add-to-benchmarks

[drisspg]

Stacked PRs:

• ->add-to-benchmarks #2427

---

add-to-benchmarks

❯ python benchmarks/float8/bench_matmul.py --recipe "nvfp4"
TMA benchmarks will be running with experimental grid constant TMA descriptor.
gpu_name: NVIDIA B200
peak tops: bf16 2.25e+15, fp8 4.50e+15, fp4 9.00e+15
M, K, N: 1024 1024 1024 tops: 2.15E+09
torch.bfloat16 time_sec 5.08E-06, tops/sec 4.23E+14, pct_peak 0.188
time_sec 5.87E-06, tops/sec 3.66E+14, pct_peak 0.041
M, K, N: 1536 1536 1536 tops: 7.25E+09
torch.bfloat16 time_sec 7.87E-06, tops/sec 9.21E+14, pct_peak 0.410
time_sec 6.73E-06, tops/sec 1.08E+15, pct_peak 0.120
M, K, N: 2048 2048 2048 tops: 1.72E+10
torch.bfloat16 time_sec 1.52E-05, tops/sec 1.13E+15, pct_peak 0.503
time_sec 1.02E-05, tops/sec 1.68E+15, pct_peak 0.187
M, K, N: 3072 3072 3072 tops: 5.80E+10
torch.bfloat16 time_sec 3.84E-05, tops/sec 1.51E+15, pct_peak 0.672
time_sec 1.65E-05, tops/sec 3.51E+15, pct_peak 0.390
M, K, N: 4096 4096 4096 tops: 1.37E+11
torch.bfloat16 time_sec 9.74E-05, tops/sec 1.41E+15, pct_peak 0.627
time_sec 3.08E-05, tops/sec 4.46E+15, pct_peak 0.495
M, K, N: 6144 6144 6144 tops: 4.64E+11
torch.bfloat16 time_sec 3.19E-04, tops/sec 1.45E+15, pct_peak 0.646
time_sec 8.44E-05, tops/sec 5.50E+15, pct_peak 0.611
M, K, N: 8192 8192 8192 tops: 1.10E+12
torch.bfloat16 time_sec 7.74E-04, tops/sec 1.42E+15, pct_peak 0.631
time_sec 2.05E-04, tops/sec 5.37E+15, pct_peak 0.597
M, K, N: 12288 12288 12288 tops: 3.71E+12
torch.bfloat16 time_sec 2.68E-03, tops/sec 1.38E+15, pct_peak 0.615
time_sec 9.06E-04, tops/sec 4.10E+15, pct_peak 0.455
M, K, N: 16384 16384 16384 tops: 8.80E+12
torch.bfloat16 time_sec 6.26E-03, tops/sec 1.41E+15, pct_peak 0.625
time_sec 2.18E-03, tops/sec 4.04E+15, pct_peak 0.449
M, K, N: 24576 24576 24576 tops: 2.97E+13
torch.bfloat16 time_sec 2.30E-02, tops/sec 1.29E+15, pct_peak 0.574
time_sec 7.59E-03, tops/sec 3.91E+15, pct_peak 0.435
   fast_accum  name      M      K      N  ref_time_s    time_s   speedup
0       False     0   1024   1024   1024    0.000005  0.000006  0.865867
1       False     1   1536   1536   1536    0.000008  0.000007  1.168251
2       False     2   2048   2048   2048    0.000015  0.000010  1.485314
3       False     3   3072   3072   3072    0.000038  0.000016  2.325474
4       False     4   4096   4096   4096    0.000097  0.000031  3.158025
5       False     5   6144   6144   6144    0.000319  0.000084  3.781864
6       False     6   8192   8192   8192    0.000774  0.000205  3.784100
7       False     7  12288  12288  12288    0.002681  0.000906  2.960314
8       False     8  16384  16384  16384    0.006258  0.002178  2.872720
9       False     9  24576  24576  24576    0.022977  0.007589  3.027770

[pytorch-bot]

🔗 Helpful Links

🧪 See artifacts and rendered test results at hud.pytorch.org/pr/pytorch/ao/2427

• 📄 Preview Python docs built from this PR

Note: Links to docs will display an error until the docs builds have been completed.

⏳ No Failures, 7 Pending

As of commit 586b59b with merge base 7ca9f10 ():
💚 Looks good so far! There are no failures yet. 💚

This comment was automatically generated by Dr. CI and updates every 15 minutes.