Age | Commit message (Collapse) | Author | Files | Lines |
|
appropriate, part 2: the non-trivial cases
|
|
This just moves the package to the current head in the official repo,
which has since seen patches that unbreak the ghc-9 build and work
with our version of TensorFlow.
|
|
The package was introduced when our source package did not expose
libtensorflow, which it does these days. Additionally the -bin package
has not been well maintained and is stuck on a version from 2020/12.
|
|
|
|
Its tensorflow-mnist-data dependency is not on Hackage, and the
license conditions for the data are unclear. Also, it's basically an
example package.
|
|
Updating to the current HEAD of the Tensorflow-Haskell bindings allows
us to also update the dependencies, specifically proto-lens, and avoid
having to retain their outdated versions.
|
|
|
|
- get the current Git head of haskell-tensorflow
- adjust dependencies
|
|
The configuration-tensorflow.nix file specified several overrides for version
0.2.x packages, but those packages are no longer included in our package set
because they are so old. (Current versions seem to be in the range of 0.6.x.)
I fixed the evalution errors, but I did not verify whether these packages
actually build with the newer versions.
|
|
|
|
|
|
tensorflow-mnist-input-data
|
|
This prevents duplication in the nix store.
|
|
the data files
This is more understandable and safer than using a nested list.
|
|
Prefetch data so that the package doesn't try to connect to the internet.
|
|
Also enable building on GHC-8.4.
|