Flux.jl/examples/MNIST.jl

using Flux, MNIST

data = [(trainfeatures(i), Vector{Float32}(onehot(trainlabel(i), 0:9))) for i = 1:60_000]
train = data[1:50_000]
test = data[50_001:60_000]

m = Chain(
  Input(784),
  Dense(128), relu,
  Dense( 64), relu,
  Dense( 10), softmax)

# Convert to TensorFlow
model = tf(m)

# An example prediction pre-training
model(data[1][1])

@time Flux.train!(model, train, test, η = 1e-3)

# An example prediction post-training
model(data[1][1])
a bunch of stuff 2016-08-25 21:49:21 +00:00			`using Flux, MNIST`
init 2016-04-01 21:11:42 +00:00
Float32 by default 2016-10-04 21:36:56 +00:00			`data = [(trainfeatures(i), Vector{Float32}(onehot(trainlabel(i), 0:9))) for i = 1:60_000]`
sort-of working mnist example 2016-09-29 20:28:53 +00:00			`train = data[1:50_000]`
			`test = data[50_001:60_000]`
init 2016-04-01 21:11:42 +00:00
run with what we have 2016-08-25 16:26:52 +00:00			`m = Chain(`
init 2016-04-01 21:11:42 +00:00			`Input(784),`
a bunch of stuff 2016-08-25 21:49:21 +00:00			`Dense(128), relu,`
			`Dense( 64), relu,`
			`Dense( 10), softmax)`
init 2016-04-01 21:11:42 +00:00
sort-of working mnist example 2016-09-29 20:28:53 +00:00			`# Convert to TensorFlow`
			`model = tf(m)`
run with what we have 2016-08-25 16:26:52 +00:00
example predictions 2016-10-04 20:11:03 +00:00			`# An example prediction pre-training`
			`model(data[1][1])`

sort-of working mnist example 2016-09-29 20:28:53 +00:00			`@time Flux.train!(model, train, test, η = 1e-3)`
example predictions 2016-10-04 20:11:03 +00:00
			`# An example prediction post-training`
			`model(data[1][1])`