Bug description

Other marian-decoder functions work, but using --force-decode results in this error

[2024-02-15 18:55:11] Error: CopyCastTo to type uint32 not implemented [2024-02-15 18:55:11] Error: Aborted from void marian::cpu::CopyCastFrom(marian::Tensor, const T*, int) [with bool add = false; T = float; marian::Tensor = IntrusivePtr<marian::TensorBase>] in /marian/src/tensors/cpu/tensor_operators.cpp:59 

How to reproduce

Model: https://github.com/Helsinki-NLP/Tatoeba-Challenge/tree/master/models/kor-eng
Using a CPU container with Marian 1.12, no CUDA, and Open BLAS
Prefixes.txt file and src.txt file are same amount of lines
/marian/build/marian-decoder -c decoder.yml --force-decode --beam-size 10 --cpu-threads=16 --n-best -o output.txt
log.txt

Context

• Marian version: Paste the output of --version here
  1.12

• CMake command: Type the cmake command you used and attach the output of --build-info all
  cmake -DUSE_SENTENCEPIECE=on -DCOMPILE_CUDA=OFF -DBUILD_ARCH=x86-64 -DUSE_OPENBLAS=on -DCOMPILE_CPU=on -DPROTOBUF_LIBRARY=/usr/lib/x86_64-linux-gnu/libprotobuf.so.9 -DPROTOBUF_INCLUDE_DIR=/usr/include/google/protobuf -DPROTOBUF_PROTOC_EXECUTABLE=/usr/bin/protoc -DUSE_MKL=off && make -j4

• Log file: Attach your training/decoding logs

Add any other information about the problem here.