Lantern
Lantern copied to clipboard
Hello, I was wondering if you had an example of how to serve a ONNX model trained in Python in Scala. Basically I'm looking to deploy a number of models...
Thanks for this very interesting project! :) I am the developer of [TensorFlow Scala](https://github.com/eaplatanios/tensorflow_scala) and I'm wondering how easy it would be to add support for TF Scala in Lantern...
High priority: - [x] cuBLAS - [x] cuDNN CUDA backend todos: - [ ] Implement cuDNN ops (forward and backward). Prioritize ones used in models. - [x] [Done.](https://github.com/feiwang3311/Lantern/pull/32) 2-D convolution....
[The design of `withBackend` is explained here.](https://github.com/feiwang3311/Lantern/issues/8#issuecomment-426563742) Non-nested invocations `withBackend` should work as expected. (Test `matrix-matrix-dot-with-backend` added in this PR is verified to work on GPU.) Nested invocations of `withBackend`...
Clean up
A big list of clean up opportunities, vaguely ordered by importance. Feel free to add/edit. - [ ] Revamp tensor initializers, especially ones that take both "data" and "shape" arguments....
- [x] SqueezeNet - [x] ResNet - [x] TreeLSTM - [ ] DeepSpeech2 - [ ] WaveNet - [ ] Transformer
Implement basic, non-strided indexing as described [here](https://github.com/feiwang3311/Lantern/commit/53fa930bb697c9e0617c905c17ef91bafd92e2da#r30776349). - [ ] Element tensor indexing: `tensor(i)`. - This returns an element tensor, whose shape is `tensor.shape.tail`. - [ ] Subtensor indexing (i.e....
Mathematically, scalars are 0-dimensional tensors. However, 0-dimensional tensors are [currently disallowed](https://github.com/feiwang3311/Lantern/blob/1bd2130fc0fbd08c68fb59b83eb9ea09915a3476/src/main/scala/lantern/ad_lms_vector.scala#L261): ```scala class Tensor(val data: Rep[Array[Float]], val dimensions: NSeq[Int]) extends Serializable { val strides = (dimensions :\ NSeq[Int]()) { case...