Closed alvarozamora closed 3 years ago
Hi,
I seem to build successfully via cmake -DCMAKE_BUILD_TYPE=Release -DCUDA_ARCH=70 -DDISABLE_R=y . && makebut I've run into two issues that suggest otherwise. Firstly, when I run tests.py, It seems to get stuck after what I think is the first test:
cmake -DCMAKE_BUILD_TYPE=Release -DCUDA_ARCH=70 -DDISABLE_R=y . && make
>>> python test.py arguments: 1 0x7ffdac27d940 0.010 0.10 1 1000 256 10 3 0 0 3 0x55d0af0577d0 0x55d0aed8f320 0x55d0aed91b30 (nil) reassignments threshold: 10 yinyang groups: 1 [0] *dest: 0x7f55d5c00000 - 0x7f55d5cfa000 (1024000) [0] device_centroids: 0x7f55d5cfa000 - 0x7f55d5cfc800 (10240) [0] device_assignments: 0x7f55d5cfc800 - 0x7f55d5cfd7a0 (4000) [0] device_assignments_prev: 0x7f55d5cfd800 - 0x7f55d5cfe7a0 (4000) [0] device_ccounts: 0x7f55d5cfe800 - 0x7f55d5cfe828 (40) [0] device_assignments_yy: 0x7f55d5cfea00 - 0x7f55d5cfea28 (40) [0] device_bounds_yy: 0x7f55d5cfec00 - 0x7f55d5d00b40 (8000) [0] device_drifts_yy: 0x7f55d5d00c00 - 0x7f55d5d03428 (10280) [0] device_passed_yy: 0x7f55d5d03600 - 0x7f55d5d045a0 (4000) reusing passed_yy for centroids_yy GPU #0 memory: used 185466880 bytes (1.6%), free 11369250816 bytes, total 11554717696 bytes GPU #0 has 49152 bytes of shared memory per block transposing the samples... transpose <<<(32, 8), (8, 32)>>> 1000, 256, xyswap performing kmeans++... kmeans++: dump 1000 256 0x55d0af994d70 kmeans++: dev #0: 0x7f55d5c00000 0x7f55d5cfa000 0x7f55d5cfc800 step 1[0] dev_dists: 0x7f55d5d04600 - 0x7f55d5d04640 (64) step 2[0] dev_dists: 0x7f55d5d04600 - 0x7f55d5d04640 (64) step 3[0] dev_dists: 0x7f55d5d04600 - 0x7f55d5d04640 (64) step 4[0] dev_dists: 0x7f55d5d04600 - 0x7f55d5d04640 (64) step 5[0] dev_dists: 0x7f55d5d04600 - 0x7f55d5d04640 (64) step 6[0] dev_dists: 0x7f55d5d04600 - 0x7f55d5d04640 (64) step 7[0] dev_dists: 0x7f55d5d04600 - 0x7f55d5d04640 (64) step 8[0] dev_dists: 0x7f55d5d04600 - 0x7f55d5d04640 (64) step 9[0] dev_dists: 0x7f55d5d04600 - 0x7f55d5d04640 (64) done running Lloyd until reassignments drop below 110 plans: [(0, 1000)] planc: [(0, 10)] iteration 1: 1000 reassignments iteration 2: 142 reassignments iteration 3: 40 reassignments transposing the samples... transpose <<<(8, 1), (32, 8)>>> 10, 256 performing kmeans++... kmeans++: dump 10 256 0x55d0af8d43f0 kmeans++: dev #0: 0x7f55d5cfa000 0x7f55d5d03600 0x7f55d5d04574 done plans: [(0, 10)] planc: [(0, 1)] iteration 1: 10 reassignments iteration 2: 0 reassignments transposing the samples... transpose <<<(8, 1), (8, 32)>>> 256, 10, xyswap plans: [(0, 1000)] planc: [(0, 10)] plang: [(0, 1)] refreshing Yinyang bounds... iteration 4: 27 reassignments passed number: 1000 refreshing Yinyang bounds... iteration 5: 11 reassignments passed number: 1000 refreshing Yinyang bounds... iteration 6: 11 reassignments passed number: 1000 refreshing Yinyang bounds... iteration 7: 7 reassignments return kmcudaSuccess .
Secondly, when I run the kNN python test on the main GitHub page, I get null results:
>>> python kNNtest.py reassignments threshold: 100 transposing the samples... performing kmeans++... done too few clusters for this yinyang_t => Lloyd iteration 1: 10000 reassignments iteration 2: 926 reassignments iteration 3: 416 reassignments iteration 4: 187 reassignments iteration 5: 87 reassignments transposing the samples... initializing the inverse assignments... calculating the cluster radiuses... calculating the centroid distance matrix... searching for the nearest neighbors... calculated 0.000000 of all the distances [0 0 0 0 0 0 0 0 0 0]
Any suggestions?
Hi,
I seem to build successfully via
cmake -DCMAKE_BUILD_TYPE=Release -DCUDA_ARCH=70 -DDISABLE_R=y . && make
but I've run into two issues that suggest otherwise. Firstly, when I run tests.py, It seems to get stuck after what I think is the first test:Secondly, when I run the kNN python test on the main GitHub page, I get null results:
Any suggestions?