Merge branch 'master' of github.com:aarati-K/one_access

benchmarks/cifar10/plots/plot_measurements.py

@@ -0,0 +1,32 @@
+from pathlib import Path
+import math
+import matplotlib.pyplot as plt
+import numpy as np
+import pylab
+
+one_access_500 = np.load(Path("../measurements/one_access_500.npy").open('rb'))
+one_access_1000 = np.load(
+    Path("../measurements/one_access_1000.npy").open('rb'))
+pt = np.load(Path("../measurements/pytorch.npy").open('rb'))
+
+fig, ax = plt.subplots()
+fig.set_figheight(5)
+fig.set_figwidth(7)
+
+plt.xticks(fontsize=14)
+plt.yticks(fontsize=14)
+
+line, = ax.plot(one_access_500, color='blue', lw=2)
+line, = ax.plot(one_access_1000, color='lightblue', lw=2)
+line, = ax.plot(pt, color='red', lw=2)
+ax.set_yscale('log')
+ax.legend([
+    'One Access with Sample Size 500', 'One Access with Sample Size 1000',
+    'PyTorch'
+],
+          loc=2,
+          prop={'size': 14})
+ax.set_xlabel('Iterations', fontsize=18)
+ax.set_ylabel('Data Loading Time (s)', fontsize=18)
+
+plt.savefig("cifar10_measurements.png")

benchmarks/cifar10/plots/plot_sample_creator_benchmark.py

@@ -0,0 +1,34 @@
+import matplotlib.pyplot as plt
+
+ss1 = [0.349, 0.306, 0.431, 0.303, 0.18, 0.69, 0.557, 0.681, 0.424, 0.300]
+ss10 = [0.742, 0.685, 0.679, 0.676, 0.673, 0.676, 0.551, 0.673, 0.669, 0.670]
+ss100 = [0.713, 0.672, 0.668, 0.671, 0.668, 0.680, 0.682, 0.675, 0.673, 0.669]
+ss1000 = [0.738, 0.689, 0.704, 0.693, 0.684, 0.683, 0.678, 0.677, 0.700, 0.687]
+ss10000 = [
+    0.765, 0.727, 0.717, 0.740, 0.723, 0.774, 0.720, 0.868, 0.724, 0.771
+]
+
+fig, ax = plt.subplots()
+fig.set_figheight(5)
+fig.set_figwidth(7)
+
+plt.xticks(fontsize=14)
+plt.yticks(fontsize=14)
+
+bplot = ax.boxplot(
+    [ss1, ss10, ss100, ss1000, ss10000],
+    vert=True,  # vertical box alignment
+    patch_artist=True,  # fill with color
+    labels=[1, 10, 100, 1000, 10000])  # will be used to label x-ticks
+ax.set_title('Cifar-10 Sample Creation Times', fontsize=18)
+
+# fill with colors
+for patch in bplot['boxes']:
+    patch.set(facecolor='lightblue')
+
+# adding horizontal grid lines
+ax.yaxis.grid(True)
+ax.set_xlabel('Sample Size', fontsize=18)
+ax.set_ylabel('Sample Creation Time (s)', fontsize=18)
+
+plt.savefig("cifar10_sample_creation.png")

benchmarks/coco/plots/plot_measurements.py

@@ -0,0 +1,25 @@
+from pathlib import Path
+import math
+import matplotlib.pyplot as plt
+import numpy as np
+import pylab
+
+oa = np.load(Path("../measurements/one_access_400_full_epoch.npy").open('rb'))
+pt = np.load(Path("../measurements/pytorch_full_epoch_w1.npy").open('rb'))
+
+fig, ax = plt.subplots()
+fig.set_figheight(5)
+fig.set_figwidth(7)
+
+plt.xticks(fontsize=14)
+plt.yticks(fontsize=14)
+
+line, = ax.plot(oa, color='blue', lw=2)
+line, = ax.plot(pt, color='red', lw=2)
+ax.set_yscale('log')
+ax.legend(['One Access with Sample Size 400', 'PyTorch'],
+          loc=2,
+          prop={'size': 14})
+ax.set_xlabel('Iterations', fontsize=18)
+ax.set_ylabel('Data Loading Time (s)', fontsize=18)
+plt.savefig("coco_measurements.png")

benchmarks/coco/plots/plot_sample_creator_benchmark.py

@@ -0,0 +1,34 @@
+import matplotlib.pyplot as plt
+
+ss1 = [1.12, 1.37, 2.32, 1.31, 2.42, 3.29, 2.92, 0.29, 1.53, 1.08]
+ss10 = [7.6, 11.75, 7.53, 12.64, 10.5, 13.93, 8.56, 10.16, 7.23, 13.1]
+ss100 = [24.1, 24.96, 24.9, 25.04, 25.07, 25.06, 25.25, 24.1, 22.52, 25.6]
+ss1000 = [26.35, 26.56, 26.44, 26.27, 26.31, 26.39, 27.25, 27.18, 26.43, 26.34]
+ss10000 = [28.3, 29.56, 29.55, 29.42, 29.09, 29.09, 29.36, 29.34, 29.8, 29.36]
+
+fig, ax = plt.subplots()
+fig.set_figheight(5)
+fig.set_figwidth(7)
+
+plt.xticks(fontsize=14)
+plt.yticks(fontsize=14)
+
+bplot = ax.boxplot(
+    [ss1, ss10, ss100, ss1000, ss10000],
+    vert=True,  # vertical box alignment
+    patch_artist=True,  # fill with color
+    labels=[1, 10, 100, 1000, 10000])  # will be used to label x-ticks
+ax.set_title('MS COCO Sample Creation Times', fontsize=18)
+
+# fill with colors
+colors = ['lightblue', 'lightblue', 'lightblue']
+
+for patch, color in zip(bplot['boxes'], colors):
+    patch.set_facecolor(color)
+
+# adding horizontal grid lines
+ax.yaxis.grid(True)
+ax.set_xlabel('Sample Size', fontsize=18)
+ax.set_ylabel('Sample Creation Time (s)', fontsize=18)
+
+plt.savefig("sample_creator_benchmark.png")

benchmarks/hogwild/plots/plot_convergence.py

@@ -0,0 +1,130 @@
+from pathlib import Path
+import math
+import matplotlib.pyplot as plt
+import numpy as np
+import pylab
+
+oa = [
+    2.310431718826294, 2.3044593334198, 2.2999351024627686, 2.302645206451416,
+    2.305140256881714, 2.304278612136841, 2.305345058441162, 2.305643081665039,
+    2.2953379154205322, 2.2945961952209473, 2.2974445819854736,
+    2.3022661209106445, 2.291269302368164, 2.290717601776123,
+    2.2888007164001465, 2.296182632446289, 2.286400556564331,
+    2.2910146713256836, 2.295991897583008, 2.272655963897705,
+    2.2934072017669678, 2.274665355682373, 2.2305684089660645,
+    2.2382869720458984, 2.3204245567321777, 2.342759847640991,
+    2.2820348739624023, 2.2226617336273193, 2.151423215866089,
+    2.2459020614624023, 2.2275006771087646, 2.142763137817383,
+    2.0960915088653564, 2.1954452991485596, 2.103869915008545,
+    2.0974860191345215, 2.3693454265594482, 2.0779247283935547,
+    2.248171806335449, 2.091695785522461, 2.1171560287475586,
+    2.0281434059143066, 1.9547799825668335, 2.0486371517181396,
+    1.907325267791748, 2.043123245239258, 2.025918483734131,
+    2.1420865058898926, 1.922833800315857, 2.028428554534912,
+    2.234170436859131, 1.711983561515808, 1.855607509613037,
+    2.1356916427612305, 2.058394193649292, 1.8293321132659912,
+    1.7653907537460327, 1.6710327863693237, 2.120668649673462,
+    1.977416753768921, 2.01466703414917, 1.7095742225646973,
+    1.7972968816757202, 1.835667371749878, 1.813649296760559,
+    1.6506456136703491, 1.8212352991104126, 1.7631133794784546,
+    2.006159543991089, 1.8940590620040894, 1.77767014503479,
+    1.8165675401687622, 1.5790164470672607, 1.9957212209701538,
+    1.8157849311828613, 1.8369755744934082, 1.8062738180160522,
+    2.02036452293396, 1.53107750415802, 1.464531660079956, 1.9184999465942383,
+    1.8831769227981567, 1.7480698823928833, 1.7008342742919922,
+    2.1173808574676514, 1.8246912956237793, 1.6952927112579346,
+    1.7756145000457764, 1.5981590747833252, 1.7205015420913696,
+    1.6052865982055664, 1.9077681303024292, 1.741557002067566,
+    1.4362589120864868, 1.4271641969680786, 1.7727869749069214,
+    1.6802165508270264, 1.6700048446655273, 1.764768123626709,
+    1.6664189100265503, 1.7431089878082275, 1.841097116470337,
+    1.609131097793579, 1.692051649093628, 1.9490209817886353,
+    1.3016743659973145, 1.6564037799835205, 1.5003793239593506,
+    1.577265977859497, 1.5513148307800293, 2.015777826309204,
+    1.7387953996658325, 1.5755693912506104, 1.517099142074585,
+    1.600704312324524, 1.681156873703003, 1.3915257453918457,
+    1.8454186916351318, 1.5253746509552002, 1.6675136089324951,
+    1.904675841331482, 1.3001149892807007, 1.3500734567642212,
+    1.421111822128296, 1.338475227355957, 1.8976962566375732,
+    1.4634175300598145, 1.1438446044921875, 1.8094542026519775,
+    1.421639323234558, 1.6332736015319824, 1.4899511337280273,
+    1.2700839042663574, 1.6670219898223877, 1.433416724205017,
+    1.6300684213638306, 1.4137150049209595, 1.7015812397003174,
+    1.602172613143921, 1.4118467569351196, 1.6460998058319092,
+    1.5490946769714355, 1.3902286291122437, 1.7114341259002686,
+    1.493067741394043, 1.5293599367141724, 1.1891926527023315,
+    1.2002012729644775, 1.4254976511001587, 1.4333221912384033,
+    1.6626935005187988, 1.3381460905075073, 1.631758213043213,
+    1.527137279510498, 1.3946272134780884, 1.617354393005371
+]
+pt = [
+    2.309093475341797, 2.29903244972229, 2.2979722023010254,
+    2.3146700859069824, 2.2988228797912598, 2.3095779418945312,
+    2.315662145614624, 2.291715145111084, 2.2961325645446777,
+    2.304250955581665, 2.297032356262207, 2.30222749710083, 2.297666072845459,
+    2.2927210330963135, 2.3093206882476807, 2.299983501434326,
+    2.2999722957611084, 2.2888784408569336, 2.290457248687744,
+    2.276545286178589, 2.2864463329315186, 2.2621915340423584,
+    2.2831177711486816, 2.206403970718384, 2.2706034183502197,
+    2.3174095153808594, 2.2077319622039795, 2.2165980339050293,
+    2.135857105255127, 2.130533218383789, 2.046015739440918,
+    2.1862354278564453, 2.1601943969726562, 1.9698059558868408,
+    2.119527578353882, 1.9596431255340576, 2.035116672515869,
+    1.9414715766906738, 2.0950546264648438, 2.1059789657592773,
+    1.9875943660736084, 2.246439218521118, 1.98484206199646, 2.056269645690918,
+    1.925750970840454, 2.095628023147583, 2.1306421756744385,
+    1.8586400747299194, 1.9911742210388184, 2.2050232887268066,
+    2.0071871280670166, 2.000347137451172, 1.6103558540344238,
+    2.0377726554870605, 1.9713151454925537, 1.9737203121185303,
+    1.7756128311157227, 1.776196837425232, 1.886519193649292,
+    2.053093433380127, 1.9902201890945435, 1.901827335357666,
+    1.784909963607788, 1.879426121711731, 1.6710565090179443,
+    1.809483528137207, 1.9051872491836548, 1.8594785928726196,
+    1.840088129043579, 1.7972636222839355, 1.8046557903289795,
+    1.7713732719421387, 1.716488003730774, 2.0590882301330566, 1.7366863489151,
+    1.638418436050415, 1.6430528163909912, 2.031158685684204,
+    1.8202089071273804, 1.4382100105285645, 1.7852904796600342,
+    1.8027703762054443, 2.0446808338165283, 1.6792430877685547,
+    1.5751527547836304, 1.6834783554077148, 2.0464272499084473,
+    1.575252652168274, 1.4901397228240967, 1.5762988328933716,
+    1.8482239246368408, 1.7038806676864624, 1.5288643836975098,
+    1.7010613679885864, 1.7013449668884277, 1.5429691076278687,
+    1.3439911603927612, 1.7403404712677002, 1.709581732749939,
+    1.4834586381912231, 1.4511845111846924, 1.5556445121765137,
+    1.7796345949172974, 1.7114813327789307, 1.5121259689331055,
+    1.3385242223739624, 1.8766744136810303, 1.7403178215026855,
+    1.4059736728668213, 1.5615800619125366, 1.3987661600112915,
+    1.3916569948196411, 1.8435180187225342, 1.5172559022903442,
+    1.523059368133545, 1.7093738317489624, 1.8336877822875977,
+    1.6907463073730469, 1.6465243101119995, 1.8130486011505127,
+    1.8594300746917725, 1.40418541431427, 1.8459088802337646,
+    1.5057564973831177, 1.273330807685852, 1.7335593700408936,
+    1.4547404050827026, 1.671050786972046, 1.8709566593170166,
+    1.4086576700210571, 1.6071624755859375, 1.3607245683670044,
+    1.6000163555145264, 1.9216904640197754, 1.281594157218933,
+    1.7730766534805298, 1.4627121686935425, 1.586707592010498,
+    1.4127451181411743, 1.4334553480148315, 1.3599401712417603,
+    1.4565072059631348, 1.37852942943573, 1.6175483465194702,
+    1.3432127237319946, 1.3964488506317139, 1.5373486280441284,
+    1.3805601596832275, 1.2533472776412964, 1.440510630607605,
+    1.4937564134597778, 1.4182919263839722, 1.447078824043274,
+    1.3968679904937744, 1.2916183471679688, 1.3860799074172974,
+    1.3477157354354858
+]
+
+fig, ax = plt.subplots()
+fig.set_figheight(5)
+fig.set_figwidth(7)
+
+plt.xticks(fontsize=14)
+plt.yticks(fontsize=14)
+
+line, = ax.plot(oa, color='blue', lw=2)
+line, = ax.plot(pt, color='red', lw=2)
+# ax.set_yscale('log')
+ax.legend(['One Access', 'PyTorch'],
+          loc=2,
+          prop={'size': 14})
+ax.set_xlabel('Iterations', fontsize=18)
+ax.set_ylabel('Loss', fontsize=18)
+plt.savefig("hogwild_convergence.png")