search

microsoft / antares

Antares: an automatic engine for multi-platform kernel generation and optimization. Supporting CPU, CUDA, ROCm, DirectX12, GraphCore, SYCL for CPU/GPU, OpenCL for AMD/NVIDIA, Android CPU/GPU backends.
Other
435 stars 45 forks source link

Is this project based on AI? What is the goal of this project? #372

Closed yhyu13 closed 10 months ago

yhyu13 commented 12 months ago

Hi,

Just come across this project, I tried to run 

# Search an efficient CUDA code for MatMul, using 2000 steps for trial:
BACKEND=c-cuda STEP=2000 COMPUTE_V1='- S = 512; einstein_v2(input_dict={"input0": {"dtype": "float32", "shape": [S, S]}, "input1": {"dtype": "float32", "shape": [S, S]}}, exprss="output0[N, M] +=! input0[N, K] * input1[K, M]")' antares

And it seems to generate and evaluate code on my GPU for 2000 runs.

I am not sure if this project involve LLM generating code and reinforcement learning in code optimization. Maybe a stupid question, could some explain breifly what is the end game goal of this project?

### Tasks
ghostplant commented 11 months ago

Hi, does the current README.md answer your questions?

yhyu13 commented 11 months ago

It's pretty much like what mojo's ML IR does? And this project mainly focus creating an adpater that enhance pytorch performance on specific compute backend?

ghostplant commented 11 months ago

Enhancing pytorch performance is one of the jobs. It also targets devices not supported by ML IR.