LLM inference in C/C++
C++ ggml
convert : accept "ExaoneMoeForCausalLM" arch spelling (#26660)
Csaba Kecskemetie700bfb37f43
diff --git a/conversion/__init__.py b/conversion/__init__.pyindex 06c2c50ad2..1f781a7903 100644--- a/conversion/__init__.py+++ b/conversion/__init__.py@@ -70,6 +70,7 @@ TEXT_MODEL_MAP: dict[str, str] = {"Exaone4ForCausalLM": "exaone","ExaoneForCausalLM": "exaone","ExaoneMoEForCausalLM": "exaone",+ "ExaoneMoeForCausalLM": "exaone","FalconForCausalLM": "falcon","FalconH1ForCausalLM": "falcon_h1","FalconMambaForCausalLM": "mamba",diff --git a/conversion/exaone.py b/conversion/exaone.pyindex bc4fb3f1b1..1cd2244dbc 100644--- a/conversion/exaone.py+++ b/conversion/exaone.py@@ -123,7 +123,9 @@ class Exaone4Model(TextModel):yield (self.format_tensor_name(gguf.MODEL_TENSOR.ROPE_FREQS), torch.tensor(rope_factors, dtype=torch.float32))[email protected]("ExaoneMoEForCausalLM")+# note: transformers >= 5.1 renamed the class to "ExaoneMoeForCausalLM" (lowercase 'e'),+# so accept both spellings - LG AI have updated the configs of already-released models[email protected]("ExaoneMoEForCausalLM", "ExaoneMoeForCausalLM")class ExaoneMoEModel(Exaone4Model):model_arch = gguf.MODEL_ARCH.EXAONE_MOE