Fix some bugs

[tekknolagi]

Cast to unsigned char https://github.com/tekknolagi/tekknolagi.github.com/blob/7f3998d154f3d19a3b4c45a151237d256d851a31/_posts/2023-09-19-compiling-ml-models.md?plain=1#L1032

[tekknolagi]

Do we need to load pixels in [0,1] instead of [0,255]?

[tekknolagi]

Can we use loss fn from https://github.com/google/jax/blob/main/examples/mnist_classifier_fromscratch.py ?

[tekknolagi]

TODO: only store weight grads at top-level; drop interstitial variables to locals. also find out if there is a way to do the same thing with the weights (but they are currently used in backprop)

[tekknolagi]

Use llvmlite?

https://llvmlite.readthedocs.io/en/latest/user-guide/ir/examples.html https://llvmlite.readthedocs.io/en/latest/user-guide/binding/examples.html

or MLIR Python bindings

[tekknolagi]

Mention https://arxiv.org/pdf/1810.08061.pdf https://github.com/tensorflow/tensorflow/tree/master/tensorflow/python/autograph

[tekknolagi]

Mention https://github.com/jmd1011/snek-LMS

[tekknolagi]

Mention https://www.cs.purdue.edu/homes/rompf/papers/decker-preprint201907.pdf

[tekknolagi]

Mention https://www.cs.purdue.edu/homes/rompf/papers/wang-nips18.pdf

[tekknolagi]

Mention tinygrad & C renderer

[tekknolagi]

Mention https://arxiv.org/pdf/2102.13267.pdf

[tekknolagi]

Use https://github.com/silentbicycle/hopscotch (Tarjan strongly connected components) to parallelize parts of graphs

or https://github.com/taskflow/taskflow

[tekknolagi]

Use https://github.com/jart/matmul or https://github.com/google/highway or https://github.com/kfjahnke/zimt for gemm/matmul