Fix errors when using smoothquant to quantize Qwen2 model

[Missmiaom]

When quantizing the Qwen2 model with SmoothQuant, the mlp.proj was not correctly split; it should be split along the 0th dimension.

command:

python3 convert_checkpoint.py --model_dir ./tmp/Qwen/7B/ --output_dir ./tllm_checkpoint_1gpu_sq --dtype float16 --smoothquant 0.5
trtllm-build --checkpoint_dir ./tllm_checkpoint_1gpu_sq \
             --output_dir ./engine_outputs \
             --gemm_plugin float16

[jershi425]

Hi @Missmiaom, thank you for the contribution. We will include this fix in our internal repo and it will be reflected in the next release.