERSION) - + logger.info("***********************************************************************************************") logger.info("** Converting to `q4_0`,`q4_1`,`q5_0`, `q5_1` or `q6_0` is not equiv to using `llama-quantize`!") logger.info("** Ftype `q4_0`,`q4_1` are here converting embeddings, output, attn_k and attn_v/qkv in q5_0.") @@ -2195,8 +2195,8 @@ class Qwen3MoeModel(Qwen2MoeModel): model_arch = gguf.MODEL_ARCH.QWEN3MOE -@ModelBase.register("Ernie4_5_ForCausalLM", "Ernie4_5ForCausalLM") -class Ernie4_5Model(TextModel): +@Model.register("Ernie4_5_ForCausalLM", "Ernie4_5ForCausalLM") +class Ernie4_5Model(Model): model_arch = gguf.MODEL_ARCH.ERNIE4_5 def set_vocab(self): @@ -2242,7 +2242,7 @@ class Ernie4_5Model(TextModel): return [(self.map_tensor_name(name), data_torch)] -@ModelBase.register("Ernie4_5_MoeForCausalLM") +@Model.register("Ernie4_5_MoeForCausalLM") class Ernie4_5MoeModel(Ernie4_5Model): model_arch = gguf.MODEL_ARCH.ERNIE4_5_MOE _experts: list[dict[str, Tensor]] | None = None