RWKV 6: Fix error in ggml_cuda_op_bin_bcast

Signed-off-by: Molly Sophia <[email protected]>

Author: MollySophia

convert_hf_to_gguf.py

@@ -324,12 +324,14 @@ def prepare_tensors(self):
                             gguf.MODEL_TENSOR.TIME_MIX_FIRST,
                             gguf.MODEL_TENSOR.TIME_MIX_W1,
                             gguf.MODEL_TENSOR.TIME_MIX_W2,
+                            gguf.MODEL_TENSOR.TIME_MIX_DECAY,
                             gguf.MODEL_TENSOR.TIME_MIX_DECAY_W1,
                             gguf.MODEL_TENSOR.TIME_MIX_DECAY_W2,
                             gguf.MODEL_TENSOR.POSNET_NORM1,
                             gguf.MODEL_TENSOR.POSNET_NORM2,
                         )
                     )
+                    or "lerp" in new_name
                     or not new_name.endswith(".weight")
                 ):
                     data_qtype = gguf.GGMLQuantizationType.F32