MIOpen

MIOpen copied to clipboard

- Added Diag Forward operations - Added driver test and gtest for Diag operations The kernel is only 20% faster than ROCm if the following constraints are applied: - tensor...

cognaiger9

- Added backward Var operation and kernel. - Added driver test and gtest for Var. When comparing the newly developed miopen var kernel with ROCm, there's performance improvement for a...

cognaiger9

arai713

alexxu-amd

[gTest] Add more test coverage for Batchnorm Backward

2

More test cases added for better coverage. Here's the summary 1. N C H W: 128 256 14 14 Covers: backward_spatial_single.cpp: variant == 3 and variant == 1 (2nd) [128...

xinlipn

MIOpen(HIP): Error [EvaluateInvokers] /MIOpen/src/hipoc/hipoc_kernel.cpp:106: Failed to launch kernel: invalid configuration argument

6

Hi, I am running VLLM on my 7900XTX(gfx1100). I use ```vllm serve ./qwen2-vl-instruct-pytorch-7b --dtype auto --port 8000 --limit_mm_per_prompt image=4 --max_model_len 8784 --gpu_memory_utilization 0.9``` But then it shows errors: ``` $...

Looong01

Dynamic Kernels Assignments

5

### Purpose This project exists to minimize our reliance on compile time parameterization in MIOpen's source kernels. The goal isn't to sacrifice performance, but rather determine a ways of reducing...

daniellowell

[Conv] add ck bfp16 for fwd and bwd

2

bghimireamd

Precompiled kernels should contain compiled defaults

10

Need to assess whether this is feasible. Two types of defaults exist: 1. Some solver have hard-coded defaults -- one kernel for everything. 2. Others are configured at runtime using...

daniellowell

Output tensor size does not present in the sqlite db binding

8

Let's conder this: ``` ./bin/MIOpenDriver conv -c 192 -H 28 -W 28 -y 5 -x 5 -k 32 -n 17 -p 1 -q 1 -v 1 -u 1 -F 1...

atamazov