Skip to content

Instantly share code, notes, and snippets.

@masahi

masahi/bert_large_gelu_fusion_autotvm.txt Secret

Created November 3, 2021 18:54
Show Gist options
  • Save masahi/45ac7c45b637c2f3e4c35f8db11e9c88 to your computer and use it in GitHub Desktop.
Save masahi/45ac7c45b637c2f3e4c35f8db11e9c88 to your computer and use it in GitHub Desktop.
Download ZIP
Raw
bert_large_gelu_fusion_autotvm.txt
[[-0.3037 -0.4268 ]
[-0.073 -0.3354 ]
[-0.3523 -0.3027 ]
[-0.4697 0.1609 ]
[-0.10986 -0.3613 ]
[ 0.666 0.285 ]
[-0.1691 -0.2268 ]
[-0.03015 -0.3994 ]]
Evaluate inference time cost...
Execution time summary:
mean (ms) median (ms) max (ms) min (ms) std (ms)
23.6729 23.5176 24.7284 23.3980 0.3317
CUDA Kernel Statistics:
Time(%) Total Time (ns) Instances Average Minimum Maximum Name
------- --------------- --------- --------- ------- ------- --------------------------------------------------------------------------------------
15.9 201,802,100 4,992 40,425.1 38,593 47,617 tvmgen_default_fused_nn_dense_1_kernel0
15.2 192,699,034 1,248 154,406.3 149,761 164,929 tvmgen_default_fused_nn_dense_add_multiply_cast_erf_cast_multiply_add_multiply_kernel0
14.0 177,696,875 1,248 142,385.3 138,241 153,441 tvmgen_default_fused_nn_dense_kernel0
7.9 99,586,100 1,248 79,796.6 77,920 84,129 tvmgen_default_fused_nn_softmax_kernel0
7.9 99,558,270 1,248 79,774.3 77,856 84,065 tvmgen_default_fused_nn_softmax_kernel2
5.1 64,498,082 2,496 25,840.6 24,351 31,232 tvmgen_default_fused_reshape_add_cast_add_kernel0
4.2 52,629,168 1,248 42,170.8 41,472 44,704 tvmgen_default_fused_nn_softmax_kernel3
3.8 48,552,637 2,496 19,452.2 18,752 25,952 tvmgen_default_fused_subtract_add_sqrt_divide_multiply_add_kernel0
3.7 46,604,136 1,248 37,343.1 36,128 42,368 tvmgen_default_fused_nn_softmax_kernel1
3.2 41,113,090 1,248 32,943.2 32,128 38,880 tvmgen_default_fused_reshape_cast_1_kernel0
3.2 39,949,270 1,248 32,010.6 30,432 36,320 tvmgen_default_fused_reshape_divide_cast_add_kernel0
3.0 38,082,089 2,496 15,257.2 14,177 19,936 tvmgen_default_fused_reshape_cast_kernel0
2.1 26,883,334 1,248 21,541.1 21,024 28,256 tvmgen_default_fused_nn_batch_matmul_kernel0
2.0 25,777,344 1,248 20,654.9 19,712 25,632 tvmgen_default_fused_nn_batch_matmul_1_kernel0
2.0 25,250,108 2,548 9,909.8 8,575 15,008 tvmgen_default_fused_mean_kernel0
1.9 23,486,210 1,248 18,819.1 18,304 24,992 tvmgen_default_fused_reshape_add_reshape_transpose_reshape_1_kernel0
1.7 22,075,656 2,496 8,844.4 8,160 13,536 tvmgen_default_fused_reshape_add_reshape_transpose_reshape_kernel0
1.2 15,043,429 2,548 5,904.0 5,441 12,896 tvmgen_default_fused_variance_kernel0
1.0 12,118,032 1,248 9,710.0 9,280 14,081 tvmgen_default_fused_reshape_transpose_reshape_kernel0
0.5 5,915,269 2,548 2,321.5 2,113 7,328 tvmgen_default_fused_mean_kernel1
0.4 5,474,084 2,548 2,148.4 1,952 12,384 tvmgen_default_fused_variance_kernel1
0.1 664,872 52 12,786.0 12,161 14,176 tvmgen_default_fused_cast_take_cast_take_add_add_kernel0
0.0 527,014 52 10,134.9 9,889 10,752 tvmgen_default_fused_nn_dense_add_tanh_kernel0
0.0 149,056 52 2,866.5 2,720 3,232 tvmgen_default_fused_expand_dims_expand_dims_cast_subtract_multiply_kernel0
0.0 147,492 52 2,836.4 2,752 2,976 tvmgen_default_fused_subtract_add_sqrt_divide_multiply_add_take_cast_kernel0
0.0 141,373 52 2,718.7 2,559 3,103 tvmgen_default_fused_nn_dense_add_kernel0
Sign up for free to join this conversation on GitHub. Already have an account? Sign in to comment