Skip to content

Instantly share code, notes, and snippets.

@shunting314

shunting314/gist:609eb9d6b60a84379f8c3b743a82d38a

Created July 3, 2024 22:41
Show Gist options
  • Save shunting314/609eb9d6b60a84379f8c3b743a82d38a to your computer and use it in GitHub Desktop.
Save shunting314/609eb9d6b60a84379f8c3b743a82d38a to your computer and use it in GitHub Desktop.
Download ZIP
Raw
gistfile0.txt
(pytorch) [shunting@devgpu005.nha1 ~/ws/pytorch (acc-sebotnet33ts_256)]$ TORCHINDUCTOR_MAX_AUTOTUNE=1 time python benchmarks/dynamo/huggingface.py --accuracy --no-translation-validation --training --amp --backend inductor --device cuda --only LayoutLMForMaskedLM
loading model: 0it [00:06, ?it/s]
cuda train LayoutLMForMaskedLM
AUTOTUNE addmm(512x3072, 512x768, 768x3072)
triton_mm_130 0.0209 ms 100.0% ACC_TYPE='tl.float32', ALLOW_TF32=False, BLOCK_K=64, BLOCK_M=128, BLOCK_N=128, B_PROLOGUE_CAST_TYPE=None, EVEN_K=True, GROUP_M=8, num_stages=3, num_warps=4
triton_mm_131 0.0217 ms 96.5% ACC_TYPE='tl.float32', ALLOW_TF32=False, BLOCK_K=64, BLOCK_M=128, BLOCK_N=128, B_PROLOGUE_CAST_TYPE=None, EVEN_K=True, GROUP_M=8, num_stages=5, num_warps=8
triton_mm_124 0.0227 ms 92.2% ACC_TYPE='tl.float32', ALLOW_TF32=False, BLOCK_K=64, BLOCK_M=64, BLOCK_N=128, B_PROLOGUE_CAST_TYPE=None, EVEN_K=True, GROUP_M=8, num_stages=3, num_warps=4
triton_mm_129 0.0240 ms 87.2% ACC_TYPE='tl.float32', ALLOW_TF32=False, BLOCK_K=32, BLOCK_M=128, BLOCK_N=128, B_PROLOGUE_CAST_TYPE=None, EVEN_K=True, GROUP_M=8, num_stages=3, num_warps=4
triton_mm_122 0.0242 ms 86.5% ACC_TYPE='tl.float32', ALLOW_TF32=False, BLOCK_K=32, BLOCK_M=64, BLOCK_N=128, B_PROLOGUE_CAST_TYPE=None, EVEN_K=True, GROUP_M=8, num_stages=3, num_warps=4
triton_mm_126 0.0252 ms 83.0% ACC_TYPE='tl.float32', ALLOW_TF32=False, BLOCK_K=32, BLOCK_M=128, BLOCK_N=64, B_PROLOGUE_CAST_TYPE=None, EVEN_K=True, GROUP_M=8, num_stages=3, num_warps=4
triton_mm_127 0.0252 ms 83.0% ACC_TYPE='tl.float32', ALLOW_TF32=False, BLOCK_K=32, BLOCK_M=128, BLOCK_N=64, B_PROLOGUE_CAST_TYPE=None, EVEN_K=True, GROUP_M=8, num_stages=4, num_warps=8
triton_mm_123 0.0257 ms 81.5% ACC_TYPE='tl.float32', ALLOW_TF32=False, BLOCK_K=32, BLOCK_M=64, BLOCK_N=128, B_PROLOGUE_CAST_TYPE=None, EVEN_K=True, GROUP_M=8, num_stages=4, num_warps=8
triton_mm_125 0.0284 ms 73.6% ACC_TYPE='tl.float32', ALLOW_TF32=False, BLOCK_K=128, BLOCK_M=64, BLOCK_N=128, B_PROLOGUE_CAST_TYPE=None, EVEN_K=True, GROUP_M=8, num_stages=4, num_warps=4
triton_mm_120 0.0285 ms 73.4% ACC_TYPE='tl.float32', ALLOW_TF32=False, BLOCK_K=64, BLOCK_M=64, BLOCK_N=64, B_PROLOGUE_CAST_TYPE=None, EVEN_K=True, GROUP_M=8, num_stages=3, num_warps=8
SingleProcess AUTOTUNE benchmarking takes 2.5020 seconds and 0.0020 seconds precompiling
AUTOTUNE addmm(512x768, 512x768, 768x768)
triton_mm_8 0.0134 ms 100.0% ACC_TYPE='tl.float32', ALLOW_TF32=False, BLOCK_K=128, BLOCK_M=64, BLOCK_N=64, B_PROLOGUE_CAST_TYPE=None, EVEN_K=True, GROUP_M=8, num_stages=5, num_warps=4
triton_mm_7 0.0152 ms 88.0% ACC_TYPE='tl.float32', ALLOW_TF32=False, BLOCK_K=64, BLOCK_M=64, BLOCK_N=64, B_PROLOGUE_CAST_TYPE=None, EVEN_K=True, GROUP_M=8, num_stages=3, num_warps=8
triton_mm_12 0.0157 ms 85.5% ACC_TYPE='tl.float32', ALLOW_TF32=False, BLOCK_K=128, BLOCK_M=64, BLOCK_N=128, B_PROLOGUE_CAST_TYPE=None, EVEN_K=True, GROUP_M=8, num_stages=4, num_warps=4
triton_mm_11 0.0160 ms 83.6% ACC_TYPE='tl.float32', ALLOW_TF32=False, BLOCK_K=64, BLOCK_M=64, BLOCK_N=128, B_PROLOGUE_CAST_TYPE=None, EVEN_K=True, GROUP_M=8, num_stages=3, num_warps=4
triton_mm_4 0.0170 ms 78.8% ACC_TYPE='tl.float32', ALLOW_TF32=False, BLOCK_K=128, BLOCK_M=64, BLOCK_N=32, B_PROLOGUE_CAST_TYPE=None, EVEN_K=True, GROUP_M=8, num_stages=5, num_warps=4
triton_mm_3 0.0171 ms 78.5% ACC_TYPE='tl.float32', ALLOW_TF32=False, BLOCK_K=32, BLOCK_M=64, BLOCK_N=32, B_PROLOGUE_CAST_TYPE=None, EVEN_K=True, GROUP_M=8, num_stages=5, num_warps=8
triton_mm_10 0.0172 ms 77.7% ACC_TYPE='tl.float32', ALLOW_TF32=False, BLOCK_K=32, BLOCK_M=64, BLOCK_N=128, B_PROLOGUE_CAST_TYPE=None, EVEN_K=True, GROUP_M=8, num_stages=4, num_warps=8
triton_mm_14 0.0172 ms 77.7% ACC_TYPE='tl.float32', ALLOW_TF32=False, BLOCK_K=32, BLOCK_M=128, BLOCK_N=64, B_PROLOGUE_CAST_TYPE=None, EVEN_K=True, GROUP_M=8, num_stages=4, num_warps=8
triton_mm_1 0.0175 ms 76.6% ACC_TYPE='tl.float32', ALLOW_TF32=False, BLOCK_K=128, BLOCK_M=32, BLOCK_N=32, B_PROLOGUE_CAST_TYPE=None, EVEN_K=True, GROUP_M=8, num_stages=2, num_warps=4
triton_mm_2 0.0177 ms 75.6% ACC_TYPE='tl.float32', ALLOW_TF32=False, BLOCK_K=32, BLOCK_M=32, BLOCK_N=64, B_PROLOGUE_CAST_TYPE=None, EVEN_K=True, GROUP_M=8, num_stages=5, num_warps=8
SingleProcess AUTOTUNE benchmarking takes 2.4985 seconds and 0.0017 seconds precompiling
AUTOTUNE bmm(12x512x64, 12x64x512)
triton_bmm_73 0.0123 ms 100.0% ACC_TYPE='tl.float32', ALLOW_TF32=False, BLOCK_K=32, BLOCK_M=128, BLOCK_N=128, B_PROLOGUE_CAST_TYPE=None, EVEN_K=True, GROUP_M=8, num_stages=3, num_warps=4
triton_bmm_69 0.0130 ms 94.6% ACC_TYPE='tl.float32', ALLOW_TF32=False, BLOCK_K=64, BLOCK_M=64, BLOCK_N=128, B_PROLOGUE_CAST_TYPE=None, EVEN_K=True, GROUP_M=8, num_stages=4, num_warps=4
triton_bmm_70 0.0132 ms 92.8% ACC_TYPE='tl.float32', ALLOW_TF32=False, BLOCK_K=32, BLOCK_M=128, BLOCK_N=64, B_PROLOGUE_CAST_TYPE=None, EVEN_K=True, GROUP_M=8, num_stages=3, num_warps=4
triton_bmm_66 0.0133 ms 92.3% ACC_TYPE='tl.float32', ALLOW_TF32=False, BLOCK_K=32, BLOCK_M=64, BLOCK_N=128, B_PROLOGUE_CAST_TYPE=None, EVEN_K=True, GROUP_M=8, num_stages=3, num_warps=4
triton_bmm_74 0.0133 ms 92.3% ACC_TYPE='tl.float32', ALLOW_TF32=False, BLOCK_K=64, BLOCK_M=128, BLOCK_N=128, B_PROLOGUE_CAST_TYPE=None, EVEN_K=True, GROUP_M=8, num_stages=3, num_warps=4
triton_bmm_72 0.0133 ms 92.1% ACC_TYPE='tl.float32', ALLOW_TF32=False, BLOCK_K=32, BLOCK_M=128, BLOCK_N=128, B_PROLOGUE_CAST_TYPE=None, EVEN_K=True, GROUP_M=8, num_stages=2, num_warps=8
triton_bmm_75 0.0134 ms 91.9% ACC_TYPE='tl.float32', ALLOW_TF32=False, BLOCK_K=64, BLOCK_M=128, BLOCK_N=128, B_PROLOGUE_CAST_TYPE=None, EVEN_K=True, GROUP_M=8, num_stages=5, num_warps=8
bmm 0.0136 ms 90.4%
triton_bmm_68 0.0137 ms 89.5% ACC_TYPE='tl.float32', ALLOW_TF32=False, BLOCK_K=64, BLOCK_M=64, BLOCK_N=128, B_PROLOGUE_CAST_TYPE=None, EVEN_K=True, GROUP_M=8, num_stages=3, num_warps=4
triton_bmm_64 0.0139 ms 88.7% ACC_TYPE='tl.float32', ALLOW_TF32=False, BLOCK_K=64, BLOCK_M=64, BLOCK_N=64, B_PROLOGUE_CAST_TYPE=None, EVEN_K=True, GROUP_M=8, num_stages=3, num_warps=8
SingleProcess AUTOTUNE benchmarking takes 2.3073 seconds and 0.0019 seconds precompiling
AUTOTUNE bmm(12x512x512, 12x512x64)
triton_bmm_87 0.0139 ms 100.0% ACC_TYPE='tl.float32', ALLOW_TF32=False, BLOCK_K=64, BLOCK_M=64, BLOCK_N=64, B_PROLOGUE_CAST_TYPE=None, EVEN_K=True, GROUP_M=8, num_stages=3, num_warps=4
triton_bmm_84 0.0141 ms 98.6% ACC_TYPE='tl.float32', ALLOW_TF32=False, BLOCK_K=128, BLOCK_M=64, BLOCK_N=64, B_PROLOGUE_CAST_TYPE=None, EVEN_K=True, GROUP_M=8, num_stages=5, num_warps=4
triton_bmm_83 0.0147 ms 94.8% ACC_TYPE='tl.float32', ALLOW_TF32=False, BLOCK_K=64, BLOCK_M=64, BLOCK_N=64, B_PROLOGUE_CAST_TYPE=None, EVEN_K=True, GROUP_M=8, num_stages=3, num_warps=8
triton_bmm_93 0.0147 ms 94.8% ACC_TYPE='tl.float32', ALLOW_TF32=False, BLOCK_K=64, BLOCK_M=128, BLOCK_N=64, B_PROLOGUE_CAST_TYPE=None, EVEN_K=True, GROUP_M=8, num_stages=5, num_warps=8
triton_bmm_88 0.0147 ms 94.3% ACC_TYPE='tl.float32', ALLOW_TF32=False, BLOCK_K=128, BLOCK_M=64, BLOCK_N=64, B_PROLOGUE_CAST_TYPE=None, EVEN_K=True, GROUP_M=8, num_stages=4, num_warps=4
triton_bmm_77 0.0149 ms 93.1% ACC_TYPE='tl.float32', ALLOW_TF32=False, BLOCK_K=128, BLOCK_M=32, BLOCK_N=32, B_PROLOGUE_CAST_TYPE=None, EVEN_K=True, GROUP_M=8, num_stages=2, num_warps=4
bmm 0.0152 ms 91.6%
triton_bmm_78 0.0152 ms 91.2% ACC_TYPE='tl.float32', ALLOW_TF32=False, BLOCK_K=32, BLOCK_M=32, BLOCK_N=64, B_PROLOGUE_CAST_TYPE=None, EVEN_K=True, GROUP_M=8, num_stages=5, num_warps=8
triton_bmm_86 0.0158 ms 87.8% ACC_TYPE='tl.float32', ALLOW_TF32=False, BLOCK_K=32, BLOCK_M=64, BLOCK_N=64, B_PROLOGUE_CAST_TYPE=None, EVEN_K=True, GROUP_M=8, num_stages=4, num_warps=8
triton_bmm_92 0.0160 ms 86.8% ACC_TYPE='tl.float32', ALLOW_TF32=False, BLOCK_K=64, BLOCK_M=128, BLOCK_N=64, B_PROLOGUE_CAST_TYPE=None, EVEN_K=True, GROUP_M=8, num_stages=3, num_warps=4
SingleProcess AUTOTUNE benchmarking takes 2.1837 seconds and 0.0017 seconds precompiling
AUTOTUNE addmm(512x768, 512x3072, 3072x768)
triton_mm_140 0.0255 ms 100.0% ACC_TYPE='tl.float32', ALLOW_TF32=False, BLOCK_K=128, BLOCK_M=64, BLOCK_N=64, B_PROLOGUE_CAST_TYPE=None, EVEN_K=True, GROUP_M=8, num_stages=5, num_warps=4
triton_mm_144 0.0316 ms 80.7% ACC_TYPE='tl.float32', ALLOW_TF32=False, BLOCK_K=128, BLOCK_M=64, BLOCK_N=128, B_PROLOGUE_CAST_TYPE=None, EVEN_K=True, GROUP_M=8, num_stages=4, num_warps=4
addmm 0.0322 ms 79.1%
triton_mm_136 0.0349 ms 73.1% ACC_TYPE='tl.float32', ALLOW_TF32=False, BLOCK_K=128, BLOCK_M=64, BLOCK_N=32, B_PROLOGUE_CAST_TYPE=None, EVEN_K=True, GROUP_M=8, num_stages=5, num_warps=4
triton_mm_139 0.0355 ms 71.9% ACC_TYPE='tl.float32', ALLOW_TF32=False, BLOCK_K=64, BLOCK_M=64, BLOCK_N=64, B_PROLOGUE_CAST_TYPE=None, EVEN_K=True, GROUP_M=8, num_stages=3, num_warps=8
triton_mm_143 0.0372 ms 68.6% ACC_TYPE='tl.float32', ALLOW_TF32=False, BLOCK_K=64, BLOCK_M=64, BLOCK_N=128, B_PROLOGUE_CAST_TYPE=None, EVEN_K=True, GROUP_M=8, num_stages=3, num_warps=4
triton_mm_146 0.0419 ms 60.8% ACC_TYPE='tl.float32', ALLOW_TF32=False, BLOCK_K=32, BLOCK_M=128, BLOCK_N=64, B_PROLOGUE_CAST_TYPE=None, EVEN_K=True, GROUP_M=8, num_stages=4, num_warps=8
triton_mm_142 0.0425 ms 60.0% ACC_TYPE='tl.float32', ALLOW_TF32=False, BLOCK_K=32, BLOCK_M=64, BLOCK_N=128, B_PROLOGUE_CAST_TYPE=None, EVEN_K=True, GROUP_M=8, num_stages=4, num_warps=8
triton_mm_133 0.0449 ms 56.8% ACC_TYPE='tl.float32', ALLOW_TF32=False, BLOCK_K=128, BLOCK_M=32, BLOCK_N=32, B_PROLOGUE_CAST_TYPE=None, EVEN_K=True, GROUP_M=8, num_stages=2, num_warps=4
triton_mm_134 0.0454 ms 56.2% ACC_TYPE='tl.float32', ALLOW_TF32=False, BLOCK_K=32, BLOCK_M=32, BLOCK_N=64, B_PROLOGUE_CAST_TYPE=None, EVEN_K=True, GROUP_M=8, num_stages=5, num_warps=8
SingleProcess AUTOTUNE benchmarking takes 2.5222 seconds and 0.0015 seconds precompiling
AUTOTUNE addmm(512x30522, 512x768, 768x30522)
triton_mm_1848 0.1421 ms 100.0% ACC_TYPE='tl.float32', ALLOW_TF32=False, BLOCK_K=64, BLOCK_M=128, BLOCK_N=128, B_PROLOGUE_CAST_TYPE=None, EVEN_K=True, GROUP_M=8, num_stages=3, num_warps=4
triton_mm_1847 0.1476 ms 96.3% ACC_TYPE='tl.float32', ALLOW_TF32=False, BLOCK_K=32, BLOCK_M=128, BLOCK_N=128, B_PROLOGUE_CAST_TYPE=None, EVEN_K=True, GROUP_M=8, num_stages=3, num_warps=4
triton_mm_1842 0.1589 ms 89.4% ACC_TYPE='tl.float32', ALLOW_TF32=False, BLOCK_K=64, BLOCK_M=64, BLOCK_N=128, B_PROLOGUE_CAST_TYPE=None, EVEN_K=True, GROUP_M=8, num_stages=3, num_warps=4
triton_mm_1849 0.1692 ms 84.0% ACC_TYPE='tl.float32', ALLOW_TF32=False, BLOCK_K=64, BLOCK_M=128, BLOCK_N=128, B_PROLOGUE_CAST_TYPE=None, EVEN_K=True, GROUP_M=8, num_stages=5, num_warps=8
triton_mm_1840 0.1735 ms 81.9% ACC_TYPE='tl.float32', ALLOW_TF32=False, BLOCK_K=32, BLOCK_M=64, BLOCK_N=128, B_PROLOGUE_CAST_TYPE=None, EVEN_K=True, GROUP_M=8, num_stages=3, num_warps=4
triton_mm_1844 0.1771 ms 80.2% ACC_TYPE='tl.float32', ALLOW_TF32=False, BLOCK_K=32, BLOCK_M=128, BLOCK_N=64, B_PROLOGUE_CAST_TYPE=None, EVEN_K=True, GROUP_M=8, num_stages=3, num_warps=4
triton_mm_1841 0.1903 ms 74.7% ACC_TYPE='tl.float32', ALLOW_TF32=False, BLOCK_K=32, BLOCK_M=64, BLOCK_N=128, B_PROLOGUE_CAST_TYPE=None, EVEN_K=True, GROUP_M=8, num_stages=4, num_warps=8
triton_mm_1845 0.1954 ms 72.7% ACC_TYPE='tl.float32', ALLOW_TF32=False, BLOCK_K=32, BLOCK_M=128, BLOCK_N=64, B_PROLOGUE_CAST_TYPE=None, EVEN_K=True, GROUP_M=8, num_stages=4, num_warps=8
bias_addmm 0.2017 ms 70.4%
triton_mm_1838 0.2179 ms 65.2% ACC_TYPE='tl.float32', ALLOW_TF32=False, BLOCK_K=64, BLOCK_M=64, BLOCK_N=64, B_PROLOGUE_CAST_TYPE=None, EVEN_K=True, GROUP_M=8, num_stages=3, num_warps=8
SingleProcess AUTOTUNE benchmarking takes 2.6045 seconds and 0.0019 seconds precompiling
Compiled module path: /tmp/torchinductor_shunting/tmpoy_zxppq/h2/ch24uha2pfcha7xkywhblrufh65dc4lqu3ujsvweiael2haisl5q.py
AUTOTUNE mm(30522x512, 512x768)
mm 0.1228 ms 100.0%
triton_mm_1885 0.1455 ms 84.3% ACC_TYPE='tl.float32', ALLOW_TF32=False, BLOCK_K=32, BLOCK_M=128, BLOCK_N=128, B_PROLOGUE_CAST_TYPE=None, EVEN_K=True, GROUP_M=8, num_stages=3, num_warps=4
triton_mm_1886 0.1643 ms 74.7% ACC_TYPE='tl.float32', ALLOW_TF32=False, BLOCK_K=64, BLOCK_M=128, BLOCK_N=128, B_PROLOGUE_CAST_TYPE=None, EVEN_K=True, GROUP_M=8, num_stages=3, num_warps=4
triton_mm_1878 0.1659 ms 74.0% ACC_TYPE='tl.float32', ALLOW_TF32=False, BLOCK_K=32, BLOCK_M=64, BLOCK_N=128, B_PROLOGUE_CAST_TYPE=None, EVEN_K=True, GROUP_M=8, num_stages=3, num_warps=4
triton_mm_1880 0.1721 ms 71.3% ACC_TYPE='tl.float32', ALLOW_TF32=False, BLOCK_K=64, BLOCK_M=64, BLOCK_N=128, B_PROLOGUE_CAST_TYPE=None, EVEN_K=True, GROUP_M=8, num_stages=3, num_warps=4
triton_mm_1879 0.1838 ms 66.8% ACC_TYPE='tl.float32', ALLOW_TF32=False, BLOCK_K=32, BLOCK_M=64, BLOCK_N=128, B_PROLOGUE_CAST_TYPE=None, EVEN_K=True, GROUP_M=8, num_stages=4, num_warps=8
triton_mm_1882 0.1933 ms 63.5% ACC_TYPE='tl.float32', ALLOW_TF32=False, BLOCK_K=32, BLOCK_M=128, BLOCK_N=64, B_PROLOGUE_CAST_TYPE=None, EVEN_K=True, GROUP_M=8, num_stages=3, num_warps=4
triton_mm_1887 0.2014 ms 60.9% ACC_TYPE='tl.float32', ALLOW_TF32=False, BLOCK_K=64, BLOCK_M=128, BLOCK_N=128, B_PROLOGUE_CAST_TYPE=None, EVEN_K=True, GROUP_M=8, num_stages=5, num_warps=8
triton_mm_1883 0.2176 ms 56.4% ACC_TYPE='tl.float32', ALLOW_TF32=False, BLOCK_K=32, BLOCK_M=128, BLOCK_N=64, B_PROLOGUE_CAST_TYPE=None, EVEN_K=True, GROUP_M=8, num_stages=4, num_warps=8
triton_mm_1875 0.2292 ms 53.6% ACC_TYPE='tl.float32', ALLOW_TF32=False, BLOCK_K=32, BLOCK_M=64, BLOCK_N=64, B_PROLOGUE_CAST_TYPE=None, EVEN_K=True, GROUP_M=8, num_stages=2, num_warps=4
SingleProcess AUTOTUNE benchmarking takes 2.4646 seconds and 0.0018 seconds precompiling
AUTOTUNE mm(768x512, 512x3072)
triton_mm_1954 0.0221 ms 100.0% ACC_TYPE='tl.float32', ALLOW_TF32=False, BLOCK_K=32, BLOCK_M=64, BLOCK_N=128, B_PROLOGUE_CAST_TYPE=None, EVEN_K=True, GROUP_M=8, num_stages=3, num_warps=4
triton_mm_1956 0.0223 ms 98.9% ACC_TYPE='tl.float32', ALLOW_TF32=False, BLOCK_K=64, BLOCK_M=64, BLOCK_N=128, B_PROLOGUE_CAST_TYPE=None, EVEN_K=True, GROUP_M=8, num_stages=3, num_warps=4
mm 0.0225 ms 98.0%
triton_mm_1961 0.0234 ms 94.3% ACC_TYPE='tl.float32', ALLOW_TF32=False, BLOCK_K=32, BLOCK_M=128, BLOCK_N=128, B_PROLOGUE_CAST_TYPE=None, EVEN_K=True, GROUP_M=8, num_stages=3, num_warps=4
triton_mm_1958 0.0242 ms 91.3% ACC_TYPE='tl.float32', ALLOW_TF32=False, BLOCK_K=32, BLOCK_M=128, BLOCK_N=64, B_PROLOGUE_CAST_TYPE=None, EVEN_K=True, GROUP_M=8, num_stages=3, num_warps=4
triton_mm_1955 0.0242 ms 91.1% ACC_TYPE='tl.float32', ALLOW_TF32=False, BLOCK_K=32, BLOCK_M=64, BLOCK_N=128, B_PROLOGUE_CAST_TYPE=None, EVEN_K=True, GROUP_M=8, num_stages=4, num_warps=8
triton_mm_1959 0.0247 ms 89.5% ACC_TYPE='tl.float32', ALLOW_TF32=False, BLOCK_K=32, BLOCK_M=128, BLOCK_N=64, B_PROLOGUE_CAST_TYPE=None, EVEN_K=True, GROUP_M=8, num_stages=4, num_warps=8
triton_mm_1962 0.0249 ms 88.6% ACC_TYPE='tl.float32', ALLOW_TF32=False, BLOCK_K=64, BLOCK_M=128, BLOCK_N=128, B_PROLOGUE_CAST_TYPE=None, EVEN_K=True, GROUP_M=8, num_stages=3, num_warps=4
triton_mm_1952 0.0287 ms 76.8% ACC_TYPE='tl.float32', ALLOW_TF32=False, BLOCK_K=64, BLOCK_M=64, BLOCK_N=64, B_PROLOGUE_CAST_TYPE=None, EVEN_K=True, GROUP_M=8, num_stages=3, num_warps=8
triton_mm_1963 0.0291 ms 75.8% ACC_TYPE='tl.float32', ALLOW_TF32=False, BLOCK_K=64, BLOCK_M=128, BLOCK_N=128, B_PROLOGUE_CAST_TYPE=None, EVEN_K=True, GROUP_M=8, num_stages=5, num_warps=8
SingleProcess AUTOTUNE benchmarking takes 2.3313 seconds and 0.0017 seconds precompiling
AUTOTUNE mm(3072x512, 512x768)
triton_mm_1996 0.0226 ms 100.0% ACC_TYPE='tl.float32', ALLOW_TF32=False, BLOCK_K=32, BLOCK_M=128, BLOCK_N=64, B_PROLOGUE_CAST_TYPE=None, EVEN_K=True, GROUP_M=8, num_stages=3, num_warps=4
triton_mm_1999 0.0226 ms 99.7% ACC_TYPE='tl.float32', ALLOW_TF32=False, BLOCK_K=32, BLOCK_M=128, BLOCK_N=128, B_PROLOGUE_CAST_TYPE=None, EVEN_K=True, GROUP_M=8, num_stages=3, num_warps=4
mm 0.0228 ms 99.0%
triton_mm_1994 0.0232 ms 97.4% ACC_TYPE='tl.float32', ALLOW_TF32=False, BLOCK_K=64, BLOCK_M=64, BLOCK_N=128, B_PROLOGUE_CAST_TYPE=None, EVEN_K=True, GROUP_M=8, num_stages=3, num_warps=4
triton_mm_1992 0.0236 ms 95.8% ACC_TYPE='tl.float32', ALLOW_TF32=False, BLOCK_K=32, BLOCK_M=64, BLOCK_N=128, B_PROLOGUE_CAST_TYPE=None, EVEN_K=True, GROUP_M=8, num_stages=3, num_warps=4
triton_mm_1997 0.0237 ms 95.1% ACC_TYPE='tl.float32', ALLOW_TF32=False, BLOCK_K=32, BLOCK_M=128, BLOCK_N=64, B_PROLOGUE_CAST_TYPE=None, EVEN_K=True, GROUP_M=8, num_stages=4, num_warps=8
triton_mm_1993 0.0239 ms 94.4% ACC_TYPE='tl.float32', ALLOW_TF32=False, BLOCK_K=32, BLOCK_M=64, BLOCK_N=128, B_PROLOGUE_CAST_TYPE=None, EVEN_K=True, GROUP_M=8, num_stages=4, num_warps=8
triton_mm_2000 0.0241 ms 93.6% ACC_TYPE='tl.float32', ALLOW_TF32=False, BLOCK_K=64, BLOCK_M=128, BLOCK_N=128, B_PROLOGUE_CAST_TYPE=None, EVEN_K=True, GROUP_M=8, num_stages=3, num_warps=4
triton_mm_1990 0.0286 ms 78.9% ACC_TYPE='tl.float32', ALLOW_TF32=False, BLOCK_K=64, BLOCK_M=64, BLOCK_N=64, B_PROLOGUE_CAST_TYPE=None, EVEN_K=True, GROUP_M=8, num_stages=3, num_warps=8
triton_mm_2001 0.0289 ms 78.1% ACC_TYPE='tl.float32', ALLOW_TF32=False, BLOCK_K=64, BLOCK_M=128, BLOCK_N=128, B_PROLOGUE_CAST_TYPE=None, EVEN_K=True, GROUP_M=8, num_stages=5, num_warps=8
SingleProcess AUTOTUNE benchmarking takes 2.3344 seconds and 0.0017 seconds precompiling
AUTOTUNE mm(512x768, 768x3072)
triton_mm_1943 0.0200 ms 100.0% ACC_TYPE='tl.float32', ALLOW_TF32=False, BLOCK_K=64, BLOCK_M=128, BLOCK_N=128, B_PROLOGUE_CAST_TYPE=None, EVEN_K=True, GROUP_M=8, num_stages=3, num_warps=4
mm 0.0200 ms 99.7%
triton_mm_1944 0.0208 ms 96.1% ACC_TYPE='tl.float32', ALLOW_TF32=False, BLOCK_K=64, BLOCK_M=128, BLOCK_N=128, B_PROLOGUE_CAST_TYPE=None, EVEN_K=True, GROUP_M=8, num_stages=5, num_warps=8
triton_mm_1937 0.0212 ms 94.3% ACC_TYPE='tl.float32', ALLOW_TF32=False, BLOCK_K=64, BLOCK_M=64, BLOCK_N=128, B_PROLOGUE_CAST_TYPE=None, EVEN_K=True, GROUP_M=8, num_stages=3, num_warps=4
triton_mm_1942 0.0226 ms 88.3% ACC_TYPE='tl.float32', ALLOW_TF32=False, BLOCK_K=32, BLOCK_M=128, BLOCK_N=128, B_PROLOGUE_CAST_TYPE=None, EVEN_K=True, GROUP_M=8, num_stages=3, num_warps=4
triton_mm_1935 0.0228 ms 87.5% ACC_TYPE='tl.float32', ALLOW_TF32=False, BLOCK_K=32, BLOCK_M=64, BLOCK_N=128, B_PROLOGUE_CAST_TYPE=None, EVEN_K=True, GROUP_M=8, num_stages=3, num_warps=4
triton_mm_1939 0.0234 ms 85.4% ACC_TYPE='tl.float32', ALLOW_TF32=False, BLOCK_K=32, BLOCK_M=128, BLOCK_N=64, B_PROLOGUE_CAST_TYPE=None, EVEN_K=True, GROUP_M=8, num_stages=3, num_warps=4
triton_mm_1936 0.0238 ms 83.9% ACC_TYPE='tl.float32', ALLOW_TF32=False, BLOCK_K=32, BLOCK_M=64, BLOCK_N=128, B_PROLOGUE_CAST_TYPE=None, EVEN_K=True, GROUP_M=8, num_stages=4, num_warps=8
triton_mm_1940 0.0244 ms 81.7% ACC_TYPE='tl.float32', ALLOW_TF32=False, BLOCK_K=32, BLOCK_M=128, BLOCK_N=64, B_PROLOGUE_CAST_TYPE=None, EVEN_K=True, GROUP_M=8, num_stages=4, num_warps=8
triton_mm_1938 0.0267 ms 74.7% ACC_TYPE='tl.float32', ALLOW_TF32=False, BLOCK_K=128, BLOCK_M=64, BLOCK_N=128, B_PROLOGUE_CAST_TYPE=None, EVEN_K=True, GROUP_M=8, num_stages=4, num_warps=4
SingleProcess AUTOTUNE benchmarking takes 2.3231 seconds and 0.0018 seconds precompiling
AUTOTUNE mm(768x512, 512x768)
mm 0.0130 ms 100.0%
triton_mm_2181 0.0132 ms 97.9% ACC_TYPE='tl.float32', ALLOW_TF32=False, BLOCK_K=128, BLOCK_M=64, BLOCK_N=128, B_PROLOGUE_CAST_TYPE=None, EVEN_K=True, GROUP_M=8, num_stages=4, num_warps=4
triton_mm_2180 0.0136 ms 95.6% ACC_TYPE='tl.float32', ALLOW_TF32=False, BLOCK_K=64, BLOCK_M=64, BLOCK_N=128, B_PROLOGUE_CAST_TYPE=None, EVEN_K=True, GROUP_M=8, num_stages=3, num_warps=4
triton_mm_2179 0.0137 ms 95.0% ACC_TYPE='tl.float32', ALLOW_TF32=False, BLOCK_K=32, BLOCK_M=64, BLOCK_N=128, B_PROLOGUE_CAST_TYPE=None, EVEN_K=True, GROUP_M=8, num_stages=4, num_warps=8
triton_mm_2176 0.0137 ms 94.5% ACC_TYPE='tl.float32', ALLOW_TF32=False, BLOCK_K=64, BLOCK_M=64, BLOCK_N=64, B_PROLOGUE_CAST_TYPE=None, EVEN_K=True, GROUP_M=8, num_stages=3, num_warps=8
triton_mm_2183 0.0141 ms 91.8% ACC_TYPE='tl.float32', ALLOW_TF32=False, BLOCK_K=32, BLOCK_M=128, BLOCK_N=64, B_PROLOGUE_CAST_TYPE=None, EVEN_K=True, GROUP_M=8, num_stages=4, num_warps=8
triton_mm_2186 0.0152 ms 85.4% ACC_TYPE='tl.float32', ALLOW_TF32=False, BLOCK_K=64, BLOCK_M=128, BLOCK_N=128, B_PROLOGUE_CAST_TYPE=None, EVEN_K=True, GROUP_M=8, num_stages=3, num_warps=4
triton_mm_2178 0.0152 ms 85.2% ACC_TYPE='tl.float32', ALLOW_TF32=False, BLOCK_K=32, BLOCK_M=64, BLOCK_N=128, B_PROLOGUE_CAST_TYPE=None, EVEN_K=True, GROUP_M=8, num_stages=3, num_warps=4
triton_mm_2182 0.0159 ms 81.6% ACC_TYPE='tl.float32', ALLOW_TF32=False, BLOCK_K=32, BLOCK_M=128, BLOCK_N=64, B_PROLOGUE_CAST_TYPE=None, EVEN_K=True, GROUP_M=8, num_stages=3, num_warps=4
triton_mm_2177 0.0161 ms 80.8% ACC_TYPE='tl.float32', ALLOW_TF32=False, BLOCK_K=128, BLOCK_M=64, BLOCK_N=64, B_PROLOGUE_CAST_TYPE=None, EVEN_K=True, GROUP_M=8, num_stages=5, num_warps=4
SingleProcess AUTOTUNE benchmarking takes 2.3058 seconds and 0.0013 seconds precompiling
AUTOTUNE mm(768x512, 512x768)
mm 0.0124 ms 100.0%
triton_mm_1918 0.0134 ms 92.1% ACC_TYPE='tl.float32', ALLOW_TF32=False, BLOCK_K=64, BLOCK_M=64, BLOCK_N=128, B_PROLOGUE_CAST_TYPE=None, EVEN_K=True, GROUP_M=8, num_stages=3, num_warps=4
triton_mm_1919 0.0135 ms 91.7% ACC_TYPE='tl.float32', ALLOW_TF32=False, BLOCK_K=128, BLOCK_M=64, BLOCK_N=128, B_PROLOGUE_CAST_TYPE=None, EVEN_K=True, GROUP_M=8, num_stages=4, num_warps=4
triton_mm_1914 0.0140 ms 88.2% ACC_TYPE='tl.float32', ALLOW_TF32=False, BLOCK_K=64, BLOCK_M=64, BLOCK_N=64, B_PROLOGUE_CAST_TYPE=None, EVEN_K=True, GROUP_M=8, num_stages=3, num_warps=8
triton_mm_1921 0.0142 ms 87.2% ACC_TYPE='tl.float32', ALLOW_TF32=False, BLOCK_K=32, BLOCK_M=128, BLOCK_N=64, B_PROLOGUE_CAST_TYPE=None, EVEN_K=True, GROUP_M=8, num_stages=4, num_warps=8
triton_mm_1917 0.0145 ms 85.1% ACC_TYPE='tl.float32', ALLOW_TF32=False, BLOCK_K=32, BLOCK_M=64, BLOCK_N=128, B_PROLOGUE_CAST_TYPE=None, EVEN_K=True, GROUP_M=8, num_stages=4, num_warps=8
triton_mm_1916 0.0153 ms 81.1% ACC_TYPE='tl.float32', ALLOW_TF32=False, BLOCK_K=32, BLOCK_M=64, BLOCK_N=128, B_PROLOGUE_CAST_TYPE=None, EVEN_K=True, GROUP_M=8, num_stages=3, num_warps=4
triton_mm_1920 0.0154 ms 80.5% ACC_TYPE='tl.float32', ALLOW_TF32=False, BLOCK_K=32, BLOCK_M=128, BLOCK_N=64, B_PROLOGUE_CAST_TYPE=None, EVEN_K=True, GROUP_M=8, num_stages=3, num_warps=4
triton_mm_1924 0.0155 ms 80.0% ACC_TYPE='tl.float32', ALLOW_TF32=False, BLOCK_K=64, BLOCK_M=128, BLOCK_N=128, B_PROLOGUE_CAST_TYPE=None, EVEN_K=True, GROUP_M=8, num_stages=3, num_warps=4
triton_mm_1925 0.0167 ms 74.0% ACC_TYPE='tl.float32', ALLOW_TF32=False, BLOCK_K=64, BLOCK_M=128, BLOCK_N=128, B_PROLOGUE_CAST_TYPE=None, EVEN_K=True, GROUP_M=8, num_stages=5, num_warps=8
SingleProcess AUTOTUNE benchmarking takes 2.3090 seconds and 0.0018 seconds precompiling
AUTOTUNE mm(512x30522, 30522x768)
triton_mm_1858 0.1764 ms 100.0% ACC_TYPE='tl.float32', ALLOW_TF32=False, BLOCK_K=128, BLOCK_M=64, BLOCK_N=64, B_PROLOGUE_CAST_TYPE=None, EVEN_K=False, GROUP_M=8, num_stages=5, num_warps=4
mm 0.1980 ms 89.1%
triton_mm_1862 0.2309 ms 76.4% ACC_TYPE='tl.float32', ALLOW_TF32=False, BLOCK_K=128, BLOCK_M=64, BLOCK_N=128, B_PROLOGUE_CAST_TYPE=None, EVEN_K=False, GROUP_M=8, num_stages=4, num_warps=4
triton_mm_1857 0.2598 ms 67.9% ACC_TYPE='tl.float32', ALLOW_TF32=False, BLOCK_K=64, BLOCK_M=64, BLOCK_N=64, B_PROLOGUE_CAST_TYPE=None, EVEN_K=False, GROUP_M=8, num_stages=3, num_warps=8
triton_mm_1854 0.2616 ms 67.4% ACC_TYPE='tl.float32', ALLOW_TF32=False, BLOCK_K=128, BLOCK_M=64, BLOCK_N=32, B_PROLOGUE_CAST_TYPE=None, EVEN_K=False, GROUP_M=8, num_stages=5, num_warps=4
triton_mm_1861 0.2714 ms 65.0% ACC_TYPE='tl.float32', ALLOW_TF32=False, BLOCK_K=64, BLOCK_M=64, BLOCK_N=128, B_PROLOGUE_CAST_TYPE=None, EVEN_K=False, GROUP_M=8, num_stages=3, num_warps=4
triton_mm_1860 0.3149 ms 56.0% ACC_TYPE='tl.float32', ALLOW_TF32=False, BLOCK_K=32, BLOCK_M=64, BLOCK_N=128, B_PROLOGUE_CAST_TYPE=None, EVEN_K=False, GROUP_M=8, num_stages=4, num_warps=8
triton_mm_1864 0.3256 ms 54.2% ACC_TYPE='tl.float32', ALLOW_TF32=False, BLOCK_K=32, BLOCK_M=128, BLOCK_N=64, B_PROLOGUE_CAST_TYPE=None, EVEN_K=False, GROUP_M=8, num_stages=4, num_warps=8
triton_mm_1851 0.3303 ms 53.4% ACC_TYPE='tl.float32', ALLOW_TF32=False, BLOCK_K=128, BLOCK_M=32, BLOCK_N=32, B_PROLOGUE_CAST_TYPE=None, EVEN_K=False, GROUP_M=8, num_stages=2, num_warps=4
triton_mm_1852 0.3794 ms 46.5% ACC_TYPE='tl.float32', ALLOW_TF32=False, BLOCK_K=32, BLOCK_M=32, BLOCK_N=64, B_PROLOGUE_CAST_TYPE=None, EVEN_K=False, GROUP_M=8, num_stages=5, num_warps=8
SingleProcess AUTOTUNE benchmarking takes 2.4976 seconds and 0.0021 seconds precompiling
AUTOTUNE mm(512x768, 768x768)
triton_mm_1896 0.0127 ms 100.0% ACC_TYPE='tl.float32', ALLOW_TF32=False, BLOCK_K=128, BLOCK_M=64, BLOCK_N=64, B_PROLOGUE_CAST_TYPE=None, EVEN_K=True, GROUP_M=8, num_stages=5, num_warps=4
mm 0.0133 ms 95.0%
triton_mm_1895 0.0139 ms 91.0% ACC_TYPE='tl.float32', ALLOW_TF32=False, BLOCK_K=64, BLOCK_M=64, BLOCK_N=64, B_PROLOGUE_CAST_TYPE=None, EVEN_K=True, GROUP_M=8, num_stages=3, num_warps=8
triton_mm_1900 0.0147 ms 86.3% ACC_TYPE='tl.float32', ALLOW_TF32=False, BLOCK_K=128, BLOCK_M=64, BLOCK_N=128, B_PROLOGUE_CAST_TYPE=None, EVEN_K=True, GROUP_M=8, num_stages=4, num_warps=4
triton_mm_1899 0.0152 ms 83.5% ACC_TYPE='tl.float32', ALLOW_TF32=False, BLOCK_K=64, BLOCK_M=64, BLOCK_N=128, B_PROLOGUE_CAST_TYPE=None, EVEN_K=True, GROUP_M=8, num_stages=3, num_warps=4
triton_mm_1898 0.0162 ms 78.1% ACC_TYPE='tl.float32', ALLOW_TF32=False, BLOCK_K=32, BLOCK_M=64, BLOCK_N=128, B_PROLOGUE_CAST_TYPE=None, EVEN_K=True, GROUP_M=8, num_stages=4, num_warps=8
triton_mm_1902 0.0165 ms 76.9% ACC_TYPE='tl.float32', ALLOW_TF32=False, BLOCK_K=32, BLOCK_M=128, BLOCK_N=64, B_PROLOGUE_CAST_TYPE=None, EVEN_K=True, GROUP_M=8, num_stages=4, num_warps=8
triton_mm_1891 0.0169 ms 75.0% ACC_TYPE='tl.float32', ALLOW_TF32=False, BLOCK_K=32, BLOCK_M=64, BLOCK_N=32, B_PROLOGUE_CAST_TYPE=None, EVEN_K=True, GROUP_M=8, num_stages=5, num_warps=8
triton_mm_1889 0.0172 ms 73.6% ACC_TYPE='tl.float32', ALLOW_TF32=False, BLOCK_K=128, BLOCK_M=32, BLOCK_N=32, B_PROLOGUE_CAST_TYPE=None, EVEN_K=True, GROUP_M=8, num_stages=2, num_warps=4
triton_mm_1890 0.0175 ms 72.5% ACC_TYPE='tl.float32', ALLOW_TF32=False, BLOCK_K=32, BLOCK_M=32, BLOCK_N=64, B_PROLOGUE_CAST_TYPE=None, EVEN_K=True, GROUP_M=8, num_stages=5, num_warps=8
SingleProcess AUTOTUNE benchmarking takes 2.3219 seconds and 0.0018 seconds precompiling
AUTOTUNE mm(512x3072, 3072x768)
triton_mm_1972 0.0252 ms 100.0% ACC_TYPE='tl.float32', ALLOW_TF32=False, BLOCK_K=128, BLOCK_M=64, BLOCK_N=64, B_PROLOGUE_CAST_TYPE=None, EVEN_K=True, GROUP_M=8, num_stages=5, num_warps=4
mm 0.0256 ms 98.6%
triton_mm_1976 0.0315 ms 80.0% ACC_TYPE='tl.float32', ALLOW_TF32=False, BLOCK_K=128, BLOCK_M=64, BLOCK_N=128, B_PROLOGUE_CAST_TYPE=None, EVEN_K=True, GROUP_M=8, num_stages=4, num_warps=4
triton_mm_1971 0.0347 ms 72.7% ACC_TYPE='tl.float32', ALLOW_TF32=False, BLOCK_K=64, BLOCK_M=64, BLOCK_N=64, B_PROLOGUE_CAST_TYPE=None, EVEN_K=True, GROUP_M=8, num_stages=3, num_warps=8
triton_mm_1968 0.0348 ms 72.6% ACC_TYPE='tl.float32', ALLOW_TF32=False, BLOCK_K=128, BLOCK_M=64, BLOCK_N=32, B_PROLOGUE_CAST_TYPE=None, EVEN_K=True, GROUP_M=8, num_stages=5, num_warps=4
triton_mm_1975 0.0362 ms 69.7% ACC_TYPE='tl.float32', ALLOW_TF32=False, BLOCK_K=64, BLOCK_M=64, BLOCK_N=128, B_PROLOGUE_CAST_TYPE=None, EVEN_K=True, GROUP_M=8, num_stages=3, num_warps=4
triton_mm_1974 0.0397 ms 63.5% ACC_TYPE='tl.float32', ALLOW_TF32=False, BLOCK_K=32, BLOCK_M=64, BLOCK_N=128, B_PROLOGUE_CAST_TYPE=None, EVEN_K=True, GROUP_M=8, num_stages=4, num_warps=8
triton_mm_1978 0.0405 ms 62.2% ACC_TYPE='tl.float32', ALLOW_TF32=False, BLOCK_K=32, BLOCK_M=128, BLOCK_N=64, B_PROLOGUE_CAST_TYPE=None, EVEN_K=True, GROUP_M=8, num_stages=4, num_warps=8
triton_mm_1965 0.0422 ms 59.7% ACC_TYPE='tl.float32', ALLOW_TF32=False, BLOCK_K=128, BLOCK_M=32, BLOCK_N=32, B_PROLOGUE_CAST_TYPE=None, EVEN_K=True, GROUP_M=8, num_stages=2, num_warps=4
triton_mm_1966 0.0445 ms 56.7% ACC_TYPE='tl.float32', ALLOW_TF32=False, BLOCK_K=32, BLOCK_M=32, BLOCK_N=64, B_PROLOGUE_CAST_TYPE=None, EVEN_K=True, GROUP_M=8, num_stages=5, num_warps=8
SingleProcess AUTOTUNE benchmarking takes 2.3305 seconds and 0.0013 seconds precompiling
AUTOTUNE bmm(12x512x512, 12x512x64)
triton_bmm_2047 0.0145 ms 100.0% ACC_TYPE='tl.float32', ALLOW_TF32=False, BLOCK_K=64, BLOCK_M=64, BLOCK_N=64, B_PROLOGUE_CAST_TYPE=None, EVEN_K=True, GROUP_M=8, num_stages=3, num_warps=8
triton_bmm_2050 0.0146 ms 99.1% ACC_TYPE='tl.float32', ALLOW_TF32=False, BLOCK_K=32, BLOCK_M=64, BLOCK_N=64, B_PROLOGUE_CAST_TYPE=None, EVEN_K=True, GROUP_M=8, num_stages=4, num_warps=8
triton_bmm_2052 0.0148 ms 97.8% ACC_TYPE='tl.float32', ALLOW_TF32=False, BLOCK_K=128, BLOCK_M=64, BLOCK_N=64, B_PROLOGUE_CAST_TYPE=None, EVEN_K=True, GROUP_M=8, num_stages=4, num_warps=4
triton_bmm_2057 0.0151 ms 96.0% ACC_TYPE='tl.float32', ALLOW_TF32=False, BLOCK_K=64, BLOCK_M=128, BLOCK_N=64, B_PROLOGUE_CAST_TYPE=None, EVEN_K=True, GROUP_M=8, num_stages=5, num_warps=8
triton_bmm_2051 0.0151 ms 95.6% ACC_TYPE='tl.float32', ALLOW_TF32=False, BLOCK_K=64, BLOCK_M=64, BLOCK_N=64, B_PROLOGUE_CAST_TYPE=None, EVEN_K=True, GROUP_M=8, num_stages=3, num_warps=4
triton_bmm_2042 0.0153 ms 94.8% ACC_TYPE='tl.float32', ALLOW_TF32=False, BLOCK_K=32, BLOCK_M=32, BLOCK_N=64, B_PROLOGUE_CAST_TYPE=None, EVEN_K=True, GROUP_M=8, num_stages=5, num_warps=8
triton_bmm_2056 0.0153 ms 94.8% ACC_TYPE='tl.float32', ALLOW_TF32=False, BLOCK_K=64, BLOCK_M=128, BLOCK_N=64, B_PROLOGUE_CAST_TYPE=None, EVEN_K=True, GROUP_M=8, num_stages=3, num_warps=4
bmm 0.0154 ms 93.8%
triton_bmm_2048 0.0156 ms 92.6% ACC_TYPE='tl.float32', ALLOW_TF32=False, BLOCK_K=128, BLOCK_M=64, BLOCK_N=64, B_PROLOGUE_CAST_TYPE=None, EVEN_K=True, GROUP_M=8, num_stages=5, num_warps=4
triton_bmm_2041 0.0158 ms 91.7% ACC_TYPE='tl.float32', ALLOW_TF32=False, BLOCK_K=128, BLOCK_M=32, BLOCK_N=32, B_PROLOGUE_CAST_TYPE=None, EVEN_K=True, GROUP_M=8, num_stages=2, num_warps=4
SingleProcess AUTOTUNE benchmarking takes 2.1815 seconds and 0.0019 seconds precompiling
AUTOTUNE bmm(12x512x64, 12x64x512)
triton_bmm_2074 0.0123 ms 100.0% ACC_TYPE='tl.float32', ALLOW_TF32=False, BLOCK_K=32, BLOCK_M=128, BLOCK_N=128, B_PROLOGUE_CAST_TYPE=None, EVEN_K=True, GROUP_M=8, num_stages=3, num_warps=4
triton_bmm_2070 0.0130 ms 94.6% ACC_TYPE='tl.float32', ALLOW_TF32=False, BLOCK_K=64, BLOCK_M=64, BLOCK_N=128, B_PROLOGUE_CAST_TYPE=None, EVEN_K=True, GROUP_M=8, num_stages=4, num_warps=4
triton_bmm_2071 0.0133 ms 92.5% ACC_TYPE='tl.float32', ALLOW_TF32=False, BLOCK_K=32, BLOCK_M=128, BLOCK_N=64, B_PROLOGUE_CAST_TYPE=None, EVEN_K=True, GROUP_M=8, num_stages=3, num_warps=4
triton_bmm_2064 0.0136 ms 90.8% ACC_TYPE='tl.float32', ALLOW_TF32=False, BLOCK_K=32, BLOCK_M=64, BLOCK_N=64, B_PROLOGUE_CAST_TYPE=None, EVEN_K=True, GROUP_M=8, num_stages=2, num_warps=4
triton_bmm_2069 0.0137 ms 89.6% ACC_TYPE='tl.float32', ALLOW_TF32=False, BLOCK_K=64, BLOCK_M=64, BLOCK_N=128, B_PROLOGUE_CAST_TYPE=None, EVEN_K=True, GROUP_M=8, num_stages=3, num_warps=4
triton_bmm_2067 0.0140 ms 87.9% ACC_TYPE='tl.float32', ALLOW_TF32=False, BLOCK_K=32, BLOCK_M=64, BLOCK_N=128, B_PROLOGUE_CAST_TYPE=None, EVEN_K=True, GROUP_M=8, num_stages=3, num_warps=4
triton_bmm_2073 0.0140 ms 87.7% ACC_TYPE='tl.float32', ALLOW_TF32=False, BLOCK_K=32, BLOCK_M=128, BLOCK_N=128, B_PROLOGUE_CAST_TYPE=None, EVEN_K=True, GROUP_M=8, num_stages=2, num_warps=8
triton_bmm_2075 0.0140 ms 87.7% ACC_TYPE='tl.float32', ALLOW_TF32=False, BLOCK_K=64, BLOCK_M=128, BLOCK_N=128, B_PROLOGUE_CAST_TYPE=None, EVEN_K=True, GROUP_M=8, num_stages=3, num_warps=4
triton_bmm_2076 0.0141 ms 87.3% ACC_TYPE='tl.float32', ALLOW_TF32=False, BLOCK_K=64, BLOCK_M=128, BLOCK_N=128, B_PROLOGUE_CAST_TYPE=None, EVEN_K=True, GROUP_M=8, num_stages=5, num_warps=8
bmm 0.0143 ms 86.0%
SingleProcess AUTOTUNE benchmarking takes 8.5642 seconds and 0.0000 seconds precompiling
AUTOTUNE bmm(12x64x512, 12x512x512)
triton_bmm_2084 0.0144 ms 100.0% ACC_TYPE='tl.float32', ALLOW_TF32=False, BLOCK_K=64, BLOCK_M=64, BLOCK_N=64, B_PROLOGUE_CAST_TYPE=None, EVEN_K=True, GROUP_M=8, num_stages=3, num_warps=8
triton_bmm_2091 0.0146 ms 98.7% ACC_TYPE='tl.float32', ALLOW_TF32=False, BLOCK_K=32, BLOCK_M=64, BLOCK_N=64, B_PROLOGUE_CAST_TYPE=None, EVEN_K=True, GROUP_M=8, num_stages=4, num_warps=8
triton_bmm_2085 0.0151 ms 95.5% ACC_TYPE='tl.float32', ALLOW_TF32=False, BLOCK_K=128, BLOCK_M=64, BLOCK_N=64, B_PROLOGUE_CAST_TYPE=None, EVEN_K=True, GROUP_M=8, num_stages=5, num_warps=4
triton_bmm_2088 0.0152 ms 94.9% ACC_TYPE='tl.float32', ALLOW_TF32=False, BLOCK_K=64, BLOCK_M=64, BLOCK_N=128, B_PROLOGUE_CAST_TYPE=None, EVEN_K=True, GROUP_M=8, num_stages=3, num_warps=4
triton_bmm_2079 0.0153 ms 94.3% ACC_TYPE='tl.float32', ALLOW_TF32=False, BLOCK_K=32, BLOCK_M=32, BLOCK_N=64, B_PROLOGUE_CAST_TYPE=None, EVEN_K=True, GROUP_M=8, num_stages=5, num_warps=8
triton_bmm_2080 0.0154 ms 93.6% ACC_TYPE='tl.float32', ALLOW_TF32=False, BLOCK_K=32, BLOCK_M=64, BLOCK_N=32, B_PROLOGUE_CAST_TYPE=None, EVEN_K=True, GROUP_M=8, num_stages=5, num_warps=8
triton_bmm_2078 0.0155 ms 92.9% ACC_TYPE='tl.float32', ALLOW_TF32=False, BLOCK_K=128, BLOCK_M=32, BLOCK_N=32, B_PROLOGUE_CAST_TYPE=None, EVEN_K=True, GROUP_M=8, num_stages=2, num_warps=4
triton_bmm_2087 0.0158 ms 91.1% ACC_TYPE='tl.float32', ALLOW_TF32=False, BLOCK_K=32, BLOCK_M=64, BLOCK_N=128, B_PROLOGUE_CAST_TYPE=None, EVEN_K=True, GROUP_M=8, num_stages=4, num_warps=8
triton_bmm_2089 0.0158 ms 91.1% ACC_TYPE='tl.float32', ALLOW_TF32=False, BLOCK_K=128, BLOCK_M=64, BLOCK_N=128, B_PROLOGUE_CAST_TYPE=None, EVEN_K=True, GROUP_M=8, num_stages=4, num_warps=4
triton_bmm_2093 0.0159 ms 90.5% ACC_TYPE='tl.float32', ALLOW_TF32=False, BLOCK_K=64, BLOCK_M=64, BLOCK_N=128, B_PROLOGUE_CAST_TYPE=None, EVEN_K=True, GROUP_M=8, num_stages=5, num_warps=8
SingleProcess AUTOTUNE benchmarking takes 2.0675 seconds and 0.0012 seconds precompiling
AUTOTUNE bmm(12x512x512, 12x512x64)
triton_bmm_2106 0.0140 ms 100.0% ACC_TYPE='tl.float32', ALLOW_TF32=False, BLOCK_K=128, BLOCK_M=64, BLOCK_N=64, B_PROLOGUE_CAST_TYPE=None, EVEN_K=True, GROUP_M=8, num_stages=4, num_warps=4
triton_bmm_2101 0.0141 ms 99.8% ACC_TYPE='tl.float32', ALLOW_TF32=False, BLOCK_K=64, BLOCK_M=64, BLOCK_N=64, B_PROLOGUE_CAST_TYPE=None, EVEN_K=True, GROUP_M=8, num_stages=3, num_warps=8
triton_bmm_2102 0.0141 ms 99.3% ACC_TYPE='tl.float32', ALLOW_TF32=False, BLOCK_K=128, BLOCK_M=64, BLOCK_N=64, B_PROLOGUE_CAST_TYPE=None, EVEN_K=True, GROUP_M=8, num_stages=5, num_warps=4
triton_bmm_2105 0.0146 ms 96.5% ACC_TYPE='tl.float32', ALLOW_TF32=False, BLOCK_K=64, BLOCK_M=64, BLOCK_N=64, B_PROLOGUE_CAST_TYPE=None, EVEN_K=True, GROUP_M=8, num_stages=3, num_warps=4
triton_bmm_2095 0.0149 ms 94.2% ACC_TYPE='tl.float32', ALLOW_TF32=False, BLOCK_K=128, BLOCK_M=32, BLOCK_N=32, B_PROLOGUE_CAST_TYPE=None, EVEN_K=True, GROUP_M=8, num_stages=2, num_warps=4
bmm 0.0152 ms 92.4%
triton_bmm_2111 0.0154 ms 91.1% ACC_TYPE='tl.float32', ALLOW_TF32=False, BLOCK_K=64, BLOCK_M=128, BLOCK_N=64, B_PROLOGUE_CAST_TYPE=None, EVEN_K=True, GROUP_M=8, num_stages=5, num_warps=8
triton_bmm_2096 0.0159 ms 88.5% ACC_TYPE='tl.float32', ALLOW_TF32=False, BLOCK_K=32, BLOCK_M=32, BLOCK_N=64, B_PROLOGUE_CAST_TYPE=None, EVEN_K=True, GROUP_M=8, num_stages=5, num_warps=8
triton_bmm_2104 0.0159 ms 88.3% ACC_TYPE='tl.float32', ALLOW_TF32=False, BLOCK_K=32, BLOCK_M=64, BLOCK_N=64, B_PROLOGUE_CAST_TYPE=None, EVEN_K=True, GROUP_M=8, num_stages=4, num_warps=8
triton_bmm_2110 0.0161 ms 87.5% ACC_TYPE='tl.float32', ALLOW_TF32=False, BLOCK_K=64, BLOCK_M=128, BLOCK_N=64, B_PROLOGUE_CAST_TYPE=None, EVEN_K=True, GROUP_M=8, num_stages=3, num_warps=4
SingleProcess AUTOTUNE benchmarking takes 8.3715 seconds and 0.0000 seconds precompiling
AUTOTUNE mm(512x768, 768x768)
triton_mm_2158 0.0140 ms 100.0% ACC_TYPE='tl.float32', ALLOW_TF32=False, BLOCK_K=128, BLOCK_M=64, BLOCK_N=64, B_PROLOGUE_CAST_TYPE=None, EVEN_K=True, GROUP_M=8, num_stages=5, num_warps=4
mm 0.0140 ms 99.5%
triton_mm_2162 0.0141 ms 99.1% ACC_TYPE='tl.float32', ALLOW_TF32=False, BLOCK_K=128, BLOCK_M=64, BLOCK_N=128, B_PROLOGUE_CAST_TYPE=None, EVEN_K=True, GROUP_M=8, num_stages=4, num_warps=4
triton_mm_2157 0.0145 ms 96.5% ACC_TYPE='tl.float32', ALLOW_TF32=False, BLOCK_K=64, BLOCK_M=64, BLOCK_N=64, B_PROLOGUE_CAST_TYPE=None, EVEN_K=True, GROUP_M=8, num_stages=3, num_warps=8
triton_mm_2161 0.0157 ms 88.8% ACC_TYPE='tl.float32', ALLOW_TF32=False, BLOCK_K=64, BLOCK_M=64, BLOCK_N=128, B_PROLOGUE_CAST_TYPE=None, EVEN_K=True, GROUP_M=8, num_stages=3, num_warps=4
triton_mm_2151 0.0164 ms 85.2% ACC_TYPE='tl.float32', ALLOW_TF32=False, BLOCK_K=128, BLOCK_M=32, BLOCK_N=32, B_PROLOGUE_CAST_TYPE=None, EVEN_K=True, GROUP_M=8, num_stages=2, num_warps=4
triton_mm_2153 0.0169 ms 82.8% ACC_TYPE='tl.float32', ALLOW_TF32=False, BLOCK_K=32, BLOCK_M=64, BLOCK_N=32, B_PROLOGUE_CAST_TYPE=None, EVEN_K=True, GROUP_M=8, num_stages=5, num_warps=8
triton_mm_2164 0.0170 ms 82.5% ACC_TYPE='tl.float32', ALLOW_TF32=False, BLOCK_K=32, BLOCK_M=128, BLOCK_N=64, B_PROLOGUE_CAST_TYPE=None, EVEN_K=True, GROUP_M=8, num_stages=4, num_warps=8
triton_mm_2160 0.0170 ms 82.1% ACC_TYPE='tl.float32', ALLOW_TF32=False, BLOCK_K=32, BLOCK_M=64, BLOCK_N=128, B_PROLOGUE_CAST_TYPE=None, EVEN_K=True, GROUP_M=8, num_stages=4, num_warps=8
triton_mm_2152 0.0178 ms 78.7% ACC_TYPE='tl.float32', ALLOW_TF32=False, BLOCK_K=32, BLOCK_M=32, BLOCK_N=64, B_PROLOGUE_CAST_TYPE=None, EVEN_K=True, GROUP_M=8, num_stages=5, num_warps=8
SingleProcess AUTOTUNE benchmarking takes 2.3193 seconds and 0.0017 seconds precompiling
Compiled module path: /tmp/torchinductor_shunting/tmpamk_crpd/sg/csg3bkw7nf37uhyhyonztj5px6lt7uzngrokblhbkeg2cb2dlmba.py
skipping cudagraphs due to deterministic index put. Found from :
File "/home/shunting/ws/pytorch/benchmarks/dynamo/huggingface.py", line 593, in torch_dynamo_resume_in_forward_and_backward_pass_at_591
pred = mod(**cloned_inputs)
File "/home/shunting/ws/pytorch/torch/nn/modules/module.py", line 1725, in _call_impl
return forward_call(*args, **kwargs)
File "/home/shunting/ws/miniconda3/envs/pytorch/lib/python3.10/site-packages/transformers/models/layoutlm/modeling_layoutlm.py", line 938, in forward
outputs = self.layoutlm(
File "/home/shunting/ws/pytorch/torch/nn/modules/module.py", line 1725, in _call_impl
return forward_call(*args, **kwargs)
File "/home/shunting/ws/miniconda3/envs/pytorch/lib/python3.10/site-packages/transformers/models/layoutlm/modeling_layoutlm.py", line 821, in forward
embedding_output = self.embeddings(
File "/home/shunting/ws/pytorch/torch/nn/modules/module.py", line 1725, in _call_impl
return forward_call(*args, **kwargs)
File "/home/shunting/ws/miniconda3/envs/pytorch/lib/python3.10/site-packages/transformers/models/layoutlm/modeling_layoutlm.py", line 99, in forward
inputs_embeds = self.word_embeddings(input_ids)
W0703 15:39:45.112000 140186256524288 torch/_logging/_internal.py:1040] [4/0] Profiler function <class 'torch.autograd.profiler.record_function'> will be ignored
Compiled module path: /tmp/torchinductor_shunting/tmp4i4sojlp/ba/cba5motcvjedyasdxygcxmq5iec3cn5d2iz5g47pk7a3ocncrkxn.py
E0703 15:40:53.447000 140186256524288 torch/_dynamo/utils.py:1486] RMSE (res-fp64): nan, (ref-fp64): 0.00240 and shape=torch.Size([768]). res.dtype: torch.float32, multiplier: 3.000000, tol: 0.010000
E0703 15:40:53.447000 140186256524288 torch/_dynamo/utils.py:1385] Accuracy failed for key name cls.predictions.transform.LayerNorm.bias.grad
fail_accuracy
491.12user 48.22system 5:40.23elapsed 158%CPU (0avgtext+0avgdata 9119476maxresident)k
0inputs+260512outputs (22major+4654007minor)pagefaults 0swaps
Sign up for free to join this conversation on GitHub. Already have an account? Sign in to comment