Ilya Dimov comments

Results 5 comments of


                                            Ilya Dimov

Multi-instance inference fails in (n-1)/n runs (where n is a number gpus/instances)

Hi @PerkzZheng ! I'm also facing the same issue and my setup is similiar to @timofeev1995 with several 3090. My topo is ``` GPU0 GPU1 GPU2 GPU3 GPU4 CPU Affinity...

Multi-instance inference fails in (n-1)/n runs (where n is a number gpus/instances)

I've tested your fix and unfortunately it did not resolve the problem for me. I double checked that the code indeed contains your fix in the docker image but I...

Multi-instance inference fails in (n-1)/n runs (where n is a number gpus/instances)

Hi, @PerkzZheng , I'm apologize I couldn't answer you sooner. I've once again tested your fix on setup with 2 to 5 gpus and it didn't work (I had my...

Multi-instance inference fails in (n-1)/n runs (where n is a number gpus/instances)

@byshiue Hi! I'm sorry it took me so long to answer, but I finally got to check your fix and it works on my setup! (RTX3090) Is it possible to...

Multi-instance inference fails in (n-1)/n runs (where n is a number gpus/instances)

@byshiue @PerkzZheng Hi! The fix from [this comment](https://github.com/triton-inference-server/fastertransformer_backend/issues/63#issuecomment-1491419307) works, are there any plans to merge it?