search

kartik4949 / AutoDeploy

AutoDeploy is a single configuration deployment library
MIT License
41 stars 12 forks source link

Onnx Loader #11

Closed nilavghosh closed 3 years ago

nilavghosh commented 3 years ago

Add onnx loader into the autodeploy

import numpy as np
import onnxruntime as ort
sess = ort.InferenceSession('horse_zebra.onnx')

input_name = sess.get_inputs()[0].name
label_name = sess.get_outputs()[0].name
pred_onx = sess.run(
    [label_name], {input_name: np.zeros((1, 3, 224,224), dtype=np.float32)})[0]
print(pred_onx)
nilavghosh commented 3 years ago

horse_zebra.zip

nilavghosh commented 3 years ago

FastAPI input array

requests.post('/predict', json = {'input' : json.dumps(im.tolist())})
class InputIn(BaseModel):
    input : str

@app.post('/')
asyc def predict(inp : InputIn):
    ...
    inp = np.array(json.loads(inp.input))
    sess.run...()
kartik4949 commented 3 years ago

solved #12