Unable to use predict_array on image due to CUDA not enabled

Part 1 (2018)
1

Hi everyone,
I have trained cats vs dogs model on fastai GPU instance on aws and then when i tried to use the model on my machine (mac OS) to predict an image class predict_array doesn’t seem to work due to this error Torch not compiled with CUDA enabled

 img = cv2.imread('dog.jpg')
 _, img_encoded = cv2.imencode('.jpg', img)
 nparr = np.fromstring(img_encoded.tostring(), np.uint8)
 flags = cv2.IMREAD_UNCHANGED+cv2.IMREAD_ANYDEPTH+cv2.IMREAD_ANYCOLOR
 image = cv2.cvtColor(cv2.imdecode(nparr, flags).astype(np.float32) / 255, cv2.COLOR_BGR2RGB)
 image = val_tfms(image)
 preds = learn.predict_array(image[None])

Does anyone know how to solve this in able to run prediction on cpu ?

2

Sounds like you may not have the latest fastai from git.

3

I installed it via pip ! should i upgrade the library with pip ?

4

The fastai version installed via pip will trail that available via Github. So I’d recommend (and Jeremy did somewhere else) to install and pull directly from Github.

5

It worked with cpu environment installation thanks :wink: