elixir-nx/axon

Issues

• The loss increases until it become NaN on XOR example

  #592 opened 21 days ago by jn-jairo
  1

• Remove namespaces

  #580 opened a month ago by seanmor5
  0

• Unable to replace existing model layers

  #529 opened a month ago by wtedw
  3

• Tagged versions missing in the repo compared to the ones in hexdocs.pm

  #585 opened 2 months ago by NJichev
  1

• Can't pass multiple inputs in custom trainer functions

  #582 opened 3 months ago by bartekupartek
  3

• Ambiguous inputs should raise

  #583 opened 3 months ago by seanmor5
  0

• Crash with mixed precision with new Nx

  #578 opened 3 months ago by seanmor5
  0

• Optionally return gradients / gradient state in train step

  #577 opened 4 months ago by seanmor5
  0

• Merge model state updates into a single `Axon.ModelState` call

  #576 opened 4 months ago by seanmor5
  0

• Most `Axon.ModelState` functions should be `defn` or `deftransform`

  #575 opened 4 months ago by seanmor5
  0

• Loss increasing with binary_cross_entropy above version 0.4.1

  #567 opened 4 months ago by preciz
  6

• Block with multiple inputs fails to compile

  #540 opened 4 months ago by mntns
  2

• Passing f32 data into LSTM with Axon.Loop trainer+run causes while shape mismatch error

  #490 opened 4 months ago by polvalente
  9

• (FunctionClauseError) no function clause matching in Nx.BinaryBackend.to_binary/1

  #497 opened 4 months ago by vishal-h
  6

• Bug in training loops batch count

  #531 opened 4 months ago by seanmor5
  1

• Req usage in OTP 26

  #570 opened 4 months ago by mwhitworth
  6

• Allow overriding layer metadata such as `:op` and `:op_name` for built-ins

  #549 opened 4 months ago by seanmor5
  1

• Optimize frozen parameter memory usage

  #548 opened 4 months ago by seanmor5
  1

• Separate state parameters and actual parameters in parameter map

  #485 opened 4 months ago by seanmor5
  3

• batch_norm states "mean" and "var" never updated

  #546 opened 4 months ago by robinmonjo
  1

• Compiled model will always result in `%Axon.None{}`

  #538 opened 4 months ago by mntns
  0

• Compile Error due to a type mismatch

  #568 opened 5 months ago by Mostafa86
  2

• a

  #564 opened 6 months ago by flowpoint
  0

• XLA Slow operation alarms: "The operation took ..."

  #558 opened 7 months ago by Szetty
  1

• Initializer does not respect Nx default backend

  #517 opened 9 months ago by alisinabh
  6

• Performance diversion between RNN in Keras and Axon

  #530 opened 10 months ago by seanmor5
  0

• `Nx.top_k` doesn't seem to support backpropagation.

  #509 opened 10 months ago by chaoqu
  3

• Cannot return map from Axon.nx

  #545 opened 10 months ago by jonatanklosko
  0

• Preserve integer input types

  #464 opened 10 months ago by seanmor5
  0

• mixed precision policies cause bumblebee models to fail

  #544 opened 10 months ago by ityonemo
  2

• Trouble implementing models in Axon

  #522 opened a year ago by krstopro
  30

• Explain how to use embedding layer

  #532 opened a year ago by a-alhusaini
  2

• Online deep learning in Axon

  #533 opened a year ago by sutgeorge
  2

• kino_vega_lite_plot: step number is wrong when plotting multiple metrics during a same loop

  #523 opened a year ago by robinmonjo
  1

• Axon ignoring the activation function?

  #526 opened a year ago by a-alhusaini
  5

• `Axon.map_nodes`documentation examples are wrong

  #518 opened a year ago by robinmonjo
  1

• Typo: `:loss_cale` but should be `:loss_scale`

  #516 opened a year ago by marcinkoziej
  0

• Add a guide for writing a simple recurrent network

  #511 opened a year ago by marcinkoziej
  7

• Add tests for loss_scale

  #465 opened a year ago by seanmor5
  1

• Example Livebooks have deps that error

  #493 opened a year ago by GenericJam
  2

• Template errors when using f64 inputs for training

  #494 opened a year ago by fabiokr
  8

• How to verify mixed precision is working for language models

  #496 opened a year ago by toranb
  3

• Different sentence embeddings for same input when using Axon.predict/3

  #488 opened a year ago by jannikbecher
  4

• MNIST example not working

  #489 opened a year ago by mvkvc
  0

• Bump ex_doc

  #482 opened a year ago by jonatanklosko
  0

• Object Tracker

  #481 opened a year ago by tiagodavi
  1

• Provide a better error message on shape mismatches in training loops

  #480 opened 2 years ago by seanmor5
  0

• Implement Kino.Render for Axon

  #478 opened 2 years ago by seanmor5
  0

• Experiences from porting YOLOv8 to Axon

  #475 opened 2 years ago by hansihe
  10

• Remove use of deprecated transforms

  #463 opened 2 years ago by seanmor5
  0