Merge branch 'master' into compilade/mamba2

ggml-org · compilade · Jul 2, 2025 · Aug 1, 2024 · Aug 19, 2024 · Aug 19, 2024
commit 35d06fac5af8f85903d6ffe14c53c16aad90dc73
diff --git a/convert_hf_to_gguf.py b/convert_hf_to_gguf.py
@@ -4175,7 +4175,6 @@ def set_gguf_parameters(self):
     _tok_embd = None
 
     def modify_tensors(self, data_torch: Tensor, name: str, bid: int | None) -> Iterable[tuple[str, Tensor]]:
-
         output_name = self.format_tensor_name(gguf.MODEL_TENSOR.OUTPUT)
         tok_embd_name = self.format_tensor_name(gguf.MODEL_TENSOR.TOKEN_EMBD)
 
@@ -4185,6 +4184,7 @@ def modify_tensors(self, data_torch: Tensor, name: str, bid: int | None) -> Iter
             logger.debug("A_log --> A ==> " + new_name)
             data_torch = -torch.exp(data_torch)
 
+        # [4 1 8192 1] -> [4 8192 1 1]
         if self.match_model_tensor_name(new_name, gguf.MODEL_TENSOR.SSM_CONV1D, bid):
             data_torch = data_torch.squeeze()
 
@@ -4199,8 +4199,8 @@ def modify_tensors(self, data_torch: Tensor, name: str, bid: int | None) -> Iter
         return [(new_name, data_torch)]
 
 
-@Model.register("Mamba2ForCausalLM")
-class Mamba2Model(Model):
+@ModelBase.register("Mamba2ForCausalLM")
+class Mamba2Model(TextModel):
     model_arch = gguf.MODEL_ARCH.MAMBA2
 
     def set_vocab(self):
@@ -4284,8 +4284,8 @@ def modify_tensors(self, data_torch: Tensor, name: str, bid: int | None) -> Iter
         yield (new_name, data_torch)
 
 
-@Model.register("CohereForCausalLM")
-class CommandR2Model(Model):
+@ModelBase.register("CohereForCausalLM")
+class CommandR2Model(TextModel):
     model_arch = gguf.MODEL_ARCH.COMMAND_R
 
     def __init__(self, *args, **kwargs):