axon
axon copied to clipboard
elixir-nx
Nx-powered Neural Networks
```elixir input = Axon.input("input_series", shape: put_elem(Nx.shape(time_x), 0, nil)) model = input |> Axon.lstm(128, activation: :relu) |> elem(0) |> Axon.dropout(rate: 0.25) |> Axon.lstm(64, activation: :relu) |> elem(0) |> Axon.dropout(rate: 0.25) |>...
I'm currently trying to implement the LoRA algorithm in Axon. It involves freezing the original model's weights, adding new weights inside an existing layer, and feeding the input into both...
When passing multiple inputs to a block, the model fails to compile: ```elixir Mix.install( [ {:exla, ">= 0.0.0"}, {:axon, path: "./axon", overwrite: true}, {:table_rex, "~> 3.1.1"} ], config: [nx: [default_backend:...
I'm trying to load a pre-trained Sentence Transformers model using `bumblebee` in order to use it with `axon` (using a custom layer for mean pooling instead of a postprocessing step)....
This allows us to override the hook if necessary. I was searching for a solution to provide Layer metadata in the hook function signature, but I don't think it's possible...
Step counts are a bit off. They should be consistent across batches, e.g.: ``` Epoch: 0, Batch: 350, accuracy: 0.5190082 loss: 0.6815987 Epoch: 1, Batch: 341, accuracy: 0.7035355 loss: 0.6067376...
Hi! Axon beginner here. I struggle to figure out how to write a very simple RNN network. Basically I want to rewrite this pytorch example from [a tutorial](https://pytorch.org/tutorials/intermediate/char_rnn_classification_tutorial.html). However, the...
Trying out the example in https://github.com/elixir-nx/axon/blob/main/examples/generative/text_generator.exs throws the error `(FunctionClauseError) no function clause matching in Nx.BinaryBackend.to_binary/1` ``` Attempted function clauses (showing 1 out of 1): defp to_binary(%Nx.Tensor{data: %{state: data}}) (nx...
Related to #453 by grouping these together in the same part of the map we've run into lots of issues with state getting updated when it was not meant to...
Right now we just raise the Nx error, but we should do a better job of making it clear why the mismatch error occurs and possible options to fix it