I'm trying to speed up the solution time for a dynamic programming problem in Julia (v. 0.5.0), via parallel processing. The problem involves choosing the optimal values for every element of a 1073 x 19 matrix at every iteration, until successive matrix differences fall within a tolerance. I thought that, within each iteration, filling in the values for each element of the matrix could be parallelized. However, I'm seeing a huge performance degradation using SharedArray, and I'm wondering if there's a better way to approach parallel processing for this problem.
I construct the arguments for the function below:
est_params = [.788,.288,.0034,.1519,.1615,.0041,.0077,.2,0.005,.7196]
r = 0.015
tau = 0.35
rho =est_params[1]
sigma =est_params[2]
delta = 0.15
gamma =est_params[3]
a_capital =est_params[4]
lambda1 =est_params[5]
lambda2 =est_params[6]
s =est_params[7]
theta =est_params[8]
mu =est_params[9]
p_bar_k_ss =est_params[10]
beta = (1+r)^(-1)
sigma_range = 4
gz = 19
gp = 29
gk = 37
lnz=collect(linspace(-sigma_range*sigma,sigma_range*sigma,gz))
z=exp(lnz)
gk_m = fld(gk,2)
# Need to add mu somewhere to k_ss
k_ss = (theta*(1-tau)/(r+delta))^(1/(1-theta))
k=cat(1,map(i->k_ss*((1-delta)^i),collect(1:gk_m)),map(i->k_ss/((1-delta)^i),collect(1:gk_m)))
insert!(k,gk_m+1,k_ss)
sort!(k)
p_bar=p_bar_k_ss*k_ss
p = collect(linspace(-p_bar/2,p_bar,gp))
#Tauchen
N = length(z)
Z = zeros(N,1)
Zprob = zeros(Float32,N,N)
Z[N] = lnz[length(z)]
Z[1] = lnz[1]
zstep = (Z[N] - Z[1]) / (N - 1)
for i=2:(N-1)
Z[i] = Z[1] + zstep * (i - 1)
end
for a = 1 : N
for b = 1 : N
if b == 1
Zprob[a,b] = 0.5*erfc(-((Z[1] - mu - rho * Z[a] + zstep / 2) / sigma)/sqrt(2))
elseif b == N
Zprob[a,b] = 1 - 0.5*erfc(-((Z[N] - mu - rho * Z[a] - zstep / 2) / sigma)/sqrt(2))
else
Zprob[a,b] = 0.5*erfc(-((Z[b] - mu - rho * Z[a] + zstep / 2) / sigma)/sqrt(2)) -
0.5*erfc(-((Z[b] - mu - rho * Z[a] - zstep / 2) / sigma)/sqrt(2))
end
end
end
# Collecting tauchen results in a 2 element array of linspace and array; [2] gets array
# Zprob=collect(tauchen(gz, rho, sigma, mu, sigma_range))[2]
Zcumprob=zeros(Float32,gz,gz)
# 2 in cumsum! denotes the 2nd dimension, i.e. columns
cumsum!(Zcumprob, Zprob,2)
gm = gk * gp
control=zeros(gm,2)
for i=1:gk
control[(1+gp*(i-1)):(gp*i),1]=fill(k[i],(gp,1))
control[(1+gp*(i-1)):(gp*i),2]=p
end
endog=copy(control)
E=Array(Float32,gm,gm,gz)
for h=1:gm
for m=1:gm
for j=1:gz
# set the nonzero net debt indicator
if endog[h,2]<0
p_ind=1
else
p_ind=0
end
# set the investment indicator
if (control[m,1]-(1-delta)*endog[h,1])!=0
i_ind=1
else
i_ind=0
end
E[m,h,j] = (1-tau)*z[j]*(endog[h,1]^theta) + control[m,2]-endog[h,2]*(1+r*(1-tau)) +
delta*endog[h,1]*tau-(control[m,1]-(1-delta)*endog[h,1]) -
(i_ind*gamma*endog[h,1]+endog[h,1]*(a_capital/2)*(((control[m,1]-(1-delta)*endog[h,1])/endog[h,1])^2)) +
s*endog[h,2]*p_ind
elem = E[m,h,j]
if E[m,h,j]<0
E[m,h,j]=elem+lambda1*elem-.5*lambda2*elem^2
else
E[m,h,j]=elem
end
end
end
end
I then constructed the function with serial processing. The two for loops iterate through each element to find the largest value in a 1072-sized (=the gm scalar argument in the function) array:
function dynam_serial(E,gm,gz,beta,Zprob)
v = Array(Float32,gm,gz )
fill!(v,E[cld(gm,2),cld(gm,2),cld(gz,2)])
Tv = Array(Float32,gm,gz)
# Set parameters for the loop
convcrit = 0.0001 # chosen convergence criterion
diff = 1 # arbitrary initial value greater than convcrit
while diff>convcrit
exp_v=v*Zprob'
for h=1:gm
for j=1:gz
Tv[h,j]=findmax(E[:,h,j] + beta*exp_v[:,j])[1]
end
end
diff = maxabs(Tv - v)
v=copy(Tv)
end
end
Timing this, I get:
#time dynam_serial(E,gm,gz,beta,Zprob)
> 106.880008 seconds (91.70 M allocations: 203.233 GB, 15.22% gc time)
Now, I try using Shared Arrays to benefit from parallel processing. Note that I reconfigured the iteration so that I only have one for loop, rather than two. I also use v=deepcopy(Tv); otherwise, v is copied as an Array object, rather than a SharedArray:
function dynam_parallel(E,gm,gz,beta,Zprob)
v = SharedArray(Float32,(gm,gz),init = S -> S[Base.localindexes(S)] = myid() )
fill!(v,E[cld(gm,2),cld(gm,2),cld(gz,2)])
# Set parameters for the loop
convcrit = 0.0001 # chosen convergence criterion
diff = 1 # arbitrary initial value greater than convcrit
while diff>convcrit
exp_v=v*Zprob'
Tv = SharedArray(Float32,gm,gz,init = S -> S[Base.localindexes(S)] = myid() )
#sync #parallel for hj=1:(gm*gz)
j=cld(hj,gm)
h=mod(hj,gm)
if h==0;h=gm;end;
#async Tv[h,j]=findmax(E[:,h,j] + beta*exp_v[:,j])[1]
end
diff = maxabs(Tv - v)
v=deepcopy(Tv)
end
end
Timing the parallel version; and using a 4-core 2.5 GHz I7 processor with 16GB of memory, I get:
addprocs(3)
#time dynam_parallel(E,gm,gz,beta,Zprob)
> 164.237208 seconds (2.64 M allocations: 201.812 MB, 0.04% gc time)
Am I doing something incorrect here? Or is there a better way to approach parallel processing in Julia for this particular problem? I've considered using Distributed Arrays, but it's difficult for me to see how to apply them to the present problem.
UPDATE:
Per #DanGetz and his helpful comments, I turned instead to trying to speed up the serial processing version. I was able to get performance down to 53.469780 seconds (67.36 M allocations: 103.419 GiB, 19.12% gc time) through:
1) Upgrading to 0.6.0 (saved about 25 seconds), which includes the helpful #views macro.
2) Preallocating the main array I'm trying to fill in (Tv), per the section on Preallocating Outputs in the Julia Performance Tips: https://docs.julialang.org/en/latest/manual/performance-tips/. (saved another 25 or so seconds)
The biggest remaining slow-down seems to be coming from the add_vecs function, which sums together subarrays of two larger matrices. I've tried devectorizing and using BLAS functions, but haven't been able to produce better performance.
In any event, the improved code for dynam_serial is below:
function add_vecs(r::Array{Float32},h::Int,j::Int,E::Array{Float32},exp_v::Array{Float32},beta::Float32)
#views r=E[:,h,j] + beta*exp_v[:,j]
return r
end
function dynam_serial(E::Array{Float32},gm::Int,gz::Int,beta::Float32,Zprob::Array{Float32})
v = Array{Float32}(gm,gz)
fill!(v,E[cld(gm,2),cld(gm,2),cld(gz,2)])
Tv = Array{Float32}(gm,gz)
r = Array{Float32}(gm)
# Set parameters for the loop
convcrit = 0.0001 # chosen convergence criterion
diff = 1 # arbitrary initial value greater than convcrit
while diff>convcrit
exp_v=v*Zprob'
for h=1:gm
for j=1:gz
#views Tv[h,j]=findmax(add_vecs(r,h,j,E,exp_v,beta))[1]
end
end
diff = maximum(abs,Tv - v)
v=copy(Tv)
end
return Tv
end
If add_vecs seems to be the critical function, writing an explicit for loop could offer more optimization. How does the following benchmark:
function add_vecs!(r::Array{Float32},h::Int,j::Int,E::Array{Float32},
exp_v::Array{Float32},beta::Float32)
#inbounds for i=1:size(E,1)
r[i]=E[i,h,j] + beta*exp_v[i,j]
end
return r
end
UPDATE
To continue optimizing dynam_serial I have tried to remove more allocations. The result is:
function add_vecs_and_max!(gm::Int,r::Array{Float64},h::Int,j::Int,E::Array{Float64},
exp_v::Array{Float64},beta::Float64)
#inbounds for i=1:gm
r[i] = E[i,h,j]+beta*exp_v[i,j]
end
return findmax(r)[1]
end
function dynam_serial(E::Array{Float64},gm::Int,gz::Int,
beta::Float64,Zprob::Array{Float64})
v = Array{Float64}(gm,gz)
fill!(v,E[cld(gm,2),cld(gm,2),cld(gz,2)])
r = Array{Float64}(gm)
exp_v = Array{Float64}(gm,gz)
# Set parameters for the loop
convcrit = 0.0001 # chosen convergence criterion
diff = 1.0 # arbitrary initial value greater than convcrit
while diff>convcrit
A_mul_Bt!(exp_v,v,Zprob)
diff = -Inf
for h=1:gm
for j=1:gz
oldv = v[h,j]
newv = add_vecs_and_max!(gm,r,h,j,E,exp_v,beta)
v[h,j]= newv
diff = max(diff, oldv-newv, newv-oldv)
end
end
end
return v
end
Switching the functions to use Float64 should increase speed (as CPUs are inherently optimized for 64-bit word lengths). Also, using the mutating A_mul_Bt! directly saves another allocation. Avoiding the copy(...) by switching the arrays v and Tv.
How do these optimizations improve your running time?
2nd UPDATE
Updated the code in the UPDATE section to use findmax. Also, changed dynam_serial to use v without Tv, as there was no need to save the old version except for the diff calculation, which is now done inside the loop.
Here's the code I copied-and-pasted, provided by Dan Getz above. I include the array and scalar definitions exactly as I ran them. Performance was: 39.507005 seconds (11 allocations: 486.891 KiB) when running #time dynam_serial(E,gm,gz,beta,Zprob).
using SpecialFunctions
est_params = [.788,.288,.0034,.1519,.1615,.0041,.0077,.2,0.005,.7196]
r = 0.015
tau = 0.35
rho =est_params[1]
sigma =est_params[2]
delta = 0.15
gamma =est_params[3]
a_capital =est_params[4]
lambda1 =est_params[5]
lambda2 =est_params[6]
s =est_params[7]
theta =est_params[8]
mu =est_params[9]
p_bar_k_ss =est_params[10]
beta = (1+r)^(-1)
sigma_range = 4
gz = 19 #15 #19
gp = 29 #19 #29
gk = 37 #25 #37
lnz=collect(linspace(-sigma_range*sigma,sigma_range*sigma,gz))
z=exp.(lnz)
gk_m = fld(gk,2)
# Need to add mu somewhere to k_ss
k_ss = (theta*(1-tau)/(r+delta))^(1/(1-theta))
k=cat(1,map(i->k_ss*((1-delta)^i),collect(1:gk_m)),map(i->k_ss/((1-delta)^i),collect(1:gk_m)))
insert!(k,gk_m+1,k_ss)
sort!(k)
p_bar=p_bar_k_ss*k_ss
p = collect(linspace(-p_bar/2,p_bar,gp))
#Tauchen
N = length(z)
Z = zeros(N,1)
Zprob = zeros(Float64,N,N)
Z[N] = lnz[length(z)]
Z[1] = lnz[1]
zstep = (Z[N] - Z[1]) / (N - 1)
for i=2:(N-1)
Z[i] = Z[1] + zstep * (i - 1)
end
for a = 1 : N
for b = 1 : N
if b == 1
Zprob[a,b] = 0.5*erfc(-((Z[1] - mu - rho * Z[a] + zstep / 2) / sigma)/sqrt(2))
elseif b == N
Zprob[a,b] = 1 - 0.5*erfc(-((Z[N] - mu - rho * Z[a] - zstep / 2) / sigma)/sqrt(2))
else
Zprob[a,b] = 0.5*erfc(-((Z[b] - mu - rho * Z[a] + zstep / 2) / sigma)/sqrt(2)) -
0.5*erfc(-((Z[b] - mu - rho * Z[a] - zstep / 2) / sigma)/sqrt(2))
end
end
end
# Collecting tauchen results in a 2 element array of linspace and array; [2] gets array
# Zprob=collect(tauchen(gz, rho, sigma, mu, sigma_range))[2]
Zcumprob=zeros(Float64,gz,gz)
# 2 in cumsum! denotes the 2nd dimension, i.e. columns
cumsum!(Zcumprob, Zprob,2)
gm = gk * gp
control=zeros(gm,2)
for i=1:gk
control[(1+gp*(i-1)):(gp*i),1]=fill(k[i],(gp,1))
control[(1+gp*(i-1)):(gp*i),2]=p
end
endog=copy(control)
E=Array(Float64,gm,gm,gz)
for h=1:gm
for m=1:gm
for j=1:gz
# set the nonzero net debt indicator
if endog[h,2]<0
p_ind=1
else
p_ind=0
end
# set the investment indicator
if (control[m,1]-(1-delta)*endog[h,1])!=0
i_ind=1
else
i_ind=0
end
E[m,h,j] = (1-tau)*z[j]*(endog[h,1]^theta) + control[m,2]-endog[h,2]*(1+r*(1-tau)) +
delta*endog[h,1]*tau-(control[m,1]-(1-delta)*endog[h,1]) -
(i_ind*gamma*endog[h,1]+endog[h,1]*(a_capital/2)*(((control[m,1]-(1-delta)*endog[h,1])/endog[h,1])^2)) +
s*endog[h,2]*p_ind
elem = E[m,h,j]
if E[m,h,j]<0
E[m,h,j]=elem+lambda1*elem-.5*lambda2*elem^2
else
E[m,h,j]=elem
end
end
end
end
function add_vecs_and_max!(gm::Int,r::Array{Float64},h::Int,j::Int,E::Array{Float64},
exp_v::Array{Float64},beta::Float64)
maxr = -Inf
#inbounds for i=1:gm r[i] = E[i,h,j]+beta*exp_v[i,j]
maxr = max(r[i],maxr)
end
return maxr
end
function dynam_serial(E::Array{Float64},gm::Int,gz::Int,
beta::Float64,Zprob::Array{Float64})
v = Array{Float64}(gm,gz)
fill!(v,E[cld(gm,2),cld(gm,2),cld(gz,2)])
Tv = Array{Float64}(gm,gz)
r = Array{Float64}(gm)
exp_v = Array{Float64}(gm,gz)
# Set parameters for the loop
convcrit = 0.0001 # chosen convergence criterion
diff = 1.0 # arbitrary initial value greater than convcrit
while diff>convcrit
A_mul_Bt!(exp_v,v,Zprob)
diff = -Inf
for h=1:gm
for j=1:gz
Tv[h,j]=add_vecs_and_max!(gm,r,h,j,E,exp_v,beta)
diff = max(abs(Tv[h,j]-v[h,j]),diff)
end
end
(v,Tv)=(Tv,v)
end
return v
end
Now, here's another version of the algorithm and inputs. The functions are similar to what Dan Getz suggested, except that I use findmax rather than an iterated max function to find the array maximum. In the input construction, I am using both Float32 and mixing different bit-types together. However, I've consistently achieved better performance this way: 24.905569 seconds (1.81 k allocations: 46.829 MiB, 0.01% gc time). But it's not clear at all why.
using SpecialFunctions
est_params = [.788,.288,.0034,.1519,.1615,.0041,.0077,.2,0.005,.7196]
r = 0.015
tau = 0.35
rho =est_params[1]
sigma =est_params[2]
delta = 0.15
gamma =est_params[3]
a_capital =est_params[4]
lambda1 =est_params[5]
lambda2 =est_params[6]
s =est_params[7]
theta =est_params[8]
mu =est_params[9]
p_bar_k_ss =est_params[10]
beta = Float32((1+r)^(-1))
sigma_range = 4
gz = 19
gp = 29
gk = 37
lnz=collect(linspace(-sigma_range*sigma,sigma_range*sigma,gz))
z=exp(lnz)
gk_m = fld(gk,2)
# Need to add mu somewhere to k_ss
k_ss = (theta*(1-tau)/(r+delta))^(1/(1-theta))
k=cat(1,map(i->k_ss*((1-delta)^i),collect(1:gk_m)),map(i->k_ss/((1-delta)^i),collect(1:gk_m)))
insert!(k,gk_m+1,k_ss)
sort!(k)
p_bar=p_bar_k_ss*k_ss
p = collect(linspace(-p_bar/2,p_bar,gp))
#Tauchen
N = length(z)
Z = zeros(N,1)
Zprob = zeros(Float32,N,N)
Z[N] = lnz[length(z)]
Z[1] = lnz[1]
zstep = (Z[N] - Z[1]) / (N - 1)
for i=2:(N-1)
Z[i] = Z[1] + zstep * (i - 1)
end
for a = 1 : N
for b = 1 : N
if b == 1
Zprob[a,b] = 0.5*erfc(-((Z[1] - mu - rho * Z[a] + zstep / 2) / sigma)/sqrt(2))
elseif b == N
Zprob[a,b] = 1 - 0.5*erfc(-((Z[N] - mu - rho * Z[a] - zstep / 2) / sigma)/sqrt(2))
else
Zprob[a,b] = 0.5*erfc(-((Z[b] - mu - rho * Z[a] + zstep / 2) / sigma)/sqrt(2)) -
0.5*erfc(-((Z[b] - mu - rho * Z[a] - zstep / 2) / sigma)/sqrt(2))
end
end
end
# Collecting tauchen results in a 2 element array of linspace and array; [2] gets array
# Zprob=collect(tauchen(gz, rho, sigma, mu, sigma_range))[2]
Zcumprob=zeros(Float32,gz,gz)
# 2 in cumsum! denotes the 2nd dimension, i.e. columns
cumsum!(Zcumprob, Zprob,2)
gm = gk * gp
control=zeros(gm,2)
for i=1:gk
control[(1+gp*(i-1)):(gp*i),1]=fill(k[i],(gp,1))
control[(1+gp*(i-1)):(gp*i),2]=p
end
endog=copy(control)
E=Array(Float32,gm,gm,gz)
for h=1:gm
for m=1:gm
for j=1:gz
# set the nonzero net debt indicator
if endog[h,2]<0
p_ind=1
else
p_ind=0
end
# set the investment indicator
if (control[m,1]-(1-delta)*endog[h,1])!=0
i_ind=1
else
i_ind=0
end
E[m,h,j] = (1-tau)*z[j]*(endog[h,1]^theta) + control[m,2]-endog[h,2]*(1+r*(1-tau)) +
delta*endog[h,1]*tau-(control[m,1]-(1-delta)*endog[h,1]) -
(i_ind*gamma*endog[h,1]+endog[h,1]*(a_capital/2)*(((control[m,1]-(1-delta)*endog[h,1])/endog[h,1])^2)) +
s*endog[h,2]*p_ind
elem = E[m,h,j]
if E[m,h,j]<0
E[m,h,j]=elem+lambda1*elem-.5*lambda2*elem^2
else
E[m,h,j]=elem
end
end
end
end
function add_vecs!(gm::Int,r::Array{Float32},h::Int,j::Int,E::Array{Float32},
exp_v::Array{Float32},beta::Float32)
#inbounds #views for i=1:gm
r[i]=E[i,h,j] + beta*exp_v[i,j]
end
return r
end
function dynam_serial(E::Array{Float32},gm::Int,gz::Int,beta::Float32,Zprob::Array{Float32})
v = Array{Float32}(gm,gz)
fill!(v,E[cld(gm,2),cld(gm,2),cld(gz,2)])
Tv = Array{Float32}(gm,gz)
# Set parameters for the loop
convcrit = 0.0001 # chosen convergence criterion
diff = 1.00000 # arbitrary initial value greater than convcrit
iter=0
exp_v=Array{Float32}(gm,gz)
r=Array{Float32}(gm)
while diff>convcrit
A_mul_Bt!(exp_v,v,Zprob)
for h=1:gm
for j=1:gz
Tv[h,j]=findmax(add_vecs!(gm,r,h,j,E,exp_v,beta))[1]
end
end
diff = maximum(abs,Tv - v)
(v,Tv)=(Tv,v)
end
return v
end
I built 5-layer neural network by using tensorflow.
I have a problem to get reproducible results (or stable results).
I found similar questions regarding reproducibility of tensorflow and the corresponding answers, such as How to get stable results with TensorFlow, setting random seed
But the problem is not solved yet.
I also set random seed like the following
tf.set_random_seed(1)
Furthermore, I added seed options to every random function such as
b1 = tf.Variable(tf.random_normal([nHidden1], seed=1234))
I confirmed that the first epoch shows the identical results, but not identical from the second epoch little by little.
How can I get the reproducible results?
Am I missing something?
Here is a code block I use.
def xavier_init(n_inputs, n_outputs, uniform=True):
if uniform:
init_range = tf.sqrt(6.0 / (n_inputs + n_outputs))
return tf.random_uniform_initializer(-init_range, init_range, seed=1234)
else:
stddev = tf.sqrt(3.0 / (n_inputs + n_outputs))
return tf.truncated_normal_initializer(stddev=stddev, seed=1234)
import numpy as np
import tensorflow as tf
import dataSetup
from scipy.stats.stats import pearsonr
tf.set_random_seed(1)
x_train, y_train, x_test, y_test = dataSetup.input_data()
# Parameters
learningRate = 0.01
trainingEpochs = 1000000
batchSize = 64
displayStep = 100
thresholdReduce = 1e-6
thresholdNow = 0.6
#dropoutRate = tf.constant(0.7)
# Network Parameter
nHidden1 = 128 # number of 1st layer nodes
nHidden2 = 64 # number of 2nd layer nodes
nInput = 24 #
nOutput = 1 # Predicted score: 1 output for regression
# save parameter
modelPath = 'model/model_layer5_%d_%d_mini%d_lr%.3f_noDrop_rollBack.ckpt' %(nHidden1, nHidden2, batchSize, learningRate)
# tf Graph input
X = tf.placeholder("float", [None, nInput])
Y = tf.placeholder("float", [None, nOutput])
# Weight
W1 = tf.get_variable("W1", shape=[nInput, nHidden1], initializer=xavier_init(nInput, nHidden1))
W2 = tf.get_variable("W2", shape=[nHidden1, nHidden2], initializer=xavier_init(nHidden1, nHidden2))
W3 = tf.get_variable("W3", shape=[nHidden2, nHidden2], initializer=xavier_init(nHidden2, nHidden2))
W4 = tf.get_variable("W4", shape=[nHidden2, nHidden2], initializer=xavier_init(nHidden2, nHidden2))
WFinal = tf.get_variable("WFinal", shape=[nHidden2, nOutput], initializer=xavier_init(nHidden2, nOutput))
# biases
b1 = tf.Variable(tf.random_normal([nHidden1], seed=1234))
b2 = tf.Variable(tf.random_normal([nHidden2], seed=1234))
b3 = tf.Variable(tf.random_normal([nHidden2], seed=1234))
b4 = tf.Variable(tf.random_normal([nHidden2], seed=1234))
bFinal = tf.Variable(tf.random_normal([nOutput], seed=1234))
# Layers for dropout
L1 = tf.nn.relu(tf.add(tf.matmul(X, W1), b1))
L2 = tf.nn.relu(tf.add(tf.matmul(L1, W2), b2))
L3 = tf.nn.relu(tf.add(tf.matmul(L2, W3), b3))
L4 = tf.nn.relu(tf.add(tf.matmul(L3, W4), b4))
hypothesis = tf.add(tf.matmul(L4, WFinal), bFinal)
print "Layer setting DONE..."
# define loss and optimizer
cost = tf.reduce_mean(tf.square(hypothesis - Y))
optimizer = tf.train.AdamOptimizer(learning_rate=learningRate).minimize(cost)
# Initialize the variable
init = tf.initialize_all_variables()
# save op to save and restore all the variables
saver = tf.train.Saver()
with tf.Session() as sess:
# initialize
sess.run(init)
print "Initialize DONE..."
# Training
costPrevious = 100000000000000.0
best = float("INF")
totalBatch = int(len(x_train)/batchSize)
print "Total Batch: %d" %totalBatch
for epoch in range(trainingEpochs):
#print "EPOCH: %04d" %epoch
avgCost = 0.
for i in range(totalBatch):
np.random.seed(i+epoch)
randidx = np.random.randint(len(x_train), size=batchSize)
batch_xs = x_train[randidx,:]
batch_ys = y_train[randidx,:]
# Fit traiing using batch data
sess.run(optimizer, feed_dict={X:batch_xs, Y:batch_ys})
# compute average loss
avgCost += sess.run(cost, feed_dict={X:batch_xs, Y:batch_ys})/totalBatch
# compare the current cost and the previous
# if current cost > the previous
# just continue and make the learning rate half
#print "Cost: %1.8f --> %1.8f at epoch %05d" %(costPrevious, avgCost, epoch+1)
if avgCost > costPrevious + .5:
#sess.run(init)
load_path = saver.restore(sess, modelPath)
print "Cost increases at the epoch %05d" %(epoch+1)
print "Cost: %1.8f --> %1.8f" %(costPrevious, avgCost)
continue
costNow = avgCost
reduceCost = abs(costPrevious - costNow)
costPrevious = costNow
#Display logs per epoch step
if costNow < best:
best = costNow
bestMatch = sess.run(hypothesis, feed_dict={X:x_test})
# model save
save_path = saver.save(sess, modelPath)
if epoch % displayStep == 0:
print "step {}".format(epoch)
pearson = np.corrcoef(bestMatch.flatten(), y_test.flatten())
print 'train loss = {}, current loss = {}, test corrcoef={}'.format(best, costNow, pearson[0][1])
if reduceCost < thresholdReduce or costNow < thresholdNow:
print "Epoch: %04d, Cost: %.9f, Prev: %.9f, Reduce: %.9f" %(epoch+1, costNow, costPrevious, reduceCost)
break
print "Optimization Finished"
It seems that your results are perhaps not reproducible because you are using Saver to write/restore from checkpoint each time? (i.e. the second time that you run the code, the variable values aren't initialized using your random seed -- they are restored from your previous checkpoint)
Please trim down your code example to just the code necessary to reproduce irreproducibility.