narySt commited on
Commit
034fd6a
·
verified ·
1 Parent(s): 2c171a4

Add files using upload-large-folder tool

Browse files
Files changed (48) hide show
  1. .gitattributes +2 -0
  2. generation_bench_results/.ipynb_checkpoints/bolmo_code_1024B_128new_throughput_vs_batch-checkpoint.png +0 -0
  3. generation_bench_results/.ipynb_checkpoints/bolmo_code_1024B_64new_throughput_vs_batch-checkpoint.png +0 -0
  4. generation_bench_results/.ipynb_checkpoints/hnet_code_1024B_128new_speedup_vs_batch-checkpoint.png +0 -0
  5. generation_bench_results/.ipynb_checkpoints/hnet_code_1024B_128new_throughput_vs_batch-checkpoint.png +0 -0
  6. generation_bench_results/.ipynb_checkpoints/hnet_serial_code_1024B_128new_throughput_vs_batch-checkpoint.png +0 -0
  7. generation_bench_results/.ipynb_checkpoints/pythia_code_1024B_64new_speedup_vs_batch-checkpoint.png +0 -0
  8. generation_bench_results/.ipynb_checkpoints/pythia_code_1024B_64new_throughput_vs_batch-checkpoint.png +0 -0
  9. generation_bench_results/.ipynb_checkpoints/summary-checkpoint.md +256 -0
  10. generation_bench_results/bolmo_code_1024B_128new_speedup_vs_batch.png +0 -0
  11. generation_bench_results/bolmo_code_1024B_128new_throughput_vs_batch.png +0 -0
  12. generation_bench_results/bolmo_code_1024B_64new_speedup_vs_batch.png +0 -0
  13. generation_bench_results/bolmo_code_1024B_64new_throughput_vs_batch.png +0 -0
  14. generation_bench_results/bolmo_code_256B_128new_speedup_vs_batch.png +0 -0
  15. generation_bench_results/bolmo_code_256B_128new_throughput_vs_batch.png +0 -0
  16. generation_bench_results/bolmo_code_256B_64new_speedup_vs_batch.png +0 -0
  17. generation_bench_results/bolmo_code_256B_64new_throughput_vs_batch.png +0 -0
  18. generation_bench_results/hnet_code_1024B_128new_speedup_vs_batch.png +0 -0
  19. generation_bench_results/hnet_code_1024B_128new_throughput_vs_batch.png +0 -0
  20. generation_bench_results/hnet_code_1024B_64new_speedup_vs_batch.png +0 -0
  21. generation_bench_results/hnet_code_1024B_64new_throughput_vs_batch.png +0 -0
  22. generation_bench_results/hnet_code_256B_128new_speedup_vs_batch.png +0 -0
  23. generation_bench_results/hnet_code_256B_128new_throughput_vs_batch.png +0 -0
  24. generation_bench_results/hnet_code_256B_64new_speedup_vs_batch.png +0 -0
  25. generation_bench_results/hnet_code_256B_64new_throughput_vs_batch.png +0 -0
  26. generation_bench_results/hnet_serial_code_1024B_128new_speedup_vs_batch.png +0 -0
  27. generation_bench_results/hnet_serial_code_1024B_128new_throughput_vs_batch.png +0 -0
  28. generation_bench_results/hnet_serial_code_1024B_64new_speedup_vs_batch.png +0 -0
  29. generation_bench_results/hnet_serial_code_1024B_64new_throughput_vs_batch.png +0 -0
  30. generation_bench_results/hnet_serial_code_256B_128new_speedup_vs_batch.png +0 -0
  31. generation_bench_results/hnet_serial_code_256B_128new_throughput_vs_batch.png +0 -0
  32. generation_bench_results/hnet_serial_code_256B_64new_speedup_vs_batch.png +0 -0
  33. generation_bench_results/hnet_serial_code_256B_64new_throughput_vs_batch.png +0 -0
  34. generation_bench_results/metadata.json +114 -0
  35. generation_bench_results/pythia_code_1024B_128new_speedup_vs_batch.png +0 -0
  36. generation_bench_results/pythia_code_1024B_128new_throughput_vs_batch.png +0 -0
  37. generation_bench_results/pythia_code_1024B_64new_speedup_vs_batch.png +0 -0
  38. generation_bench_results/pythia_code_1024B_64new_throughput_vs_batch.png +0 -0
  39. generation_bench_results/pythia_code_256B_128new_speedup_vs_batch.png +0 -0
  40. generation_bench_results/pythia_code_256B_128new_throughput_vs_batch.png +0 -0
  41. generation_bench_results/pythia_code_256B_64new_speedup_vs_batch.png +0 -0
  42. generation_bench_results/pythia_code_256B_64new_throughput_vs_batch.png +0 -0
  43. generation_bench_results/results.json +3 -0
  44. generation_bench_results/runs.partial.jsonl +3 -0
  45. generation_bench_results/summary.csv +81 -0
  46. generation_bench_results/summary.md +256 -0
  47. generation_bench_results/warmup_runs.partial.jsonl +0 -0
  48. generation_bench_results/warmup_summary.csv +81 -0
.gitattributes CHANGED
@@ -47,3 +47,5 @@ lr_sweep/pythia_1b_lr_1e-4/wandb/run-20260425_201333-p8ozhgpm/run-p8ozhgpm.wandb
47
  lr_sweep/hnet_xl_code_lr_2e-4/wandb/run-20260425_220653-ln6tfunh/run-ln6tfunh.wandb filter=lfs diff=lfs merge=lfs -text
48
  lr_sweep/hnet_xl_code_lr_2e-4/wandb/run-20260425_222011-khn25dwv/run-khn25dwv.wandb filter=lfs diff=lfs merge=lfs -text
49
  pythia_1_4b_rerun/pythia_1_4b_lr_2e-5/wandb/run-20260425_215604-jhqe4qjw/run-jhqe4qjw.wandb filter=lfs diff=lfs merge=lfs -text
 
 
 
47
  lr_sweep/hnet_xl_code_lr_2e-4/wandb/run-20260425_220653-ln6tfunh/run-ln6tfunh.wandb filter=lfs diff=lfs merge=lfs -text
48
  lr_sweep/hnet_xl_code_lr_2e-4/wandb/run-20260425_222011-khn25dwv/run-khn25dwv.wandb filter=lfs diff=lfs merge=lfs -text
49
  pythia_1_4b_rerun/pythia_1_4b_lr_2e-5/wandb/run-20260425_215604-jhqe4qjw/run-jhqe4qjw.wandb filter=lfs diff=lfs merge=lfs -text
50
+ generation_bench_results/runs.partial.jsonl filter=lfs diff=lfs merge=lfs -text
51
+ generation_bench_results/results.json filter=lfs diff=lfs merge=lfs -text
generation_bench_results/.ipynb_checkpoints/bolmo_code_1024B_128new_throughput_vs_batch-checkpoint.png ADDED
generation_bench_results/.ipynb_checkpoints/bolmo_code_1024B_64new_throughput_vs_batch-checkpoint.png ADDED
generation_bench_results/.ipynb_checkpoints/hnet_code_1024B_128new_speedup_vs_batch-checkpoint.png ADDED
generation_bench_results/.ipynb_checkpoints/hnet_code_1024B_128new_throughput_vs_batch-checkpoint.png ADDED
generation_bench_results/.ipynb_checkpoints/hnet_serial_code_1024B_128new_throughput_vs_batch-checkpoint.png ADDED
generation_bench_results/.ipynb_checkpoints/pythia_code_1024B_64new_speedup_vs_batch-checkpoint.png ADDED
generation_bench_results/.ipynb_checkpoints/pythia_code_1024B_64new_throughput_vs_batch-checkpoint.png ADDED
generation_bench_results/.ipynb_checkpoints/summary-checkpoint.md ADDED
@@ -0,0 +1,256 @@
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
1
+ # Generation Speed Benchmark
2
+
3
+ ## Speed
4
+
5
+ | Model | Workload | Batch | Max new | Samples | Status | Prompt bytes | Prefill ms/chunk | Prefill bytes/s | Prefill units/s | Decode ms/chunk | Decode bytes/s | Decode units/s | Total bytes/s | Speedup vs B=1 | Peak VRAM MB |
6
+ |---|---|---:|---:|---:|---|---:|---:|---:|---:|---:|---:|---:|---:|---:|---:|
7
+ | bolmo | code/1024B | 1 | 64 | 384 | ok | 472.1 | 68.8 | 6866.1 | 6880.6 | 4836.2 | 110.4 | 110.4 | 100.4 | 1.0 | 2938.8 |
8
+ | bolmo | code/1024B | 1 | 128 | 384 | ok | 472.1 | 44.8 | 10547.7 | 10570.1 | 5402.8 | 110.6 | 110.6 | 101.0 | 1.0 | 2938.8 |
9
+ | bolmo | code/1024B | 2 | 64 | 384 | ok | 472.1 | 144.4 | 6539.5 | 6553.4 | 8605.5 | 146.7 | 146.7 | 129.5 | 1.29 | 3027.4 |
10
+ | bolmo | code/1024B | 2 | 128 | 384 | ok | 472.1 | 47.3 | 19952.9 | 19995.1 | 9472.7 | 146.7 | 146.7 | 130.8 | 1.29 | 3027.4 |
11
+ | bolmo | code/1024B | 4 | 64 | 384 | ok | 472.1 | 188.2 | 10030.6 | 10051.9 | 13762.9 | 213.0 | 213.0 | 178.3 | 1.78 | 3229.9 |
12
+ | bolmo | code/1024B | 4 | 128 | 384 | ok | 472.1 | 51.5 | 36681.4 | 36759.1 | 14978.1 | 212.1 | 212.1 | 178.8 | 1.77 | 3229.9 |
13
+ | bolmo | code/1024B | 8 | 64 | 384 | ok | 472.1 | 387.7 | 9741.6 | 9762.2 | 20813.7 | 319.9 | 319.9 | 238.0 | 2.37 | 3616.3 |
14
+ | bolmo | code/1024B | 8 | 128 | 384 | ok | 472.1 | 76.8 | 49167.3 | 49271.5 | 22267.3 | 320.7 | 320.7 | 240.1 | 2.38 | 3616.3 |
15
+ | bolmo | code/1024B | 16 | 64 | 384 | ok | 472.1 | 452.1 | 16706.1 | 16741.5 | 28969.9 | 501.5 | 501.5 | 298.8 | 2.98 | 4407.8 |
16
+ | bolmo | code/1024B | 16 | 128 | 384 | ok | 472.1 | 138.9 | 54365.5 | 54480.7 | 31091.6 | 499.2 | 499.2 | 296.2 | 2.93 | 4407.8 |
17
+ | bolmo | code/256B | 1 | 64 | 384 | ok | 196.7 | 43.8 | 4490.7 | 4513.6 | 2372.5 | 109.0 | 109.0 | 99.3 | 1.0 | 2854.9 |
18
+ | bolmo | code/256B | 1 | 128 | 384 | ok | 196.7 | 44.0 | 4465.3 | 4488.0 | 2964.5 | 108.8 | 108.8 | 99.4 | 1.0 | 2858.6 |
19
+ | bolmo | code/256B | 2 | 64 | 384 | ok | 196.7 | 72.1 | 5455.3 | 5483.0 | 3815.8 | 145.9 | 145.9 | 130.2 | 1.31 | 2896.6 |
20
+ | bolmo | code/256B | 2 | 128 | 384 | ok | 196.7 | 46.9 | 8392.3 | 8435.0 | 4676.2 | 146.2 | 146.2 | 131.3 | 1.32 | 2896.6 |
21
+ | bolmo | code/256B | 4 | 64 | 384 | ok | 196.7 | 70.6 | 11143.7 | 11200.4 | 5184.4 | 220.7 | 220.7 | 191.2 | 1.93 | 2965.1 |
22
+ | bolmo | code/256B | 4 | 128 | 384 | ok | 196.7 | 52.1 | 15110.2 | 15187.0 | 6306.6 | 221.1 | 221.1 | 192.1 | 1.93 | 2966.3 |
23
+ | bolmo | code/256B | 8 | 64 | 384 | ok | 196.7 | 118.2 | 13311.2 | 13378.9 | 6677.2 | 355.1 | 355.1 | 287.1 | 2.89 | 3064.1 |
24
+ | bolmo | code/256B | 8 | 128 | 384 | ok | 196.7 | 53.5 | 29410.2 | 29559.7 | 8085.7 | 354.6 | 354.6 | 288.0 | 2.9 | 3102.8 |
25
+ | bolmo | code/256B | 16 | 64 | 384 | ok | 196.7 | 181.9 | 17296.9 | 17384.9 | 8502.9 | 578.0 | 578.0 | 410.5 | 4.13 | 3308.9 |
26
+ | bolmo | code/256B | 16 | 128 | 384 | ok | 196.7 | 68.3 | 46070.8 | 46305.0 | 10256.4 | 578.3 | 578.3 | 411.2 | 4.14 | 3374.6 |
27
+ | hnet | code/1024B | 1 | 64 | 384 | ok | 472.1 | 62.2 | 7595.3 | 7611.4 | 774.4 | 82.6 | 82.6 | 76.1 | 1.0 | 3518.6 |
28
+ | hnet | code/1024B | 1 | 128 | 384 | ok | 472.1 | 62.1 | 7601.1 | 7617.2 | 1574.6 | 81.3 | 81.3 | 78.0 | 1.0 | 3533.7 |
29
+ | hnet | code/1024B | 2 | 64 | 384 | ok | 472.1 | 64.7 | 14589.9 | 14620.8 | 1080.6 | 118.5 | 118.5 | 110.8 | 1.46 | 3801.8 |
30
+ | hnet | code/1024B | 2 | 128 | 384 | ok | 472.1 | 64.0 | 14754.4 | 14785.6 | 2190.2 | 116.9 | 116.9 | 113.0 | 1.45 | 3849.3 |
31
+ | hnet | code/1024B | 4 | 64 | 384 | ok | 472.1 | 68.2 | 27690.1 | 27748.7 | 1481.8 | 172.8 | 172.8 | 163.2 | 2.14 | 4378.3 |
32
+ | hnet | code/1024B | 4 | 128 | 384 | ok | 472.1 | 67.2 | 28092.9 | 28152.4 | 2985.5 | 171.5 | 171.5 | 166.6 | 2.14 | 4437.3 |
33
+ | hnet | code/1024B | 8 | 64 | 384 | ok | 472.1 | 82.0 | 46036.4 | 46133.9 | 1892.0 | 270.6 | 270.6 | 254.8 | 3.35 | 5463.0 |
34
+ | hnet | code/1024B | 8 | 128 | 384 | ok | 472.1 | 78.1 | 48370.5 | 48473.0 | 3832.7 | 267.2 | 267.2 | 259.5 | 3.33 | 5579.6 |
35
+ | hnet | code/1024B | 16 | 64 | 384 | ok | 472.1 | 103.5 | 72962.8 | 73117.3 | 2234.0 | 458.4 | 458.4 | 429.2 | 5.64 | 7649.6 |
36
+ | hnet | code/1024B | 16 | 128 | 384 | ok | 472.1 | 101.8 | 74199.7 | 74356.9 | 4465.0 | 458.7 | 458.7 | 443.7 | 5.69 | 7885.6 |
37
+ | hnet | code/256B | 1 | 64 | 384 | ok | 196.7 | 65.4 | 3005.8 | 3021.1 | 805.0 | 79.5 | 79.5 | 73.3 | 1.0 | 3316.6 |
38
+ | hnet | code/256B | 1 | 128 | 384 | ok | 196.7 | 61.7 | 3187.4 | 3203.6 | 1610.2 | 79.5 | 79.5 | 76.4 | 1.0 | 3339.2 |
39
+ | hnet | code/256B | 2 | 64 | 384 | ok | 196.7 | 64.0 | 6143.2 | 6174.4 | 1128.4 | 113.4 | 113.4 | 107.0 | 1.46 | 3409.3 |
40
+ | hnet | code/256B | 2 | 128 | 384 | ok | 196.7 | 62.9 | 6258.4 | 6290.2 | 2248.0 | 113.9 | 113.9 | 110.5 | 1.45 | 3441.1 |
41
+ | hnet | code/256B | 4 | 64 | 384 | ok | 196.7 | 65.9 | 11943.4 | 12004.1 | 1509.6 | 169.6 | 169.6 | 161.6 | 2.2 | 3606.9 |
42
+ | hnet | code/256B | 4 | 128 | 384 | ok | 196.7 | 66.2 | 11882.8 | 11943.2 | 3046.2 | 168.1 | 168.1 | 163.9 | 2.14 | 3665.3 |
43
+ | hnet | code/256B | 8 | 64 | 384 | ok | 196.7 | 72.9 | 21582.5 | 21692.3 | 1907.4 | 268.4 | 268.4 | 256.3 | 3.5 | 3976.9 |
44
+ | hnet | code/256B | 8 | 128 | 384 | ok | 196.7 | 72.7 | 21647.2 | 21757.3 | 3853.7 | 265.7 | 265.7 | 259.4 | 3.39 | 4092.1 |
45
+ | hnet | code/256B | 16 | 64 | 384 | ok | 196.7 | 86.7 | 36276.7 | 36461.2 | 2250.8 | 454.9 | 454.9 | 431.9 | 5.89 | 4716.5 |
46
+ | hnet | code/256B | 16 | 128 | 384 | ok | 196.7 | 86.7 | 36304.2 | 36488.8 | 4488.7 | 456.3 | 456.3 | 443.9 | 5.81 | 4952.5 |
47
+ | hnet_serial | code/1024B | 1 | 64 | 384 | ok | 472.1 | 60.2 | 7838.1 | 7854.7 | 769.5 | 83.2 | 83.2 | 76.6 | 1.0 | 3518.0 |
48
+ | hnet_serial | code/1024B | 1 | 128 | 384 | ok | 472.1 | 59.7 | 7905.9 | 7922.7 | 1556.9 | 82.2 | 82.2 | 78.9 | 1.0 | 3532.7 |
49
+ | hnet_serial | code/1024B | 2 | 64 | 384 | ok | 472.1 | 117.4 | 8042.9 | 8059.9 | 1546.0 | 82.8 | 82.8 | 76.5 | 1.0 | 3698.3 |
50
+ | hnet_serial | code/1024B | 2 | 128 | 384 | ok | 472.1 | 116.7 | 8090.6 | 8107.8 | 3108.5 | 82.4 | 82.4 | 79.1 | 1.0 | 3735.8 |
51
+ | hnet_serial | code/1024B | 4 | 64 | 384 | ok | 472.1 | 227.7 | 8291.5 | 8309.1 | 3056.5 | 83.8 | 83.8 | 77.6 | 1.01 | 3698.3 |
52
+ | hnet_serial | code/1024B | 4 | 128 | 384 | ok | 472.1 | 229.7 | 8219.7 | 8237.1 | 6225.2 | 82.2 | 82.2 | 79.1 | 1.0 | 3735.8 |
53
+ | hnet_serial | code/1024B | 8 | 64 | 384 | ok | 472.1 | 451.3 | 8367.9 | 8385.6 | 6135.1 | 83.5 | 83.5 | 77.4 | 1.01 | 3698.3 |
54
+ | hnet_serial | code/1024B | 8 | 128 | 384 | ok | 472.1 | 451.5 | 8364.2 | 8381.9 | 12405.5 | 82.5 | 82.5 | 79.5 | 1.01 | 3735.8 |
55
+ | hnet_serial | code/1024B | 16 | 64 | 384 | ok | 472.1 | 895.0 | 8438.8 | 8456.7 | 12254.5 | 83.6 | 83.6 | 77.6 | 1.01 | 3699.0 |
56
+ | hnet_serial | code/1024B | 16 | 128 | 384 | ok | 472.1 | 895.7 | 8432.5 | 8450.3 | 24806.4 | 82.6 | 82.6 | 79.5 | 1.01 | 3735.2 |
57
+ | hnet_serial | code/256B | 1 | 64 | 384 | ok | 196.7 | 65.8 | 2989.4 | 3004.6 | 798.1 | 80.2 | 80.2 | 73.7 | 1.0 | 3314.1 |
58
+ | hnet_serial | code/256B | 1 | 128 | 384 | ok | 196.7 | 59.2 | 3324.8 | 3341.7 | 1594.4 | 80.3 | 80.3 | 77.2 | 1.0 | 3338.8 |
59
+ | hnet_serial | code/256B | 2 | 64 | 384 | ok | 196.7 | 115.7 | 3400.6 | 3417.9 | 1587.2 | 80.6 | 80.6 | 74.8 | 1.01 | 3397.0 |
60
+ | hnet_serial | code/256B | 2 | 128 | 384 | ok | 196.7 | 116.1 | 3386.8 | 3404.0 | 3191.0 | 80.2 | 80.2 | 77.2 | 1.0 | 3426.6 |
61
+ | hnet_serial | code/256B | 4 | 64 | 384 | ok | 196.7 | 226.5 | 3474.1 | 3491.7 | 3186.1 | 80.3 | 80.3 | 74.8 | 1.01 | 3397.0 |
62
+ | hnet_serial | code/256B | 4 | 128 | 384 | ok | 196.7 | 226.3 | 3476.8 | 3494.5 | 6365.5 | 80.4 | 80.4 | 77.5 | 1.0 | 3426.6 |
63
+ | hnet_serial | code/256B | 8 | 64 | 384 | ok | 196.7 | 448.5 | 3508.4 | 3526.2 | 6371.8 | 80.4 | 80.4 | 74.9 | 1.02 | 3397.0 |
64
+ | hnet_serial | code/256B | 8 | 128 | 384 | ok | 196.7 | 448.8 | 3506.1 | 3524.0 | 12761.3 | 80.2 | 80.2 | 77.4 | 1.0 | 3426.6 |
65
+ | hnet_serial | code/256B | 16 | 64 | 384 | ok | 196.7 | 887.2 | 3546.8 | 3564.8 | 12710.5 | 80.6 | 80.6 | 75.1 | 1.02 | 3397.0 |
66
+ | hnet_serial | code/256B | 16 | 128 | 384 | ok | 196.7 | 888.2 | 3543.1 | 3561.2 | 25442.9 | 80.5 | 80.5 | 77.7 | 1.01 | 3426.6 |
67
+ | pythia | code/1024B | 1 | 64 | 384 | ok | 472.1 | 15.4 | 30579.0 | 8838.6 | 587.1 | 389.3 | 109.0 | 379.3 | 1.0 | 2014.9 |
68
+ | pythia | code/1024B | 1 | 128 | 384 | ok | 472.1 | 15.2 | 31063.0 | 8978.5 | 1172.3 | 387.7 | 109.2 | 382.7 | 1.0 | 2014.9 |
69
+ | pythia | code/1024B | 2 | 64 | 384 | ok | 472.1 | 18.8 | 50344.4 | 14551.6 | 593.8 | 697.0 | 208.7 | 675.6 | 1.78 | 2088.9 |
70
+ | pythia | code/1024B | 2 | 128 | 384 | ok | 472.1 | 18.4 | 51420.9 | 14862.7 | 1199.3 | 676.4 | 202.5 | 666.2 | 1.74 | 2088.9 |
71
+ | pythia | code/1024B | 4 | 64 | 384 | ok | 472.1 | 19.6 | 96476.0 | 27885.5 | 604.9 | 1293.3 | 409.4 | 1252.7 | 3.3 | 2244.1 |
72
+ | pythia | code/1024B | 4 | 128 | 384 | ok | 472.1 | 19.8 | 95362.1 | 27563.5 | 1230.2 | 1243.9 | 397.4 | 1224.2 | 3.2 | 2244.1 |
73
+ | pythia | code/1024B | 8 | 64 | 384 | ok | 472.1 | 31.4 | 120427.6 | 34808.5 | 653.0 | 2353.4 | 770.8 | 2245.5 | 5.92 | 2544.6 |
74
+ | pythia | code/1024B | 8 | 128 | 384 | ok | 472.1 | 31.9 | 118556.1 | 34267.5 | 1295.0 | 2306.7 | 763.7 | 2251.2 | 5.88 | 2544.6 |
75
+ | pythia | code/1024B | 16 | 64 | 384 | ok | 472.1 | 63.2 | 119473.2 | 34532.6 | 638.2 | 4639.1 | 1570.3 | 4220.7 | 11.13 | 3153.5 |
76
+ | pythia | code/1024B | 16 | 128 | 384 | ok | 472.1 | 62.4 | 121004.4 | 34975.2 | 1273.0 | 4533.8 | 1543.7 | 4321.7 | 11.29 | 3153.5 |
77
+ | pythia | code/256B | 1 | 64 | 384 | ok | 196.7 | 12.9 | 15205.4 | 4704.4 | 585.0 | 365.5 | 108.1 | 357.6 | 1.0 | 1959.7 |
78
+ | pythia | code/256B | 1 | 128 | 384 | ok | 196.7 | 13.0 | 15153.2 | 4688.3 | 1157.3 | 359.9 | 108.7 | 355.9 | 1.0 | 1966.7 |
79
+ | pythia | code/256B | 2 | 64 | 384 | ok | 196.7 | 18.9 | 20759.3 | 6422.8 | 597.1 | 698.3 | 204.9 | 676.8 | 1.89 | 1996.4 |
80
+ | pythia | code/256B | 2 | 128 | 384 | ok | 196.7 | 18.6 | 21139.2 | 6540.3 | 1186.2 | 684.0 | 202.2 | 673.5 | 1.89 | 2001.3 |
81
+ | pythia | code/256B | 4 | 64 | 384 | ok | 196.7 | 18.3 | 42908.1 | 13275.5 | 600.9 | 1291.7 | 401.0 | 1253.4 | 3.51 | 2023.1 |
82
+ | pythia | code/256B | 4 | 128 | 384 | ok | 196.7 | 17.3 | 45482.3 | 14072.0 | 1210.3 | 1232.1 | 387.5 | 1214.7 | 3.41 | 2057.8 |
83
+ | pythia | code/256B | 8 | 64 | 384 | ok | 196.7 | 19.6 | 80095.9 | 24781.2 | 621.1 | 2436.6 | 775.5 | 2361.8 | 6.61 | 2109.2 |
84
+ | pythia | code/256B | 8 | 128 | 384 | ok | 196.7 | 19.4 | 81135.3 | 25102.8 | 1236.0 | 2357.2 | 759.3 | 2320.8 | 6.52 | 2167.8 |
85
+ | pythia | code/256B | 16 | 64 | 384 | ok | 196.7 | 24.8 | 126664.5 | 39189.3 | 623.6 | 4867.8 | 1503.2 | 4681.0 | 13.09 | 2279.2 |
86
+ | pythia | code/256B | 16 | 128 | 384 | ok | 196.7 | 25.3 | 124556.5 | 38537.1 | 1266.8 | 4524.9 | 1406.5 | 4436.3 | 12.47 | 2398.3 |
87
+
88
+ ## Warmup
89
+
90
+ | Model | Workload | Batch | Max new | Samples | Status | Prompt bytes | Prefill ms/chunk | Decode ms/chunk | Total ms/chunk | Peak VRAM MB |
91
+ |---|---|---:|---:|---:|---|---:|---:|---:|---:|---:|
92
+ | bolmo | code/1024B | 1 | 64 | 1 | ok | 265.0 | 43.8 | 3052.0 | 3341.1 | 2866.2 |
93
+ | bolmo | code/1024B | 1 | 128 | 1 | ok | 265.0 | 42.4 | 3651.7 | 3987.2 | 2857.4 |
94
+ | bolmo | code/1024B | 2 | 64 | 2 | ok | 613.5 | 1894.2 | 13495.6 | 17069.0 | 3026.0 |
95
+ | bolmo | code/1024B | 2 | 128 | 2 | ok | 613.5 | 38.1 | 14306.3 | 16128.1 | 3026.0 |
96
+ | bolmo | code/1024B | 4 | 64 | 4 | ok | 531.5 | 56.9 | 18632.9 | 22371.1 | 3217.5 |
97
+ | bolmo | code/1024B | 4 | 128 | 4 | ok | 531.5 | 50.8 | 19793.1 | 23814.4 | 3217.5 |
98
+ | bolmo | code/1024B | 8 | 64 | 8 | ok | 425.2 | 1960.2 | 24538.3 | 35218.8 | 3572.6 |
99
+ | bolmo | code/1024B | 8 | 128 | 8 | ok | 425.2 | 83.9 | 26349.0 | 35921.7 | 3572.6 |
100
+ | bolmo | code/1024B | 16 | 64 | 16 | ok | 423.9 | 2017.6 | 25250.7 | 43045.2 | 4319.6 |
101
+ | bolmo | code/1024B | 16 | 128 | 16 | ok | 423.9 | 146.6 | 27043.3 | 44540.5 | 4319.6 |
102
+ | bolmo | code/256B | 1 | 64 | 1 | ok | 212.0 | 80.8 | 2573.0 | 5345.6 | 2848.9 |
103
+ | bolmo | code/256B | 1 | 128 | 1 | ok | 212.0 | 38.8 | 3186.3 | 3472.0 | 2848.9 |
104
+ | bolmo | code/256B | 2 | 64 | 2 | ok | 196.0 | 2709.1 | 3626.6 | 6705.1 | 2876.3 |
105
+ | bolmo | code/256B | 2 | 128 | 2 | ok | 196.0 | 40.4 | 4453.0 | 4954.2 | 2876.3 |
106
+ | bolmo | code/256B | 4 | 64 | 4 | ok | 187.8 | 55.1 | 4618.7 | 5335.7 | 2924.8 |
107
+ | bolmo | code/256B | 4 | 128 | 4 | ok | 187.8 | 117.3 | 5647.9 | 6577.9 | 2924.8 |
108
+ | bolmo | code/256B | 8 | 64 | 8 | ok | 190.2 | 3023.3 | 6481.8 | 10735.6 | 3026.6 |
109
+ | bolmo | code/256B | 8 | 128 | 8 | ok | 190.2 | 49.6 | 6992.3 | 8624.1 | 3026.6 |
110
+ | bolmo | code/256B | 16 | 64 | 16 | ok | 179.8 | 3332.2 | 9968.8 | 16233.8 | 3239.8 |
111
+ | bolmo | code/256B | 16 | 128 | 16 | ok | 179.8 | 60.9 | 9696.9 | 13438.3 | 3252.2 |
112
+ | hnet | code/1024B | 1 | 64 | 1 | ok | 265.0 | 63.9 | 884.9 | 953.0 | 3319.5 |
113
+ | hnet | code/1024B | 1 | 128 | 1 | ok | 265.0 | 61.1 | 1819.8 | 1883.8 | 3343.8 |
114
+ | hnet | code/1024B | 2 | 64 | 2 | ok | 613.5 | 64.6 | 1288.2 | 1362.7 | 3773.8 |
115
+ | hnet | code/1024B | 2 | 128 | 2 | ok | 613.5 | 64.0 | 2531.4 | 2608.4 | 3826.2 |
116
+ | hnet | code/1024B | 4 | 64 | 4 | ok | 531.5 | 69.6 | 1531.7 | 1614.6 | 4310.4 |
117
+ | hnet | code/1024B | 4 | 128 | 4 | ok | 531.5 | 66.9 | 3024.8 | 3114.0 | 4370.2 |
118
+ | hnet | code/1024B | 8 | 64 | 8 | ok | 425.2 | 74.7 | 1765.8 | 1865.1 | 5350.4 |
119
+ | hnet | code/1024B | 8 | 128 | 8 | ok | 425.2 | 77.9 | 3662.8 | 3776.5 | 5470.1 |
120
+ | hnet | code/1024B | 16 | 64 | 16 | ok | 423.9 | 94.4 | 2196.2 | 2341.5 | 7480.3 |
121
+ | hnet | code/1024B | 16 | 128 | 16 | ok | 423.9 | 103.2 | 4417.9 | 4566.6 | 7716.3 |
122
+ | hnet | code/256B | 1 | 64 | 1 | ok | 212.0 | 20030.1 | 1047.3 | 21080.4 | 3556.9 |
123
+ | hnet | code/256B | 1 | 128 | 1 | ok | 212.0 | 61.1 | 1683.1 | 1747.1 | 3318.4 |
124
+ | hnet | code/256B | 2 | 64 | 2 | ok | 196.0 | 207.4 | 1191.1 | 1402.4 | 3384.2 |
125
+ | hnet | code/256B | 2 | 128 | 2 | ok | 196.0 | 63.0 | 2332.0 | 2399.4 | 3414.9 |
126
+ | hnet | code/256B | 4 | 64 | 4 | ok | 187.8 | 68.3 | 1503.7 | 1578.2 | 3559.3 |
127
+ | hnet | code/256B | 4 | 128 | 4 | ok | 187.8 | 66.7 | 3104.5 | 3182.8 | 3616.0 |
128
+ | hnet | code/256B | 8 | 64 | 8 | ok | 190.2 | 76.7 | 1872.6 | 1960.9 | 3903.5 |
129
+ | hnet | code/256B | 8 | 128 | 8 | ok | 190.2 | 73.2 | 3763.2 | 3853.8 | 4020.3 |
130
+ | hnet | code/256B | 16 | 64 | 16 | ok | 179.8 | 85.6 | 2464.0 | 2571.8 | 4640.5 |
131
+ | hnet | code/256B | 16 | 128 | 16 | ok | 179.8 | 85.2 | 4508.5 | 4627.2 | 4876.5 |
132
+ | hnet_serial | code/1024B | 1 | 64 | 1 | ok | 265.0 | 752.1 | 890.2 | 1646.6 | 3317.7 |
133
+ | hnet_serial | code/1024B | 1 | 128 | 1 | ok | 265.0 | 59.6 | 1813.3 | 1877.3 | 3343.6 |
134
+ | hnet_serial | code/1024B | 2 | 64 | 2 | ok | 613.5 | 117.1 | 1838.9 | 1967.3 | 3598.2 |
135
+ | hnet_serial | code/1024B | 2 | 128 | 2 | ok | 613.5 | 115.7 | 3463.3 | 3591.9 | 3635.6 |
136
+ | hnet_serial | code/1024B | 4 | 64 | 4 | ok | 531.5 | 225.7 | 3236.4 | 3479.5 | 3608.5 |
137
+ | hnet_serial | code/1024B | 4 | 128 | 4 | ok | 531.5 | 229.7 | 6386.6 | 6634.0 | 3635.6 |
138
+ | hnet_serial | code/1024B | 8 | 64 | 8 | ok | 425.2 | 452.2 | 6272.0 | 6749.1 | 3608.5 |
139
+ | hnet_serial | code/1024B | 8 | 128 | 8 | ok | 425.2 | 450.3 | 12604.3 | 13080.9 | 3635.6 |
140
+ | hnet_serial | code/1024B | 16 | 64 | 16 | ok | 423.9 | 894.5 | 12325.2 | 13261.0 | 3608.5 |
141
+ | hnet_serial | code/1024B | 16 | 128 | 16 | ok | 423.9 | 900.3 | 25133.6 | 26075.5 | 3635.6 |
142
+ | hnet_serial | code/256B | 1 | 64 | 1 | ok | 212.0 | 30196.8 | 847.5 | 31047.6 | 3554.1 |
143
+ | hnet_serial | code/256B | 1 | 128 | 1 | ok | 212.0 | 58.6 | 1740.8 | 1803.3 | 3317.0 |
144
+ | hnet_serial | code/256B | 2 | 64 | 2 | ok | 196.0 | 116.2 | 1691.3 | 1814.3 | 3370.4 |
145
+ | hnet_serial | code/256B | 2 | 128 | 2 | ok | 196.0 | 116.4 | 3369.0 | 3499.2 | 3398.4 |
146
+ | hnet_serial | code/256B | 4 | 64 | 4 | ok | 187.8 | 226.9 | 3276.3 | 3513.7 | 3370.4 |
147
+ | hnet_serial | code/256B | 4 | 128 | 4 | ok | 187.8 | 225.6 | 6515.2 | 6751.9 | 3398.4 |
148
+ | hnet_serial | code/256B | 8 | 64 | 8 | ok | 190.2 | 447.1 | 6504.2 | 6969.9 | 3376.2 |
149
+ | hnet_serial | code/256B | 8 | 128 | 8 | ok | 190.2 | 449.7 | 12838.5 | 13306.9 | 3404.0 |
150
+ | hnet_serial | code/256B | 16 | 64 | 16 | ok | 179.8 | 884.6 | 13329.9 | 14247.4 | 3382.4 |
151
+ | hnet_serial | code/256B | 16 | 128 | 16 | ok | 179.8 | 887.1 | 26288.8 | 27209.4 | 3413.2 |
152
+ | pythia | code/1024B | 1 | 64 | 1 | ok | 265.0 | 16.3 | 589.7 | 606.0 | 1959.9 |
153
+ | pythia | code/1024B | 1 | 128 | 1 | ok | 265.0 | 13.7 | 1162.3 | 1176.1 | 1966.8 |
154
+ | pythia | code/1024B | 2 | 64 | 2 | ok | 613.5 | 34.4 | 610.7 | 645.2 | 2065.1 |
155
+ | pythia | code/1024B | 2 | 128 | 2 | ok | 613.5 | 19.9 | 1176.4 | 1196.4 | 2065.1 |
156
+ | pythia | code/1024B | 4 | 64 | 4 | ok | 531.5 | 20.4 | 602.2 | 622.7 | 2192.2 |
157
+ | pythia | code/1024B | 4 | 128 | 4 | ok | 531.5 | 20.1 | 1202.6 | 1222.8 | 2192.2 |
158
+ | pythia | code/1024B | 8 | 64 | 8 | ok | 425.2 | 32.5 | 615.5 | 648.0 | 2438.1 |
159
+ | pythia | code/1024B | 8 | 128 | 8 | ok | 425.2 | 37.0 | 1230.5 | 1267.5 | 2438.1 |
160
+ | pythia | code/1024B | 16 | 64 | 16 | ok | 423.9 | 59.5 | 639.4 | 698.9 | 2944.8 |
161
+ | pythia | code/1024B | 16 | 128 | 16 | ok | 423.9 | 65.0 | 1277.3 | 1342.3 | 2944.8 |
162
+ | pythia | code/256B | 1 | 64 | 1 | ok | 212.0 | 18.7 | 597.2 | 616.0 | 1957.2 |
163
+ | pythia | code/256B | 1 | 128 | 1 | ok | 212.0 | 12.8 | 1176.5 | 1189.4 | 1965.2 |
164
+ | pythia | code/256B | 2 | 64 | 2 | ok | 196.0 | 21.6 | 599.2 | 620.8 | 1996.1 |
165
+ | pythia | code/256B | 2 | 128 | 2 | ok | 196.0 | 18.8 | 1179.8 | 1198.6 | 1998.0 |
166
+ | pythia | code/256B | 4 | 64 | 4 | ok | 187.8 | 16.3 | 597.9 | 614.2 | 2014.2 |
167
+ | pythia | code/256B | 4 | 128 | 4 | ok | 187.8 | 16.2 | 1197.6 | 1213.8 | 2053.4 |
168
+ | pythia | code/256B | 8 | 64 | 8 | ok | 190.2 | 19.3 | 615.4 | 634.7 | 2088.5 |
169
+ | pythia | code/256B | 8 | 128 | 8 | ok | 190.2 | 19.0 | 1233.3 | 1252.3 | 2156.6 |
170
+ | pythia | code/256B | 16 | 64 | 16 | ok | 179.8 | 22.3 | 624.2 | 646.5 | 2239.5 |
171
+ | pythia | code/256B | 16 | 128 | 16 | ok | 179.8 | 23.7 | 1272.3 | 1296.0 | 2382.3 |
172
+
173
+ ## Compression
174
+
175
+ | Model | Workload | Raw prompt bytes | Model input units | Stage0 / patches | Stage1 | Bytes per stage0 | Stage0 per stage1 |
176
+ |---|---|---:|---:|---:|---:|---:|---:|
177
+ | bolmo | code/1024B | 181272.0 | 181656.0 | 40221.0 | None | 4.51 | None |
178
+ | bolmo | code/1024B | 181272.0 | 181656.0 | 40221.0 | None | 4.51 | None |
179
+ | bolmo | code/1024B | 181272.0 | 181656.0 | 40206.0 | None | 4.51 | None |
180
+ | bolmo | code/1024B | 181272.0 | 181656.0 | 40206.0 | None | 4.51 | None |
181
+ | bolmo | code/1024B | 181272.0 | 181656.0 | 40194.0 | None | 4.51 | None |
182
+ | bolmo | code/1024B | 181272.0 | 181656.0 | 40194.0 | None | 4.51 | None |
183
+ | bolmo | code/1024B | 181272.0 | 181656.0 | 40212.0 | None | 4.51 | None |
184
+ | bolmo | code/1024B | 181272.0 | 181656.0 | 40212.0 | None | 4.51 | None |
185
+ | bolmo | code/1024B | 181272.0 | 181656.0 | 40209.0 | None | 4.51 | None |
186
+ | bolmo | code/1024B | 181272.0 | 181656.0 | 40209.0 | None | 4.51 | None |
187
+ | bolmo | code/256B | 75525.0 | 75909.0 | 17772.0 | None | 4.25 | None |
188
+ | bolmo | code/256B | 75525.0 | 75909.0 | 17772.0 | None | 4.25 | None |
189
+ | bolmo | code/256B | 75525.0 | 75909.0 | 17769.0 | None | 4.25 | None |
190
+ | bolmo | code/256B | 75525.0 | 75909.0 | 17769.0 | None | 4.25 | None |
191
+ | bolmo | code/256B | 75525.0 | 75909.0 | 17769.0 | None | 4.25 | None |
192
+ | bolmo | code/256B | 75525.0 | 75909.0 | 17769.0 | None | 4.25 | None |
193
+ | bolmo | code/256B | 75525.0 | 75909.0 | 17775.0 | None | 4.25 | None |
194
+ | bolmo | code/256B | 75525.0 | 75909.0 | 17775.0 | None | 4.25 | None |
195
+ | bolmo | code/256B | 75525.0 | 75909.0 | 17763.0 | None | 4.25 | None |
196
+ | bolmo | code/256B | 75525.0 | 75909.0 | 17763.0 | None | 4.25 | None |
197
+ | hnet | code/1024B | 181272.0 | 181656.0 | 63561.0 | 30735.0 | 2.85 | 2.07 |
198
+ | hnet | code/1024B | 181272.0 | 181656.0 | 63561.0 | 30735.0 | 2.85 | 2.07 |
199
+ | hnet | code/1024B | 181272.0 | 181656.0 | 63564.0 | 30726.0 | 2.85 | 2.07 |
200
+ | hnet | code/1024B | 181272.0 | 181656.0 | 63564.0 | 30726.0 | 2.85 | 2.07 |
201
+ | hnet | code/1024B | 181272.0 | 181656.0 | 63576.0 | 30720.0 | 2.85 | 2.07 |
202
+ | hnet | code/1024B | 181272.0 | 181656.0 | 63576.0 | 30720.0 | 2.85 | 2.07 |
203
+ | hnet | code/1024B | 181272.0 | 181656.0 | 63573.0 | 30723.0 | 2.85 | 2.07 |
204
+ | hnet | code/1024B | 181272.0 | 181656.0 | 63573.0 | 30723.0 | 2.85 | 2.07 |
205
+ | hnet | code/1024B | 181272.0 | 181656.0 | 63570.0 | 30729.0 | 2.85 | 2.07 |
206
+ | hnet | code/1024B | 181272.0 | 181656.0 | 63570.0 | 30729.0 | 2.85 | 2.07 |
207
+ | hnet | code/256B | 75525.0 | 75909.0 | 28476.0 | 14922.0 | 2.65 | 1.91 |
208
+ | hnet | code/256B | 75525.0 | 75909.0 | 28476.0 | 14922.0 | 2.65 | 1.91 |
209
+ | hnet | code/256B | 75525.0 | 75909.0 | 28479.0 | 14916.0 | 2.65 | 1.91 |
210
+ | hnet | code/256B | 75525.0 | 75909.0 | 28479.0 | 14916.0 | 2.65 | 1.91 |
211
+ | hnet | code/256B | 75525.0 | 75909.0 | 28473.0 | 14919.0 | 2.65 | 1.91 |
212
+ | hnet | code/256B | 75525.0 | 75909.0 | 28473.0 | 14919.0 | 2.65 | 1.91 |
213
+ | hnet | code/256B | 75525.0 | 75909.0 | 28473.0 | 14922.0 | 2.65 | 1.91 |
214
+ | hnet | code/256B | 75525.0 | 75909.0 | 28473.0 | 14922.0 | 2.65 | 1.91 |
215
+ | hnet | code/256B | 75525.0 | 75909.0 | 28473.0 | 14928.0 | 2.65 | 1.91 |
216
+ | hnet | code/256B | 75525.0 | 75909.0 | 28473.0 | 14928.0 | 2.65 | 1.91 |
217
+ | hnet_serial | code/1024B | 181272.0 | 181656.0 | 63561.0 | 30726.0 | 2.85 | 2.07 |
218
+ | hnet_serial | code/1024B | 181272.0 | 181656.0 | 63561.0 | 30726.0 | 2.85 | 2.07 |
219
+ | hnet_serial | code/1024B | 181272.0 | 181656.0 | 63561.0 | 30726.0 | 2.85 | 2.07 |
220
+ | hnet_serial | code/1024B | 181272.0 | 181656.0 | 63561.0 | 30726.0 | 2.85 | 2.07 |
221
+ | hnet_serial | code/1024B | 181272.0 | 181656.0 | 63561.0 | 30726.0 | 2.85 | 2.07 |
222
+ | hnet_serial | code/1024B | 181272.0 | 181656.0 | 63561.0 | 30726.0 | 2.85 | 2.07 |
223
+ | hnet_serial | code/1024B | 181272.0 | 181656.0 | 63561.0 | 30726.0 | 2.85 | 2.07 |
224
+ | hnet_serial | code/1024B | 181272.0 | 181656.0 | 63561.0 | 30726.0 | 2.85 | 2.07 |
225
+ | hnet_serial | code/1024B | 181272.0 | 181656.0 | 63561.0 | 30726.0 | 2.85 | 2.07 |
226
+ | hnet_serial | code/1024B | 181272.0 | 181656.0 | 63561.0 | 30726.0 | 2.85 | 2.07 |
227
+ | hnet_serial | code/256B | 75525.0 | 75909.0 | 28476.0 | 14922.0 | 2.65 | 1.91 |
228
+ | hnet_serial | code/256B | 75525.0 | 75909.0 | 28476.0 | 14922.0 | 2.65 | 1.91 |
229
+ | hnet_serial | code/256B | 75525.0 | 75909.0 | 28476.0 | 14922.0 | 2.65 | 1.91 |
230
+ | hnet_serial | code/256B | 75525.0 | 75909.0 | 28476.0 | 14922.0 | 2.65 | 1.91 |
231
+ | hnet_serial | code/256B | 75525.0 | 75909.0 | 28476.0 | 14922.0 | 2.65 | 1.91 |
232
+ | hnet_serial | code/256B | 75525.0 | 75909.0 | 28476.0 | 14922.0 | 2.65 | 1.91 |
233
+ | hnet_serial | code/256B | 75525.0 | 75909.0 | 28476.0 | 14922.0 | 2.65 | 1.91 |
234
+ | hnet_serial | code/256B | 75525.0 | 75909.0 | 28476.0 | 14922.0 | 2.65 | 1.91 |
235
+ | hnet_serial | code/256B | 75525.0 | 75909.0 | 28476.0 | 14922.0 | 2.65 | 1.91 |
236
+ | hnet_serial | code/256B | 75525.0 | 75909.0 | 28476.0 | 14922.0 | 2.65 | 1.91 |
237
+ | pythia | code/1024B | 181272.0 | 52395.0 | None | None | None | None |
238
+ | pythia | code/1024B | 181272.0 | 52395.0 | None | None | None | None |
239
+ | pythia | code/1024B | 181272.0 | 52395.0 | None | None | None | None |
240
+ | pythia | code/1024B | 181272.0 | 52395.0 | None | None | None | None |
241
+ | pythia | code/1024B | 181272.0 | 52395.0 | None | None | None | None |
242
+ | pythia | code/1024B | 181272.0 | 52395.0 | None | None | None | None |
243
+ | pythia | code/1024B | 181272.0 | 52395.0 | None | None | None | None |
244
+ | pythia | code/1024B | 181272.0 | 52395.0 | None | None | None | None |
245
+ | pythia | code/1024B | 181272.0 | 52395.0 | None | None | None | None |
246
+ | pythia | code/1024B | 181272.0 | 52395.0 | None | None | None | None |
247
+ | pythia | code/256B | 75525.0 | 23367.0 | None | None | None | None |
248
+ | pythia | code/256B | 75525.0 | 23367.0 | None | None | None | None |
249
+ | pythia | code/256B | 75525.0 | 23367.0 | None | None | None | None |
250
+ | pythia | code/256B | 75525.0 | 23367.0 | None | None | None | None |
251
+ | pythia | code/256B | 75525.0 | 23367.0 | None | None | None | None |
252
+ | pythia | code/256B | 75525.0 | 23367.0 | None | None | None | None |
253
+ | pythia | code/256B | 75525.0 | 23367.0 | None | None | None | None |
254
+ | pythia | code/256B | 75525.0 | 23367.0 | None | None | None | None |
255
+ | pythia | code/256B | 75525.0 | 23367.0 | None | None | None | None |
256
+ | pythia | code/256B | 75525.0 | 23367.0 | None | None | None | None |
generation_bench_results/bolmo_code_1024B_128new_speedup_vs_batch.png ADDED
generation_bench_results/bolmo_code_1024B_128new_throughput_vs_batch.png ADDED
generation_bench_results/bolmo_code_1024B_64new_speedup_vs_batch.png ADDED
generation_bench_results/bolmo_code_1024B_64new_throughput_vs_batch.png ADDED
generation_bench_results/bolmo_code_256B_128new_speedup_vs_batch.png ADDED
generation_bench_results/bolmo_code_256B_128new_throughput_vs_batch.png ADDED
generation_bench_results/bolmo_code_256B_64new_speedup_vs_batch.png ADDED
generation_bench_results/bolmo_code_256B_64new_throughput_vs_batch.png ADDED
generation_bench_results/hnet_code_1024B_128new_speedup_vs_batch.png ADDED
generation_bench_results/hnet_code_1024B_128new_throughput_vs_batch.png ADDED
generation_bench_results/hnet_code_1024B_64new_speedup_vs_batch.png ADDED
generation_bench_results/hnet_code_1024B_64new_throughput_vs_batch.png ADDED
generation_bench_results/hnet_code_256B_128new_speedup_vs_batch.png ADDED
generation_bench_results/hnet_code_256B_128new_throughput_vs_batch.png ADDED
generation_bench_results/hnet_code_256B_64new_speedup_vs_batch.png ADDED
generation_bench_results/hnet_code_256B_64new_throughput_vs_batch.png ADDED
generation_bench_results/hnet_serial_code_1024B_128new_speedup_vs_batch.png ADDED
generation_bench_results/hnet_serial_code_1024B_128new_throughput_vs_batch.png ADDED
generation_bench_results/hnet_serial_code_1024B_64new_speedup_vs_batch.png ADDED
generation_bench_results/hnet_serial_code_1024B_64new_throughput_vs_batch.png ADDED
generation_bench_results/hnet_serial_code_256B_128new_speedup_vs_batch.png ADDED
generation_bench_results/hnet_serial_code_256B_128new_throughput_vs_batch.png ADDED
generation_bench_results/hnet_serial_code_256B_64new_speedup_vs_batch.png ADDED
generation_bench_results/hnet_serial_code_256B_64new_throughput_vs_batch.png ADDED
generation_bench_results/metadata.json ADDED
@@ -0,0 +1,114 @@
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
1
+ {
2
+ "config": {
3
+ "seed": 42,
4
+ "device": "cuda",
5
+ "dtype": "bfloat16",
6
+ "benchmark": {
7
+ "repeats": 3,
8
+ "batch_sizes": [
9
+ 1,
10
+ 2,
11
+ 4,
12
+ 8,
13
+ 16
14
+ ],
15
+ "max_new_tokens": [
16
+ 64,
17
+ 128
18
+ ],
19
+ "stop_on_eos": true,
20
+ "measure_memory": true,
21
+ "collect_traces": true,
22
+ "continue_on_error": true,
23
+ "warmup": {
24
+ "enabled": true
25
+ }
26
+ },
27
+ "workloads": {
28
+ "synthetic": {
29
+ "enabled": false,
30
+ "prompt_byte_buckets": [
31
+ 256,
32
+ 1024
33
+ ],
34
+ "samples_per_bucket": 128
35
+ },
36
+ "code": {
37
+ "enabled": true,
38
+ "data_path": "code_completion_exp/datasets/data_V4_full",
39
+ "split": "validation",
40
+ "prompt_byte_buckets": [
41
+ 256,
42
+ 1024
43
+ ],
44
+ "samples_per_bucket": 128,
45
+ "max_samples_to_scan": 5000
46
+ }
47
+ },
48
+ "models": {
49
+ "enabled": [
50
+ "hnet",
51
+ "hnet_serial",
52
+ "bolmo",
53
+ "pythia"
54
+ ],
55
+ "hnet": {
56
+ "config_path": "hnet_project/configs/hnet_2stage_XL_code.json",
57
+ "checkpoint_path": "hnet_project/checkpoints/hnet_2stage_XL_code.pt"
58
+ },
59
+ "hnet_serial": {
60
+ "config_path": "hnet_project/configs/hnet_2stage_XL_code.json",
61
+ "checkpoint_path": "hnet_project/checkpoints/hnet_2stage_XL_code.pt"
62
+ },
63
+ "bolmo": {
64
+ "model_name": "allenai/Bolmo-1B",
65
+ "cache_dir": "cache/bolmo",
66
+ "local_files_only": false
67
+ },
68
+ "pythia": {
69
+ "model_name": "EleutherAI/pythia-1b",
70
+ "cache_dir": "cache/hf",
71
+ "local_files_only": false,
72
+ "max_context_len": 4096
73
+ }
74
+ },
75
+ "paths": {
76
+ "output_root": "generation_speed_exp/outputs"
77
+ }
78
+ },
79
+ "repo_root": "/workspace/byte-llms-code",
80
+ "workloads": {
81
+ "code/256B": 128,
82
+ "code/1024B": 128
83
+ },
84
+ "environment": {
85
+ "timestamp": "2026-05-10T23:04:30",
86
+ "hostname": "e41665783a40",
87
+ "platform": "Linux-6.8.0-110-generic-x86_64-with-glibc2.39",
88
+ "python": "3.12.0 | packaged by Anaconda, Inc. | (main, Oct 2 2023, 17:29:18) [GCC 11.2.0]",
89
+ "git_commit": "21e26deacf90f77563e3834cfb46e6dc64f52d6e",
90
+ "git_branch": "feature/generation-bench",
91
+ "git_dirty": false,
92
+ "torch": "2.6.0+cu124",
93
+ "torch_cuda": "12.4",
94
+ "cuda_available": true,
95
+ "transformers": "4.57.6",
96
+ "cuda_devices": [
97
+ {
98
+ "index": 0,
99
+ "name": "NVIDIA A100X",
100
+ "total_memory_mb": 81341.8125,
101
+ "major": 8,
102
+ "minor": 0
103
+ },
104
+ {
105
+ "index": 1,
106
+ "name": "NVIDIA A100X",
107
+ "total_memory_mb": 81341.8125,
108
+ "major": 8,
109
+ "minor": 0
110
+ }
111
+ ],
112
+ "nvidia_smi": "NVIDIA A100X, 81920 MiB, 570.86.10\nNVIDIA A100X, 81920 MiB, 570.86.10"
113
+ }
114
+ }
generation_bench_results/pythia_code_1024B_128new_speedup_vs_batch.png ADDED
generation_bench_results/pythia_code_1024B_128new_throughput_vs_batch.png ADDED
generation_bench_results/pythia_code_1024B_64new_speedup_vs_batch.png ADDED
generation_bench_results/pythia_code_1024B_64new_throughput_vs_batch.png ADDED
generation_bench_results/pythia_code_256B_128new_speedup_vs_batch.png ADDED
generation_bench_results/pythia_code_256B_128new_throughput_vs_batch.png ADDED
generation_bench_results/pythia_code_256B_64new_speedup_vs_batch.png ADDED
generation_bench_results/pythia_code_256B_64new_throughput_vs_batch.png ADDED
generation_bench_results/results.json ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:4397f619307472f1c9dbfefd8ed6b2e08bc06c619e546575c591710c316134fe
3
+ size 116684441
generation_bench_results/runs.partial.jsonl ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:51a5a198d0d6d84b0193fd5dcae6293e450070bdfb53d3ba6d25bc70ebb00128
3
+ size 36031940
generation_bench_results/summary.csv ADDED
@@ -0,0 +1,81 @@
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
1
+ model,workload,batch_size,max_new_tokens,status,num_samples,num_chunks,num_failures,prompt_bytes_mean,generated_units_mean,generated_bytes_mean,prefill_time_sec,decode_time_sec,total_time_sec,prefill_bytes_per_sec,prefill_units_per_sec,decode_bytes_per_sec,decode_units_per_sec,total_bytes_per_sec,total_units_per_sec,speedup_vs_batch1,peak_allocated_mb,peak_reserved_mb
2
+ bolmo,code/1024B,1,64,ok,384,384,0,472.0625,534.0625,534.0625,0.068753,4.836162,5.321893,6866.053934,6880.598732,110.431064,110.431064,100.351981,100.351981,1.0,2938.752441,3044.0
3
+ bolmo,code/1024B,1,128,ok,384,384,0,472.0625,597.617188,597.617188,0.044755,5.402814,5.915648,10547.721543,10570.065452,110.612215,110.612215,101.02312,101.02312,1.0,2938.752441,3044.0
4
+ bolmo,code/1024B,2,64,ok,384,192,0,472.0625,631.15625,631.15625,0.144373,8.605539,9.747681,6539.503151,6553.356196,146.686051,146.686051,129.498754,129.498754,1.290445,3027.405273,3146.0
5
+ bolmo,code/1024B,2,128,ok,384,192,0,472.0625,694.710938,694.710938,0.047318,9.472696,10.623823,19952.876036,19995.143481,146.676497,146.676497,130.783605,130.783605,1.294591,3027.405273,3146.0
6
+ bolmo,code/1024B,4,64,ok,384,96,0,472.0625,732.835938,732.835938,0.188248,13.762859,16.441378,10030.629056,10051.877575,212.98945,212.98945,178.290639,178.290639,1.776653,3229.916504,3424.0
7
+ bolmo,code/1024B,4,128,ok,384,96,0,472.0625,794.335938,794.335938,0.051477,14.978097,17.770405,36681.427648,36759.132248,212.132668,212.132668,178.799738,178.799738,1.769889,3229.916504,3418.0
8
+ bolmo,code/1024B,8,64,ok,384,48,0,472.0625,832.1875,832.1875,0.387669,20.813725,27.976574,9741.561057,9762.197225,319.861046,319.861046,237.966953,237.966953,2.371323,3616.254395,3854.0
9
+ bolmo,code/1024B,8,128,ok,384,48,0,472.0625,892.6875,892.6875,0.076809,22.26734,29.739563,49167.31725,49271.471503,320.716348,320.716348,240.134664,240.134664,2.377027,3616.254395,3854.0
10
+ bolmo,code/1024B,16,64,ok,384,24,0,472.0625,908.085938,908.085938,0.452111,28.969853,48.624499,16706.064851,16741.45437,501.534301,501.534301,298.807704,298.807704,2.977596,4407.841797,4968.0
11
+ bolmo,code/1024B,16,128,ok,384,24,0,472.0625,970.109375,970.109375,0.13893,31.091566,52.396775,54365.508593,54480.674506,499.227024,499.227024,296.234833,296.234833,2.932347,4407.841797,4906.0
12
+ bolmo,code/256B,1,64,ok,384,384,0,196.679688,258.679688,258.679688,0.043797,2.372468,2.605415,4490.73205,4513.564769,109.033984,109.033984,99.285411,99.285411,1.0,2854.858887,3002.0
13
+ bolmo,code/256B,1,128,ok,384,384,0,196.679688,322.679688,322.679688,0.044046,2.964494,3.244852,4465.306638,4488.010083,108.848155,108.848155,99.443587,99.443587,1.0,2858.609375,3002.0
14
+ bolmo,code/256B,2,64,ok,384,192,0,196.679688,278.328125,278.328125,0.072106,3.81578,4.275807,5455.311963,5483.049001,145.882697,145.882697,130.187419,130.187419,1.311244,2896.588867,3024.0
15
+ bolmo,code/256B,2,128,ok,384,192,0,196.679688,341.75,341.75,0.046872,4.676194,5.203776,8392.282488,8434.952286,146.165868,146.165868,131.346923,131.346923,1.320818,2896.589844,3026.0
16
+ bolmo,code/256B,4,64,ok,384,96,0,196.679688,286.078125,286.078125,0.070597,5.184386,5.983706,11143.738831,11200.398159,220.722849,220.722849,191.238097,191.238097,1.926145,2965.133301,3040.0
17
+ bolmo,code/256B,4,128,ok,384,96,0,196.679688,348.578125,348.578125,0.052066,6.306583,7.257811,15110.171749,15186.998045,221.088426,221.088426,192.111994,192.111994,1.931869,2966.34082,3040.0
18
+ bolmo,code/256B,8,64,ok,384,48,0,196.679688,296.40625,296.40625,0.118204,6.677236,8.259403,13311.216975,13378.896648,355.124486,355.124486,287.097024,287.097024,2.891634,3064.116699,3252.0
19
+ bolmo,code/256B,8,128,ok,384,48,0,196.679688,358.40625,358.40625,0.0535,8.085679,9.955838,29410.213202,29559.746759,354.608419,354.608419,287.996841,287.996841,2.896083,3102.790039,3290.0
20
+ bolmo,code/256B,16,64,ok,384,24,0,196.679688,307.1875,307.1875,0.181932,8.502904,11.973908,17296.94618,17384.890932,578.037789,578.037789,410.475832,410.475832,4.134302,3308.902832,3750.0
21
+ bolmo,code/256B,16,128,ok,384,24,0,196.679688,370.6875,370.6875,0.068305,10.256425,14.423305,46070.759626,46305.002217,578.271683,578.271683,411.209502,411.209502,4.135103,3374.62793,3864.0
22
+ hnet,code/1024B,1,64,ok,384,384,0,472.0625,64.0,64.0,0.062152,0.774414,0.840556,7595.334888,7611.424568,82.643175,82.643175,76.140115,76.140115,1.0,3518.566406,3618.0
23
+ hnet,code/1024B,1,128,ok,384,384,0,472.0625,128.0,128.0,0.062105,1.57457,1.641366,7601.055009,7617.156807,81.292013,81.292013,77.98382,77.98382,1.0,3533.662598,3618.0
24
+ hnet,code/1024B,2,64,ok,384,192,0,472.0625,64.0,64.0,0.064711,1.080596,1.155029,14589.929687,14620.836463,118.453185,118.453185,110.819748,110.819748,1.455471,3801.771973,3924.0
25
+ hnet,code/1024B,2,128,ok,384,192,0,472.0625,128.0,128.0,0.06399,2.19019,2.264525,14754.350786,14785.605865,116.884838,116.884838,113.04797,113.04797,1.449634,3849.330566,3924.0
26
+ hnet,code/1024B,4,64,ok,384,96,0,472.0625,64.0,64.0,0.068192,1.481849,1.568884,27690.090107,27748.747785,172.757098,172.757098,163.173329,163.173329,2.143066,4378.296387,4456.0
27
+ hnet,code/1024B,4,128,ok,384,96,0,472.0625,128.0,128.0,0.067215,2.985488,3.073012,28092.872708,28152.383626,171.496275,171.496275,166.611755,166.611755,2.136491,4437.296387,4516.0
28
+ hnet,code/1024B,8,64,ok,384,48,0,472.0625,64.0,64.0,0.082033,1.892013,2.009045,46036.40171,46133.923546,270.611242,270.611242,254.847484,254.847484,3.347086,5463.019531,5594.0
29
+ hnet,code/1024B,8,128,ok,384,48,0,472.0625,128.0,128.0,0.078074,3.832693,3.946751,48370.545143,48473.011544,267.175056,267.175056,259.453905,259.453905,3.327022,5579.550781,5714.0
30
+ hnet,code/1024B,16,64,ok,384,24,0,472.0625,64.0,64.0,0.103519,2.233956,2.385581,72962.773649,73117.33533,458.379707,458.379707,429.245476,429.245476,5.637573,7649.574707,7842.0
31
+ hnet,code/1024B,16,128,ok,384,24,0,472.0625,128.0,128.0,0.101793,4.46502,4.616226,74199.727789,74356.909789,458.676538,458.676538,443.652437,443.652437,5.689032,7885.574707,8078.0
32
+ hnet,code/256B,1,64,ok,384,384,0,196.679688,64.0,64.0,0.065434,0.805041,0.87303,3005.772131,3021.054707,79.49902,79.49902,73.307922,73.307922,1.0,3316.642578,3358.0
33
+ hnet,code/256B,1,128,ok,384,384,0,196.679688,128.0,128.0,0.061706,1.610171,1.674526,3187.36849,3203.574375,79.494648,79.494648,76.439552,76.439552,1.0,3339.232422,3378.0
34
+ hnet,code/256B,2,64,ok,384,192,0,196.679688,64.0,64.0,0.064032,1.128386,1.196365,6143.158254,6174.392584,113.436401,113.436401,106.990791,106.990791,1.459471,3409.324219,3456.0
35
+ hnet,code/256B,2,128,ok,384,192,0,196.679688,128.0,128.0,0.062853,2.247984,2.315847,6258.412654,6290.232985,113.879827,113.879827,110.542704,110.542704,1.446145,3441.13623,3516.0
36
+ hnet,code/256B,4,64,ok,384,96,0,196.679688,64.0,64.0,0.065871,1.509599,1.583962,11943.40729,12004.132459,169.581429,169.581429,161.620015,161.620015,2.204673,3606.863281,3646.0
37
+ hnet,code/256B,4,128,ok,384,96,0,196.679688,128.0,128.0,0.066207,3.046167,3.123346,11882.781698,11943.198622,168.080064,168.080064,163.92676,163.92676,2.144528,3665.328613,3702.0
38
+ hnet,code/256B,8,64,ok,384,48,0,196.679688,64.0,64.0,0.072903,1.907366,1.997639,21582.546929,21692.281428,268.433013,268.433013,256.302595,256.302595,3.496247,3976.942383,4054.0
39
+ hnet,code/256B,8,128,ok,384,48,0,196.679688,128.0,128.0,0.072685,3.853694,3.947047,21647.204718,21757.267963,265.719082,265.719082,259.434432,259.434432,3.393982,4092.133789,4170.0
40
+ hnet,code/256B,16,64,ok,384,24,0,196.679688,64.0,64.0,0.086746,2.250826,2.370719,36276.728751,36461.174482,454.944028,454.944028,431.936555,431.936555,5.892086,4716.527832,4836.0
41
+ hnet,code/256B,16,128,ok,384,24,0,196.679688,128.0,128.0,0.086681,4.488675,4.614004,36304.189831,36488.775186,456.259358,456.259358,443.866119,443.866119,5.80676,4952.527832,5072.0
42
+ hnet_serial,code/1024B,1,64,ok,384,384,0,472.0625,64.0,64.0,0.060226,0.769539,0.835311,7838.145943,7854.749986,83.166632,83.166632,76.61816,76.61816,1.0,3518.044922,3630.0
43
+ hnet_serial,code/1024B,1,128,ok,384,384,0,472.0625,128.0,128.0,0.05971,1.556894,1.622588,7905.929166,7922.676798,82.21497,82.21497,78.88632,78.88632,1.0,3532.688965,3630.0
44
+ hnet_serial,code/1024B,2,64,ok,384,192,0,472.0625,64.0,64.0,0.117386,1.546019,1.673656,8042.884415,8059.922169,82.793292,82.793292,76.479278,76.479278,0.998187,3698.293457,3838.0
45
+ hnet_serial,code/1024B,2,128,ok,384,192,0,472.0625,128.0,128.0,0.116693,3.108496,3.236786,8090.643795,8107.78272,82.354948,82.354948,79.090797,79.090797,1.002592,3735.762207,3838.0
46
+ hnet_serial,code/1024B,4,64,ok,384,96,0,472.0625,64.0,64.0,0.227732,3.05651,3.300198,8291.542819,8309.107321,83.755664,83.755664,77.571103,77.571103,1.012438,3698.293457,3838.0
47
+ hnet_serial,code/1024B,4,128,ok,384,96,0,472.0625,128.0,128.0,0.229722,6.225154,6.472352,8219.728507,8237.140881,82.246964,82.246964,79.105716,79.105716,1.002781,3735.762207,3840.0
48
+ hnet_serial,code/1024B,8,64,ok,384,48,0,472.0625,64.0,64.0,0.451309,6.135115,6.613503,8367.881261,8385.607476,83.454012,83.454012,77.417369,77.417369,1.010431,3698.293457,3842.0
49
+ hnet_serial,code/1024B,8,128,ok,384,48,0,472.0625,128.0,128.0,0.451507,12.405507,12.885156,8364.208257,8381.926691,82.543985,82.543985,79.47129,79.47129,1.007415,3735.762207,3842.0
50
+ hnet_serial,code/1024B,16,64,ok,384,24,0,472.0625,64.0,64.0,0.89503,12.254463,13.193673,8438.819396,8456.695884,83.561394,83.561394,77.612958,77.612958,1.012984,3698.963867,3842.0
51
+ hnet_serial,code/1024B,16,128,ok,384,24,0,472.0625,128.0,128.0,0.895705,24.806422,25.745994,8432.459537,8450.322552,82.559267,82.559267,79.546355,79.546355,1.008367,3735.199219,3880.0
52
+ hnet_serial,code/256B,1,64,ok,384,384,0,196.679688,64.0,64.0,0.065792,0.798052,0.867803,2989.431654,3004.631148,80.195298,80.195298,73.749488,73.749488,1.0,3314.123535,3350.0
53
+ hnet_serial,code/256B,1,128,ok,384,384,0,196.679688,128.0,128.0,0.059155,1.594404,1.657715,3324.803433,3341.708094,80.280771,80.280771,77.21474,77.21474,1.0,3338.775879,3368.0
54
+ hnet_serial,code/256B,2,64,ok,384,192,0,196.679688,64.0,64.0,0.115675,1.587237,1.710293,3400.56481,3417.854672,80.643301,80.643301,74.840965,74.840965,1.0148,3396.96582,3440.0
55
+ hnet_serial,code/256B,2,128,ok,384,192,0,196.679688,128.0,128.0,0.116146,3.191044,3.315233,3386.765826,3403.985529,80.224518,80.224518,77.219307,77.219307,1.000059,3426.63623,3480.0
56
+ hnet_serial,code/256B,4,64,ok,384,96,0,196.679688,64.0,64.0,0.226455,3.186142,3.423891,3474.066682,3491.730259,80.347962,80.347962,74.76873,74.76873,1.01382,3396.96582,3440.0
57
+ hnet_serial,code/256B,4,128,ok,384,96,0,196.679688,128.0,128.0,0.226278,6.365509,6.604025,3476.78274,3494.460126,80.433468,80.433468,77.52848,77.52848,1.004063,3426.63623,3480.0
58
+ hnet_serial,code/256B,8,64,ok,384,48,0,196.679688,64.0,64.0,0.448479,6.371802,6.839489,3508.386612,3526.224685,80.354031,80.354031,74.859389,74.859389,1.01505,3396.96582,3440.0
59
+ hnet_serial,code/256B,8,128,ok,384,48,0,196.679688,128.0,128.0,0.448768,12.76133,13.230026,3506.127639,3523.954227,80.242418,80.242418,77.399698,77.399698,1.002395,3426.63623,3480.0
60
+ hnet_serial,code/256B,16,64,ok,384,24,0,196.679688,64.0,64.0,0.887241,12.710502,13.63144,3546.81,3564.843434,80.5633,80.5633,75.120456,75.120456,1.01859,3396.96582,3442.0
61
+ hnet_serial,code/256B,16,128,ok,384,24,0,196.679688,128.0,128.0,0.888159,25.442948,26.365968,3543.142093,3561.156877,80.493818,80.493818,77.675888,77.675888,1.005972,3426.63623,3480.0
62
+ pythia,code/1024B,1,64,ok,384,384,0,472.0625,64.0,228.570312,0.015437,0.587075,0.602548,30578.97123,8838.56965,389.337169,109.01494,379.339326,106.21553,1.0,2014.904297,2066.0
63
+ pythia,code/1024B,1,128,ok,384,384,0,472.0625,128.0,454.46875,0.015197,1.172286,1.187519,31063.047172,8978.487337,387.677231,109.188334,382.704555,107.787792,1.0,2014.904297,2066.0
64
+ pythia,code/1024B,2,64,ok,384,192,0,472.0625,61.96875,206.921875,0.018753,0.59378,0.612567,50344.438048,14551.595566,696.96499,208.726357,675.58929,202.324785,1.780963,2088.924316,2180.0
65
+ pythia,code/1024B,2,128,ok,384,192,0,472.0625,121.4375,405.625,0.018361,1.199283,1.217679,51420.928537,14862.745216,676.445868,202.516845,666.226587,199.457359,1.740838,2088.924316,2180.0
66
+ pythia,code/1024B,4,64,ok,384,96,0,472.0625,61.914062,195.570312,0.019572,0.604865,0.624472,96475.999327,27885.497952,1293.315184,409.440452,1252.708837,396.585209,3.302344,2244.148438,2402.0
67
+ pythia,code/1024B,4,128,ok,384,96,0,472.0625,122.21875,382.570312,0.019801,1.230192,1.250028,95362.074273,27563.528187,1243.936583,397.397208,1224.197386,391.091178,3.198805,2244.148438,2402.0
68
+ pythia,code/1024B,8,64,ok,384,48,0,472.0625,62.914062,192.09375,0.031359,0.652979,0.684379,120427.569936,34808.478567,2353.445574,770.794583,2245.465339,735.429168,5.919411,2544.552734,2788.0
69
+ pythia,code/1024B,8,128,ok,384,48,0,472.0625,123.632812,373.398438,0.031854,1.295015,1.32691,118556.106233,34267.549241,2306.680966,763.745712,2251.236047,745.387811,5.882439,2544.552734,2788.0
70
+ pythia,code/1024B,16,64,ok,384,24,0,472.0625,62.632812,185.03125,0.063219,0.638165,0.701421,119473.186915,34532.622956,4639.07925,1570.321666,4220.71474,1428.70588,11.126489,3153.496094,3598.0
71
+ pythia,code/1024B,16,128,ok,384,24,0,472.0625,122.820312,360.710938,0.062419,1.272973,1.335429,121004.362154,34975.195039,4533.776083,1543.728613,4321.738335,1471.530796,11.292623,3153.496094,3598.0
72
+ pythia,code/256B,1,64,ok,384,384,0,196.679688,63.242188,213.828125,0.012935,0.585042,0.59801,15205.354431,4704.449083,365.491977,108.098559,357.5662,105.754417,1.0,1959.659668,2028.0
73
+ pythia,code/256B,1,128,ok,384,384,0,196.679688,125.742188,416.46875,0.012979,1.157294,1.170306,15153.211048,4688.316221,359.864351,108.651923,355.863096,107.443846,1.0,1966.70752,2038.0
74
+ pythia,code/256B,2,64,ok,384,192,0,196.679688,61.179688,208.492188,0.018949,0.597125,0.616108,20759.302462,6422.808615,698.319652,204.914048,676.804135,198.600561,1.892808,1996.35498,2042.0
75
+ pythia,code/256B,2,128,ok,384,192,0,196.679688,119.953125,405.71875,0.018608,1.186235,1.204876,21139.196906,6540.345768,684.044443,202.241746,673.461271,199.11277,1.892473,2001.300293,2062.0
76
+ pythia,code/256B,4,64,ok,384,96,0,196.679688,60.242188,194.03125,0.018335,0.60085,0.61922,42908.124346,13275.526535,1291.711003,401.046205,1253.392167,389.149098,3.505343,2023.126953,2066.0
77
+ pythia,code/256B,4,128,ok,384,96,0,196.679688,117.25,372.8125,0.017297,1.210337,1.227669,45482.280289,14071.955558,1232.095007,387.495429,1214.700192,382.024738,3.413392,2057.760742,2106.0
78
+ pythia,code/256B,8,64,ok,384,48,0,196.679688,60.210938,189.1875,0.019644,0.621142,0.640821,80095.915997,24781.215083,2436.640028,775.486649,2361.812336,751.671939,6.605245,2109.179688,2186.0
79
+ pythia,code/256B,8,128,ok,384,48,0,196.679688,117.3125,364.195312,0.019393,1.236002,1.255429,81135.255982,25102.780888,2357.247057,759.302868,2320.769548,747.552943,6.521524,2167.822266,2326.0
80
+ pythia,code/256B,16,64,ok,384,24,0,196.679688,58.585938,189.710938,0.024844,0.623566,0.648445,126664.524281,39189.274265,4867.766204,1503.248312,4681.007925,1445.574205,13.091304,2279.179688,2438.0
81
+ pythia,code/256B,16,128,ok,384,24,0,196.679688,111.359375,358.257812,0.025265,1.266783,1.292083,124556.538536,38537.075617,4524.946928,1406.515767,4436.342603,1378.974365,12.466431,2398.257812,2712.0
generation_bench_results/summary.md ADDED
@@ -0,0 +1,256 @@
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
1
+ # Generation Speed Benchmark
2
+
3
+ ## Speed
4
+
5
+ | Model | Workload | Batch | Max new | Samples | Status | Prompt bytes | Prefill ms/chunk | Prefill bytes/s | Prefill units/s | Decode ms/chunk | Decode bytes/s | Decode units/s | Total bytes/s | Speedup vs B=1 | Peak VRAM MB |
6
+ |---|---|---:|---:|---:|---|---:|---:|---:|---:|---:|---:|---:|---:|---:|---:|
7
+ | bolmo | code/1024B | 1 | 64 | 384 | ok | 472.1 | 68.8 | 6866.1 | 6880.6 | 4836.2 | 110.4 | 110.4 | 100.4 | 1.0 | 2938.8 |
8
+ | bolmo | code/1024B | 1 | 128 | 384 | ok | 472.1 | 44.8 | 10547.7 | 10570.1 | 5402.8 | 110.6 | 110.6 | 101.0 | 1.0 | 2938.8 |
9
+ | bolmo | code/1024B | 2 | 64 | 384 | ok | 472.1 | 144.4 | 6539.5 | 6553.4 | 8605.5 | 146.7 | 146.7 | 129.5 | 1.29 | 3027.4 |
10
+ | bolmo | code/1024B | 2 | 128 | 384 | ok | 472.1 | 47.3 | 19952.9 | 19995.1 | 9472.7 | 146.7 | 146.7 | 130.8 | 1.29 | 3027.4 |
11
+ | bolmo | code/1024B | 4 | 64 | 384 | ok | 472.1 | 188.2 | 10030.6 | 10051.9 | 13762.9 | 213.0 | 213.0 | 178.3 | 1.78 | 3229.9 |
12
+ | bolmo | code/1024B | 4 | 128 | 384 | ok | 472.1 | 51.5 | 36681.4 | 36759.1 | 14978.1 | 212.1 | 212.1 | 178.8 | 1.77 | 3229.9 |
13
+ | bolmo | code/1024B | 8 | 64 | 384 | ok | 472.1 | 387.7 | 9741.6 | 9762.2 | 20813.7 | 319.9 | 319.9 | 238.0 | 2.37 | 3616.3 |
14
+ | bolmo | code/1024B | 8 | 128 | 384 | ok | 472.1 | 76.8 | 49167.3 | 49271.5 | 22267.3 | 320.7 | 320.7 | 240.1 | 2.38 | 3616.3 |
15
+ | bolmo | code/1024B | 16 | 64 | 384 | ok | 472.1 | 452.1 | 16706.1 | 16741.5 | 28969.9 | 501.5 | 501.5 | 298.8 | 2.98 | 4407.8 |
16
+ | bolmo | code/1024B | 16 | 128 | 384 | ok | 472.1 | 138.9 | 54365.5 | 54480.7 | 31091.6 | 499.2 | 499.2 | 296.2 | 2.93 | 4407.8 |
17
+ | bolmo | code/256B | 1 | 64 | 384 | ok | 196.7 | 43.8 | 4490.7 | 4513.6 | 2372.5 | 109.0 | 109.0 | 99.3 | 1.0 | 2854.9 |
18
+ | bolmo | code/256B | 1 | 128 | 384 | ok | 196.7 | 44.0 | 4465.3 | 4488.0 | 2964.5 | 108.8 | 108.8 | 99.4 | 1.0 | 2858.6 |
19
+ | bolmo | code/256B | 2 | 64 | 384 | ok | 196.7 | 72.1 | 5455.3 | 5483.0 | 3815.8 | 145.9 | 145.9 | 130.2 | 1.31 | 2896.6 |
20
+ | bolmo | code/256B | 2 | 128 | 384 | ok | 196.7 | 46.9 | 8392.3 | 8435.0 | 4676.2 | 146.2 | 146.2 | 131.3 | 1.32 | 2896.6 |
21
+ | bolmo | code/256B | 4 | 64 | 384 | ok | 196.7 | 70.6 | 11143.7 | 11200.4 | 5184.4 | 220.7 | 220.7 | 191.2 | 1.93 | 2965.1 |
22
+ | bolmo | code/256B | 4 | 128 | 384 | ok | 196.7 | 52.1 | 15110.2 | 15187.0 | 6306.6 | 221.1 | 221.1 | 192.1 | 1.93 | 2966.3 |
23
+ | bolmo | code/256B | 8 | 64 | 384 | ok | 196.7 | 118.2 | 13311.2 | 13378.9 | 6677.2 | 355.1 | 355.1 | 287.1 | 2.89 | 3064.1 |
24
+ | bolmo | code/256B | 8 | 128 | 384 | ok | 196.7 | 53.5 | 29410.2 | 29559.7 | 8085.7 | 354.6 | 354.6 | 288.0 | 2.9 | 3102.8 |
25
+ | bolmo | code/256B | 16 | 64 | 384 | ok | 196.7 | 181.9 | 17296.9 | 17384.9 | 8502.9 | 578.0 | 578.0 | 410.5 | 4.13 | 3308.9 |
26
+ | bolmo | code/256B | 16 | 128 | 384 | ok | 196.7 | 68.3 | 46070.8 | 46305.0 | 10256.4 | 578.3 | 578.3 | 411.2 | 4.14 | 3374.6 |
27
+ | hnet | code/1024B | 1 | 64 | 384 | ok | 472.1 | 62.2 | 7595.3 | 7611.4 | 774.4 | 82.6 | 82.6 | 76.1 | 1.0 | 3518.6 |
28
+ | hnet | code/1024B | 1 | 128 | 384 | ok | 472.1 | 62.1 | 7601.1 | 7617.2 | 1574.6 | 81.3 | 81.3 | 78.0 | 1.0 | 3533.7 |
29
+ | hnet | code/1024B | 2 | 64 | 384 | ok | 472.1 | 64.7 | 14589.9 | 14620.8 | 1080.6 | 118.5 | 118.5 | 110.8 | 1.46 | 3801.8 |
30
+ | hnet | code/1024B | 2 | 128 | 384 | ok | 472.1 | 64.0 | 14754.4 | 14785.6 | 2190.2 | 116.9 | 116.9 | 113.0 | 1.45 | 3849.3 |
31
+ | hnet | code/1024B | 4 | 64 | 384 | ok | 472.1 | 68.2 | 27690.1 | 27748.7 | 1481.8 | 172.8 | 172.8 | 163.2 | 2.14 | 4378.3 |
32
+ | hnet | code/1024B | 4 | 128 | 384 | ok | 472.1 | 67.2 | 28092.9 | 28152.4 | 2985.5 | 171.5 | 171.5 | 166.6 | 2.14 | 4437.3 |
33
+ | hnet | code/1024B | 8 | 64 | 384 | ok | 472.1 | 82.0 | 46036.4 | 46133.9 | 1892.0 | 270.6 | 270.6 | 254.8 | 3.35 | 5463.0 |
34
+ | hnet | code/1024B | 8 | 128 | 384 | ok | 472.1 | 78.1 | 48370.5 | 48473.0 | 3832.7 | 267.2 | 267.2 | 259.5 | 3.33 | 5579.6 |
35
+ | hnet | code/1024B | 16 | 64 | 384 | ok | 472.1 | 103.5 | 72962.8 | 73117.3 | 2234.0 | 458.4 | 458.4 | 429.2 | 5.64 | 7649.6 |
36
+ | hnet | code/1024B | 16 | 128 | 384 | ok | 472.1 | 101.8 | 74199.7 | 74356.9 | 4465.0 | 458.7 | 458.7 | 443.7 | 5.69 | 7885.6 |
37
+ | hnet | code/256B | 1 | 64 | 384 | ok | 196.7 | 65.4 | 3005.8 | 3021.1 | 805.0 | 79.5 | 79.5 | 73.3 | 1.0 | 3316.6 |
38
+ | hnet | code/256B | 1 | 128 | 384 | ok | 196.7 | 61.7 | 3187.4 | 3203.6 | 1610.2 | 79.5 | 79.5 | 76.4 | 1.0 | 3339.2 |
39
+ | hnet | code/256B | 2 | 64 | 384 | ok | 196.7 | 64.0 | 6143.2 | 6174.4 | 1128.4 | 113.4 | 113.4 | 107.0 | 1.46 | 3409.3 |
40
+ | hnet | code/256B | 2 | 128 | 384 | ok | 196.7 | 62.9 | 6258.4 | 6290.2 | 2248.0 | 113.9 | 113.9 | 110.5 | 1.45 | 3441.1 |
41
+ | hnet | code/256B | 4 | 64 | 384 | ok | 196.7 | 65.9 | 11943.4 | 12004.1 | 1509.6 | 169.6 | 169.6 | 161.6 | 2.2 | 3606.9 |
42
+ | hnet | code/256B | 4 | 128 | 384 | ok | 196.7 | 66.2 | 11882.8 | 11943.2 | 3046.2 | 168.1 | 168.1 | 163.9 | 2.14 | 3665.3 |
43
+ | hnet | code/256B | 8 | 64 | 384 | ok | 196.7 | 72.9 | 21582.5 | 21692.3 | 1907.4 | 268.4 | 268.4 | 256.3 | 3.5 | 3976.9 |
44
+ | hnet | code/256B | 8 | 128 | 384 | ok | 196.7 | 72.7 | 21647.2 | 21757.3 | 3853.7 | 265.7 | 265.7 | 259.4 | 3.39 | 4092.1 |
45
+ | hnet | code/256B | 16 | 64 | 384 | ok | 196.7 | 86.7 | 36276.7 | 36461.2 | 2250.8 | 454.9 | 454.9 | 431.9 | 5.89 | 4716.5 |
46
+ | hnet | code/256B | 16 | 128 | 384 | ok | 196.7 | 86.7 | 36304.2 | 36488.8 | 4488.7 | 456.3 | 456.3 | 443.9 | 5.81 | 4952.5 |
47
+ | hnet_serial | code/1024B | 1 | 64 | 384 | ok | 472.1 | 60.2 | 7838.1 | 7854.7 | 769.5 | 83.2 | 83.2 | 76.6 | 1.0 | 3518.0 |
48
+ | hnet_serial | code/1024B | 1 | 128 | 384 | ok | 472.1 | 59.7 | 7905.9 | 7922.7 | 1556.9 | 82.2 | 82.2 | 78.9 | 1.0 | 3532.7 |
49
+ | hnet_serial | code/1024B | 2 | 64 | 384 | ok | 472.1 | 117.4 | 8042.9 | 8059.9 | 1546.0 | 82.8 | 82.8 | 76.5 | 1.0 | 3698.3 |
50
+ | hnet_serial | code/1024B | 2 | 128 | 384 | ok | 472.1 | 116.7 | 8090.6 | 8107.8 | 3108.5 | 82.4 | 82.4 | 79.1 | 1.0 | 3735.8 |
51
+ | hnet_serial | code/1024B | 4 | 64 | 384 | ok | 472.1 | 227.7 | 8291.5 | 8309.1 | 3056.5 | 83.8 | 83.8 | 77.6 | 1.01 | 3698.3 |
52
+ | hnet_serial | code/1024B | 4 | 128 | 384 | ok | 472.1 | 229.7 | 8219.7 | 8237.1 | 6225.2 | 82.2 | 82.2 | 79.1 | 1.0 | 3735.8 |
53
+ | hnet_serial | code/1024B | 8 | 64 | 384 | ok | 472.1 | 451.3 | 8367.9 | 8385.6 | 6135.1 | 83.5 | 83.5 | 77.4 | 1.01 | 3698.3 |
54
+ | hnet_serial | code/1024B | 8 | 128 | 384 | ok | 472.1 | 451.5 | 8364.2 | 8381.9 | 12405.5 | 82.5 | 82.5 | 79.5 | 1.01 | 3735.8 |
55
+ | hnet_serial | code/1024B | 16 | 64 | 384 | ok | 472.1 | 895.0 | 8438.8 | 8456.7 | 12254.5 | 83.6 | 83.6 | 77.6 | 1.01 | 3699.0 |
56
+ | hnet_serial | code/1024B | 16 | 128 | 384 | ok | 472.1 | 895.7 | 8432.5 | 8450.3 | 24806.4 | 82.6 | 82.6 | 79.5 | 1.01 | 3735.2 |
57
+ | hnet_serial | code/256B | 1 | 64 | 384 | ok | 196.7 | 65.8 | 2989.4 | 3004.6 | 798.1 | 80.2 | 80.2 | 73.7 | 1.0 | 3314.1 |
58
+ | hnet_serial | code/256B | 1 | 128 | 384 | ok | 196.7 | 59.2 | 3324.8 | 3341.7 | 1594.4 | 80.3 | 80.3 | 77.2 | 1.0 | 3338.8 |
59
+ | hnet_serial | code/256B | 2 | 64 | 384 | ok | 196.7 | 115.7 | 3400.6 | 3417.9 | 1587.2 | 80.6 | 80.6 | 74.8 | 1.01 | 3397.0 |
60
+ | hnet_serial | code/256B | 2 | 128 | 384 | ok | 196.7 | 116.1 | 3386.8 | 3404.0 | 3191.0 | 80.2 | 80.2 | 77.2 | 1.0 | 3426.6 |
61
+ | hnet_serial | code/256B | 4 | 64 | 384 | ok | 196.7 | 226.5 | 3474.1 | 3491.7 | 3186.1 | 80.3 | 80.3 | 74.8 | 1.01 | 3397.0 |
62
+ | hnet_serial | code/256B | 4 | 128 | 384 | ok | 196.7 | 226.3 | 3476.8 | 3494.5 | 6365.5 | 80.4 | 80.4 | 77.5 | 1.0 | 3426.6 |
63
+ | hnet_serial | code/256B | 8 | 64 | 384 | ok | 196.7 | 448.5 | 3508.4 | 3526.2 | 6371.8 | 80.4 | 80.4 | 74.9 | 1.02 | 3397.0 |
64
+ | hnet_serial | code/256B | 8 | 128 | 384 | ok | 196.7 | 448.8 | 3506.1 | 3524.0 | 12761.3 | 80.2 | 80.2 | 77.4 | 1.0 | 3426.6 |
65
+ | hnet_serial | code/256B | 16 | 64 | 384 | ok | 196.7 | 887.2 | 3546.8 | 3564.8 | 12710.5 | 80.6 | 80.6 | 75.1 | 1.02 | 3397.0 |
66
+ | hnet_serial | code/256B | 16 | 128 | 384 | ok | 196.7 | 888.2 | 3543.1 | 3561.2 | 25442.9 | 80.5 | 80.5 | 77.7 | 1.01 | 3426.6 |
67
+ | pythia | code/1024B | 1 | 64 | 384 | ok | 472.1 | 15.4 | 30579.0 | 8838.6 | 587.1 | 389.3 | 109.0 | 379.3 | 1.0 | 2014.9 |
68
+ | pythia | code/1024B | 1 | 128 | 384 | ok | 472.1 | 15.2 | 31063.0 | 8978.5 | 1172.3 | 387.7 | 109.2 | 382.7 | 1.0 | 2014.9 |
69
+ | pythia | code/1024B | 2 | 64 | 384 | ok | 472.1 | 18.8 | 50344.4 | 14551.6 | 593.8 | 697.0 | 208.7 | 675.6 | 1.78 | 2088.9 |
70
+ | pythia | code/1024B | 2 | 128 | 384 | ok | 472.1 | 18.4 | 51420.9 | 14862.7 | 1199.3 | 676.4 | 202.5 | 666.2 | 1.74 | 2088.9 |
71
+ | pythia | code/1024B | 4 | 64 | 384 | ok | 472.1 | 19.6 | 96476.0 | 27885.5 | 604.9 | 1293.3 | 409.4 | 1252.7 | 3.3 | 2244.1 |
72
+ | pythia | code/1024B | 4 | 128 | 384 | ok | 472.1 | 19.8 | 95362.1 | 27563.5 | 1230.2 | 1243.9 | 397.4 | 1224.2 | 3.2 | 2244.1 |
73
+ | pythia | code/1024B | 8 | 64 | 384 | ok | 472.1 | 31.4 | 120427.6 | 34808.5 | 653.0 | 2353.4 | 770.8 | 2245.5 | 5.92 | 2544.6 |
74
+ | pythia | code/1024B | 8 | 128 | 384 | ok | 472.1 | 31.9 | 118556.1 | 34267.5 | 1295.0 | 2306.7 | 763.7 | 2251.2 | 5.88 | 2544.6 |
75
+ | pythia | code/1024B | 16 | 64 | 384 | ok | 472.1 | 63.2 | 119473.2 | 34532.6 | 638.2 | 4639.1 | 1570.3 | 4220.7 | 11.13 | 3153.5 |
76
+ | pythia | code/1024B | 16 | 128 | 384 | ok | 472.1 | 62.4 | 121004.4 | 34975.2 | 1273.0 | 4533.8 | 1543.7 | 4321.7 | 11.29 | 3153.5 |
77
+ | pythia | code/256B | 1 | 64 | 384 | ok | 196.7 | 12.9 | 15205.4 | 4704.4 | 585.0 | 365.5 | 108.1 | 357.6 | 1.0 | 1959.7 |
78
+ | pythia | code/256B | 1 | 128 | 384 | ok | 196.7 | 13.0 | 15153.2 | 4688.3 | 1157.3 | 359.9 | 108.7 | 355.9 | 1.0 | 1966.7 |
79
+ | pythia | code/256B | 2 | 64 | 384 | ok | 196.7 | 18.9 | 20759.3 | 6422.8 | 597.1 | 698.3 | 204.9 | 676.8 | 1.89 | 1996.4 |
80
+ | pythia | code/256B | 2 | 128 | 384 | ok | 196.7 | 18.6 | 21139.2 | 6540.3 | 1186.2 | 684.0 | 202.2 | 673.5 | 1.89 | 2001.3 |
81
+ | pythia | code/256B | 4 | 64 | 384 | ok | 196.7 | 18.3 | 42908.1 | 13275.5 | 600.9 | 1291.7 | 401.0 | 1253.4 | 3.51 | 2023.1 |
82
+ | pythia | code/256B | 4 | 128 | 384 | ok | 196.7 | 17.3 | 45482.3 | 14072.0 | 1210.3 | 1232.1 | 387.5 | 1214.7 | 3.41 | 2057.8 |
83
+ | pythia | code/256B | 8 | 64 | 384 | ok | 196.7 | 19.6 | 80095.9 | 24781.2 | 621.1 | 2436.6 | 775.5 | 2361.8 | 6.61 | 2109.2 |
84
+ | pythia | code/256B | 8 | 128 | 384 | ok | 196.7 | 19.4 | 81135.3 | 25102.8 | 1236.0 | 2357.2 | 759.3 | 2320.8 | 6.52 | 2167.8 |
85
+ | pythia | code/256B | 16 | 64 | 384 | ok | 196.7 | 24.8 | 126664.5 | 39189.3 | 623.6 | 4867.8 | 1503.2 | 4681.0 | 13.09 | 2279.2 |
86
+ | pythia | code/256B | 16 | 128 | 384 | ok | 196.7 | 25.3 | 124556.5 | 38537.1 | 1266.8 | 4524.9 | 1406.5 | 4436.3 | 12.47 | 2398.3 |
87
+
88
+ ## Warmup
89
+
90
+ | Model | Workload | Batch | Max new | Samples | Status | Prompt bytes | Prefill ms/chunk | Decode ms/chunk | Total ms/chunk | Peak VRAM MB |
91
+ |---|---|---:|---:|---:|---|---:|---:|---:|---:|---:|
92
+ | bolmo | code/1024B | 1 | 64 | 1 | ok | 265.0 | 43.8 | 3052.0 | 3341.1 | 2866.2 |
93
+ | bolmo | code/1024B | 1 | 128 | 1 | ok | 265.0 | 42.4 | 3651.7 | 3987.2 | 2857.4 |
94
+ | bolmo | code/1024B | 2 | 64 | 2 | ok | 613.5 | 1894.2 | 13495.6 | 17069.0 | 3026.0 |
95
+ | bolmo | code/1024B | 2 | 128 | 2 | ok | 613.5 | 38.1 | 14306.3 | 16128.1 | 3026.0 |
96
+ | bolmo | code/1024B | 4 | 64 | 4 | ok | 531.5 | 56.9 | 18632.9 | 22371.1 | 3217.5 |
97
+ | bolmo | code/1024B | 4 | 128 | 4 | ok | 531.5 | 50.8 | 19793.1 | 23814.4 | 3217.5 |
98
+ | bolmo | code/1024B | 8 | 64 | 8 | ok | 425.2 | 1960.2 | 24538.3 | 35218.8 | 3572.6 |
99
+ | bolmo | code/1024B | 8 | 128 | 8 | ok | 425.2 | 83.9 | 26349.0 | 35921.7 | 3572.6 |
100
+ | bolmo | code/1024B | 16 | 64 | 16 | ok | 423.9 | 2017.6 | 25250.7 | 43045.2 | 4319.6 |
101
+ | bolmo | code/1024B | 16 | 128 | 16 | ok | 423.9 | 146.6 | 27043.3 | 44540.5 | 4319.6 |
102
+ | bolmo | code/256B | 1 | 64 | 1 | ok | 212.0 | 80.8 | 2573.0 | 5345.6 | 2848.9 |
103
+ | bolmo | code/256B | 1 | 128 | 1 | ok | 212.0 | 38.8 | 3186.3 | 3472.0 | 2848.9 |
104
+ | bolmo | code/256B | 2 | 64 | 2 | ok | 196.0 | 2709.1 | 3626.6 | 6705.1 | 2876.3 |
105
+ | bolmo | code/256B | 2 | 128 | 2 | ok | 196.0 | 40.4 | 4453.0 | 4954.2 | 2876.3 |
106
+ | bolmo | code/256B | 4 | 64 | 4 | ok | 187.8 | 55.1 | 4618.7 | 5335.7 | 2924.8 |
107
+ | bolmo | code/256B | 4 | 128 | 4 | ok | 187.8 | 117.3 | 5647.9 | 6577.9 | 2924.8 |
108
+ | bolmo | code/256B | 8 | 64 | 8 | ok | 190.2 | 3023.3 | 6481.8 | 10735.6 | 3026.6 |
109
+ | bolmo | code/256B | 8 | 128 | 8 | ok | 190.2 | 49.6 | 6992.3 | 8624.1 | 3026.6 |
110
+ | bolmo | code/256B | 16 | 64 | 16 | ok | 179.8 | 3332.2 | 9968.8 | 16233.8 | 3239.8 |
111
+ | bolmo | code/256B | 16 | 128 | 16 | ok | 179.8 | 60.9 | 9696.9 | 13438.3 | 3252.2 |
112
+ | hnet | code/1024B | 1 | 64 | 1 | ok | 265.0 | 63.9 | 884.9 | 953.0 | 3319.5 |
113
+ | hnet | code/1024B | 1 | 128 | 1 | ok | 265.0 | 61.1 | 1819.8 | 1883.8 | 3343.8 |
114
+ | hnet | code/1024B | 2 | 64 | 2 | ok | 613.5 | 64.6 | 1288.2 | 1362.7 | 3773.8 |
115
+ | hnet | code/1024B | 2 | 128 | 2 | ok | 613.5 | 64.0 | 2531.4 | 2608.4 | 3826.2 |
116
+ | hnet | code/1024B | 4 | 64 | 4 | ok | 531.5 | 69.6 | 1531.7 | 1614.6 | 4310.4 |
117
+ | hnet | code/1024B | 4 | 128 | 4 | ok | 531.5 | 66.9 | 3024.8 | 3114.0 | 4370.2 |
118
+ | hnet | code/1024B | 8 | 64 | 8 | ok | 425.2 | 74.7 | 1765.8 | 1865.1 | 5350.4 |
119
+ | hnet | code/1024B | 8 | 128 | 8 | ok | 425.2 | 77.9 | 3662.8 | 3776.5 | 5470.1 |
120
+ | hnet | code/1024B | 16 | 64 | 16 | ok | 423.9 | 94.4 | 2196.2 | 2341.5 | 7480.3 |
121
+ | hnet | code/1024B | 16 | 128 | 16 | ok | 423.9 | 103.2 | 4417.9 | 4566.6 | 7716.3 |
122
+ | hnet | code/256B | 1 | 64 | 1 | ok | 212.0 | 20030.1 | 1047.3 | 21080.4 | 3556.9 |
123
+ | hnet | code/256B | 1 | 128 | 1 | ok | 212.0 | 61.1 | 1683.1 | 1747.1 | 3318.4 |
124
+ | hnet | code/256B | 2 | 64 | 2 | ok | 196.0 | 207.4 | 1191.1 | 1402.4 | 3384.2 |
125
+ | hnet | code/256B | 2 | 128 | 2 | ok | 196.0 | 63.0 | 2332.0 | 2399.4 | 3414.9 |
126
+ | hnet | code/256B | 4 | 64 | 4 | ok | 187.8 | 68.3 | 1503.7 | 1578.2 | 3559.3 |
127
+ | hnet | code/256B | 4 | 128 | 4 | ok | 187.8 | 66.7 | 3104.5 | 3182.8 | 3616.0 |
128
+ | hnet | code/256B | 8 | 64 | 8 | ok | 190.2 | 76.7 | 1872.6 | 1960.9 | 3903.5 |
129
+ | hnet | code/256B | 8 | 128 | 8 | ok | 190.2 | 73.2 | 3763.2 | 3853.8 | 4020.3 |
130
+ | hnet | code/256B | 16 | 64 | 16 | ok | 179.8 | 85.6 | 2464.0 | 2571.8 | 4640.5 |
131
+ | hnet | code/256B | 16 | 128 | 16 | ok | 179.8 | 85.2 | 4508.5 | 4627.2 | 4876.5 |
132
+ | hnet_serial | code/1024B | 1 | 64 | 1 | ok | 265.0 | 752.1 | 890.2 | 1646.6 | 3317.7 |
133
+ | hnet_serial | code/1024B | 1 | 128 | 1 | ok | 265.0 | 59.6 | 1813.3 | 1877.3 | 3343.6 |
134
+ | hnet_serial | code/1024B | 2 | 64 | 2 | ok | 613.5 | 117.1 | 1838.9 | 1967.3 | 3598.2 |
135
+ | hnet_serial | code/1024B | 2 | 128 | 2 | ok | 613.5 | 115.7 | 3463.3 | 3591.9 | 3635.6 |
136
+ | hnet_serial | code/1024B | 4 | 64 | 4 | ok | 531.5 | 225.7 | 3236.4 | 3479.5 | 3608.5 |
137
+ | hnet_serial | code/1024B | 4 | 128 | 4 | ok | 531.5 | 229.7 | 6386.6 | 6634.0 | 3635.6 |
138
+ | hnet_serial | code/1024B | 8 | 64 | 8 | ok | 425.2 | 452.2 | 6272.0 | 6749.1 | 3608.5 |
139
+ | hnet_serial | code/1024B | 8 | 128 | 8 | ok | 425.2 | 450.3 | 12604.3 | 13080.9 | 3635.6 |
140
+ | hnet_serial | code/1024B | 16 | 64 | 16 | ok | 423.9 | 894.5 | 12325.2 | 13261.0 | 3608.5 |
141
+ | hnet_serial | code/1024B | 16 | 128 | 16 | ok | 423.9 | 900.3 | 25133.6 | 26075.5 | 3635.6 |
142
+ | hnet_serial | code/256B | 1 | 64 | 1 | ok | 212.0 | 30196.8 | 847.5 | 31047.6 | 3554.1 |
143
+ | hnet_serial | code/256B | 1 | 128 | 1 | ok | 212.0 | 58.6 | 1740.8 | 1803.3 | 3317.0 |
144
+ | hnet_serial | code/256B | 2 | 64 | 2 | ok | 196.0 | 116.2 | 1691.3 | 1814.3 | 3370.4 |
145
+ | hnet_serial | code/256B | 2 | 128 | 2 | ok | 196.0 | 116.4 | 3369.0 | 3499.2 | 3398.4 |
146
+ | hnet_serial | code/256B | 4 | 64 | 4 | ok | 187.8 | 226.9 | 3276.3 | 3513.7 | 3370.4 |
147
+ | hnet_serial | code/256B | 4 | 128 | 4 | ok | 187.8 | 225.6 | 6515.2 | 6751.9 | 3398.4 |
148
+ | hnet_serial | code/256B | 8 | 64 | 8 | ok | 190.2 | 447.1 | 6504.2 | 6969.9 | 3376.2 |
149
+ | hnet_serial | code/256B | 8 | 128 | 8 | ok | 190.2 | 449.7 | 12838.5 | 13306.9 | 3404.0 |
150
+ | hnet_serial | code/256B | 16 | 64 | 16 | ok | 179.8 | 884.6 | 13329.9 | 14247.4 | 3382.4 |
151
+ | hnet_serial | code/256B | 16 | 128 | 16 | ok | 179.8 | 887.1 | 26288.8 | 27209.4 | 3413.2 |
152
+ | pythia | code/1024B | 1 | 64 | 1 | ok | 265.0 | 16.3 | 589.7 | 606.0 | 1959.9 |
153
+ | pythia | code/1024B | 1 | 128 | 1 | ok | 265.0 | 13.7 | 1162.3 | 1176.1 | 1966.8 |
154
+ | pythia | code/1024B | 2 | 64 | 2 | ok | 613.5 | 34.4 | 610.7 | 645.2 | 2065.1 |
155
+ | pythia | code/1024B | 2 | 128 | 2 | ok | 613.5 | 19.9 | 1176.4 | 1196.4 | 2065.1 |
156
+ | pythia | code/1024B | 4 | 64 | 4 | ok | 531.5 | 20.4 | 602.2 | 622.7 | 2192.2 |
157
+ | pythia | code/1024B | 4 | 128 | 4 | ok | 531.5 | 20.1 | 1202.6 | 1222.8 | 2192.2 |
158
+ | pythia | code/1024B | 8 | 64 | 8 | ok | 425.2 | 32.5 | 615.5 | 648.0 | 2438.1 |
159
+ | pythia | code/1024B | 8 | 128 | 8 | ok | 425.2 | 37.0 | 1230.5 | 1267.5 | 2438.1 |
160
+ | pythia | code/1024B | 16 | 64 | 16 | ok | 423.9 | 59.5 | 639.4 | 698.9 | 2944.8 |
161
+ | pythia | code/1024B | 16 | 128 | 16 | ok | 423.9 | 65.0 | 1277.3 | 1342.3 | 2944.8 |
162
+ | pythia | code/256B | 1 | 64 | 1 | ok | 212.0 | 18.7 | 597.2 | 616.0 | 1957.2 |
163
+ | pythia | code/256B | 1 | 128 | 1 | ok | 212.0 | 12.8 | 1176.5 | 1189.4 | 1965.2 |
164
+ | pythia | code/256B | 2 | 64 | 2 | ok | 196.0 | 21.6 | 599.2 | 620.8 | 1996.1 |
165
+ | pythia | code/256B | 2 | 128 | 2 | ok | 196.0 | 18.8 | 1179.8 | 1198.6 | 1998.0 |
166
+ | pythia | code/256B | 4 | 64 | 4 | ok | 187.8 | 16.3 | 597.9 | 614.2 | 2014.2 |
167
+ | pythia | code/256B | 4 | 128 | 4 | ok | 187.8 | 16.2 | 1197.6 | 1213.8 | 2053.4 |
168
+ | pythia | code/256B | 8 | 64 | 8 | ok | 190.2 | 19.3 | 615.4 | 634.7 | 2088.5 |
169
+ | pythia | code/256B | 8 | 128 | 8 | ok | 190.2 | 19.0 | 1233.3 | 1252.3 | 2156.6 |
170
+ | pythia | code/256B | 16 | 64 | 16 | ok | 179.8 | 22.3 | 624.2 | 646.5 | 2239.5 |
171
+ | pythia | code/256B | 16 | 128 | 16 | ok | 179.8 | 23.7 | 1272.3 | 1296.0 | 2382.3 |
172
+
173
+ ## Compression
174
+
175
+ | Model | Workload | Raw prompt bytes | Model input units | Stage0 / patches | Stage1 | Bytes per stage0 | Stage0 per stage1 |
176
+ |---|---|---:|---:|---:|---:|---:|---:|
177
+ | bolmo | code/1024B | 181272.0 | 181656.0 | 40221.0 | None | 4.51 | None |
178
+ | bolmo | code/1024B | 181272.0 | 181656.0 | 40221.0 | None | 4.51 | None |
179
+ | bolmo | code/1024B | 181272.0 | 181656.0 | 40206.0 | None | 4.51 | None |
180
+ | bolmo | code/1024B | 181272.0 | 181656.0 | 40206.0 | None | 4.51 | None |
181
+ | bolmo | code/1024B | 181272.0 | 181656.0 | 40194.0 | None | 4.51 | None |
182
+ | bolmo | code/1024B | 181272.0 | 181656.0 | 40194.0 | None | 4.51 | None |
183
+ | bolmo | code/1024B | 181272.0 | 181656.0 | 40212.0 | None | 4.51 | None |
184
+ | bolmo | code/1024B | 181272.0 | 181656.0 | 40212.0 | None | 4.51 | None |
185
+ | bolmo | code/1024B | 181272.0 | 181656.0 | 40209.0 | None | 4.51 | None |
186
+ | bolmo | code/1024B | 181272.0 | 181656.0 | 40209.0 | None | 4.51 | None |
187
+ | bolmo | code/256B | 75525.0 | 75909.0 | 17772.0 | None | 4.25 | None |
188
+ | bolmo | code/256B | 75525.0 | 75909.0 | 17772.0 | None | 4.25 | None |
189
+ | bolmo | code/256B | 75525.0 | 75909.0 | 17769.0 | None | 4.25 | None |
190
+ | bolmo | code/256B | 75525.0 | 75909.0 | 17769.0 | None | 4.25 | None |
191
+ | bolmo | code/256B | 75525.0 | 75909.0 | 17769.0 | None | 4.25 | None |
192
+ | bolmo | code/256B | 75525.0 | 75909.0 | 17769.0 | None | 4.25 | None |
193
+ | bolmo | code/256B | 75525.0 | 75909.0 | 17775.0 | None | 4.25 | None |
194
+ | bolmo | code/256B | 75525.0 | 75909.0 | 17775.0 | None | 4.25 | None |
195
+ | bolmo | code/256B | 75525.0 | 75909.0 | 17763.0 | None | 4.25 | None |
196
+ | bolmo | code/256B | 75525.0 | 75909.0 | 17763.0 | None | 4.25 | None |
197
+ | hnet | code/1024B | 181272.0 | 181656.0 | 63561.0 | 30735.0 | 2.85 | 2.07 |
198
+ | hnet | code/1024B | 181272.0 | 181656.0 | 63561.0 | 30735.0 | 2.85 | 2.07 |
199
+ | hnet | code/1024B | 181272.0 | 181656.0 | 63564.0 | 30726.0 | 2.85 | 2.07 |
200
+ | hnet | code/1024B | 181272.0 | 181656.0 | 63564.0 | 30726.0 | 2.85 | 2.07 |
201
+ | hnet | code/1024B | 181272.0 | 181656.0 | 63576.0 | 30720.0 | 2.85 | 2.07 |
202
+ | hnet | code/1024B | 181272.0 | 181656.0 | 63576.0 | 30720.0 | 2.85 | 2.07 |
203
+ | hnet | code/1024B | 181272.0 | 181656.0 | 63573.0 | 30723.0 | 2.85 | 2.07 |
204
+ | hnet | code/1024B | 181272.0 | 181656.0 | 63573.0 | 30723.0 | 2.85 | 2.07 |
205
+ | hnet | code/1024B | 181272.0 | 181656.0 | 63570.0 | 30729.0 | 2.85 | 2.07 |
206
+ | hnet | code/1024B | 181272.0 | 181656.0 | 63570.0 | 30729.0 | 2.85 | 2.07 |
207
+ | hnet | code/256B | 75525.0 | 75909.0 | 28476.0 | 14922.0 | 2.65 | 1.91 |
208
+ | hnet | code/256B | 75525.0 | 75909.0 | 28476.0 | 14922.0 | 2.65 | 1.91 |
209
+ | hnet | code/256B | 75525.0 | 75909.0 | 28479.0 | 14916.0 | 2.65 | 1.91 |
210
+ | hnet | code/256B | 75525.0 | 75909.0 | 28479.0 | 14916.0 | 2.65 | 1.91 |
211
+ | hnet | code/256B | 75525.0 | 75909.0 | 28473.0 | 14919.0 | 2.65 | 1.91 |
212
+ | hnet | code/256B | 75525.0 | 75909.0 | 28473.0 | 14919.0 | 2.65 | 1.91 |
213
+ | hnet | code/256B | 75525.0 | 75909.0 | 28473.0 | 14922.0 | 2.65 | 1.91 |
214
+ | hnet | code/256B | 75525.0 | 75909.0 | 28473.0 | 14922.0 | 2.65 | 1.91 |
215
+ | hnet | code/256B | 75525.0 | 75909.0 | 28473.0 | 14928.0 | 2.65 | 1.91 |
216
+ | hnet | code/256B | 75525.0 | 75909.0 | 28473.0 | 14928.0 | 2.65 | 1.91 |
217
+ | hnet_serial | code/1024B | 181272.0 | 181656.0 | 63561.0 | 30726.0 | 2.85 | 2.07 |
218
+ | hnet_serial | code/1024B | 181272.0 | 181656.0 | 63561.0 | 30726.0 | 2.85 | 2.07 |
219
+ | hnet_serial | code/1024B | 181272.0 | 181656.0 | 63561.0 | 30726.0 | 2.85 | 2.07 |
220
+ | hnet_serial | code/1024B | 181272.0 | 181656.0 | 63561.0 | 30726.0 | 2.85 | 2.07 |
221
+ | hnet_serial | code/1024B | 181272.0 | 181656.0 | 63561.0 | 30726.0 | 2.85 | 2.07 |
222
+ | hnet_serial | code/1024B | 181272.0 | 181656.0 | 63561.0 | 30726.0 | 2.85 | 2.07 |
223
+ | hnet_serial | code/1024B | 181272.0 | 181656.0 | 63561.0 | 30726.0 | 2.85 | 2.07 |
224
+ | hnet_serial | code/1024B | 181272.0 | 181656.0 | 63561.0 | 30726.0 | 2.85 | 2.07 |
225
+ | hnet_serial | code/1024B | 181272.0 | 181656.0 | 63561.0 | 30726.0 | 2.85 | 2.07 |
226
+ | hnet_serial | code/1024B | 181272.0 | 181656.0 | 63561.0 | 30726.0 | 2.85 | 2.07 |
227
+ | hnet_serial | code/256B | 75525.0 | 75909.0 | 28476.0 | 14922.0 | 2.65 | 1.91 |
228
+ | hnet_serial | code/256B | 75525.0 | 75909.0 | 28476.0 | 14922.0 | 2.65 | 1.91 |
229
+ | hnet_serial | code/256B | 75525.0 | 75909.0 | 28476.0 | 14922.0 | 2.65 | 1.91 |
230
+ | hnet_serial | code/256B | 75525.0 | 75909.0 | 28476.0 | 14922.0 | 2.65 | 1.91 |
231
+ | hnet_serial | code/256B | 75525.0 | 75909.0 | 28476.0 | 14922.0 | 2.65 | 1.91 |
232
+ | hnet_serial | code/256B | 75525.0 | 75909.0 | 28476.0 | 14922.0 | 2.65 | 1.91 |
233
+ | hnet_serial | code/256B | 75525.0 | 75909.0 | 28476.0 | 14922.0 | 2.65 | 1.91 |
234
+ | hnet_serial | code/256B | 75525.0 | 75909.0 | 28476.0 | 14922.0 | 2.65 | 1.91 |
235
+ | hnet_serial | code/256B | 75525.0 | 75909.0 | 28476.0 | 14922.0 | 2.65 | 1.91 |
236
+ | hnet_serial | code/256B | 75525.0 | 75909.0 | 28476.0 | 14922.0 | 2.65 | 1.91 |
237
+ | pythia | code/1024B | 181272.0 | 52395.0 | None | None | None | None |
238
+ | pythia | code/1024B | 181272.0 | 52395.0 | None | None | None | None |
239
+ | pythia | code/1024B | 181272.0 | 52395.0 | None | None | None | None |
240
+ | pythia | code/1024B | 181272.0 | 52395.0 | None | None | None | None |
241
+ | pythia | code/1024B | 181272.0 | 52395.0 | None | None | None | None |
242
+ | pythia | code/1024B | 181272.0 | 52395.0 | None | None | None | None |
243
+ | pythia | code/1024B | 181272.0 | 52395.0 | None | None | None | None |
244
+ | pythia | code/1024B | 181272.0 | 52395.0 | None | None | None | None |
245
+ | pythia | code/1024B | 181272.0 | 52395.0 | None | None | None | None |
246
+ | pythia | code/1024B | 181272.0 | 52395.0 | None | None | None | None |
247
+ | pythia | code/256B | 75525.0 | 23367.0 | None | None | None | None |
248
+ | pythia | code/256B | 75525.0 | 23367.0 | None | None | None | None |
249
+ | pythia | code/256B | 75525.0 | 23367.0 | None | None | None | None |
250
+ | pythia | code/256B | 75525.0 | 23367.0 | None | None | None | None |
251
+ | pythia | code/256B | 75525.0 | 23367.0 | None | None | None | None |
252
+ | pythia | code/256B | 75525.0 | 23367.0 | None | None | None | None |
253
+ | pythia | code/256B | 75525.0 | 23367.0 | None | None | None | None |
254
+ | pythia | code/256B | 75525.0 | 23367.0 | None | None | None | None |
255
+ | pythia | code/256B | 75525.0 | 23367.0 | None | None | None | None |
256
+ | pythia | code/256B | 75525.0 | 23367.0 | None | None | None | None |
generation_bench_results/warmup_runs.partial.jsonl ADDED
The diff for this file is too large to render. See raw diff
 
generation_bench_results/warmup_summary.csv ADDED
@@ -0,0 +1,81 @@
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
1
+ model,workload,batch_size,max_new_tokens,status,num_samples,num_chunks,num_failures,prompt_bytes_mean,generated_units_mean,generated_bytes_mean,prefill_time_sec,decode_time_sec,total_time_sec,prefill_bytes_per_sec,prefill_units_per_sec,decode_bytes_per_sec,decode_units_per_sec,total_bytes_per_sec,total_units_per_sec,speedup_vs_batch1,peak_allocated_mb,peak_reserved_mb
2
+ bolmo,code/1024B,1,64,ok,1,1,0,265.0,327.0,327.0,0.04379,3.051978,3.341087,6051.566017,6074.402115,107.143639,107.143639,97.872342,97.872342,1.0,2866.227051,3072.0
3
+ bolmo,code/1024B,1,128,ok,1,1,0,265.0,391.0,391.0,0.042398,3.651743,3.98722,6250.363814,6273.950092,107.072168,107.072168,98.063313,98.063313,1.0,2857.4375,2986.0
4
+ bolmo,code/1024B,2,64,ok,2,1,0,613.5,1024.0,1024.0,1.894246,13.495629,17.069034,647.75116,648.806989,151.752833,151.752833,119.983356,119.983356,1.225917,3025.992676,3108.0
5
+ bolmo,code/1024B,2,128,ok,2,1,0,613.5,1088.0,1088.0,0.038093,14.306308,16.128132,32210.236461,32262.738884,152.100737,152.100737,134.919528,134.919528,1.375841,3025.992676,3106.0
6
+ bolmo,code/1024B,4,64,ok,4,1,0,531.5,1024.0,1024.0,0.056905,18.632943,22.371078,37360.658277,37430.951143,219.825716,219.825716,183.093543,183.093543,1.870738,3217.512695,3348.0
7
+ bolmo,code/1024B,4,128,ok,4,1,0,531.5,1088.0,1088.0,0.050812,19.793111,23.814403,41840.244151,41918.965212,219.874484,219.874484,182.746551,182.746551,1.863557,3217.512695,3344.0
8
+ bolmo,code/1024B,8,64,ok,8,1,0,425.25,928.625,928.625,1.960168,24.538283,35.218823,1735.565389,1739.646672,302.75142,302.75142,210.938339,210.938339,2.15524,3572.615234,3838.0
9
+ bolmo,code/1024B,8,128,ok,8,1,0,425.25,984.625,984.625,0.083857,26.348955,35.921669,40568.993831,40664.39417,298.949237,298.949237,219.282687,219.282687,2.236134,3572.615234,3838.0
10
+ bolmo,code/1024B,16,64,ok,16,1,0,423.875,976.3125,976.3125,2.017594,25.250654,43.045219,3361.429629,3369.359867,618.637453,618.637453,362.897445,362.897445,3.707865,4319.635742,4690.0
11
+ bolmo,code/1024B,16,128,ok,16,1,0,423.875,1036.3125,1036.3125,0.146639,27.043296,44.540518,46249.668799,46358.780374,613.127938,613.127938,372.267787,372.267787,3.796198,4319.635742,4672.0
12
+ bolmo,code/256B,1,64,ok,1,1,0,212.0,274.0,274.0,0.08084,2.573032,5.345569,2622.476225,2634.846395,106.489148,106.489148,51.257404,51.257404,1.0,2848.945312,2982.0
13
+ bolmo,code/256B,1,128,ok,1,1,0,212.0,338.0,338.0,0.038767,3.186286,3.472025,5468.543804,5494.338822,106.079631,106.079631,97.349533,97.349533,1.0,2848.945312,2984.0
14
+ bolmo,code/256B,2,64,ok,2,1,0,196.0,274.0,274.0,2.709114,3.62656,6.70509,144.696754,145.435003,151.107376,151.107376,81.728954,81.728954,1.594481,2876.251465,2998.0
15
+ bolmo,code/256B,2,128,ok,2,1,0,196.0,338.0,338.0,0.040447,4.453029,4.954154,9691.761205,9741.208967,151.806793,151.806793,136.451135,136.451135,1.401662,2876.251465,3004.0
16
+ bolmo,code/256B,4,64,ok,4,1,0,187.75,274.0,274.0,0.055123,4.618685,5.335736,13624.138596,13696.703914,237.296963,237.296963,205.407474,205.407474,4.007372,2924.803711,3018.0
17
+ bolmo,code/256B,4,128,ok,4,1,0,187.75,338.0,338.0,0.117345,5.647875,6.577894,6399.94181,6434.029383,239.382056,239.382056,205.53691,205.53691,2.111329,2924.803711,3016.0
18
+ bolmo,code/256B,8,64,ok,8,1,0,190.25,279.0,279.0,3.023288,6.481793,10.735635,503.425464,506.07159,344.349153,344.349153,207.905725,207.905725,4.056111,3026.631348,3092.0
19
+ bolmo,code/256B,8,128,ok,8,1,0,190.25,343.0,343.0,0.049587,6.992286,8.624101,30693.38685,30854.718713,392.432482,392.432482,318.178077,318.178077,3.268409,3026.631348,3152.0
20
+ bolmo,code/256B,16,64,ok,16,1,0,179.75,294.0,294.0,3.332186,9.968759,16.233793,863.097136,867.898789,471.874186,471.874186,289.765918,289.765918,5.653152,3239.841797,3402.0
21
+ bolmo,code/256B,16,128,ok,16,1,0,179.75,358.0,358.0,0.060938,9.696858,13.438344,47195.174085,47457.734163,590.70682,590.70682,426.242987,426.242987,4.37848,3252.166992,3592.0
22
+ hnet,code/1024B,1,64,ok,1,1,0,265.0,64.0,64.0,0.06391,0.884931,0.952953,4146.449207,4162.096185,72.322016,72.322016,67.159627,67.159627,1.0,3319.489258,3356.0
23
+ hnet,code/1024B,1,128,ok,1,1,0,265.0,128.0,128.0,0.061092,1.819794,1.883795,4337.689842,4354.058483,70.33762,70.33762,67.947929,67.947929,1.0,3343.760742,3378.0
24
+ hnet,code/1024B,2,64,ok,2,1,0,613.5,64.0,64.0,0.064624,1.288248,1.362692,18986.89134,19017.839818,99.359761,99.359761,93.931687,93.931687,1.398633,3773.793457,3902.0
25
+ hnet,code/1024B,2,128,ok,2,1,0,613.5,128.0,128.0,0.063998,2.531424,2.608432,19172.331535,19203.582279,101.128844,101.128844,98.14327,98.14327,1.444389,3826.20166,3902.0
26
+ hnet,code/1024B,4,64,ok,4,1,0,531.5,64.0,64.0,0.06962,1.531696,1.614571,30537.244453,30594.699287,167.134986,167.134986,158.556024,158.556024,2.360883,4310.385254,4432.0
27
+ hnet,code/1024B,4,128,ok,4,1,0,531.5,128.0,128.0,0.066946,3.024764,3.113983,31756.78282,31816.532176,169.269417,169.269417,164.419656,164.419656,2.419789,4370.244629,4488.0
28
+ hnet,code/1024B,8,64,ok,8,1,0,425.25,64.0,64.0,0.074711,1.765848,1.865068,45535.648034,45642.727747,289.945747,289.945747,274.520801,274.520801,4.087587,5350.383789,5472.0
29
+ hnet,code/1024B,8,128,ok,8,1,0,425.25,128.0,128.0,0.077896,3.662848,3.776504,43673.720404,43776.421686,279.563854,279.563854,271.150244,271.150244,3.990559,5470.102539,5588.0
30
+ hnet,code/1024B,16,64,ok,16,1,0,423.875,64.0,64.0,0.094416,2.196197,2.341482,71831.338431,72000.801925,466.260444,466.260444,437.329774,437.329774,6.511796,7480.275879,7680.0
31
+ hnet,code/1024B,16,128,ok,16,1,0,423.875,128.0,128.0,0.103236,4.417948,4.566563,65693.925133,65848.909326,463.563602,463.563602,448.477375,448.477375,6.60031,7716.275879,7916.0
32
+ hnet,code/256B,1,64,ok,1,1,0,212.0,64.0,64.0,20.030064,1.04729,21.080372,10.58409,10.634015,61.110126,61.110126,3.036,3.036,1.0,3556.90332,9920.0
33
+ hnet,code/256B,1,128,ok,1,1,0,212.0,128.0,128.0,0.061119,1.683118,1.747076,3468.642289,3485.003809,76.049328,76.049328,73.26526,73.26526,1.0,3318.39209,3358.0
34
+ hnet,code/256B,2,64,ok,2,1,0,196.0,64.0,64.0,0.20736,1.191087,1.402412,1890.433512,1900.078581,107.464868,107.464868,91.271295,91.271295,30.063014,3384.153809,3456.0
35
+ hnet,code/256B,2,128,ok,2,1,0,196.0,128.0,128.0,0.062983,2.331989,2.399365,6223.898425,6255.653009,109.777531,109.777531,106.69488,106.69488,1.456282,3414.874512,3496.0
36
+ hnet,code/256B,4,64,ok,4,1,0,187.75,64.0,64.0,0.068301,1.50372,1.57824,10995.397685,11053.961721,170.244449,170.244449,162.206045,162.206045,53.427559,3559.291016,3606.0
37
+ hnet,code/256B,4,128,ok,4,1,0,187.75,128.0,128.0,0.066677,3.104469,3.182806,11263.324475,11323.315551,164.923546,164.923546,160.864354,160.864354,2.195643,3615.97168,3690.0
38
+ hnet,code/256B,8,64,ok,8,1,0,190.25,64.0,64.0,0.076728,1.872641,1.960873,19836.195269,19940.459108,273.410681,273.410681,261.108161,261.108161,86.004019,3903.505859,3964.0
39
+ hnet,code/256B,8,128,ok,8,1,0,190.25,128.0,128.0,0.073157,3.763189,3.853792,20804.506167,20913.859682,272.109619,272.109619,265.71232,265.71232,3.626716,4020.332031,4072.0
40
+ hnet,code/256B,16,64,ok,16,1,0,179.75,64.0,64.0,0.085578,2.463951,2.571779,33606.598053,33793.561046,415.592674,415.592674,398.167884,398.167884,131.148862,4640.475098,4750.0
41
+ hnet,code/256B,16,128,ok,16,1,0,179.75,128.0,128.0,0.085244,4.508492,4.627245,33738.579493,33926.276736,454.253873,454.253873,442.595984,442.595984,6.041007,4876.475098,4986.0
42
+ hnet_serial,code/1024B,1,64,ok,1,1,0,265.0,64.0,64.0,0.752092,0.890161,1.646582,352.350467,353.680091,71.897116,71.897116,38.868399,38.868399,1.0,3317.672852,3346.0
43
+ hnet_serial,code/1024B,1,128,ok,1,1,0,265.0,128.0,128.0,0.059563,1.81326,1.8773,4449.066768,4465.855699,70.591098,70.591098,68.183034,68.183034,1.0,3343.647461,3368.0
44
+ hnet_serial,code/1024B,2,64,ok,2,1,0,613.5,64.0,64.0,0.117054,1.838932,1.967271,10482.350978,10499.437125,69.605609,69.605609,65.064749,65.064749,1.673976,3598.213867,3680.0
45
+ hnet_serial,code/1024B,2,128,ok,2,1,0,613.5,128.0,128.0,0.115725,3.46331,3.591909,10602.702277,10619.984596,73.917723,73.917723,71.271303,71.271303,1.045294,3635.569824,3720.0
46
+ hnet_serial,code/1024B,4,64,ok,4,1,0,531.5,64.0,64.0,0.225711,3.236391,3.479509,9419.140029,9436.861836,79.100455,79.100455,73.573593,73.573593,1.89289,3608.458008,3680.0
47
+ hnet_serial,code/1024B,4,128,ok,4,1,0,531.5,128.0,128.0,0.229733,6.386635,6.633962,9254.23937,9271.650921,80.167412,80.167412,77.178616,77.178616,1.131933,3635.569824,3720.0
48
+ hnet_serial,code/1024B,8,64,ok,8,1,0,425.25,64.0,64.0,0.452215,6.271957,6.749106,7522.962699,7540.653382,81.633216,81.633216,75.861895,75.861895,1.951763,3608.458008,3682.0
49
+ hnet_serial,code/1024B,8,128,ok,8,1,0,425.25,128.0,128.0,0.450344,12.604267,13.080947,7554.222629,7571.986821,81.242328,81.242328,78.281794,78.281794,1.148113,3635.569824,3722.0
50
+ hnet_serial,code/1024B,16,64,ok,16,1,0,423.875,64.0,64.0,0.894514,12.325218,13.261026,7581.773133,7599.659947,83.081696,83.081696,77.218763,77.218763,1.986672,3608.458008,3682.0
51
+ hnet_serial,code/1024B,16,128,ok,16,1,0,423.875,128.0,128.0,0.900315,25.133584,26.075524,7532.922791,7550.694358,81.484598,81.484598,78.541088,78.541088,1.151915,3635.569824,3722.0
52
+ hnet_serial,code/256B,1,64,ok,1,1,0,212.0,64.0,64.0,30.196807,0.847523,31.047593,7.02061,7.053726,75.514192,75.514192,2.061351,2.061351,1.0,3554.09375,9928.0
53
+ hnet_serial,code/256B,1,128,ok,1,1,0,212.0,128.0,128.0,0.058596,1.740761,1.803314,3617.963719,3635.029586,73.53107,73.53107,70.980447,70.980447,1.0,3317.029297,3350.0
54
+ hnet_serial,code/256B,2,64,ok,2,1,0,196.0,64.0,64.0,0.116233,1.691341,1.814323,3372.546587,3389.753457,75.679572,75.679572,70.549717,70.549717,34.224983,3370.398438,3398.0
55
+ hnet_serial,code/256B,2,128,ok,2,1,0,196.0,128.0,128.0,0.116396,3.369004,3.499221,3367.813006,3384.995726,75.986851,75.986851,73.159136,73.159136,1.030694,3398.428223,3438.0
56
+ hnet_serial,code/256B,4,64,ok,4,1,0,187.75,64.0,64.0,0.226863,3.276344,3.513685,3310.371333,3328.003137,78.135871,78.135871,72.857983,72.857983,35.344766,3370.398438,3398.0
57
+ hnet_serial,code/256B,4,128,ok,4,1,0,187.75,128.0,128.0,0.22564,6.515192,6.751937,3328.307352,3346.034688,78.585564,78.585564,75.830095,75.830095,1.068324,3398.428223,3438.0
58
+ hnet_serial,code/256B,8,64,ok,8,1,0,190.25,64.0,64.0,0.447135,6.504238,6.969891,3403.894143,3421.785833,78.717909,78.717909,73.458828,73.458828,35.636247,3376.219238,3420.0
59
+ hnet_serial,code/256B,8,128,ok,8,1,0,190.25,128.0,128.0,0.449676,12.838486,13.306864,3384.660027,3402.450619,79.760184,79.760184,76.952766,76.952766,1.08414,3404.047363,3440.0
60
+ hnet_serial,code/256B,16,64,ok,16,1,0,179.75,64.0,64.0,0.884576,13.329869,14.247372,3251.274385,3269.362143,76.819957,76.819957,71.872905,71.872905,34.866887,3382.36084,3420.0
61
+ hnet_serial,code/256B,16,128,ok,16,1,0,179.75,128.0,128.0,0.887136,26.288754,27.209399,3241.891632,3259.927191,77.904035,77.904035,75.268109,75.268109,1.060406,3413.165527,3440.0
62
+ pythia,code/1024B,1,64,ok,1,1,0,265.0,64.0,192.0,0.016295,0.589667,0.606002,16263.020594,5768.769569,325.607237,108.535746,316.83071,105.610237,1.0,1959.880371,2028.0
63
+ pythia,code/1024B,1,128,ok,1,1,0,265.0,128.0,377.0,0.013746,1.162323,1.176103,19279.033989,6838.600736,324.350472,110.124298,320.550147,108.834002,1.0,1966.83252,2038.0
64
+ pythia,code/1024B,2,64,ok,2,1,0,613.5,64.0,174.5,0.03442,0.610733,0.645218,35647.652311,10778.548498,571.444514,209.584234,540.902108,198.382435,1.707228,2065.075195,2138.0
65
+ pythia,code/1024B,2,128,ok,2,1,0,613.5,128.0,340.5,0.019911,1.176438,1.196383,61623.345681,18632.649754,578.866099,217.606052,569.215498,213.97822,1.775746,2065.075195,2138.0
66
+ pythia,code/1024B,4,64,ok,4,1,0,531.5,64.0,137.0,0.020396,0.602228,0.622657,104235.526768,30594.058656,909.955025,425.088479,880.099022,411.141149,2.777821,2192.229492,2314.0
67
+ pythia,code/1024B,4,128,ok,4,1,0,531.5,128.0,269.75,0.020142,1.202574,1.222751,105550.402026,30979.986295,897.24172,425.753253,882.436743,418.728093,2.752882,2192.229492,2314.0
68
+ pythia,code/1024B,8,64,ok,8,1,0,425.25,63.5,144.375,0.032533,0.615469,0.648039,104569.277888,30030.624485,1876.617272,825.386645,1782.298866,783.902878,5.625398,2438.133789,2662.0
69
+ pythia,code/1024B,8,128,ok,8,1,0,425.25,119.5,267.875,0.037028,1.230472,1.267535,91877.256717,26385.678957,1741.607585,776.937401,1690.683171,754.219837,5.274317,2438.133789,2662.0
70
+ pythia,code/1024B,16,64,ok,16,1,0,423.875,63.75,132.75,0.059468,0.639361,0.698863,114045.406289,32286.520212,3322.068222,1595.343496,3039.221397,1459.5131,9.592572,2944.830078,3308.0
71
+ pythia,code/1024B,16,128,ok,16,1,0,423.875,123.75,251.9375,0.065003,1.277255,1.342294,104334.304294,29537.284613,3155.987646,1550.199836,3003.067703,1475.086592,9.36848,2944.830078,3308.0
72
+ pythia,code/256B,1,64,ok,1,1,0,212.0,64.0,203.0,0.018742,0.597224,0.616,11311.197754,4321.731217,339.905808,107.162422,329.545711,103.896185,1.0,1957.20752,2008.0
73
+ pythia,code/256B,1,128,ok,1,1,0,212.0,128.0,366.0,0.012845,1.176523,1.1894,16504.877313,6306.108785,311.086092,108.795136,307.718177,107.617286,1.0,1965.20752,2016.0
74
+ pythia,code/256B,2,64,ok,2,1,0,196.0,64.0,196.0,0.021619,0.599184,0.620836,18132.168255,6198.241189,654.223164,213.62389,631.406978,206.173707,1.915992,1996.058105,2034.0
75
+ pythia,code/256B,2,128,ok,2,1,0,196.0,128.0,374.0,0.018807,1.179804,1.198644,20843.431514,7125.050568,634.00346,216.985142,624.038317,213.574611,2.027954,1997.972168,2054.0
76
+ pythia,code/256B,4,64,ok,4,1,0,187.75,57.75,171.25,0.016291,0.597874,0.614199,46099.828633,15161.994237,1145.726217,386.368987,1115.273917,376.099671,3.384277,2014.158203,2056.0
77
+ pythia,code/256B,4,128,ok,4,1,0,187.75,105.75,316.0,0.016186,1.197611,1.21383,46397.213258,15259.802496,1055.434217,353.203065,1041.331676,348.483623,3.384043,2053.385742,2076.0
78
+ pythia,code/256B,8,64,ok,8,1,0,190.25,54.5,154.0,0.019256,0.615446,0.634739,79038.973353,24615.291307,2001.800532,708.429409,1940.955063,686.896435,5.88979,2088.477539,2176.0
79
+ pythia,code/256B,8,128,ok,8,1,0,190.25,102.5,286.875,0.019011,1.233271,1.252318,80058.155115,24932.697454,1860.904293,664.898266,1832.601265,654.785637,5.955453,2156.571289,2256.0
80
+ pythia,code/256B,16,64,ok,16,1,0,179.75,57.0625,173.1875,0.022288,0.624164,0.646487,129040.565601,39887.71308,4439.53928,1462.756897,4286.242083,1412.247933,13.006518,2239.532227,2398.0
81
+ pythia,code/256B,16,128,ok,16,1,0,179.75,105.0625,319.375,0.02366,1.272321,1.296037,121554.936154,37573.831099,4016.280495,1321.20695,3942.789851,1297.03126,12.81299,2382.317383,2698.0