Metadata-Version: 2.1
Name: spexmlp
Version: 2.0.0.post23
Summary: fused modules of EPRI(SGRI) visual models
Home-page: https://github.com/flishwang/spexmlp
Author: flish_wang
Author-email: flish_wang@sina.com
Description-Content-Type: text/markdown
License-File: LICENSE
Requires-Dist: tqdm>=4.28.1
Requires-Dist: numpy>=1.15.3
Requires-Dist: PyYAML>=5.1
Requires-Dist: packaging>=20.6
Requires-Dist: torch<2.4.0,>=2.3.0

# USAGE

```
from torch import nn
from spexmlp.eswiglu import load_license,sFusedSwiglu

load_license(path="mylicense.dat",device="cuda:0")
model=sFusedSwiglu(1280,1280*3,1024)
x=torch.zeros((256,1280),dtype=torch.float16,device='cuda:0')
dy=torch.randn((256,1024),dtype=torch.float16,device='cuda:0')
y=model(x)
y.backward(dy)
print(x.grad)

```

