gaspersic/triton
1
0
Fork 0
Code Issues Pull Requests Packages Projects Releases Wiki Activity

nn?

Browse Source
This commit is contained in:
Philippe Tillet
2014-10-01 04:44:16 +02:00
parent 3523a3756f
commit 2f6d41f661
2 changed files with 79 additions and 84 deletions
Download Patch File Download Diff File Expand all files Collapse all files

97
autotune/python/dataset.py
View File

@@ -17,7 +17,7 @@ def resample(X, tbincount, densities, step):
r = random.random()
while(True):
if(len(tbincount)==0 or len(densities)==0 or r<=1.0/len(densities)):
x = np.array([step*random.randint(1,40), step*random.randint(1,40), step*random.randint(1,40)]);
x = np.array([step*random.randint(1,40), step*random.randint(1,40), step*random.randint(1,40)])
else:
probs = [1.0/x if x>0 else 0 for x in tbincount]
distr = np.random.choice(range(tbincount.size), p = probs/np.sum(probs))
@@ -28,67 +28,46 @@ def resample(X, tbincount, densities, step):
return x.astype(int)
def generate_dataset(TemplateType, execution_handler):
I = 0
I = 10
step = 64
max_size = 4000
path = "./data"
#Tries to resume
try:
X = np.loadtxt(open(os.path.join(path, "X.csv"),"rb"))
t = np.loadtxt(open(os.path.join(path, "t.csv"),"rb"))
profiles = np.loadtxt(open(os.path.join(path, "profiles.csv"),"rb")).tolist()
if not isinstance(profiles[0], list):
profiles = [profiles]
N = t.size
X.resize((N+I, 3), refcheck=False)
t.resize(N+I, refcheck=False)
print 'Resuming dataset generation...'
except:
X = np.empty((I,I))
t = np.empty(I)
profiles = []
N = 0
pass
#Generates new data
print "Getting some good profiles..."
densities = [KernelDensity(kernel='gaussian', bandwidth=2*step).fit(X[t==i,:]) for i in range(int(max(t))+1)] if N else [];
X.resize((N+I, 3), refcheck=False)
t.resize(N+I, refcheck=False)
for i in range(I):
tbincount = np.bincount(t[0:i+1].astype(int))
x = resample(X, tbincount, densities, step)
y = execution_handler(x)
if y not in profiles:
profiles.append(y)
densities.append(KernelDensity(kernel='gaussian', bandwidth=2*step))
idx = profiles.index(y)
X[N+i,:] = x
t[N+i] = idx
densities[idx].fit(X[t[0:N+i+1]==idx,:])
np.savetxt(os.path.join(path,"X.csv"), X)
np.savetxt(os.path.join(path,"t.csv"), t)
np.savetxt(os.path.join(path,"profiles.csv"), profiles)
print "Generating the dataset..."
N = 500
Y = np.empty((N, len(profiles)))
X = np.empty((N,3))
t = []
for i in range(N):
x = resample(X, np.bincount(t), densities, step)
for j,y in enumerate(profiles):
T = execution_handler(x, os.devnull, decode(map(int, y)))
Y[i,j] = 2*1e-9*x[0]*x[1]*x[2]/T
idx = np.argmax(Y[i,:])
X[i,:] = x
t = np.argmax(Y[:i+1,], axis=1)
densities[idx].fit(X[t==idx,:])
np.savetxt(os.path.join(path,"Y.csv"), Y)
# print "Getting some good profiles..."
# X = np.empty((I, 3))
# t = np.empty(I)
# profiles = []
# for i in range(I):
# x = resample(X, [], [], step)
# y = execution_handler(x)
# if y not in profiles:
# profiles.append(y)
# idx = profiles.index(y)
# X[i,:] = x
# t[i] = idx
# densities = [KernelDensity(kernel='gaussian', bandwidth=2*step).fit(X[t==i,:]) for i in range(int(max(t))+1)];
#
# print "Generating the dataset..."
# N = 1000
# Y = np.empty((N, len(profiles)))
# X = np.empty((N,3))
# t = []
#
# for i in range(N):
# x = resample(X, np.bincount(t), densities, step)
# for j,y in enumerate(profiles):
# T = execution_handler(x, os.devnull, decode(map(int, y)))
# Y[i,j] = 2*1e-9*x[0]*x[1]*x[2]/T
# idx = np.argmax(Y[i,:])
# X[i,:] = x
# t = np.argmax(Y[:i+1,], axis=1)
# densities[idx].fit(X[t==idx,:])
#
# np.savetxt(os.path.join(path,"profiles.csv"), profiles)
# np.savetxt(os.path.join(path,"X.csv"), X)
# np.savetxt(os.path.join(path,"Y.csv"), Y)
profiles = np.loadtxt(os.path.join(path,"profiles.csv"))
X = np.loadtxt(os.path.join(path,"X.csv"))
Y = np.loadtxt(os.path.join(path,"Y.csv"))
return X, Y, profiles

66
autotune/python/model.py
View File

@@ -2,41 +2,57 @@ from sklearn import *;
from sklearn import ensemble;
import numpy as np
import scipy as sp
from pybrain.datasets import SupervisedDataSet
from pybrain.tools.shortcuts import buildNetwork
from pybrain.supervised.trainers import BackpropTrainer
from pybrain.structure import LinearLayer, TanhLayer, SigmoidLayer, SoftmaxLayer, FeedForwardNetwork, BiasUnit
from pybrain.tools.neuralnets import NNregression, Trainer
def train_model(X, Y, profiles):
#Preprocessing
scaler = preprocessing.StandardScaler().fit(X);
X = scaler.transform(X);
ref = np.argmax(np.bincount(np.argmax(Y, axis=1))) #most common profile
Xmean = np.mean(X, axis=0)
Xstd = np.std(X, axis=0)
X = (X - Xmean)/Xstd
Ymax = np.max(Y)
Y = Y/Ymax
ref = np.argmax(np.bincount(np.argmax(Y, axis=1))) #most common profile
#Cross-validation data-sets
cut = int(0.5*X.shape[0]+1);
XTr = X[0:cut, :];
YTr = Y[0:cut, :];
XTe = X[cut:,:];
YTe = Y[cut:,:];
cut = int(0.1*X.shape[0]+1)
XTr = X[0:cut, :]
YTr = Y[0:cut, :]
XTe = X[cut:,:]
YTe = Y[cut:,:]
#Train the model
print("Training the model...");
clf = linear_model.LinearRegression().fit(XTr,YTr);
print("Training the model...")
ds = SupervisedDataSet(X.shape[1], Y.shape[1])
for idx, x in enumerate(X):
ds.addSample(x, Y[idx,:])
clf = buildNetwork(*[X.shape[1], 100, Y.shape[1]], hiddenclass = TanhLayer, outclass = LinearLayer)
#print fnn;
#trainer = RPropMinusTrainer( fnn, dataset=ds, verbose=True);
trainer = BackpropTrainer( clf, dataset=ds, verbose=True, momentum=0.01, weightdecay=0.01, learningrate=0.002, batchlearning=False)
trainer.trainUntilConvergence(maxEpochs=100)
#Evaluate the model
GFlops = np.empty(XTe.shape[0]);
speedups = np.empty(XTe.shape[0]);
optspeedups = np.empty(XTe.shape[0]);
GFlops = np.empty(XTe.shape[0])
speedups = np.empty(XTe.shape[0])
optspeedups = np.empty(XTe.shape[0])
for i,x in enumerate(XTe):
predictions = clf.predict(x);
label = np.argmax(predictions);
speedups[i] = YTe[i,label]/YTe[i,ref];
optspeedups[i] = np.max(YTe[i,:])/YTe[i,ref];
GFlops[i] = YTe[i,ref];
predictions = clf.activate(x)
label = np.argmax(predictions)
# print YTe[i,label], YTe[i,ref], np.max(YTe[i,:])
speedups[i] = YTe[i,label]/YTe[i,ref]
optspeedups[i] = np.max(YTe[i,:])/YTe[i,ref]
GFlops[i] = YTe[i,ref]*Ymax
np.set_printoptions(precision=2);
print("-----------------");
print("Average testing speedup : %f (Optimal : %f)"%(sp.stats.gmean(speedups), sp.stats.gmean(optspeedups)));
print("Average GFLOP/s : %f (Default %f, Optimal %f)"%(np.mean(np.multiply(GFlops,speedups)), np.mean(GFlops), np.mean(np.multiply(GFlops,optspeedups))));
print("Minimum speedup is %f wrt %i GFlops"%(np.min(speedups), GFlops[np.argmin(speedups)]));
print("Maximum speedup is %f wrt %i GFlops"%(np.max(speedups), GFlops[np.argmax(speedups)]));
print("--------");
np.set_printoptions(precision=2)
print("-----------------")
print("Average testing speedup : %f (Optimal : %f)"%(sp.stats.gmean(speedups), sp.stats.gmean(optspeedups)))
print("Average GFLOP/s : %f (Default %f, Optimal %f)"%(np.mean(np.multiply(GFlops,speedups)), np.mean(GFlops), np.mean(np.multiply(GFlops,optspeedups))))
print("Minimum speedup is %f wrt %i GFlops"%(np.min(speedups), GFlops[np.argmin(speedups)]))
print("Maximum speedup is %f wrt %i GFlops"%(np.max(speedups), GFlops[np.argmax(speedups)]))
print("--------")
print clf