[GH-PAGES] Updated website

triton-lang · Jul 13, 2024 · 3e644ee · 3e644ee
1 parent 72bb7f3
commit 3e644ee
Show file tree

Hide file tree

Showing 52 changed files with 413 additions and 413 deletions.
diff --git a/main/.doctrees/environment.pickle b/main/.doctrees/environment.pickle
diff --git a/main/.doctrees/getting-started/tutorials/01-vector-add.doctree b/main/.doctrees/getting-started/tutorials/01-vector-add.doctree
diff --git a/main/.doctrees/getting-started/tutorials/02-fused-softmax.doctree b/main/.doctrees/getting-started/tutorials/02-fused-softmax.doctree
diff --git a/main/.doctrees/getting-started/tutorials/03-matrix-multiplication.doctree b/main/.doctrees/getting-started/tutorials/03-matrix-multiplication.doctree
diff --git a/main/.doctrees/getting-started/tutorials/04-low-memory-dropout.doctree b/main/.doctrees/getting-started/tutorials/04-low-memory-dropout.doctree
diff --git a/main/.doctrees/getting-started/tutorials/05-layer-norm.doctree b/main/.doctrees/getting-started/tutorials/05-layer-norm.doctree
diff --git a/main/.doctrees/getting-started/tutorials/06-fused-attention.doctree b/main/.doctrees/getting-started/tutorials/06-fused-attention.doctree
diff --git a/main/.doctrees/getting-started/tutorials/07-extern-functions.doctree b/main/.doctrees/getting-started/tutorials/07-extern-functions.doctree
diff --git a/main/.doctrees/getting-started/tutorials/08-grouped-gemm.doctree b/main/.doctrees/getting-started/tutorials/08-grouped-gemm.doctree
diff --git a/main/.doctrees/getting-started/tutorials/09-persistent-matmul.doctree b/main/.doctrees/getting-started/tutorials/09-persistent-matmul.doctree
diff --git a/main/.doctrees/getting-started/tutorials/sg_execution_times.doctree b/main/.doctrees/getting-started/tutorials/sg_execution_times.doctree
diff --git a/main/.doctrees/sg_execution_times.doctree b/main/.doctrees/sg_execution_times.doctree
diff --git a/main/_downloads/662999063954282841dc90b8945f85ce/tutorials_jupyter.zip b/main/_downloads/662999063954282841dc90b8945f85ce/tutorials_jupyter.zip
diff --git a/main/_downloads/763344228ae6bc253ed1a6cf586aa30d/tutorials_python.zip b/main/_downloads/763344228ae6bc253ed1a6cf586aa30d/tutorials_python.zip
diff --git a/main/_images/sphx_glr_01-vector-add_001.png b/main/_images/sphx_glr_01-vector-add_001.png
diff --git a/main/_images/sphx_glr_01-vector-add_thumb.png b/main/_images/sphx_glr_01-vector-add_thumb.png
diff --git a/main/_images/sphx_glr_02-fused-softmax_001.png b/main/_images/sphx_glr_02-fused-softmax_001.png
diff --git a/main/_images/sphx_glr_02-fused-softmax_thumb.png b/main/_images/sphx_glr_02-fused-softmax_thumb.png
diff --git a/main/_images/sphx_glr_03-matrix-multiplication_001.png b/main/_images/sphx_glr_03-matrix-multiplication_001.png
diff --git a/main/_images/sphx_glr_03-matrix-multiplication_002.png b/main/_images/sphx_glr_03-matrix-multiplication_002.png
diff --git a/main/_images/sphx_glr_03-matrix-multiplication_thumb.png b/main/_images/sphx_glr_03-matrix-multiplication_thumb.png
diff --git a/main/_images/sphx_glr_05-layer-norm_001.png b/main/_images/sphx_glr_05-layer-norm_001.png
diff --git a/main/_images/sphx_glr_05-layer-norm_thumb.png b/main/_images/sphx_glr_05-layer-norm_thumb.png
diff --git a/main/_images/sphx_glr_06-fused-attention_001.png b/main/_images/sphx_glr_06-fused-attention_001.png
diff --git a/main/_images/sphx_glr_06-fused-attention_002.png b/main/_images/sphx_glr_06-fused-attention_002.png
diff --git a/main/_images/sphx_glr_06-fused-attention_003.png b/main/_images/sphx_glr_06-fused-attention_003.png
diff --git a/main/_images/sphx_glr_06-fused-attention_thumb.png b/main/_images/sphx_glr_06-fused-attention_thumb.png
diff --git a/main/_images/sphx_glr_08-grouped-gemm_001.png b/main/_images/sphx_glr_08-grouped-gemm_001.png
diff --git a/main/_images/sphx_glr_08-grouped-gemm_thumb.png b/main/_images/sphx_glr_08-grouped-gemm_thumb.png
diff --git a/main/_sources/getting-started/tutorials/01-vector-add.rst.txt b/main/_sources/getting-started/tutorials/01-vector-add.rst.txt
@@ -230,7 +230,7 @@ We can now run the decorated function above. Pass `print_data=True` to see the p
 
     vector-add-performance:
                size       Triton        Torch
-    0        4096.0     9.600000     8.000000
+    0        4096.0     8.000000     8.000000
     1        8192.0    15.999999    15.999999
     2       16384.0    31.999999    31.999999
     3       32768.0    63.999998    63.999998
@@ -240,20 +240,20 @@ We can now run the decorated function above. Pass `print_data=True` to see the p
     7      524288.0   614.400016   614.400016
     8     1048576.0   819.200021   819.200021
     9     2097152.0  1068.521715  1023.999964
-    10    4194304.0  1260.307736  1228.800031
-    11    8388608.0  1424.695621  1404.342820
+    10    4194304.0  1228.800031  1228.800031
+    11    8388608.0  1424.695621  1424.695621
     12   16777216.0  1560.380965  1560.380965
-    13   33554432.0  1624.859540  1624.859540
+    13   33554432.0  1631.601649  1624.859540
     14   67108864.0  1669.706983  1662.646960
-    15  134217728.0  1685.813499  1678.616907
+    15  134217728.0  1685.813499  1680.410210
 
 
 
 
 
 .. rst-class:: sphx-glr-timing
 
-   **Total running time of the script:** (0 minutes 6.108 seconds)
+   **Total running time of the script:** (0 minutes 7.255 seconds)
 
 
 .. _sphx_glr_download_getting-started_tutorials_01-vector-add.py:

diff --git a/main/_sources/getting-started/tutorials/02-fused-softmax.rst.txt b/main/_sources/getting-started/tutorials/02-fused-softmax.rst.txt
@@ -330,104 +330,104 @@ We will then compare its performance against (1) :code:`torch.softmax` and (2) t
 
     softmax-performance:
               N       Triton        Torch
-    0     256.0   465.899433   695.400871
-    1     384.0   599.815408   807.020248
-    2     512.0   739.459306   924.249184
-    3     640.0   784.700074   960.990973
-    4     768.0   874.487232  1030.789241
-    5     896.0   935.698473  1077.052131
-    6    1024.0   990.685459  1116.257327
-    7    1152.0  1110.135125   613.630925
-    8    1280.0  1154.193459   666.115844
-    9    1408.0  1157.517822   725.913001
-    10   1536.0  1195.835030   779.161847
-    11   1664.0  1219.348019   817.012769
-    12   1792.0  1239.633734   854.965309
-    13   1920.0  1248.100052   911.347452
-    14   2048.0  1270.927959   961.951095
-    15   2176.0  1264.593366   975.527805
-    16   2304.0  1265.499474  1005.167158
-    17   2432.0  1295.900421  1051.420386
-    18   2560.0  1298.256730  1087.002501
-    19   2688.0  1314.040612  1101.179374
-    20   2816.0  1329.702431  1132.560909
-    21   2944.0  1330.030843  1168.833710
-    22   3072.0  1353.756264  1186.699361
-    23   3200.0  1357.158394  1195.006563
-    24   3328.0  1361.807495  1227.066438
-    25   3456.0  1375.275072  1250.813561
-    26   3584.0  1376.282091  1262.264808
-    27   3712.0  1388.432948  1268.527151
-    28   3840.0  1382.295210  1301.933161
-    29   3968.0  1386.149360  1312.473127
-    30   4096.0  1400.629104  1327.721744
-    31   4224.0  1333.831746  1159.129769
-    32   4352.0  1333.760428  1174.726516
-    33   4480.0  1353.650060  1179.176446
-    34   4608.0  1365.886186  1196.894219
-    35   4736.0  1358.863901  1200.398175
-    36   4864.0  1377.793654  1223.731439
-    37   4992.0  1373.081556  1240.313833
-    38   5120.0  1374.984041  1253.357183
-    39   5248.0  1376.048059  1257.664222
-    40   5376.0  1376.865992  1288.188045
-    41   5504.0  1382.599188  1302.161870
-    42   5632.0  1383.617232  1316.247201
-    43   5760.0  1392.948579  1322.699760
-    44   5888.0  1386.522787  1339.839405
-    45   6016.0  1398.107365  1355.338813
-    46   6144.0  1407.684281  1373.820324
-    47   6272.0  1416.271235  1376.926267
-    48   6400.0  1414.938152  1388.249114
-    49   6528.0  1413.721245  1395.184663
-    50   6656.0  1418.293562  1400.944280
-    51   6784.0  1409.025437  1412.998752
-    52   6912.0  1427.728073  1423.071760
-    53   7040.0  1416.167999  1432.971798
-    54   7168.0  1427.689891  1432.707454
-    55   7296.0  1432.571811  1441.978963
-    56   7424.0  1430.504789  1446.438448
-    57   7552.0  1426.894191  1457.129560
-    58   7680.0  1435.643420  1461.041621
-    59   7808.0  1433.714451  1466.245230
-    60   7936.0  1433.318056  1470.350468
-    61   8064.0  1436.414976  1475.123931
-    62   8192.0  1436.322963  1485.552428
-    63   8320.0  1387.271085  1400.007827
-    64   8448.0  1381.020103  1403.634729
-    65   8576.0  1395.709564  1398.096531
-    66   8704.0  1389.154186  1401.751387
-    67   8832.0  1384.919913  1405.777492
-    68   8960.0  1396.810407  1412.159886
-    69   9088.0  1410.570222  1415.506069
-    70   9216.0  1405.948627  1423.613969
-    71   9344.0  1399.243649  1424.604302
-    72   9472.0  1400.665506  1433.733197
-    73   9600.0  1397.385811  1435.917574
-    74   9728.0  1401.198731  1444.541706
-    75   9856.0  1417.113861  1441.493538
-    76   9984.0  1401.726644  1448.798733
-    77  10112.0  1414.063853  1457.280277
-    78  10240.0  1420.382570  1464.215739
-    79  10368.0  1410.894543  1465.873460
-    80  10496.0  1416.552242  1466.261371
-    81  10624.0  1408.635952  1470.933291
-    82  10752.0  1408.357213  1472.254052
-    83  10880.0  1399.799498  1478.373967
-    84  11008.0  1417.983378  1476.518710
-    85  11136.0  1420.672542  1485.479428
-    86  11264.0  1429.026090  1490.464377
-    87  11392.0  1411.306562  1491.597518
-    88  11520.0  1422.292360  1492.171333
-    89  11648.0  1426.555471  1498.184979
-    90  11776.0  1431.373098  1501.448115
-    91  11904.0  1444.678821  1507.035636
-    92  12032.0  1423.355685  1508.957071
-    93  12160.0  1419.783945  1510.828124
-    94  12288.0  1437.730313  1393.186584
-    95  12416.0  1446.912130  1391.646552
-    96  12544.0  1443.867455  1393.235392
-    97  12672.0  1447.227434  1393.185222
+    0     256.0   471.287878   689.485126
+    1     384.0   607.545625   802.766850
+    2     512.0   746.784491   918.771676
+    3     640.0   793.480413   948.611927
+    4     768.0   884.637626  1029.962504
+    5     896.0   931.653153  1063.726176
+    6    1024.0   981.759456  1122.646617
+    7    1152.0  1100.875060   615.275475
+    8    1280.0  1141.917701   666.049406
+    9    1408.0  1156.342840   725.597348
+    10   1536.0  1193.636384   783.221369
+    11   1664.0  1219.491875   814.819044
+    12   1792.0  1240.884912   855.880679
+    13   1920.0  1257.616075   904.868615
+    14   2048.0  1279.760651   959.352173
+    15   2176.0  1258.709120   972.864439
+    16   2304.0  1271.887508  1010.792638
+    17   2432.0  1299.146020  1056.466170
+    18   2560.0  1300.453243  1087.550694
+    19   2688.0  1307.427065  1100.580784
+    20   2816.0  1322.679365  1131.909523
+    21   2944.0  1323.889816  1165.149120
+    22   3072.0  1352.515854  1182.128023
+    23   3200.0  1350.947949  1191.027380
+    24   3328.0  1359.986797  1222.943176
+    25   3456.0  1376.836647  1245.452967
+    26   3584.0  1375.922883  1262.100673
+    27   3712.0  1385.137457  1269.772683
+    28   3840.0  1389.818923  1302.266160
+    29   3968.0  1390.210050  1312.792128
+    30   4096.0  1396.772291  1328.375657
+    31   4224.0  1332.085989  1162.167274
+    32   4352.0  1333.404746  1171.657760
+    33   4480.0  1354.041044  1180.499634
+    34   4608.0  1359.857405  1191.384628
+    35   4736.0  1360.878093  1199.422703
+    36   4864.0  1373.823052  1222.240366
+    37   4992.0  1372.126365  1234.433031
+    38   5120.0  1373.796548  1254.616258
+    39   5248.0  1379.672985  1257.191000
+    40   5376.0  1378.696358  1284.925185
+    41   5504.0  1381.458718  1299.361110
+    42   5632.0  1385.904545  1316.323864
+    43   5760.0  1395.077124  1327.534593
+    44   5888.0  1391.064501  1342.593437
+    45   6016.0  1402.561115  1352.571423
+    46   6144.0  1409.757962  1372.639386
+    47   6272.0  1416.463957  1372.630933
+    48   6400.0  1418.186285  1387.020223
+    49   6528.0  1414.181115  1390.448491
+    50   6656.0  1424.306186  1404.314693
+    51   6784.0  1414.053887  1412.720164
+    52   6912.0  1423.921911  1422.520113
+    53   7040.0  1424.534830  1431.532648
+    54   7168.0  1424.864201  1433.826160
+    55   7296.0  1433.386447  1443.789405
+    56   7424.0  1435.268965  1448.241512
+    57   7552.0  1429.018827  1452.301968
+    58   7680.0  1435.122377  1461.254882
+    59   7808.0  1434.843432  1466.624970
+    60   7936.0  1437.454903  1470.040703
+    61   8064.0  1438.119365  1475.599827
+    62   8192.0  1440.187531  1486.127447
+    63   8320.0  1387.637449  1402.761094
+    64   8448.0  1378.102013  1402.115887
+    65   8576.0  1393.342881  1396.319369
+    66   8704.0  1393.177889  1401.658077
+    67   8832.0  1378.801324  1403.220841
+    68   8960.0  1398.504866  1413.312918
+    69   9088.0  1406.261956  1418.134932
+    70   9216.0  1402.007434  1423.911640
+    71   9344.0  1398.774623  1424.051601
+    72   9472.0  1397.199495  1436.025831
+    73   9600.0  1395.327256  1434.863618
+    74   9728.0  1402.479216  1440.674074
+    75   9856.0  1416.135151  1444.858741
+    76   9984.0  1400.280488  1452.675258
+    77  10112.0  1413.443656  1455.099380
+    78  10240.0  1420.010134  1467.223957
+    79  10368.0  1410.372273  1462.363919
+    80  10496.0  1411.274735  1464.723823
+    81  10624.0  1405.199533  1468.494233
+    82  10752.0  1400.715482  1472.490151
+    83  10880.0  1398.350423  1481.582102
+    84  11008.0  1419.884062  1475.731901
+    85  11136.0  1420.550901  1487.223937
+    86  11264.0  1428.925516  1484.583695
+    87  11392.0  1415.668954  1490.090298
+    88  11520.0  1422.613285  1495.273189
+    89  11648.0  1426.616650  1499.586431
+    90  11776.0  1429.276053  1500.305781
+    91  11904.0  1442.513902  1504.375391
+    92  12032.0  1424.425254  1509.798144
+    93  12160.0  1416.570364  1512.312839
+    94  12288.0  1434.724163  1392.188715
+    95  12416.0  1446.856771  1389.922048
+    96  12544.0  1440.710123  1392.749310
+    97  12672.0  1447.636431  1393.181321
 
 
 
@@ -442,7 +442,7 @@ In the above plot, we can see that:
 
 .. rst-class:: sphx-glr-timing
 
-   **Total running time of the script:** (0 minutes 27.634 seconds)
+   **Total running time of the script:** (0 minutes 23.398 seconds)
 
 
 .. _sphx_glr_download_getting-started_tutorials_02-fused-softmax.py:

diff --git a/main/_sources/getting-started/tutorials/03-matrix-multiplication.rst.txt b/main/_sources/getting-started/tutorials/03-matrix-multiplication.rst.txt
@@ -569,69 +569,69 @@ but feel free to arrange this script as you wish to benchmark any other matrix s
 
     matmul-performance-fp16:
              M       N       K      cuBLAS      Triton
-    0    256.0   256.0   256.0    4.096000    4.096000
+    0    256.0   256.0   256.0    4.096000    3.640889
     1    384.0   384.0   384.0   12.288000   12.288000
     2    512.0   512.0   512.0   26.214401   26.214401
     3    640.0   640.0   640.0   42.666665   42.666665
-    4    768.0   768.0   768.0   63.195428   68.056616
+    4    768.0   768.0   768.0   63.195428   63.195428
     5    896.0   896.0   896.0   78.051553   87.808000
     6   1024.0  1024.0  1024.0  110.376426   99.864382
     7   1152.0  1152.0  1152.0  135.726544  129.825388
     8   1280.0  1280.0  1280.0  157.538463  163.840004
     9   1408.0  1408.0  1408.0  155.765024  132.970149
-    10  1536.0  1536.0  1536.0  176.947204  157.286398
+    10  1536.0  1536.0  1536.0  176.947204  153.867127
     11  1664.0  1664.0  1664.0  179.978245  179.978245
-    12  1792.0  1792.0  1792.0  172.914215  204.353162
+    12  1792.0  1792.0  1792.0  172.914215  208.137481
     13  1920.0  1920.0  1920.0  200.347822  166.554219
     14  2048.0  2048.0  2048.0  226.719125  190.650180
-    15  2176.0  2176.0  2176.0  211.827867  207.460296
-    16  2304.0  2304.0  2304.0  229.691080  227.503545
-    17  2432.0  2432.0  2432.0  205.069087  200.674737
-    18  2560.0  2560.0  2560.0  223.672340  218.453323
-    19  2688.0  2688.0  2688.0  196.544332  198.602388
-    20  2816.0  2816.0  2816.0  213.795141  211.719459
-    21  2944.0  2944.0  2944.0  220.513412  221.493479
-    22  3072.0  3072.0  3072.0  208.941345  210.494802
-    23  3200.0  3200.0  3200.0  216.216207  217.687077
-    24  3328.0  3328.0  3328.0  207.467716  208.067338
-    25  3456.0  3456.0  3456.0  217.308808  217.308808
-    26  3584.0  3584.0  3584.0  216.142772  212.565943
-    27  3712.0  3712.0  3712.0  210.753890  215.761000
-    28  3840.0  3840.0  3840.0  207.489687  208.271176
-    29  3968.0  3968.0  3968.0  210.386099  216.738793
-    30  4096.0  4096.0  4096.0  219.668951  219.310012
+    15  2176.0  2176.0  2176.0  211.827867  209.621326
+    16  2304.0  2304.0  2304.0  229.691080  229.691080
+    17  2432.0  2432.0  2432.0  203.583068  200.674737
+    18  2560.0  2560.0  2560.0  224.438347  218.453323
+    19  2688.0  2688.0  2688.0  198.602388  195.531224
+    20  2816.0  2816.0  2816.0  211.719459  209.683695
+    21  2944.0  2944.0  2944.0  218.579083  215.740400
+    22  3072.0  3072.0  3072.0  206.653671  211.280236
+    23  3200.0  3200.0  3200.0  214.765101  219.178074
+    24  3328.0  3328.0  3328.0  207.467716  205.689424
+    25  3456.0  3456.0  3456.0  214.419058  217.602074
+    26  3584.0  3584.0  3584.0  218.772251  214.595213
+    27  3712.0  3712.0  3712.0  210.310194  218.116474
+    28  3840.0  3840.0  3840.0  208.271176  210.651436
+    29  3968.0  3968.0  3968.0  210.931616  215.971570
+    30  4096.0  4096.0  4096.0  220.029067  219.130982
     matmul-performance-fp8:
              M       N       K      Triton
-    0    256.0   256.0   256.0    2.978909
+    0    256.0   256.0   256.0    3.276800
     1    384.0   384.0   384.0    9.216000
     2    512.0   512.0   512.0   20.164923
     3    640.0   640.0   640.0   34.133334
-    4    768.0   768.0   768.0   42.130286
+    4    768.0   768.0   768.0   40.215272
     5    896.0   896.0   896.0   58.538665
     6   1024.0  1024.0  1024.0   61.680940
     7   1152.0  1152.0  1152.0   80.702267
-    8   1280.0  1280.0  1280.0  102.400003
+    8   1280.0  1280.0  1280.0   99.902441
     9   1408.0  1408.0  1408.0   82.602666
     10  1536.0  1536.0  1536.0   99.688560
-    11  1664.0  1664.0  1664.0  116.868992
+    11  1664.0  1664.0  1664.0  115.370671
     12  1792.0  1792.0  1792.0  135.414749
     13  1920.0  1920.0  1920.0   99.453240
     14  2048.0  2048.0  2048.0  113.359563
-    15  2176.0  2176.0  2176.0  120.500882
+    15  2176.0  2176.0  2176.0  121.226797
     16  2304.0  2304.0  2304.0  134.959733
     17  2432.0  2432.0  2432.0  131.898888
-    18  2560.0  2560.0  2560.0  146.941707
+    18  2560.0  2560.0  2560.0  146.285712
     19  2688.0  2688.0  2688.0  117.804519
-    20  2816.0  2816.0  2816.0  129.036114
+    20  2816.0  2816.0  2816.0  129.419013
     21  2944.0  2944.0  2944.0  139.596724
-    22  3072.0  3072.0  3072.0  143.713461
-    23  3200.0  3200.0  3200.0  138.828637
-    24  3328.0  3328.0  3328.0  131.370982
-    25  3456.0  3456.0  3456.0  138.287420
-    26  3584.0  3584.0  3584.0  149.361120
-    27  3712.0  3712.0  3712.0  141.297511
+    22  3072.0  3072.0  3072.0  144.079147
+    23  3200.0  3200.0  3200.0  139.433550
+    24  3328.0  3328.0  3328.0  130.893266
+    25  3456.0  3456.0  3456.0  138.763456
+    26  3584.0  3584.0  3584.0  149.858980
+    27  3712.0  3712.0  3712.0  139.716570
     28  3840.0  3840.0  3840.0  138.067418
-    29  3968.0  3968.0  3968.0  145.439735
+    29  3968.0  3968.0  3968.0  145.613293
     30  4096.0  4096.0  4096.0  155.165002
 
 
@@ -640,7 +640,7 @@ but feel free to arrange this script as you wish to benchmark any other matrix s
 
 .. rst-class:: sphx-glr-timing
 
-   **Total running time of the script:** (2 minutes 16.816 seconds)
+   **Total running time of the script:** (2 minutes 16.568 seconds)
 
 
 .. _sphx_glr_download_getting-started_tutorials_03-matrix-multiplication.py:

diff --git a/main/_sources/getting-started/tutorials/04-low-memory-dropout.rst.txt b/main/_sources/getting-started/tutorials/04-low-memory-dropout.rst.txt
@@ -242,7 +242,7 @@ References
 
 .. rst-class:: sphx-glr-timing
 
-   **Total running time of the script:** (0 minutes 0.679 seconds)
+   **Total running time of the script:** (0 minutes 0.689 seconds)
 
 
 .. _sphx_glr_download_getting-started_tutorials_04-low-memory-dropout.py: