qurator-spk/eynollah
1
0
Fork 0
mirror of https://github.com/qurator-spk/eynollah.git synced 2026-06-16 09:59:13 +02:00
Code Issues Projects Releases Wiki Activity

ModelZoo ONNX backend for inference: support multi-input or -output

Browse source
This commit is contained in:
Robert Sachunsky 2026-06-03 20:57:02 +02:00
parent 38fe4d33ad
commit 27ca9733db
1 changed files with 18 additions and 7 deletions
Download patch file Download diff file Expand all files Collapse all files

25
src/eynollah/model_zoo/model_zoo.py
View file

@ -331,14 +331,25 @@ class EynollahModelZoo:
model_path, model_path,
providers=providers) providers=providers)
# FIXME: notify about selected provider/device # FIXME: notify about selected provider/device
input_name = model.get_inputs()[0].name model_inputs = [model_input.name
output_name = model.get_outputs()[0].name for model_input in model.get_inputs()]
model_outputs = [model_output.name
for model_output in model.get_outputs()]
def predict_onnx(inputs): def predict_onnx(inputs):
# models expect data_type() == 'tensor(float)', but np.float16 is 'tensor(float16)' if len(model_inputs) == 1:
# FIXME: do this dynamically (but how to convert .type to np.dtype?) inputs = [inputs]
inputs = inputs.astype(np.float32) outputs = model.run(model_outputs, {
return model.run( model_input:
[output_name], {input_name: inputs})[0] input_data.astype(
# models expect data_type() == 'tensor(float)', but np.float16 is 'tensor(float16)'
# FIXME: do this dynamically (but how to convert .type to np.dtype?)
np.float32 if input_data.dtype in [np.float16, np.float64] else
input_data.dtype)
for model_input, input_data in zip(model_inputs, inputs)
})
if len(model_outputs) == 1:
outputs = outputs[0]
return outputs
model.predict_on_batch = predict_onnx model.predict_on_batch = predict_onnx
model.input_shape = model.get_inputs()[0].shape model.input_shape = model.get_inputs()[0].shape