clip skip prompt token doesn't work with fp16 models

[ssube]

[2023-04-01 22:21:34,745] ERROR: 329721 139777349949248 onnx_web.worker.worker: error while running job          
Traceback (most recent call last):                                                                                                                                                               
  File "/opt/onnx-web/api/onnx_web/worker/worker.py", line 60, in worker_main                                                                                                                    
    job.fn(context, *job.args, **job.kwargs)                                                                                                                                                     
  File "/opt/onnx-web/api/onnx_web/diffusers/run.py", line 78, in run_txt2img_pipeline                                                                                                           
    result = pipe(                                                                                                                                                                               
  File "/home/ssube/miniconda3/envs/onnx-web-rocm/lib/python3.9/site-packages/diffusers/pipelines/stable_diffusion/pipeline_onnx_stable_diffusion.py", line 234, in __call__
    prompt_embeds = self._encode_prompt(
  File "/home/ssube/miniconda3/envs/onnx-web-rocm/lib/python3.9/site-packages/torch/autograd/grad_mode.py", line 27, in decorate_context
    return func(*args, **kwargs)
  File "/opt/onnx-web/api/onnx_web/diffusers/utils.py", line 88, in expand_prompt
    norm_state = layer_norm(
  File "/home/ssube/miniconda3/envs/onnx-web-rocm/lib/python3.9/site-packages/torch/nn/modules/module.py", line 1194, in _call_impl
    return forward_call(*input, **kwargs)
  File "/home/ssube/miniconda3/envs/onnx-web-rocm/lib/python3.9/site-packages/torch/nn/modules/normalization.py", line 190, in forward
    return F.layer_norm(
  File "/home/ssube/miniconda3/envs/onnx-web-rocm/lib/python3.9/site-packages/torch/nn/functional.py", line 2515, in layer_norm
    return torch.layer_norm(input, normalized_shape, weight, bias, eps, torch.backends.cudnn.enabled)
RuntimeError: "LayerNormKernelImpl" not implemented for 'Half'

This is because of the call to layer_norm = torch.nn.LayerNorm(last_state.shape[2]) on the CPU without any type conversion (CPU ops need to be fp32).