Mnist dataset wrapping success! Not sure how to do other datasets wit…

…hout nim-lang/Nim#16653 and nim-lang/Nim#16655
SciNim · Jan 9, 2021 · 45a2687 · 45a2687
1 parent ec43ff1
commit 45a2687
Show file tree

Hide file tree

Showing 3 changed files with 30 additions and 13 deletions.
diff --git a/flambeau/raw_bindings/data_api.nim b/flambeau/raw_bindings/data_api.nim
@@ -18,6 +18,14 @@ import
 # This should ease searching PyTorch and libtorch documentation,
 # and make C++ tutorials easily applicable.
 
+# Headers
+# -----------------------------------------------------------------------
+
+{.passC: "-I" & headersPath.}
+{.passC: "-I" & torchHeadersPath.}
+
+{.push header: torchHeader.}
+
 # #######################################################################
 #
 #                         Datasets
@@ -47,21 +55,27 @@ type
     ##   with T being Example[Data, Target]
     ## BatchRequest is by default ArrayRef[csize_t]
 
+  # TODO: https://github.com/nim-lang/Nim/issues/16653
+  #   generics + {.inheritable.} doesn't work
+  # TODO: https://github.com/nim-lang/Nim/issues/16655
+  # CRTP + importcpp don't work
   Dataset*
     {.bycopy, pure,
     importcpp: "torch::data::datasets::Dataset".}
-    [Self, Batch]
+    # [Self, Batch]
       = object of BatchDataset # [Self, Batch, ArrayRef[csize_t]]
     ## A Dataset type
     ## Self: is the class type that implements the Dataset API
     ##   (using the Curious Recurring Template Pattern in underlying C++)
     ## Batch is by default the type CppVector[T]
     ##   with T being Example[Data, Target]
 
+  # TODO: https://github.com/nim-lang/Nim/issues/16655
+  # CRTP + importcpp don't work
   Mnist*
     {.bycopy, pure,
     importcpp: "torch::data::datasets::MNIST".}
-    = object of Dataset[Mnist, CppVector[Example[Tensor, Tensor]]]
+    = object of Dataset # [Mnist, CppVector[Example[Tensor, Tensor]]]
     ## The MNIST dataset
     ## http://yann.lecun.com/exdb/mnist
 
@@ -71,7 +85,7 @@ type
     kTrain = 0
     kTest = 1
 
-func mnist*(rootPath: cstring, mode = kTrain): Mnist {.constructor, importcpp:"MNIST(@)".}
+func mnist*(rootPath: cstring, mode = kTrain): Mnist {.constructor, importcpp:"torch::data::datasets::MNIST(@)".}
   ## Loads the MNIST dataset from the `root` path
   ## The supplied `rootpath` should contain the *content* of the unzipped
   ## MNIST dataset, available from http://yann.lecun.com/exdb/mnist.

diff --git a/flambeau/raw_bindings/tensors.nim b/flambeau/raw_bindings/tensors.nim
@@ -72,9 +72,9 @@ else:
 # Headers
 # -----------------------------------------------------------------------
 
-const headersPath = libTorchPath & "/include"
-const torchHeadersPath = headersPath / "torch/csrc/api/include"
-const torchHeader = torchHeadersPath / "torch/torch.h"
+const headersPath* = libTorchPath & "/include"
+const torchHeadersPath* = headersPath / "torch/csrc/api/include"
+const torchHeader* = torchHeadersPath / "torch/torch.h"
 
 {.passC: "-I" & headersPath.}
 {.passC: "-I" & torchHeadersPath.}

diff --git a/proof_of_concepts/poc07_datasets.nim b/proof_of_concepts/poc07_datasets.nim
@@ -3,11 +3,14 @@ import
     data_api, tensors
   ]
 
-let mnist = mnist("build/mnist")
+proc main() =
+  let mnist = mnist("build/mnist")
 
-echo "Data"
-# mnist.get(0).data.print()
-# echo "\n-----------------------"
-# echo "Target"
-# mnist.get(0).target.print()
-# echo "\n-----------------------"
+  echo "Data"
+  mnist.get(0).data.print()
+  echo "\n-----------------------"
+  echo "Target"
+  mnist.get(0).target.print()
+  echo "\n-----------------------"
+
+main()