Updated char rnn

text/char-rnn/Sample.txt

@@ -0,0 +1,29 @@
+kt el!
+
+nPEDNRONIO:
+Plamincay likind, lawak gunse in'd the is
+rn?
+
+SORRCARYINCHNITESNZCY , heenthe ly! srius dne deatshy, looditifn tinfow, inf the rhou gicen.'ldnes'd my nounty, and was yay!
+Od; folne his the tict you, I but so up trpllary,
+If' my ohour gugooher nendce.
+
+OTLTINAS:
+Howg, comet l tn'Taeld,
+Thisss be your. Pilit?
+
+PROTOCK arit, bayoigabansk upeielss end thmngwe good be ad: st wal patcirpurpeoy Cadvalt sirprupt,
+Tiupord
+Fed, and of taestlingit stedu't lonid that we dvou wa dore.
+
+LASIUS:
+GiniI, litn! nof lirithid not be eseadlcord: Y: I Illel ookd surl! reenniondnd thint lowg'nW roff: or spothortwith. lold treotled, an,ist, with them tratco;
+That! thy mator is wit?
+Wece and k nos the and langoS:
+If Perme and at sord's sigff, abeied,
+Frethese burepul!
+Thing;
+Harla you st thBIs iuwhich det t!
+Oluld,
+Nall cimen all gurnus mour to byou coveny sion, ray, the wif, rivennnolavus? soneldemt ned lirwoyit a dohicthe to ove brordut weld of are sordwll of licumencine, alis,
+Sto ut, P

text/char-rnn/char-rnn.jl

@@ -1,45 +1,55 @@
+#Declaration of Packages used
 using Flux
-using Flux: onehot, chunk, batchseq, throttle, crossentropy
+using Flux: onehot, chunk, batchseq, throttle, crossentropy,@epochs
 using StatsBase: wsample
 using Base.Iterators: partition
 
 cd(@__DIR__)
 
+#Check for the input.txt file and if doesn't exist then downloads the text file 
 isfile("input.txt") ||
   download("https://cs.stanford.edu/people/karpathy/char-rnn/shakespeare_input.txt",
            "input.txt")
 
+#Text Preprocessing
 text = collect(String(read("input.txt")))
 alphabet = [unique(text)..., '_']
 text = map(ch -> onehot(ch, alphabet), text)
 stop = onehot('_', alphabet)
 
-N = length(alphabet)
-seqlen = 50
+N = length(alphabet) #
+seqlen = 20
 nbatch = 50
+epo = 1
 
+#Partioning the data into batches of chunks for Training
 Xs = collect(partition(batchseq(chunk(text, nbatch), stop), seqlen))
 Ys = collect(partition(batchseq(chunk(text[2:end], nbatch), stop), seqlen))
 
+#Defining the model
 m = Chain(
-  LSTM(N, 128),
-  LSTM(128, 128),
+    #Input of Shape N*M where M is the length of Input batch
+  LSTM(N, 128,relu),
+  LSTM(128, 256),
+  LSTM(256,128,relu),
   Dense(128, N),
   softmax)
 
 m = gpu(m)
 
 function loss(xs, ys)
   l = sum(crossentropy.(m.(gpu.(xs)), gpu.(ys)))
-  Flux.truncate!(m)
+  Flux.reset!(m)
   return l
 end
 
+#Compilation parameters
 opt = ADAM(0.01)
 tx, ty = (Xs[5], Ys[5])
 evalcb = () -> @show loss(tx, ty)
 
-Flux.train!(loss, params(m), zip(Xs, Ys), opt,
+#Training the model
+@epochs epo Flux.train!(loss, params(m), zip(Xs, Ys), opt,
             cb = throttle(evalcb, 30))
 
 # Sampling
@@ -51,9 +61,18 @@ function sample(m, alphabet, len)
   c = rand(alphabet)
   for i = 1:len
     write(buf, c)
-    c = wsample(alphabet, m(onehot(c, alphabet)).data)
+    c = wsample(alphabet, m(onehot(c, alphabet)))
   end
   return String(take!(buf))
 end
 
 sample(m, alphabet, 1000) |> println
+
+#Saving the text generated by our model to Sample.txt file
+open("Sample.txt", "w") do f
+    write(f, sample(m, alphabet, 1000))
+end
+
+
+
+