certik/mlc

Issues

• Do batch as the last (contiguous) dimension

  #56 opened 10 months ago by certik
  0

• HLIR: add add_32_f16

  #18 opened a year ago by certik
  0

• Operations like "+" and matmul

  #9 opened a year ago by certik
  0

• Speedup the PyTorch training

  #73 opened 10 months ago by certik
  0

• Checks and warnings in LLIR

  #58 opened 10 months ago by certik
  5

• Implement the other MNIST network with batchnorm

  #65 opened 10 months ago by certik
  0

• Finish f16 implementation

  #55 opened 10 months ago by certik
  0

• Binary format choice

  #34 opened 10 months ago by certik
  2

• Crash generating mnist-tests.gguf

  #36 opened 10 months ago by rebcabin
  3

• Design CPU LLIR

  #23 opened 10 months ago by certik
  0

• Fixed vs free dimensions

  #21 opened a year ago by certik
  0

• Idea: figure out from the NNIR what input/output dimensions are fixed/free

  #20 opened a year ago by certik
  0

• Design LLIR

  #19 opened a year ago by certik
  0

• NNIR: Dropout node for training

  #12 opened a year ago by certik
  0

• Consider not putting previous dimension in NNIR

  #11 opened a year ago by certik
  1

• TensorFlow and PyTorch backend

  #13 opened a year ago by certik
  0

• conv2d should only work for 4D arrays in HLIR

  #8 opened a year ago by certik
  0