tkonolige commented on pull request #6671: URL: https://github.com/apache/incubator-tvm/pull/6671#issuecomment-717601781
Here is the tests I've run on a 64-core machine: ``` This PR XGB iter: 0 tr-a-recall@64: 0.598716 tr-map: 0.142857 XGB iter: 25 tr-a-recall@64: 0.667324 tr-map: 0.500000 XGB stopped. Best iteration: [13] tr-a-recall@64:0.67690 tr-map:0.50000 XGB train: 1.24 obs: 64 error: 50 n_cache: 64 SA iter: 50 last_update: 49 max-0: 4.42 max-1: 4.71 temp: 0.90 elapsed: 16.14 SA iter: 100 last_update: 97 max-0: 4.55 max-1: 4.71 temp: 0.80 elapsed: 30.59 SA iter: 150 last_update: 145 max-0: 4.56 max-1: 4.74 temp: 0.70 elapsed: 45.74 SA iter: 200 last_update: 197 max-0: 4.58 max-1: 4.83 temp: 0.60 elapsed: 60.61 SA iter: 250 last_update: 249 max-0: 4.67 max-1: 4.84 temp: 0.50 elapsed: 76.09 SA iter: 300 last_update: 299 max-0: 4.69 max-1: 4.84 temp: 0.40 elapsed: 90.07 SA iter: 350 last_update: 349 max-0: 4.71 max-1: 4.84 temp: 0.30 elapsed: 104.27 SA iter: 400 last_update: 377 max-0: 4.71 max-1: 4.84 temp: 0.20 elapsed: 118.12 SA iter: 450 last_update: 448 max-0: 4.72 max-1: 4.84 temp: 0.10 elapsed: 130.86 SA iter: 500 last_update: 498 max-0: 4.76 max-1: 4.84 temp: 0.00 elapsed: 143.39 SA iter: 500 last_update: 498 elapsed: 143.39 main XGB iter: 0 tr-a-recall@64: 0.641594 tr-map: 0.500000 XGB iter: 25 tr-a-recall@64: 0.716034 tr-map: 0.500000 XGB iter: 50 tr-a-recall@64: 0.728245 tr-map: 1.000000 XGB stopped. Best iteration: [38] tr-a-recall@64:0.73020 tr-map:0.50000 XGB train: 1.10 obs: 64 error: 46 n_cache: 64 SA iter: 50 last_update: 49 max-0: 6.21 max-1: 6.94 temp: 0.90 elapsed: 3.40 SA iter: 100 last_update: 99 max-0: 6.50 max-1: 7.16 temp: 0.80 elapsed: 6.68 SA iter: 150 last_update: 149 max-0: 6.74 max-1: 7.38 temp: 0.70 elapsed: 10.24 SA iter: 200 last_update: 199 max-0: 6.86 max-1: 7.39 temp: 0.60 elapsed: 13.91 SA iter: 250 last_update: 249 max-0: 6.97 max-1: 7.39 temp: 0.50 elapsed: 17.68 SA iter: 300 last_update: 294 max-0: 7.07 max-1: 7.56 temp: 0.40 elapsed: 21.29 SA iter: 350 last_update: 348 max-0: 7.11 max-1: 7.56 temp: 0.30 elapsed: 24.52 SA iter: 400 last_update: 399 max-0: 7.15 max-1: 7.56 temp: 0.20 elapsed: 27.68 SA iter: 450 last_update: 449 max-0: 7.16 max-1: 7.56 temp: 0.10 elapsed: 30.63 SA iter: 500 last_update: 499 max-0: 7.17 max-1: 7.56 temp: 0.00 elapsed: 33.48 SA iter: 500 last_update: 499 elapsed: 33.48 ``` Seems like the issue is hit when there are a lot of cores. I've pulled out the autotvm parts and will send another pr when I figure out those. ---------------------------------------------------------------- This is an automated message from the Apache Git Service. To respond to the message, please log on to GitHub and use the URL above to go to the specific comment. For queries about this service, please contact Infrastructure at: us...@infra.apache.org