TypeError: cannot resolve operator 'LogSoftmax' with opsets: ai.onnx v9

microsoft / onnxjs

ONNX.js: run ONNX models using JavaScript

Other

1.75k stars 130 forks source link

Open iiSeymour opened 3 years ago

iiSeymour commented 3 years ago

Great project - many of the models I'm interested in have a final LogSoftmax layer which I see is not yet supported.

$\text{LogSoftmax}(x_{i}) = \log{\frac{\exp(x_i)}{\sum_j \exp(x_j)}}$

However, the principal components {Log, Softmax} are implemented so is there a mechanism where I can provide my own implementation at runtime?