reduce size to be able to upload to crate.io

Author: daniel-e

Cargo.toml

@@ -11,6 +11,18 @@ description = """
 A library for doing maching learning in Rust.
 """
 
+# Currently we can only upload up to 10MB. So we need to remove
+# everything that is not required for building the library.
+exclude = [
+	"atlas/*",
+	"build/*",
+	"build_all.sh",
+	"doc.sh",
+	"examples/*",
+	"datasets/mnist_digits/train-images-idx3-ubyte.gz",
+	"datasets/mnist_digits/train-labels-idx1-ubyte.gz",
+]
+
 [lib]
 name = "rustml"
 path = "src/lib.rs"

datasets/mnist_digits/README.md

@@ -1 +1,10 @@
 Dataset of digits from http://yann.lecun.com/exdb/mnist/
+
+reduced dataset
+
+# remove 10000 examples
+dd if=train-images-idx3-ubyte of=train-images-idx3-ubyte.50k bs=16 count=2450001
+gzip -9 train-images-idx3-ubyte.50k
+dd if=train-labels-idx1-ubyte of=train-labels-idx1-ubyte.50k bs=1 count=50008
+gzip -9 train-labels-idx1-ubyte.50k
+

datasets/mnist_digits/train-images-idx3-ubyte.50k.gz

@@ -16,6 +16,9 @@ use matrix::*;
 /// The database that is used is available at http://yann.lecun.com/exdb/mnist/
 /// and contains 60,000 training examples and 10,000 test examples of handwritten
 /// digits.
+///
+/// Currently, due to upload limits at crate.io a smaller training dataset
+/// is used with 50,000 training examples.
 pub struct MnistDigits;
 
 impl MnistDigits {
@@ -106,8 +109,8 @@ impl MnistDigits {
 
         // TODO location of dataset
         MnistDigits::from(
-            "datasets/mnist_digits/train-images-idx3-ubyte.gz",
-            "datasets/mnist_digits/train-labels-idx1-ubyte.gz"
+            "datasets/mnist_digits/train-images-idx3-ubyte.50k.gz",
+            "datasets/mnist_digits/train-labels-idx1-ubyte.50k.gz"
         )
     }