DeepSpeed
DeepSpeed copied to clipboard
Fix Bug in transform.cu
This is to fix the below error
/opt/conda/lib/python3.8/site-packages/deepspeed/ops/csrc/transformer/inference/csrc/transform.hip:688:6: error: function template partial specialization is not allowed
void launch_transform4d_0213<T>(T* out,
^ ~~~
cc: @jithunnair-amd
@rraminen - curious what you ran that you came across this?
@rraminen - curious what you ran that you came across this?
I was running the Deepspeed-Inference workloads from https://github.com/huggingface/transformers-bloom-inference.git
This error also comes up when building DeepSpeed with 'deepspeed.ops.transformer.inference.transformer_inference_op'
extension in non-JIT way