[Relay][TOPI] Gluncv SSD support on the GPU#2784
Conversation
|
Overall lgtm. One more thing is what is the situation of gpu related grammar sugars supporting in hybrid script? @were If possible we can migrate vision op to hybrid script implementation as well. |
Currently, I'm still experimenting with the hybrid script. There could still be some problems. But I'll communicate with @were during the process. |
83d1ab2 to
b66ff61
Compare
d6830a2 to
02855fa
Compare
a4789e8 to
a7b5d9e
Compare
|
LGTM. @Laurawly Can you open an issue to track the progress to migrate from ir_builder to hybrid script? |
icemelon
left a comment
There was a problem hiding this comment.
Some high level comment:
argsortdoesn't consider output dtype as int- we shouldn't place sort-related op under vision category
icemelon
left a comment
There was a problem hiding this comment.
Could you also update the doc?
| return _default_schedule(outs, False) | ||
|
|
||
| @tvm.target.generic_func | ||
| def schedule_argsort(outs): |
There was a problem hiding this comment.
move this function to topi/generic/sort.py
|
|
||
|
|
||
| **Level 6: Algorithm Operators** | ||
| ------------------- |
|
@Laurawly please act on @icemelon9 's remaining comment |
Comments addressed. |
|
Thanks everyone. This is now merged. |
* ssd gluoncv gpu op updated * ssd gluoncv gpu op updated * tutorials and testes modified * tutorials and testes modified * fix lint * fix lint * address comment * multibox bug fixed * space line added * use less threads per block * use less threads per block * less threads per block for get valid count * less threads per block for get valid count * merge with master * Revert "less threads per block for get valid count" This reverts commit 08896cf. * Revert "less threads per block for get valid count" This reverts commit 08896cf. * typo fixed * elem length made to a variable * fix lint error * fix lint error * lint fixed * bug fixed * bug fixed * lint fixed * error fixed * error fixed * test ci * test ci * seperate argsort to be an independent op * seperate argsort to be an independent op * fix lint * fix lint * remove unsupported models * typo fixed * argsort added to realy * solve conflicts with master * fix lint * fix lint * test push * Revert "test push" This reverts commit 6db0088. * fix lint error * fix more lint * cpu test_sort udpated * debug ci * nms fixed * expose argsort to relay frontend * test ci * fix lint * sort register error fixed * fix nnvm * nms type fixed * adaptive pooling added to relay * Revert "adaptive pooling added to relay" This reverts commit 1119f1f. * fix lint * expose argsort op * fix lint * fix lint * fix lint * sort test updated * sort bug fixed * nnvm error fixed * fix argsort default data type returned to be float insteaf of int * fix lint * fix lint * test fixed * fix valid count * fix titanx bug * tutorial add both targets * titanx error fixed * try to fix CI old gpu error * try to solve CI GPU error * get_valid_count added * reverse get_valid_count * get valid count optimized * address comments * fix ci error * remove unessesary block sync * add back one sync * address comments * address more comments * more comments * move sort to be indepent algorithm * typo fixed * more typos * comments addressed * doc updated * fix pylint * address final comments * apache license added
* ssd gluoncv gpu op updated * ssd gluoncv gpu op updated * tutorials and testes modified * tutorials and testes modified * fix lint * fix lint * address comment * multibox bug fixed * space line added * use less threads per block * use less threads per block * less threads per block for get valid count * less threads per block for get valid count * merge with master * Revert "less threads per block for get valid count" This reverts commit 08896cf. * Revert "less threads per block for get valid count" This reverts commit 08896cf. * typo fixed * elem length made to a variable * fix lint error * fix lint error * lint fixed * bug fixed * bug fixed * lint fixed * error fixed * error fixed * test ci * test ci * seperate argsort to be an independent op * seperate argsort to be an independent op * fix lint * fix lint * remove unsupported models * typo fixed * argsort added to realy * solve conflicts with master * fix lint * fix lint * test push * Revert "test push" This reverts commit 6db0088. * fix lint error * fix more lint * cpu test_sort udpated * debug ci * nms fixed * expose argsort to relay frontend * test ci * fix lint * sort register error fixed * fix nnvm * nms type fixed * adaptive pooling added to relay * Revert "adaptive pooling added to relay" This reverts commit 1119f1f. * fix lint * expose argsort op * fix lint * fix lint * fix lint * sort test updated * sort bug fixed * nnvm error fixed * fix argsort default data type returned to be float insteaf of int * fix lint * fix lint * test fixed * fix valid count * fix titanx bug * tutorial add both targets * titanx error fixed * try to fix CI old gpu error * try to solve CI GPU error * get_valid_count added * reverse get_valid_count * get valid count optimized * address comments * fix ci error * remove unessesary block sync * add back one sync * address comments * address more comments * more comments * move sort to be indepent algorithm * typo fixed * more typos * comments addressed * doc updated * fix pylint * address final comments * apache license added
Following PR #2353, applying apis accordingly, support glunocv ssd models on the GPU.
Also exposed
coord_start,score_indexvariables to frontend for supporting more object detection networks. Addedargsortas an operator, support both CPU and GPU. Adaptive pooling op exposed.