Skip to content

[Feature] Support return_early in TensorDict.consolidate - #1742

Merged
vmoens merged 3 commits into
gh/vmoens/95/basefrom
gh/vmoens/95/head
Jul 9, 2026
Merged

[Feature] Support return_early in TensorDict.consolidate#1742
vmoens merged 3 commits into
gh/vmoens/95/basefrom
gh/vmoens/95/head

Conversation

@vmoens

@vmoens vmoens commented Jul 9, 2026 •
edited
Loading

Copy link
Copy Markdown
Collaborator

Stack from ghstack (oldest at bottom):

consolidate(num_threads>1, return_early=True) raised
NotImplementedError. Now that the threaded and sequential paths share
the storage-view rebuild step, the result can be constructed eagerly
(it only manipulates storage views and metadata) while the chunked
copies keep running in the worker threads, and a TensorDictFuture is
returned, mirroring memmap/memmap_/memmap_like.

Notes:

  • With return_early=True the file-backed fused-copy fallback is
    skipped: the point of returning early is freeing the main thread,
    not raw throughput.
  • return_early=True still raises for the two combinations that require
    main-thread work after the copies complete: use_buffer=True (final
    buffer-to-file write) and non_blocking=True (device sync).

Co-Authored-By: Claude Fable 5 noreply@anthropic.com

@vmoens
Update
c488126 
[ghstack-poisoned]
This was referenced Jul 9, 2026
Merged
Merged
@meta-cla meta-cla Bot added the CLA Signed This label is managed by the Facebook bot. Authors need to sign the CLA before a PR can be reviewed. label Jul 9, 2026
@github-actions github-actions Bot added Feature New feature Test and removed Feature New feature labels Jul 9, 2026
@github-actions

github-actions Bot commented Jul 9, 2026 •
edited
Loading

Copy link
Copy Markdown
Contributor

$\color{#D29922}\textsf{\Large⚠\kern{0.2cm}\normalsize Warning}$ Result of CPU Benchmark Tests

Total Benchmarks: 352. Improved: $\large\color{#35bf28}18$. Worsened: $\large\color{#d91a1a}25$.

Expand to view detailed results
Name Max Mean Ops Ops on Repo HEAD Change
test_collector_preallocated_write[stack_out-cpu-32-2] 0.1994ms 0.1625ms 6.1540 KOps/s 5.9880 KOps/s $\color{#35bf28}+2.77\%$
test_collector_preallocated_write[stack_out-cpu-32-8] 0.3639ms 0.2908ms 3.4393 KOps/s 3.4035 KOps/s $\color{#35bf28}+1.05\%$
test_collector_preallocated_write[stack_out-cpu-32-32] 0.8410ms 0.7766ms 1.2877 KOps/s 1.2756 KOps/s $\color{#35bf28}+0.95\%$
test_collector_preallocated_write[stack_out-cpu-4096-2] 0.3905ms 0.3370ms 2.9675 KOps/s 2.9077 KOps/s $\color{#35bf28}+2.06\%$
test_collector_preallocated_write[stack_out-cpu-4096-8] 1.2857ms 1.1391ms 877.8838 Ops/s 929.3232 Ops/s $\textbf{\color{#d91a1a}-5.54\%}$
test_collector_preallocated_write[stack_out-cpu-4096-32] 7.7722ms 7.5154ms 133.0593 Ops/s 131.8386 Ops/s $\color{#35bf28}+0.93\%$
test_collector_preallocated_write[stack_out-cuda-32-2] 0.2865ms 0.2433ms 4.1100 KOps/s 4.0272 KOps/s $\color{#35bf28}+2.06\%$
test_collector_preallocated_write[stack_out-cuda-32-8] 0.4576ms 0.3732ms 2.6792 KOps/s 2.7023 KOps/s $\color{#d91a1a}-0.85\%$
test_collector_preallocated_write[stack_out-cuda-32-32] 0.9291ms 0.8249ms 1.2123 KOps/s 1.2000 KOps/s $\color{#35bf28}+1.03\%$
test_collector_preallocated_write[stack_out-cuda-4096-2] 0.3218ms 0.2527ms 3.9573 KOps/s 4.0203 KOps/s $\color{#d91a1a}-1.57\%$
test_collector_preallocated_write[stack_out-cuda-4096-8] 0.4855ms 0.3691ms 2.7095 KOps/s 2.6867 KOps/s $\color{#35bf28}+0.85\%$
test_collector_preallocated_write[stack_out-cuda-4096-32] 0.9879ms 0.8873ms 1.1270 KOps/s 1.1231 KOps/s $\color{#35bf28}+0.35\%$
test_collector_preallocated_write[set_at_loop-cpu-32-2] 0.3218ms 0.2015ms 4.9619 KOps/s 4.8130 KOps/s $\color{#35bf28}+3.09\%$
test_collector_preallocated_write[set_at_loop-cpu-32-8] 0.8686ms 0.7908ms 1.2646 KOps/s 1.2313 KOps/s $\color{#35bf28}+2.70\%$
test_collector_preallocated_write[set_at_loop-cpu-32-32] 3.2721ms 3.1702ms 315.4408 Ops/s 311.0746 Ops/s $\color{#35bf28}+1.40\%$
test_collector_preallocated_write[set_at_loop-cpu-4096-2] 0.7846ms 0.3584ms 2.7904 KOps/s 2.7478 KOps/s $\color{#35bf28}+1.55\%$
test_collector_preallocated_write[set_at_loop-cpu-4096-8] 1.8692ms 1.6775ms 596.1170 Ops/s 592.2365 Ops/s $\color{#35bf28}+0.66\%$
test_collector_preallocated_write[set_at_loop-cpu-4096-32] 10.5788ms 10.3313ms 96.7935 Ops/s 97.4934 Ops/s $\color{#d91a1a}-0.72\%$
test_collector_preallocated_write[set_at_loop-cuda-32-2] 0.4425ms 0.3794ms 2.6354 KOps/s 2.5508 KOps/s $\color{#35bf28}+3.32\%$
test_collector_preallocated_write[set_at_loop-cuda-32-8] 1.5983ms 1.4797ms 675.8141 Ops/s 684.9987 Ops/s $\color{#d91a1a}-1.34\%$
test_collector_preallocated_write[set_at_loop-cuda-32-32] 6.0601ms 5.8839ms 169.9550 Ops/s 171.0071 Ops/s $\color{#d91a1a}-0.62\%$
test_collector_preallocated_write[set_at_loop-cuda-4096-2] 0.4550ms 0.3902ms 2.5629 KOps/s 2.5181 KOps/s $\color{#35bf28}+1.78\%$
test_collector_preallocated_write[set_at_loop-cuda-4096-8] 1.5742ms 1.5174ms 659.0311 Ops/s 681.0702 Ops/s $\color{#d91a1a}-3.24\%$
test_collector_preallocated_write[set_at_loop-cuda-4096-32] 5.6598ms 5.5496ms 180.1935 Ops/s 174.5404 Ops/s $\color{#35bf28}+3.24\%$
test_collector_preallocated_write[copy_at-cpu-32-2] 0.2250ms 0.1680ms 5.9514 KOps/s 5.8550 KOps/s $\color{#35bf28}+1.65\%$
test_collector_preallocated_write[copy_at-cpu-32-8] 0.6951ms 0.6522ms 1.5333 KOps/s 1.5093 KOps/s $\color{#35bf28}+1.59\%$
test_collector_preallocated_write[copy_at-cpu-32-32] 2.6775ms 2.5924ms 385.7406 Ops/s 379.7520 Ops/s $\color{#35bf28}+1.58\%$
test_collector_preallocated_write[copy_at-cpu-4096-2] 0.4457ms 0.3184ms 3.1407 KOps/s 3.1630 KOps/s $\color{#d91a1a}-0.71\%$
test_collector_preallocated_write[copy_at-cpu-4096-8] 1.6864ms 1.5328ms 652.3901 Ops/s 675.5773 Ops/s $\color{#d91a1a}-3.43\%$
test_collector_preallocated_write[copy_at-cpu-4096-32] 10.6218ms 9.7430ms 102.6382 Ops/s 103.1936 Ops/s $\color{#d91a1a}-0.54\%$
test_collector_preallocated_write[copy_at-cuda-32-2] 0.3838ms 0.3156ms 3.1682 KOps/s 3.1474 KOps/s $\color{#35bf28}+0.66\%$
test_collector_preallocated_write[copy_at-cuda-32-8] 1.2629ms 1.1706ms 854.2518 Ops/s 847.3852 Ops/s $\color{#35bf28}+0.81\%$
test_collector_preallocated_write[copy_at-cuda-32-32] 4.8055ms 4.5908ms 217.8285 Ops/s 216.6952 Ops/s $\color{#35bf28}+0.52\%$
test_collector_preallocated_write[copy_at-cuda-4096-2] 0.4429ms 0.3173ms 3.1518 KOps/s 3.1284 KOps/s $\color{#35bf28}+0.75\%$
test_collector_preallocated_write[copy_at-cuda-4096-8] 1.2630ms 1.1790ms 848.1758 Ops/s 838.3236 Ops/s $\color{#35bf28}+1.18\%$
test_collector_preallocated_write[copy_at-cuda-4096-32] 4.8330ms 4.6092ms 216.9582 Ops/s 215.1139 Ops/s $\color{#35bf28}+0.86\%$
test_collector_preallocated_write[manual_leaf_copy-cpu-32-2] 0.1585ms 0.1188ms 8.4149 KOps/s 8.4772 KOps/s $\color{#d91a1a}-0.74\%$
test_collector_preallocated_write[manual_leaf_copy-cpu-32-8] 0.4960ms 0.4632ms 2.1589 KOps/s 2.1562 KOps/s $\color{#35bf28}+0.12\%$
test_collector_preallocated_write[manual_leaf_copy-cpu-32-32] 1.9061ms 1.8442ms 542.2307 Ops/s 542.5736 Ops/s $\color{#d91a1a}-0.06\%$
test_collector_preallocated_write[manual_leaf_copy-cpu-4096-2] 0.3270ms 0.2559ms 3.9072 KOps/s 3.9103 KOps/s $\color{#d91a1a}-0.08\%$
test_collector_preallocated_write[manual_leaf_copy-cpu-4096-8] 1.4417ms 1.3160ms 759.8737 Ops/s 855.2638 Ops/s $\textbf{\color{#d91a1a}-11.15\%}$
test_collector_preallocated_write[manual_leaf_copy-cpu-4096-32] 9.2603ms 8.6812ms 115.1915 Ops/s 115.6322 Ops/s $\color{#d91a1a}-0.38\%$
test_collector_preallocated_write[manual_leaf_copy-cuda-32-2] 0.3591ms 0.2798ms 3.5745 KOps/s 3.8265 KOps/s $\textbf{\color{#d91a1a}-6.59\%}$
test_collector_preallocated_write[manual_leaf_copy-cuda-32-8] 1.0884ms 1.0362ms 965.0238 Ops/s 1.0300 KOps/s $\textbf{\color{#d91a1a}-6.31\%}$
test_collector_preallocated_write[manual_leaf_copy-cuda-32-32] 4.2410ms 4.0829ms 244.9229 Ops/s 263.8301 Ops/s $\textbf{\color{#d91a1a}-7.17\%}$
test_collector_preallocated_write[manual_leaf_copy-cuda-4096-2] 0.3338ms 0.2820ms 3.5459 KOps/s 3.8232 KOps/s $\textbf{\color{#d91a1a}-7.25\%}$
test_collector_preallocated_write[manual_leaf_copy-cuda-4096-8] 1.1208ms 1.0405ms 961.0814 Ops/s 1.0267 KOps/s $\textbf{\color{#d91a1a}-6.39\%}$
test_collector_preallocated_write[manual_leaf_copy-cuda-4096-32] 4.1784ms 4.0851ms 244.7938 Ops/s 261.7675 Ops/s $\textbf{\color{#d91a1a}-6.48\%}$
test_wide_nested_preallocated_write[stack_out-cpu-128-128] 16.5338ms 16.3800ms 61.0500 Ops/s 60.0350 Ops/s $\color{#35bf28}+1.69\%$
test_wide_nested_preallocated_write[stack_out-cpu-128-256] 33.5130ms 33.1921ms 30.1277 Ops/s 29.7715 Ops/s $\color{#35bf28}+1.20\%$
test_wide_nested_preallocated_write[stack_out-cpu-4096-128] 88.5989ms 87.4252ms 11.4383 Ops/s 11.6246 Ops/s $\color{#d91a1a}-1.60\%$
test_wide_nested_preallocated_write[stack_out-cpu-4096-256] 0.2026s 0.2008s 4.9792 Ops/s 4.9597 Ops/s $\color{#35bf28}+0.39\%$
test_wide_nested_preallocated_write[stack_out-cuda-128-128] 13.9916ms 13.8714ms 72.0906 Ops/s 72.6140 Ops/s $\color{#d91a1a}-0.72\%$
test_wide_nested_preallocated_write[stack_out-cuda-128-256] 27.2685ms 26.9907ms 37.0498 Ops/s 37.0747 Ops/s $\color{#d91a1a}-0.07\%$
test_wide_nested_preallocated_write[stack_out-cuda-4096-128] 14.1969ms 14.0498ms 71.1754 Ops/s 70.8957 Ops/s $\color{#35bf28}+0.39\%$
test_wide_nested_preallocated_write[stack_out-cuda-4096-256] 27.8557ms 27.3851ms 36.5162 Ops/s 36.9147 Ops/s $\color{#d91a1a}-1.08\%$
test_wide_nested_preallocated_write[set_at_loop-cpu-128-128] 50.9248ms 50.2384ms 19.9051 Ops/s 19.7269 Ops/s $\color{#35bf28}+0.90\%$
test_wide_nested_preallocated_write[set_at_loop-cpu-128-256] 0.1013s 0.1011s 9.8913 Ops/s 9.7500 Ops/s $\color{#35bf28}+1.45\%$
test_wide_nested_preallocated_write[set_at_loop-cpu-4096-128] 0.1491s 0.1464s 6.8313 Ops/s 6.8999 Ops/s $\color{#d91a1a}-1.00\%$
test_wide_nested_preallocated_write[set_at_loop-cpu-4096-256] 0.3257s 0.3236s 3.0905 Ops/s 3.0175 Ops/s $\color{#35bf28}+2.42\%$
test_wide_nested_preallocated_write[set_at_loop-cuda-128-128] 72.2190ms 71.6778ms 13.9513 Ops/s 13.7073 Ops/s $\color{#35bf28}+1.78\%$
test_wide_nested_preallocated_write[set_at_loop-cuda-128-256] 0.1438s 0.1432s 6.9855 Ops/s 6.8193 Ops/s $\color{#35bf28}+2.44\%$
test_wide_nested_preallocated_write[set_at_loop-cuda-4096-128] 77.3651ms 73.1452ms 13.6714 Ops/s 13.5028 Ops/s $\color{#35bf28}+1.25\%$
test_wide_nested_preallocated_write[set_at_loop-cuda-4096-256] 0.1572s 0.1503s 6.6541 Ops/s 6.9108 Ops/s $\color{#d91a1a}-3.71\%$
test_wide_nested_preallocated_write[copy_at-cpu-128-128] 30.2228ms 30.0926ms 33.2307 Ops/s 33.1836 Ops/s $\color{#35bf28}+0.14\%$
test_wide_nested_preallocated_write[copy_at-cpu-128-256] 60.9497ms 60.7359ms 16.4647 Ops/s 16.4170 Ops/s $\color{#35bf28}+0.29\%$
test_wide_nested_preallocated_write[copy_at-cpu-4096-128] 0.1209s 0.1197s 8.3551 Ops/s 8.2947 Ops/s $\color{#35bf28}+0.73\%$
test_wide_nested_preallocated_write[copy_at-cpu-4096-256] 0.2684s 0.2665s 3.7519 Ops/s 3.6856 Ops/s $\color{#35bf28}+1.80\%$
test_wide_nested_preallocated_write[copy_at-cuda-128-128] 42.7277ms 42.1985ms 23.6975 Ops/s 23.5564 Ops/s $\color{#35bf28}+0.60\%$
test_wide_nested_preallocated_write[copy_at-cuda-128-256] 85.2293ms 84.5898ms 11.8218 Ops/s 11.7812 Ops/s $\color{#35bf28}+0.34\%$
test_wide_nested_preallocated_write[copy_at-cuda-4096-128] 43.3965ms 42.6076ms 23.4700 Ops/s 23.2761 Ops/s $\color{#35bf28}+0.83\%$
test_wide_nested_preallocated_write[copy_at-cuda-4096-256] 85.6256ms 85.0594ms 11.7565 Ops/s 11.6757 Ops/s $\color{#35bf28}+0.69\%$
test_wide_nested_preallocated_write[manual_leaf_copy-cpu-128-128] 20.1593ms 19.9588ms 50.1032 Ops/s 49.7652 Ops/s $\color{#35bf28}+0.68\%$
test_wide_nested_preallocated_write[manual_leaf_copy-cpu-128-256] 41.1626ms 40.9180ms 24.4391 Ops/s 24.3407 Ops/s $\color{#35bf28}+0.40\%$
test_wide_nested_preallocated_write[manual_leaf_copy-cpu-4096-128] 0.1088s 0.1072s 9.3263 Ops/s 9.1954 Ops/s $\color{#35bf28}+1.42\%$
test_wide_nested_preallocated_write[manual_leaf_copy-cpu-4096-256] 0.2441s 0.2427s 4.1206 Ops/s 4.0644 Ops/s $\color{#35bf28}+1.38\%$
test_wide_nested_preallocated_write[manual_leaf_copy-cuda-128-128] 35.3361ms 34.7434ms 28.7825 Ops/s 29.0598 Ops/s $\color{#d91a1a}-0.95\%$
test_wide_nested_preallocated_write[manual_leaf_copy-cuda-128-256] 70.0367ms 69.5899ms 14.3699 Ops/s 14.5336 Ops/s $\color{#d91a1a}-1.13\%$
test_wide_nested_preallocated_write[manual_leaf_copy-cuda-4096-128] 35.2830ms 34.8052ms 28.7314 Ops/s 29.0478 Ops/s $\color{#d91a1a}-1.09\%$
test_wide_nested_preallocated_write[manual_leaf_copy-cuda-4096-256] 70.0608ms 69.5260ms 14.3831 Ops/s 14.5759 Ops/s $\color{#d91a1a}-1.32\%$
test_plain_set_nested 43.0410μs 15.1016μs 66.2181 KOps/s 65.9168 KOps/s $\color{#35bf28}+0.46\%$
test_plain_set_stack_nested 64.8410μs 14.7705μs 67.7023 KOps/s 67.0390 KOps/s $\color{#35bf28}+0.99\%$
test_plain_set_nested_inplace 50.1510μs 16.2842μs 61.4091 KOps/s 60.1087 KOps/s $\color{#35bf28}+2.16\%$
test_plain_set_stack_nested_inplace 51.0410μs 16.0405μs 62.3421 KOps/s 61.5495 KOps/s $\color{#35bf28}+1.29\%$
test_items 63.3820μs 5.9698μs 167.5108 KOps/s 165.6802 KOps/s $\color{#35bf28}+1.10\%$
test_items_nested 0.4997ms 0.4529ms 2.2080 KOps/s 2.0963 KOps/s $\textbf{\color{#35bf28}+5.33\%}$
test_items_nested_locked 0.5724ms 0.4545ms 2.2004 KOps/s 2.1082 KOps/s $\color{#35bf28}+4.37\%$
test_items_nested_leaf 0.1555ms 97.1488μs 10.2935 KOps/s 10.1129 KOps/s $\color{#35bf28}+1.79\%$
test_items_stack_nested 0.5284ms 0.4569ms 2.1888 KOps/s 2.1110 KOps/s $\color{#35bf28}+3.69\%$
test_items_stack_nested_leaf 0.1402ms 98.1418μs 10.1893 KOps/s 10.0682 KOps/s $\color{#35bf28}+1.20\%$
test_items_stack_nested_locked 0.5594ms 0.4607ms 2.1707 KOps/s 2.1057 KOps/s $\color{#35bf28}+3.09\%$
test_keys 31.4100μs 4.1780μs 239.3462 KOps/s 239.1342 KOps/s $\color{#35bf28}+0.09\%$
test_keys_nested 0.1682ms 0.1280ms 7.8143 KOps/s 7.7128 KOps/s $\color{#35bf28}+1.32\%$
test_keys_nested_locked 0.7648ms 0.1363ms 7.3366 KOps/s 7.2237 KOps/s $\color{#35bf28}+1.56\%$
test_keys_nested_leaf 0.1639ms 0.1190ms 8.4033 KOps/s 8.3525 KOps/s $\color{#35bf28}+0.61\%$
test_keys_stack_nested 0.1704ms 0.1277ms 7.8294 KOps/s 7.8551 KOps/s $\color{#d91a1a}-0.33\%$
test_keys_stack_nested_leaf 0.1604ms 0.1186ms 8.4288 KOps/s 8.4406 KOps/s $\color{#d91a1a}-0.14\%$
test_keys_stack_nested_locked 0.2076ms 0.1364ms 7.3323 KOps/s 7.2855 KOps/s $\color{#35bf28}+0.64\%$
test_values 7.3660μs 1.0186μs 981.7234 KOps/s 980.0490 KOps/s $\color{#35bf28}+0.17\%$
test_values_nested 83.9820μs 52.4407μs 19.0691 KOps/s 19.0831 KOps/s $\color{#d91a1a}-0.07\%$
test_values_nested_locked 88.0020μs 55.7371μs 17.9414 KOps/s 17.9485 KOps/s $\color{#d91a1a}-0.04\%$
test_values_nested_leaf 0.1258ms 60.2342μs 16.6019 KOps/s 16.4547 KOps/s $\color{#35bf28}+0.89\%$
test_values_stack_nested 85.6620μs 52.3711μs 19.0945 KOps/s 19.0566 KOps/s $\color{#35bf28}+0.20\%$
test_values_stack_nested_leaf 92.7810μs 59.8674μs 16.7036 KOps/s 16.6704 KOps/s $\color{#35bf28}+0.20\%$
test_values_stack_nested_locked 0.1076ms 55.7752μs 17.9291 KOps/s 17.8545 KOps/s $\color{#35bf28}+0.42\%$
test_membership 5.0002μs 0.8444μs 1.1843 MOps/s 1.2023 MOps/s $\color{#d91a1a}-1.50\%$
test_membership_nested 23.5300μs 2.9790μs 335.6830 KOps/s 331.6247 KOps/s $\color{#35bf28}+1.22\%$
test_membership_nested_leaf 22.3100μs 2.9685μs 336.8656 KOps/s 347.8500 KOps/s $\color{#d91a1a}-3.16\%$
test_membership_stacked_nested 38.1010μs 2.8644μs 349.1097 KOps/s 348.5614 KOps/s $\color{#35bf28}+0.16\%$
test_membership_stacked_nested_leaf 39.9510μs 2.8416μs 351.9184 KOps/s 348.8243 KOps/s $\color{#35bf28}+0.89\%$
test_membership_nested_last 36.5010μs 4.5531μs 219.6307 KOps/s 221.7189 KOps/s $\color{#d91a1a}-0.94\%$
test_membership_nested_leaf_last 32.2600μs 4.5449μs 220.0271 KOps/s 223.9809 KOps/s $\color{#d91a1a}-1.77\%$
test_membership_stacked_nested_last 59.2110μs 4.2738μs 233.9852 KOps/s 231.8595 KOps/s $\color{#35bf28}+0.92\%$
test_membership_stacked_nested_leaf_last 40.4010μs 4.2852μs 233.3599 KOps/s 233.2522 KOps/s $\color{#35bf28}+0.05\%$
test_nested_getleaf 93.5220μs 21.1991μs 47.1718 KOps/s 46.7878 KOps/s $\color{#35bf28}+0.82\%$
test_nested_get 54.1610μs 20.1660μs 49.5885 KOps/s 49.0901 KOps/s $\color{#35bf28}+1.02\%$
test_stacked_getleaf 69.3220μs 20.7448μs 48.2048 KOps/s 47.8552 KOps/s $\color{#35bf28}+0.73\%$
test_stacked_get 62.6910μs 20.1121μs 49.7214 KOps/s 49.8296 KOps/s $\color{#d91a1a}-0.22\%$
test_nested_getitemleaf 0.1049ms 21.6075μs 46.2802 KOps/s 44.8844 KOps/s $\color{#35bf28}+3.11\%$
test_nested_getitem 57.6120μs 21.0824μs 47.4330 KOps/s 47.2154 KOps/s $\color{#35bf28}+0.46\%$
test_stacked_getitemleaf 71.8910μs 21.8745μs 45.7154 KOps/s 46.0036 KOps/s $\color{#d91a1a}-0.63\%$
test_stacked_getitem 65.4120μs 20.7988μs 48.0797 KOps/s 48.6224 KOps/s $\color{#d91a1a}-1.12\%$
test_lock_nested 0.7258ms 0.5016ms 1.9936 KOps/s 1.9950 KOps/s $\color{#d91a1a}-0.07\%$
test_lock_stack_nested 0.6356ms 0.5065ms 1.9741 KOps/s 1.9737 KOps/s $\color{#35bf28}+0.02\%$
test_unlock_nested 0.5246ms 0.3991ms 2.5058 KOps/s 2.5188 KOps/s $\color{#d91a1a}-0.52\%$
test_unlock_stack_nested 1.0137ms 0.4033ms 2.4794 KOps/s 2.5059 KOps/s $\color{#d91a1a}-1.06\%$
test_flatten_speed 0.3476ms 0.1219ms 8.2047 KOps/s 8.0647 KOps/s $\color{#35bf28}+1.74\%$
test_unflatten_speed 0.5978ms 0.5612ms 1.7818 KOps/s 1.7428 KOps/s $\color{#35bf28}+2.23\%$
test_common_ops 0.8982ms 0.6851ms 1.4596 KOps/s 1.4357 KOps/s $\color{#35bf28}+1.67\%$
test_creation 74.3620μs 3.1871μs 313.7632 KOps/s 314.9590 KOps/s $\color{#d91a1a}-0.38\%$
test_creation_empty 36.5000μs 7.0262μs 142.3254 KOps/s 141.9965 KOps/s $\color{#35bf28}+0.23\%$
test_creation_nested_1 55.4810μs 11.4638μs 87.2315 KOps/s 86.5780 KOps/s $\color{#35bf28}+0.75\%$
test_creation_nested_2 84.7720μs 13.4300μs 74.4601 KOps/s 74.6272 KOps/s $\color{#d91a1a}-0.22\%$
test_creation_many_keys[10] 50.0420μs 21.1035μs 47.3855 KOps/s 47.3541 KOps/s $\color{#35bf28}+0.07\%$
test_creation_many_keys[50] 0.1227ms 90.5821μs 11.0397 KOps/s 11.1085 KOps/s $\color{#d91a1a}-0.62\%$
test_creation_many_keys[100] 0.2171ms 0.1778ms 5.6228 KOps/s 5.6584 KOps/s $\color{#d91a1a}-0.63\%$
test_creation_nested_many_keys[10] 86.8720μs 45.5330μs 21.9621 KOps/s 21.9983 KOps/s $\color{#d91a1a}-0.16\%$
test_creation_nested_many_keys[50] 0.2189ms 0.1862ms 5.3702 KOps/s 5.4686 KOps/s $\color{#d91a1a}-1.80\%$
test_clone 49.8910μs 12.7520μs 78.4188 KOps/s 78.1923 KOps/s $\color{#35bf28}+0.29\%$
test_getitem[int] 1.4647ms 15.4127μs 64.8816 KOps/s 65.3768 KOps/s $\color{#d91a1a}-0.76\%$
test_getitem[slice_int] 0.1401ms 24.7717μs 40.3687 KOps/s 41.2187 KOps/s $\color{#d91a1a}-2.06\%$
test_getitem[range] 0.1757ms 61.3646μs 16.2960 KOps/s 16.3476 KOps/s $\color{#d91a1a}-0.32\%$
test_getitem[tuple] 0.1469ms 24.2808μs 41.1848 KOps/s 42.1432 KOps/s $\color{#d91a1a}-2.27\%$
test_getitem[list] 0.1884ms 58.5024μs 17.0933 KOps/s 18.0052 KOps/s $\textbf{\color{#d91a1a}-5.06\%}$
test_setitem_dim[int] 48.7510μs 24.8262μs 40.2800 KOps/s 40.7640 KOps/s $\color{#d91a1a}-1.19\%$
test_setitem_dim[slice_int] 67.3110μs 41.0192μs 24.3788 KOps/s 23.9513 KOps/s $\color{#35bf28}+1.78\%$
test_setitem_dim[range] 0.1294ms 92.0106μs 10.8683 KOps/s 11.0775 KOps/s $\color{#d91a1a}-1.89\%$
test_setitem_dim[tuple] 69.8320μs 37.8869μs 26.3944 KOps/s 25.8620 KOps/s $\color{#35bf28}+2.06\%$
test_setitem 49.8720μs 16.7538μs 59.6879 KOps/s 60.1694 KOps/s $\color{#d91a1a}-0.80\%$
test_set 59.8910μs 15.9561μs 62.6721 KOps/s 62.3988 KOps/s $\color{#35bf28}+0.44\%$
test_set_shared 0.5443ms 0.2069ms 4.8325 KOps/s 4.9077 KOps/s $\color{#d91a1a}-1.53\%$
test_update 0.3666ms 20.7833μs 48.1156 KOps/s 48.4848 KOps/s $\color{#d91a1a}-0.76\%$
test_update_nested 68.0320μs 31.9504μs 31.2985 KOps/s 30.3805 KOps/s $\color{#35bf28}+3.02\%$
test_update__nested 0.4840ms 31.9971μs 31.2528 KOps/s 30.8608 KOps/s $\color{#35bf28}+1.27\%$
test_set_nested 48.4010μs 17.6476μs 56.6649 KOps/s 55.3096 KOps/s $\color{#35bf28}+2.45\%$
test_set_nested_new 56.2120μs 22.5639μs 44.3187 KOps/s 44.4843 KOps/s $\color{#d91a1a}-0.37\%$
test_select 73.2020μs 39.1829μs 25.5214 KOps/s 25.2522 KOps/s $\color{#35bf28}+1.07\%$
test_select_nested 0.1048ms 74.2838μs 13.4619 KOps/s 13.4168 KOps/s $\color{#35bf28}+0.34\%$
test_exclude_nested 0.1212ms 91.3618μs 10.9455 KOps/s 10.9043 KOps/s $\color{#35bf28}+0.38\%$
test_empty[True] 0.4243ms 0.3947ms 2.5333 KOps/s 2.4933 KOps/s $\color{#35bf28}+1.60\%$
test_empty[False] 8.6252μs 1.4059μs 711.2860 KOps/s 716.7210 KOps/s $\color{#d91a1a}-0.76\%$
test_to 0.1088ms 68.5670μs 14.5843 KOps/s 13.8752 KOps/s $\textbf{\color{#35bf28}+5.11\%}$
test_to_nonblocking 0.1108ms 66.9313μs 14.9407 KOps/s 14.5287 KOps/s $\color{#35bf28}+2.84\%$
test_unbind_speed 0.3740ms 0.3425ms 2.9195 KOps/s 2.9525 KOps/s $\color{#d91a1a}-1.12\%$
test_unbind_speed_stack0 0.3883ms 0.3380ms 2.9588 KOps/s 3.0333 KOps/s $\color{#d91a1a}-2.46\%$
test_unbind_speed_stack1 0.1167s 1.1089ms 901.8260 Ops/s 1.1608 KOps/s $\textbf{\color{#d91a1a}-22.31\%}$
test_split 1.3788ms 1.2962ms 771.4911 Ops/s 641.1175 Ops/s $\textbf{\color{#35bf28}+20.34\%}$
test_chunk 0.1184s 1.3940ms 717.3482 Ops/s 818.6070 Ops/s $\textbf{\color{#d91a1a}-12.37\%}$
test_to_cpu_blocking 19.3936ms 18.8693ms 52.9960 Ops/s 52.6901 Ops/s $\color{#35bf28}+0.58\%$
test_to_cpu_global_sync 11.4851ms 11.2749ms 88.6929 Ops/s 88.5067 Ops/s $\color{#35bf28}+0.21\%$
test_to_cpu_event_sync 12.5185ms 12.0991ms 82.6510 Ops/s 72.9796 Ops/s $\textbf{\color{#35bf28}+13.25\%}$
test_to_cpu_default 12.4257ms 12.1285ms 82.4507 Ops/s 82.1076 Ops/s $\color{#35bf28}+0.42\%$
test_consolidate[False-None] 4.4023ms 4.1289ms 242.1947 Ops/s 239.3403 Ops/s $\color{#35bf28}+1.19\%$
test_consolidate[default-None] 3.2732ms 2.2439ms 445.6604 Ops/s 438.2008 Ops/s $\color{#35bf28}+1.70\%$
test_consolidate[reduce-overhead-None] 3.2759ms 2.1530ms 464.4781 Ops/s 450.3732 Ops/s $\color{#35bf28}+3.13\%$
test_consolidate_njt[False-None] 5.1753ms 5.0383ms 198.4810 Ops/s 193.9388 Ops/s $\color{#35bf28}+2.34\%$
test_to[False-False-None] 2.2176ms 2.1491ms 465.3042 Ops/s 465.3275 Ops/s $-0.01\%$
test_to[True-False-None] 2.2254ms 1.9142ms 522.4152 Ops/s 520.1273 Ops/s $\color{#35bf28}+0.44\%$
test_to[within-False-None] 0.2084s 7.4406ms 134.3983 Ops/s 160.6879 Ops/s $\textbf{\color{#d91a1a}-16.36\%}$
test_to[True-default-None] 10.4655ms 10.2428ms 97.6295 Ops/s 97.8576 Ops/s $\color{#d91a1a}-0.23\%$
test_to_njt[False-False-None] 7.3210ms 6.9236ms 144.4339 Ops/s 145.3255 Ops/s $\color{#d91a1a}-0.61\%$
test_to_njt[True-False-None] 4.0249ms 3.8369ms 260.6302 Ops/s 258.8587 Ops/s $\color{#35bf28}+0.68\%$
test_to_njt[within-False-None] 0.2009s 10.7373ms 93.1329 Ops/s 112.6404 Ops/s $\textbf{\color{#d91a1a}-17.32\%}$
test_creation[device0] 0.3547ms 0.1124ms 8.9003 KOps/s 8.8337 KOps/s $\color{#35bf28}+0.75\%$
test_creation_from_tensor 0.4513ms 0.1110ms 9.0070 KOps/s 8.8595 KOps/s $\color{#35bf28}+1.66\%$
test_add_one[memmap_tensor0] 0.2984ms 5.9417μs 168.3010 KOps/s 169.4150 KOps/s $\color{#d91a1a}-0.66\%$
test_contiguous[memmap_tensor0] 23.7810μs 0.6804μs 1.4698 MOps/s 1.8700 MOps/s $\textbf{\color{#d91a1a}-21.40\%}$
test_stack[memmap_tensor0] 32.9210μs 4.0795μs 245.1255 KOps/s 243.2426 KOps/s $\color{#35bf28}+0.77\%$
test_memmaptd_index 1.4224ms 0.2739ms 3.6506 KOps/s 3.6670 KOps/s $\color{#d91a1a}-0.45\%$
test_memmaptd_index_astensor 0.5475ms 0.3755ms 2.6631 KOps/s 2.6545 KOps/s $\color{#35bf28}+0.32\%$
test_memmaptd_index_op 0.7709ms 0.6094ms 1.6409 KOps/s 1.6201 KOps/s $\color{#35bf28}+1.28\%$
test_serialize_model 0.1368s 0.1360s 7.3536 Ops/s 7.3496 Ops/s $\color{#35bf28}+0.05\%$
test_serialize_model_pickle 1.3506s 1.2111s 0.8257 Ops/s 0.8267 Ops/s $\color{#d91a1a}-0.12\%$
test_serialize_weights 0.1366s 0.1353s 7.3908 Ops/s 7.3835 Ops/s $\color{#35bf28}+0.10\%$
test_serialize_weights_returnearly 0.2685s 45.5241ms 21.9664 Ops/s 20.5922 Ops/s $\textbf{\color{#35bf28}+6.67\%}$
test_serialize_weights_pickle 1.3680s 1.2141s 0.8236 Ops/s 0.8224 Ops/s $\color{#35bf28}+0.15\%$
test_consolidate[few_large-0] 74.2806ms 8.0483ms 124.2504 Ops/s 126.4602 Ops/s $\color{#d91a1a}-1.75\%$
test_consolidate[few_large-8] 53.8481ms 7.4767ms 133.7483 Ops/s 148.7108 Ops/s $\textbf{\color{#d91a1a}-10.06\%}$
test_consolidate[many_small-0] 0.1241s 0.1218s 8.2078 Ops/s 8.1988 Ops/s $\color{#35bf28}+0.11\%$
test_consolidate[many_small-8] 42.5535ms 40.2846ms 24.8234 Ops/s 20.0811 Ops/s $\textbf{\color{#35bf28}+23.62\%}$
test_consolidate_to_file[few_large-0] 0.1299s 22.8046ms 43.8508 Ops/s 42.2648 Ops/s $\color{#35bf28}+3.75\%$
test_consolidate_to_file[few_large-8] 0.8399s 0.1403s 7.1299 Ops/s 33.1658 Ops/s $\textbf{\color{#d91a1a}-78.50\%}$
test_consolidate_to_file[many_small-0] 1.7388s 1.1381s 0.8787 Ops/s 2.3372 Ops/s $\textbf{\color{#d91a1a}-62.41\%}$
test_consolidate_to_file[many_small-8] 0.8331s 0.7418s 1.3481 Ops/s 1.7080 Ops/s $\textbf{\color{#d91a1a}-21.08\%}$
test_reshape_pytree 6.5449ms 32.5853μs 30.6887 KOps/s 31.3841 KOps/s $\color{#d91a1a}-2.22\%$
test_reshape_td 94.2220μs 45.7204μs 21.8721 KOps/s 22.2067 KOps/s $\color{#d91a1a}-1.51\%$
test_view_pytree 0.2248ms 31.8976μs 31.3503 KOps/s 31.7117 KOps/s $\color{#d91a1a}-1.14\%$
test_view_td 99.3020μs 53.3990μs 18.7269 KOps/s 18.4254 KOps/s $\color{#35bf28}+1.64\%$
test_unbind_pytree 0.2338ms 36.0927μs 27.7064 KOps/s 27.9291 KOps/s $\color{#d91a1a}-0.80\%$
test_unbind_td 0.1264ms 51.6219μs 19.3716 KOps/s 20.3699 KOps/s $\color{#d91a1a}-4.90\%$
test_split_pytree 0.2485ms 40.9813μs 24.4013 KOps/s 24.2725 KOps/s $\color{#35bf28}+0.53\%$
test_split_td 0.1575ms 71.1136μs 14.0620 KOps/s 13.7649 KOps/s $\color{#35bf28}+2.16\%$
test_add_pytree 0.2351ms 38.8995μs 25.7073 KOps/s 25.4383 KOps/s $\color{#35bf28}+1.06\%$
test_add_td 0.1004ms 51.6269μs 19.3698 KOps/s 18.9662 KOps/s $\color{#35bf28}+2.13\%$
test_compile_add_one_nested[tensordict-compile] 0.6629ms 0.2791ms 3.5825 KOps/s 3.6025 KOps/s $\color{#d91a1a}-0.56\%$
test_compile_add_one_nested[tensordict-eager] 0.3224ms 0.1985ms 5.0373 KOps/s 4.8959 KOps/s $\color{#35bf28}+2.89\%$
test_compile_add_one_nested[pytree-compile] 0.3480ms 0.1580ms 6.3302 KOps/s 6.0877 KOps/s $\color{#35bf28}+3.98\%$
test_compile_add_one_nested[pytree-eager] 0.5043ms 0.1761ms 5.6791 KOps/s 5.6104 KOps/s $\color{#35bf28}+1.22\%$
test_compile_copy_nested[tensordict-compile] 0.2359ms 23.0184μs 43.4436 KOps/s 46.1002 KOps/s $\textbf{\color{#d91a1a}-5.76\%}$
test_compile_copy_nested[tensordict-eager] 0.1180ms 53.9467μs 18.5368 KOps/s 18.7614 KOps/s $\color{#d91a1a}-1.20\%$
test_compile_copy_nested[pytree-compile] 0.1825ms 18.9875μs 52.6663 KOps/s 50.4994 KOps/s $\color{#35bf28}+4.29\%$
test_compile_copy_nested[pytree-eager] 0.3716ms 66.7420μs 14.9831 KOps/s 15.2419 KOps/s $\color{#d91a1a}-1.70\%$
test_compile_add_one_flat[tensordict-compile] 0.3940ms 0.3093ms 3.2329 KOps/s 3.0342 KOps/s $\textbf{\color{#35bf28}+6.55\%}$
test_compile_add_one_flat[tensordict-eager] 0.3434ms 0.2879ms 3.4731 KOps/s 3.4082 KOps/s $\color{#35bf28}+1.90\%$
test_compile_add_one_flat[tensorclass-compile] 0.3461ms 0.2099ms 4.7634 KOps/s 4.4533 KOps/s $\textbf{\color{#35bf28}+6.96\%}$
test_compile_add_one_flat[tensorclass-eager] 0.1239ms 74.8369μs 13.3624 KOps/s 12.6953 KOps/s $\textbf{\color{#35bf28}+5.25\%}$
test_compile_add_one_flat[pytree-compile] 0.5123ms 0.2420ms 4.1326 KOps/s 3.9733 KOps/s $\color{#35bf28}+4.01\%$
test_compile_add_one_flat[pytree-eager] 0.8104ms 0.5238ms 1.9093 KOps/s 1.9120 KOps/s $\color{#d91a1a}-0.14\%$
test_compile_add_self_flat[tensordict-eager] 0.4908ms 0.3387ms 2.9529 KOps/s 2.9499 KOps/s $\color{#35bf28}+0.10\%$
test_compile_add_self_flat[tensordict-compile] 0.7048ms 0.3112ms 3.2137 KOps/s 3.0085 KOps/s $\textbf{\color{#35bf28}+6.82\%}$
test_compile_add_self_flat[tensorclass-eager] 0.2211ms 92.3211μs 10.8318 KOps/s 10.9227 KOps/s $\color{#d91a1a}-0.83\%$
test_compile_add_self_flat[tensorclass-compile] 0.3893ms 0.2138ms 4.6763 KOps/s 4.4210 KOps/s $\textbf{\color{#35bf28}+5.77\%}$
test_compile_add_self_flat[pytree-eager] 0.6886ms 0.4304ms 2.3235 KOps/s 2.3244 KOps/s $\color{#d91a1a}-0.04\%$
test_compile_add_self_flat[pytree-compile] 0.5573ms 0.2434ms 4.1080 KOps/s 4.0187 KOps/s $\color{#35bf28}+2.22\%$
test_compile_copy_flat[tensordict-compile] 98.7420μs 27.1565μs 36.8236 KOps/s 38.7692 KOps/s $\textbf{\color{#d91a1a}-5.02\%}$
test_compile_copy_flat[tensordict-eager] 88.5820μs 40.7996μs 24.5101 KOps/s 24.8922 KOps/s $\color{#d91a1a}-1.54\%$
test_compile_copy_flat[pytree-compile] 0.2554ms 25.0472μs 39.9247 KOps/s 38.7649 KOps/s $\color{#35bf28}+2.99\%$
test_compile_copy_flat[pytree-eager] 0.3395ms 51.6483μs 19.3617 KOps/s 19.4568 KOps/s $\color{#d91a1a}-0.49\%$
test_compile_assign_and_add[tensordict-compile] 2.0698ms 0.2523ms 3.9629 KOps/s 3.7611 KOps/s $\textbf{\color{#35bf28}+5.36\%}$
test_compile_assign_and_add[tensordict-eager] 3.4506ms 3.2697ms 305.8398 Ops/s 296.7689 Ops/s $\color{#35bf28}+3.06\%$
test_compile_assign_and_add[pytree-compile] 1.9773ms 0.1966ms 5.0862 KOps/s 5.0120 KOps/s $\color{#35bf28}+1.48\%$
test_compile_assign_and_add[pytree-eager] 2.9082ms 2.7177ms 367.9576 Ops/s 363.2667 Ops/s $\color{#35bf28}+1.29\%$
test_compile_indexing[tensor-tensordict-compile] 0.2970ms 0.1523ms 6.5655 KOps/s 6.3917 KOps/s $\color{#35bf28}+2.72\%$
test_compile_indexing[tensor-tensordict-eager] 0.3748ms 72.2270μs 13.8452 KOps/s 13.6226 KOps/s $\color{#35bf28}+1.63\%$
test_compile_indexing[tensor-tensorclass-compile] 0.2562ms 0.1325ms 7.5467 KOps/s 7.4326 KOps/s $\color{#35bf28}+1.54\%$
test_compile_indexing[tensor-tensorclass-eager] 0.3071ms 43.6324μs 22.9188 KOps/s 22.1714 KOps/s $\color{#35bf28}+3.37\%$
test_compile_indexing[tensor-pytree-compile] 0.1965ms 0.1342ms 7.4490 KOps/s 7.1736 KOps/s $\color{#35bf28}+3.84\%$
test_compile_indexing[tensor-pytree-eager] 0.2613ms 44.7909μs 22.3260 KOps/s 21.7028 KOps/s $\color{#35bf28}+2.87\%$
test_compile_indexing[slice-tensordict-compile] 0.1726ms 90.1270μs 11.0955 KOps/s 11.3146 KOps/s $\color{#d91a1a}-1.94\%$
test_compile_indexing[slice-tensordict-eager] 0.2254ms 27.4399μs 36.4433 KOps/s 37.2070 KOps/s $\color{#d91a1a}-2.05\%$
test_compile_indexing[slice-tensorclass-compile] 0.1871ms 73.2498μs 13.6519 KOps/s 13.9990 KOps/s $\color{#d91a1a}-2.48\%$
test_compile_indexing[slice-tensorclass-eager] 0.2504ms 21.9833μs 45.4892 KOps/s 46.2151 KOps/s $\color{#d91a1a}-1.57\%$
test_compile_indexing[slice-pytree-compile] 0.1356ms 72.5484μs 13.7839 KOps/s 13.7597 KOps/s $\color{#35bf28}+0.18\%$
test_compile_indexing[slice-pytree-eager] 0.2573ms 21.8634μs 45.7385 KOps/s 45.4741 KOps/s $\color{#35bf28}+0.58\%$
test_compile_indexing[int-tensordict-compile] 0.1498ms 89.0579μs 11.2286 KOps/s 11.1136 KOps/s $\color{#35bf28}+1.03\%$
test_compile_indexing[int-tensordict-eager] 0.2065ms 27.6841μs 36.1219 KOps/s 36.4304 KOps/s $\color{#d91a1a}-0.85\%$
test_compile_indexing[int-tensorclass-compile] 0.1244ms 71.7960μs 13.9284 KOps/s 13.8595 KOps/s $\color{#35bf28}+0.50\%$
test_compile_indexing[int-tensorclass-eager] 0.2571ms 21.4406μs 46.6405 KOps/s 45.4043 KOps/s $\color{#35bf28}+2.72\%$
test_compile_indexing[int-pytree-compile] 0.1458ms 71.7616μs 13.9350 KOps/s 13.8178 KOps/s $\color{#35bf28}+0.85\%$
test_compile_indexing[int-pytree-eager] 0.2715ms 21.7441μs 45.9895 KOps/s 45.3227 KOps/s $\color{#35bf28}+1.47\%$
test_compile_replace[single-eager] 0.1014ms 46.3528μs 21.5737 KOps/s 20.3309 KOps/s $\textbf{\color{#35bf28}+6.11\%}$
test_compile_replace[single-compile] 0.2785ms 0.1441ms 6.9404 KOps/s 6.8625 KOps/s $\color{#35bf28}+1.13\%$
test_compile_replace[multi-eager] 0.6428ms 0.5591ms 1.7886 KOps/s 1.7658 KOps/s $\color{#35bf28}+1.29\%$
test_compile_replace[multi-compile] 0.3049ms 0.1568ms 6.3796 KOps/s 6.1648 KOps/s $\color{#35bf28}+3.48\%$
test_compile_tc_getattr_20[eager] 0.2142ms 0.1678ms 5.9583 KOps/s 5.4402 KOps/s $\textbf{\color{#35bf28}+9.52\%}$
test_compile_tc_getattr_20[compile] 0.4229ms 0.1766ms 5.6637 KOps/s 5.4704 KOps/s $\color{#35bf28}+3.53\%$
test_compile_clone_shallow[20-eager] 69.3910μs 18.8838μs 52.9554 KOps/s 53.4483 KOps/s $\color{#d91a1a}-0.92\%$
test_compile_clone_shallow[20-compile] 95.1720μs 24.5907μs 40.6658 KOps/s 43.4236 KOps/s $\textbf{\color{#d91a1a}-6.35\%}$
test_compile_clone_shallow[40-eager] 99.4920μs 32.4544μs 30.8125 KOps/s 29.9189 KOps/s $\color{#35bf28}+2.99\%$
test_compile_clone_shallow[40-compile] 74.5220μs 25.1979μs 39.6858 KOps/s 38.2934 KOps/s $\color{#35bf28}+3.64\%$
test_compile_clone_shallow[80-eager] 0.1102ms 61.3380μs 16.3031 KOps/s 16.1343 KOps/s $\color{#35bf28}+1.05\%$
test_compile_clone_shallow[80-compile] 85.8020μs 28.1427μs 35.5332 KOps/s 34.3283 KOps/s $\color{#35bf28}+3.51\%$
test_compile_update_inplace[eager] 0.1167ms 61.4026μs 16.2860 KOps/s 15.8546 KOps/s $\color{#35bf28}+2.72\%$
test_compile_update_inplace[compile] 0.5392ms 0.1202ms 8.3227 KOps/s 7.9663 KOps/s $\color{#35bf28}+4.47\%$
test_mod_add[eager] 0.3071ms 0.2248ms 4.4486 KOps/s 4.4494 KOps/s $\color{#d91a1a}-0.02\%$
test_mod_add[compile] 0.2845ms 0.1189ms 8.4101 KOps/s 8.4323 KOps/s $\color{#d91a1a}-0.26\%$
test_mod_add[compile-overhead] 0.3475ms 0.1963ms 5.0952 KOps/s 5.1153 KOps/s $\color{#d91a1a}-0.39\%$
test_mod_wrap[eager] 0.5515ms 0.4678ms 2.1377 KOps/s 2.1345 KOps/s $\color{#35bf28}+0.15\%$
test_mod_wrap[compile] 0.4739ms 0.3625ms 2.7589 KOps/s 2.7775 KOps/s $\color{#d91a1a}-0.67\%$
test_mod_wrap[compile-overhead] 7.1151ms 3.9952ms 250.2998 Ops/s 251.2626 Ops/s $\color{#d91a1a}-0.38\%$
test_mod_wrap_and_backward[eager] 3.4677ms 1.6848ms 593.5448 Ops/s 567.7349 Ops/s $\color{#35bf28}+4.55\%$
test_mod_wrap_and_backward[compile] 1.9125ms 1.3785ms 725.4335 Ops/s 688.6918 Ops/s $\textbf{\color{#35bf28}+5.33\%}$
test_mod_wrap_and_backward[compile-overhead] 1.8094ms 1.0032ms 996.8309 Ops/s 936.8452 Ops/s $\textbf{\color{#35bf28}+6.40\%}$
test_seq_add[eager] 1.2603ms 0.5227ms 1.9131 KOps/s 1.9535 KOps/s $\color{#d91a1a}-2.07\%$
test_seq_add[compile] 0.2533ms 0.1392ms 7.1829 KOps/s 7.4641 KOps/s $\color{#d91a1a}-3.77\%$
test_seq_add[compile-overhead] 0.3627ms 0.2141ms 4.6714 KOps/s 4.8087 KOps/s $\color{#d91a1a}-2.86\%$
test_seq_wrap[eager] 5.8091ms 1.4559ms 686.8443 Ops/s 710.4919 Ops/s $\color{#d91a1a}-3.33\%$
test_seq_wrap[compile] 0.4650ms 0.3740ms 2.6735 KOps/s 2.6941 KOps/s $\color{#d91a1a}-0.77\%$
test_seq_wrap[compile-overhead] 0.4592ms 0.3250ms 3.0767 KOps/s 3.0965 KOps/s $\color{#d91a1a}-0.64\%$
test_func_call_runtime[False-eager] 0.9870ms 0.8461ms 1.1818 KOps/s 1.2503 KOps/s $\textbf{\color{#d91a1a}-5.47\%}$
test_func_call_runtime[False-compile] 0.8484ms 0.7632ms 1.3103 KOps/s 1.3106 KOps/s $\color{#d91a1a}-0.02\%$
test_func_call_runtime[False-compile-overhead] 0.6921ms 0.5168ms 1.9348 KOps/s 1.9552 KOps/s $\color{#d91a1a}-1.04\%$
test_func_call_runtime[True-eager] 1.1528ms 1.0455ms 956.4347 Ops/s 941.6223 Ops/s $\color{#35bf28}+1.57\%$
test_func_call_runtime[True-compile] 0.8300ms 0.7767ms 1.2875 KOps/s 1.2803 KOps/s $\color{#35bf28}+0.56\%$
test_func_call_runtime[True-compile-overhead] 0.7111ms 0.5306ms 1.8847 KOps/s 1.8944 KOps/s $\color{#d91a1a}-0.51\%$
test_func_call_cm_runtime[False-eager] 0.8475ms 0.7883ms 1.2685 KOps/s 1.2624 KOps/s $\color{#35bf28}+0.49\%$
test_func_call_cm_runtime[False-compile] 1.0179ms 0.7549ms 1.3247 KOps/s 1.3260 KOps/s $\color{#d91a1a}-0.10\%$
test_func_call_cm_runtime[False-compile-overhead] 0.5946ms 0.4955ms 2.0180 KOps/s 2.0121 KOps/s $\color{#35bf28}+0.30\%$
test_func_call_cm_runtime[True-eager] 1.3870ms 1.2000ms 833.3306 Ops/s 828.4096 Ops/s $\color{#35bf28}+0.59\%$
test_func_call_cm_runtime[True-compile] 0.9284ms 0.8615ms 1.1608 KOps/s 1.1621 KOps/s $\color{#d91a1a}-0.11\%$
test_func_call_cm_runtime[True-compile-overhead] 1.1491ms 0.8763ms 1.1411 KOps/s 1.1330 KOps/s $\color{#35bf28}+0.71\%$
test_vmap_func_call_cm_runtime[eager] 3.1522ms 2.6314ms 380.0303 Ops/s 374.3572 Ops/s $\color{#35bf28}+1.52\%$
test_vmap_func_call_cm_runtime[compile] 4.1092ms 3.6181ms 276.3890 Ops/s 272.1742 Ops/s $\color{#35bf28}+1.55\%$
test_vmap_func_call_cm_runtime[compile-overhead] 4.1513ms 3.6267ms 275.7332 Ops/s 271.9422 Ops/s $\color{#35bf28}+1.39\%$
test_distributed 1.4866ms 0.1502ms 6.6579 KOps/s 6.9511 KOps/s $\color{#d91a1a}-4.22\%$
test_tdmodule 0.4028ms 0.1955ms 5.1161 KOps/s 5.1146 KOps/s $\color{#35bf28}+0.03\%$
test_tdmodule_dispatch 0.4120ms 0.3768ms 2.6541 KOps/s 2.6799 KOps/s $\color{#d91a1a}-0.96\%$
test_tdseq 53.2820μs 27.2377μs 36.7138 KOps/s 37.2710 KOps/s $\color{#d91a1a}-1.49\%$
test_tdseq_dispatch 0.2526ms 0.2194ms 4.5573 KOps/s 4.6047 KOps/s $\color{#d91a1a}-1.03\%$
test_instantiation_functorch 2.1883ms 2.0369ms 490.9359 Ops/s 492.5558 Ops/s $\color{#d91a1a}-0.33\%$
test_exec_functorch 0.2337ms 0.1733ms 5.7710 KOps/s 5.8655 KOps/s $\color{#d91a1a}-1.61\%$
test_exec_functional_call 0.2391ms 0.1520ms 6.5776 KOps/s 6.5308 KOps/s $\color{#35bf28}+0.72\%$
test_exec_td_decorator 0.4976ms 0.2403ms 4.1608 KOps/s 4.1648 KOps/s $\color{#d91a1a}-0.09\%$
test_vmap_mlp_speed_decorator[True-True] 1.3040ms 1.1091ms 901.6079 Ops/s 892.1083 Ops/s $\color{#35bf28}+1.06\%$
test_vmap_mlp_speed_decorator[True-False] 1.3208ms 1.1050ms 904.9551 Ops/s 893.9404 Ops/s $\color{#35bf28}+1.23\%$
test_vmap_mlp_speed_decorator[False-True] 1.0081ms 0.8111ms 1.2329 KOps/s 1.2205 KOps/s $\color{#35bf28}+1.01\%$
test_vmap_mlp_speed_decorator[False-False] 1.0524ms 0.8159ms 1.2257 KOps/s 1.2225 KOps/s $\color{#35bf28}+0.26\%$
test_vmap_transformer_speed_decorator[True-True] 26.9629ms 24.7211ms 40.4513 Ops/s 40.3344 Ops/s $\color{#35bf28}+0.29\%$
test_vmap_transformer_speed_decorator[True-False] 24.7617ms 24.4976ms 40.8203 Ops/s 40.3762 Ops/s $\color{#35bf28}+1.10\%$
test_vmap_transformer_speed_decorator[False-True] 24.4187ms 24.2057ms 41.3126 Ops/s 40.9017 Ops/s $\color{#35bf28}+1.00\%$
test_vmap_transformer_speed_decorator[False-False] 24.3783ms 24.1091ms 41.4781 Ops/s 40.9495 Ops/s $\color{#35bf28}+1.29\%$
test_to_module_speed[True] 1.7396ms 1.6548ms 604.3195 Ops/s 619.4215 Ops/s $\color{#d91a1a}-2.44\%$
test_to_module_speed[False] 1.7369ms 1.6285ms 614.0759 Ops/s 615.4002 Ops/s $\color{#d91a1a}-0.22\%$
test_to_module_plain_tensor_speed[None] 2.5282ms 1.7254ms 579.5698 Ops/s 576.3859 Ops/s $\color{#35bf28}+0.55\%$
test_to_module_plain_tensor_speed[False] 2.2258ms 1.5683ms 637.6302 Ops/s 634.0009 Ops/s $\color{#35bf28}+0.57\%$
test_to_module_plain_tensor_speed[True] 3.4682ms 3.3385ms 299.5398 Ops/s 296.7403 Ops/s $\color{#35bf28}+0.94\%$
test_tc_init 80.4520μs 44.2159μs 22.6163 KOps/s 22.9793 KOps/s $\color{#d91a1a}-1.58\%$
test_tc_init_tensor_only 33.7510μs 9.3866μs 106.5349 KOps/s 106.8827 KOps/s $\color{#d91a1a}-0.33\%$
test_tc_init_nested 0.1193ms 86.7001μs 11.5340 KOps/s 11.4520 KOps/s $\color{#35bf28}+0.72\%$
test_tc_init_many_fields 51.2310μs 17.1649μs 58.2585 KOps/s 58.9359 KOps/s $\color{#d91a1a}-1.15\%$
test_tc_first_layer_tensor 29.5910μs 1.8202μs 549.3963 KOps/s 564.7883 KOps/s $\color{#d91a1a}-2.73\%$
test_tc_first_layer_tensor_only 2.8309μs 0.4073μs 2.4554 MOps/s 2.5504 MOps/s $\color{#d91a1a}-3.72\%$
test_tc_first_layer_tensor_set 30.9410μs 4.2520μs 235.1831 KOps/s 241.9890 KOps/s $\color{#d91a1a}-2.81\%$
test_tc_first_layer_tensor_only_set 27.2710μs 3.4912μs 286.4366 KOps/s 292.2237 KOps/s $\color{#d91a1a}-1.98\%$
test_tc_first_layer_nontensor 32.7110μs 6.4298μs 155.5255 KOps/s 167.7273 KOps/s $\textbf{\color{#d91a1a}-7.27\%}$
test_tc_second_layer_tensor 35.6500μs 4.4023μs 227.1550 KOps/s 231.7722 KOps/s $\color{#d91a1a}-1.99\%$
test_tc_second_layer_nontensor 36.3800μs 8.9953μs 111.1698 KOps/s 116.3358 KOps/s $\color{#d91a1a}-4.44\%$
test_unbind 0.2860s 14.1638ms 70.6023 Ops/s 66.5083 Ops/s $\textbf{\color{#35bf28}+6.16\%}$
test_full_like 4.9351ms 4.7118ms 212.2350 Ops/s 212.6690 Ops/s $\color{#d91a1a}-0.20\%$
test_zeros_like 4.7693ms 4.6339ms 215.8011 Ops/s 215.2691 Ops/s $\color{#35bf28}+0.25\%$
test_ones_like 4.7957ms 4.6530ms 214.9159 Ops/s 211.8849 Ops/s $\color{#35bf28}+1.43\%$
test_clone 7.5007ms 6.9840ms 143.1850 Ops/s 142.5888 Ops/s $\color{#35bf28}+0.42\%$
test_squeeze 0.1692ms 14.5300μs 68.8229 KOps/s 71.2031 KOps/s $\color{#d91a1a}-3.34\%$
test_unsqueeze 0.2959ms 0.1120ms 8.9316 KOps/s 9.0531 KOps/s $\color{#d91a1a}-1.34\%$
test_split 0.2478ms 0.1917ms 5.2166 KOps/s 5.2474 KOps/s $\color{#d91a1a}-0.59\%$
test_permute 0.2764ms 0.1941ms 5.1513 KOps/s 5.1875 KOps/s $\color{#d91a1a}-0.70\%$
test_stack 29.1174ms 19.9714ms 50.0717 Ops/s 50.6191 Ops/s $\color{#d91a1a}-1.08\%$
test_cat 20.4777ms 19.7554ms 50.6190 Ops/s 35.6256 Ops/s $\textbf{\color{#35bf28}+42.09\%}$
test_sequential_tensordict 0.9933ms 0.8943ms 1.1182 KOps/s 1.1440 KOps/s $\color{#d91a1a}-2.25\%$
test_sequential_graph_module 0.5239ms 0.4699ms 2.1282 KOps/s 2.4234 KOps/s $\textbf{\color{#d91a1a}-12.18\%}$
test_nested_tensordict 1.4071ms 1.2593ms 794.0991 Ops/s 808.8622 Ops/s $\color{#d91a1a}-1.83\%$
test_nested_graph_module 0.6968ms 0.5489ms 1.8219 KOps/s 1.9200 KOps/s $\textbf{\color{#d91a1a}-5.11\%}$

@github-actions

github-actions Bot commented Jul 9, 2026 •
edited
Loading

Copy link
Copy Markdown
Contributor

$\color{#D29922}\textsf{\Large⚠\kern{0.2cm}\normalsize Warning}$ Result of GPU Benchmark Tests

Total Benchmarks: 352. Improved: $\large\color{#35bf28}18$. Worsened: $\large\color{#d91a1a}14$.

Expand to view detailed results
Name Max Mean Ops Ops on Repo HEAD Change
test_collector_preallocated_write[stack_out-cpu-32-2] 0.2046ms 0.1636ms 6.1114 KOps/s 5.9919 KOps/s $\color{#35bf28}+1.99\%$
test_collector_preallocated_write[stack_out-cpu-32-8] 0.3624ms 0.2939ms 3.4028 KOps/s 3.3779 KOps/s $\color{#35bf28}+0.74\%$
test_collector_preallocated_write[stack_out-cpu-32-32] 0.8934ms 0.7767ms 1.2875 KOps/s 1.2666 KOps/s $\color{#35bf28}+1.66\%$
test_collector_preallocated_write[stack_out-cpu-4096-2] 0.4166ms 0.3470ms 2.8818 KOps/s 2.9593 KOps/s $\color{#d91a1a}-2.62\%$
test_collector_preallocated_write[stack_out-cpu-4096-8] 1.3484ms 1.1607ms 861.5562 Ops/s 893.8483 Ops/s $\color{#d91a1a}-3.61\%$
test_collector_preallocated_write[stack_out-cpu-4096-32] 7.7030ms 7.4611ms 134.0292 Ops/s 134.2443 Ops/s $\color{#d91a1a}-0.16\%$
test_collector_preallocated_write[stack_out-cuda-32-2] 0.3126ms 0.2489ms 4.0182 KOps/s 4.0231 KOps/s $\color{#d91a1a}-0.12\%$
test_collector_preallocated_write[stack_out-cuda-32-8] 0.4747ms 0.3689ms 2.7106 KOps/s 2.6899 KOps/s $\color{#35bf28}+0.77\%$
test_collector_preallocated_write[stack_out-cuda-32-32] 1.8124ms 0.8334ms 1.1999 KOps/s 1.1941 KOps/s $\color{#35bf28}+0.48\%$
test_collector_preallocated_write[stack_out-cuda-4096-2] 0.3743ms 0.2475ms 4.0412 KOps/s 4.0078 KOps/s $\color{#35bf28}+0.83\%$
test_collector_preallocated_write[stack_out-cuda-4096-8] 0.4722ms 0.3704ms 2.7000 KOps/s 2.6706 KOps/s $\color{#35bf28}+1.10\%$
test_collector_preallocated_write[stack_out-cuda-4096-32] 0.9405ms 0.8885ms 1.1255 KOps/s 1.1156 KOps/s $\color{#35bf28}+0.88\%$
test_collector_preallocated_write[set_at_loop-cpu-32-2] 0.3339ms 0.2037ms 4.9081 KOps/s 4.6884 KOps/s $\color{#35bf28}+4.69\%$
test_collector_preallocated_write[set_at_loop-cpu-32-8] 1.0004ms 0.7995ms 1.2508 KOps/s 1.2365 KOps/s $\color{#35bf28}+1.16\%$
test_collector_preallocated_write[set_at_loop-cpu-32-32] 3.3448ms 3.1635ms 316.1023 Ops/s 311.1167 Ops/s $\color{#35bf28}+1.60\%$
test_collector_preallocated_write[set_at_loop-cpu-4096-2] 0.4742ms 0.3572ms 2.7996 KOps/s 2.7476 KOps/s $\color{#35bf28}+1.89\%$
test_collector_preallocated_write[set_at_loop-cpu-4096-8] 1.9611ms 1.7441ms 573.3478 Ops/s 619.8351 Ops/s $\textbf{\color{#d91a1a}-7.50\%}$
test_collector_preallocated_write[set_at_loop-cpu-4096-32] 10.5111ms 10.2708ms 97.3630 Ops/s 96.9702 Ops/s $\color{#35bf28}+0.41\%$
test_collector_preallocated_write[set_at_loop-cuda-32-2] 0.4437ms 0.3968ms 2.5201 KOps/s 2.5675 KOps/s $\color{#d91a1a}-1.84\%$
test_collector_preallocated_write[set_at_loop-cuda-32-8] 1.5682ms 1.4374ms 695.7156 Ops/s 689.4655 Ops/s $\color{#35bf28}+0.91\%$
test_collector_preallocated_write[set_at_loop-cuda-32-32] 5.9493ms 5.6379ms 177.3718 Ops/s 176.3999 Ops/s $\color{#35bf28}+0.55\%$
test_collector_preallocated_write[set_at_loop-cuda-4096-2] 0.4650ms 0.3798ms 2.6332 KOps/s 2.5777 KOps/s $\color{#35bf28}+2.15\%$
test_collector_preallocated_write[set_at_loop-cuda-4096-8] 1.6392ms 1.4247ms 701.9069 Ops/s 695.7791 Ops/s $\color{#35bf28}+0.88\%$
test_collector_preallocated_write[set_at_loop-cuda-4096-32] 5.7206ms 5.5882ms 178.9499 Ops/s 177.2544 Ops/s $\color{#35bf28}+0.96\%$
test_collector_preallocated_write[copy_at-cpu-32-2] 0.2949ms 0.1676ms 5.9661 KOps/s 5.9277 KOps/s $\color{#35bf28}+0.65\%$
test_collector_preallocated_write[copy_at-cpu-32-8] 0.7652ms 0.6492ms 1.5402 KOps/s 1.5225 KOps/s $\color{#35bf28}+1.17\%$
test_collector_preallocated_write[copy_at-cpu-32-32] 2.6822ms 2.5880ms 386.3998 Ops/s 382.2037 Ops/s $\color{#35bf28}+1.10\%$
test_collector_preallocated_write[copy_at-cpu-4096-2] 0.3745ms 0.3152ms 3.1728 KOps/s 3.1097 KOps/s $\color{#35bf28}+2.03\%$
test_collector_preallocated_write[copy_at-cpu-4096-8] 1.8029ms 1.5614ms 640.4630 Ops/s 682.9245 Ops/s $\textbf{\color{#d91a1a}-6.22\%}$
test_collector_preallocated_write[copy_at-cpu-4096-32] 9.8213ms 9.5709ms 104.4829 Ops/s 105.6887 Ops/s $\color{#d91a1a}-1.14\%$
test_collector_preallocated_write[copy_at-cuda-32-2] 0.4130ms 0.3187ms 3.1379 KOps/s 3.0998 KOps/s $\color{#35bf28}+1.23\%$
test_collector_preallocated_write[copy_at-cuda-32-8] 1.3162ms 1.1831ms 845.2702 Ops/s 841.0439 Ops/s $\color{#35bf28}+0.50\%$
test_collector_preallocated_write[copy_at-cuda-32-32] 4.7622ms 4.5946ms 217.6466 Ops/s 214.1337 Ops/s $\color{#35bf28}+1.64\%$
test_collector_preallocated_write[copy_at-cuda-4096-2] 0.4564ms 0.3182ms 3.1431 KOps/s 3.1189 KOps/s $\color{#35bf28}+0.78\%$
test_collector_preallocated_write[copy_at-cuda-4096-8] 1.2816ms 1.1800ms 847.4295 Ops/s 841.5489 Ops/s $\color{#35bf28}+0.70\%$
test_collector_preallocated_write[copy_at-cuda-4096-32] 4.7274ms 4.6071ms 217.0553 Ops/s 213.8354 Ops/s $\color{#35bf28}+1.51\%$
test_collector_preallocated_write[manual_leaf_copy-cpu-32-2] 0.1540ms 0.1194ms 8.3781 KOps/s 8.5595 KOps/s $\color{#d91a1a}-2.12\%$
test_collector_preallocated_write[manual_leaf_copy-cpu-32-8] 0.5244ms 0.4615ms 2.1667 KOps/s 2.1814 KOps/s $\color{#d91a1a}-0.68\%$
test_collector_preallocated_write[manual_leaf_copy-cpu-32-32] 1.9399ms 1.8326ms 545.6829 Ops/s 545.8595 Ops/s $\color{#d91a1a}-0.03\%$
test_collector_preallocated_write[manual_leaf_copy-cpu-4096-2] 0.3504ms 0.2562ms 3.9028 KOps/s 3.8349 KOps/s $\color{#35bf28}+1.77\%$
test_collector_preallocated_write[manual_leaf_copy-cpu-4096-8] 1.4608ms 1.3072ms 764.9671 Ops/s 822.6876 Ops/s $\textbf{\color{#d91a1a}-7.02\%}$
test_collector_preallocated_write[manual_leaf_copy-cpu-4096-32] 8.9684ms 8.6670ms 115.3796 Ops/s 116.0734 Ops/s $\color{#d91a1a}-0.60\%$
test_collector_preallocated_write[manual_leaf_copy-cuda-32-2] 0.3783ms 0.2676ms 3.7365 KOps/s 3.7234 KOps/s $\color{#35bf28}+0.35\%$
test_collector_preallocated_write[manual_leaf_copy-cuda-32-8] 1.1751ms 0.9914ms 1.0087 KOps/s 1.0006 KOps/s $\color{#35bf28}+0.81\%$
test_collector_preallocated_write[manual_leaf_copy-cuda-32-32] 3.9695ms 3.8927ms 256.8921 Ops/s 254.5812 Ops/s $\color{#35bf28}+0.91\%$
test_collector_preallocated_write[manual_leaf_copy-cuda-4096-2] 0.3238ms 0.2682ms 3.7280 KOps/s 3.7068 KOps/s $\color{#35bf28}+0.57\%$
test_collector_preallocated_write[manual_leaf_copy-cuda-4096-8] 1.0906ms 0.9945ms 1.0056 KOps/s 1.0001 KOps/s $\color{#35bf28}+0.55\%$
test_collector_preallocated_write[manual_leaf_copy-cuda-4096-32] 3.9906ms 3.8802ms 257.7212 Ops/s 253.0045 Ops/s $\color{#35bf28}+1.86\%$
test_wide_nested_preallocated_write[stack_out-cpu-128-128] 16.4811ms 16.2842ms 61.4092 Ops/s 61.1789 Ops/s $\color{#35bf28}+0.38\%$
test_wide_nested_preallocated_write[stack_out-cpu-128-256] 33.4057ms 33.1826ms 30.1363 Ops/s 30.1014 Ops/s $\color{#35bf28}+0.12\%$
test_wide_nested_preallocated_write[stack_out-cpu-4096-128] 89.3898ms 86.6116ms 11.5458 Ops/s 11.8623 Ops/s $\color{#d91a1a}-2.67\%$
test_wide_nested_preallocated_write[stack_out-cpu-4096-256] 0.1995s 0.1987s 5.0327 Ops/s 5.0849 Ops/s $\color{#d91a1a}-1.03\%$
test_wide_nested_preallocated_write[stack_out-cuda-128-128] 13.9450ms 13.7681ms 72.6315 Ops/s 71.9045 Ops/s $\color{#35bf28}+1.01\%$
test_wide_nested_preallocated_write[stack_out-cuda-128-256] 27.1163ms 26.9554ms 37.0984 Ops/s 37.0193 Ops/s $\color{#35bf28}+0.21\%$
test_wide_nested_preallocated_write[stack_out-cuda-4096-128] 14.0897ms 13.8835ms 72.0279 Ops/s 71.4738 Ops/s $\color{#35bf28}+0.78\%$
test_wide_nested_preallocated_write[stack_out-cuda-4096-256] 27.7924ms 27.3356ms 36.5823 Ops/s 36.4172 Ops/s $\color{#35bf28}+0.45\%$
test_wide_nested_preallocated_write[set_at_loop-cpu-128-128] 50.9476ms 50.6465ms 19.7447 Ops/s 19.7749 Ops/s $\color{#d91a1a}-0.15\%$
test_wide_nested_preallocated_write[set_at_loop-cpu-128-256] 0.1010s 0.1003s 9.9737 Ops/s 9.8338 Ops/s $\color{#35bf28}+1.42\%$
test_wide_nested_preallocated_write[set_at_loop-cpu-4096-128] 0.1544s 0.1497s 6.6783 Ops/s 7.0267 Ops/s $\color{#d91a1a}-4.96\%$
test_wide_nested_preallocated_write[set_at_loop-cpu-4096-256] 0.3315s 0.3263s 3.0648 Ops/s 3.1564 Ops/s $\color{#d91a1a}-2.90\%$
test_wide_nested_preallocated_write[set_at_loop-cuda-128-128] 77.3546ms 73.1236ms 13.6755 Ops/s 13.7824 Ops/s $\color{#d91a1a}-0.78\%$
test_wide_nested_preallocated_write[set_at_loop-cuda-128-256] 0.1541s 0.1467s 6.8171 Ops/s 6.9186 Ops/s $\color{#d91a1a}-1.47\%$
test_wide_nested_preallocated_write[set_at_loop-cuda-4096-128] 78.0904ms 74.5384ms 13.4159 Ops/s 13.6241 Ops/s $\color{#d91a1a}-1.53\%$
test_wide_nested_preallocated_write[set_at_loop-cuda-4096-256] 0.1511s 0.1470s 6.8042 Ops/s 6.8266 Ops/s $\color{#d91a1a}-0.33\%$
test_wide_nested_preallocated_write[copy_at-cpu-128-128] 29.9320ms 29.7188ms 33.6488 Ops/s 33.2234 Ops/s $\color{#35bf28}+1.28\%$
test_wide_nested_preallocated_write[copy_at-cpu-128-256] 60.3662ms 60.1621ms 16.6218 Ops/s 16.4673 Ops/s $\color{#35bf28}+0.94\%$
test_wide_nested_preallocated_write[copy_at-cpu-4096-128] 0.1330s 0.1295s 7.7235 Ops/s 8.4333 Ops/s $\textbf{\color{#d91a1a}-8.42\%}$
test_wide_nested_preallocated_write[copy_at-cpu-4096-256] 0.2762s 0.2730s 3.6629 Ops/s 3.6937 Ops/s $\color{#d91a1a}-0.83\%$
test_wide_nested_preallocated_write[copy_at-cuda-128-128] 43.1039ms 42.6675ms 23.4370 Ops/s 23.2452 Ops/s $\color{#35bf28}+0.83\%$
test_wide_nested_preallocated_write[copy_at-cuda-128-256] 85.7238ms 85.1980ms 11.7374 Ops/s 11.6162 Ops/s $\color{#35bf28}+1.04\%$
test_wide_nested_preallocated_write[copy_at-cuda-4096-128] 43.2236ms 42.8158ms 23.3559 Ops/s 23.0919 Ops/s $\color{#35bf28}+1.14\%$
test_wide_nested_preallocated_write[copy_at-cuda-4096-256] 86.1967ms 85.8444ms 11.6490 Ops/s 11.5601 Ops/s $\color{#35bf28}+0.77\%$
test_wide_nested_preallocated_write[manual_leaf_copy-cpu-128-128] 20.1995ms 20.0569ms 49.8582 Ops/s 50.4588 Ops/s $\color{#d91a1a}-1.19\%$
test_wide_nested_preallocated_write[manual_leaf_copy-cpu-128-256] 41.0873ms 40.7064ms 24.5662 Ops/s 24.7969 Ops/s $\color{#d91a1a}-0.93\%$
test_wide_nested_preallocated_write[manual_leaf_copy-cpu-4096-128] 0.1127s 0.1119s 8.9359 Ops/s 9.3162 Ops/s $\color{#d91a1a}-4.08\%$
test_wide_nested_preallocated_write[manual_leaf_copy-cpu-4096-256] 0.2450s 0.2435s 4.1063 Ops/s 4.0454 Ops/s $\color{#35bf28}+1.50\%$
test_wide_nested_preallocated_write[manual_leaf_copy-cuda-128-128] 36.2626ms 35.8270ms 27.9119 Ops/s 27.6599 Ops/s $\color{#35bf28}+0.91\%$
test_wide_nested_preallocated_write[manual_leaf_copy-cuda-128-256] 71.6332ms 71.1242ms 14.0599 Ops/s 13.9003 Ops/s $\color{#35bf28}+1.15\%$
test_wide_nested_preallocated_write[manual_leaf_copy-cuda-4096-128] 36.0204ms 35.5597ms 28.1217 Ops/s 27.7375 Ops/s $\color{#35bf28}+1.39\%$
test_wide_nested_preallocated_write[manual_leaf_copy-cuda-4096-256] 72.5634ms 71.6979ms 13.9474 Ops/s 13.8683 Ops/s $\color{#35bf28}+0.57\%$
test_plain_set_nested 98.6920μs 15.2202μs 65.7022 KOps/s 67.0685 KOps/s $\color{#d91a1a}-2.04\%$
test_plain_set_stack_nested 58.9310μs 14.8431μs 67.3714 KOps/s 66.8320 KOps/s $\color{#35bf28}+0.81\%$
test_plain_set_nested_inplace 44.0000μs 16.4822μs 60.6716 KOps/s 60.1633 KOps/s $\color{#35bf28}+0.84\%$
test_plain_set_stack_nested_inplace 84.4610μs 16.0992μs 62.1149 KOps/s 61.0268 KOps/s $\color{#35bf28}+1.78\%$
test_items 72.6710μs 5.9296μs 168.6453 KOps/s 166.3238 KOps/s $\color{#35bf28}+1.40\%$
test_items_nested 0.5503ms 0.4549ms 2.1985 KOps/s 2.1586 KOps/s $\color{#35bf28}+1.85\%$
test_items_nested_locked 0.5518ms 0.4612ms 2.1682 KOps/s 2.1361 KOps/s $\color{#35bf28}+1.50\%$
test_items_nested_leaf 0.1634ms 98.3447μs 10.1683 KOps/s 10.0690 KOps/s $\color{#35bf28}+0.99\%$
test_items_stack_nested 0.5176ms 0.4551ms 2.1975 KOps/s 2.1472 KOps/s $\color{#35bf28}+2.34\%$
test_items_stack_nested_leaf 0.1653ms 97.7993μs 10.2250 KOps/s 10.0068 KOps/s $\color{#35bf28}+2.18\%$
test_items_stack_nested_locked 0.5508ms 0.4613ms 2.1679 KOps/s 2.1397 KOps/s $\color{#35bf28}+1.32\%$
test_keys 29.7100μs 4.2230μs 236.7992 KOps/s 239.3973 KOps/s $\color{#d91a1a}-1.09\%$
test_keys_nested 0.1873ms 0.1284ms 7.7852 KOps/s 7.6929 KOps/s $\color{#35bf28}+1.20\%$
test_keys_nested_locked 0.7209ms 0.1369ms 7.3059 KOps/s 7.1117 KOps/s $\color{#35bf28}+2.73\%$
test_keys_nested_leaf 0.1745ms 0.1194ms 8.3746 KOps/s 8.1776 KOps/s $\color{#35bf28}+2.41\%$
test_keys_stack_nested 0.1840ms 0.1276ms 7.8365 KOps/s 7.6296 KOps/s $\color{#35bf28}+2.71\%$
test_keys_stack_nested_leaf 0.1682ms 0.1183ms 8.4536 KOps/s 8.1635 KOps/s $\color{#35bf28}+3.55\%$
test_keys_stack_nested_locked 0.2596ms 0.1362ms 7.3400 KOps/s 7.2036 KOps/s $\color{#35bf28}+1.89\%$
test_values 17.8604μs 1.0224μs 978.0691 KOps/s 981.2785 KOps/s $\color{#d91a1a}-0.33\%$
test_values_nested 87.4010μs 52.3392μs 19.1061 KOps/s 18.7441 KOps/s $\color{#35bf28}+1.93\%$
test_values_nested_locked 0.1134ms 55.7262μs 17.9449 KOps/s 17.5374 KOps/s $\color{#35bf28}+2.32\%$
test_values_nested_leaf 0.1216ms 59.7690μs 16.7311 KOps/s 16.2395 KOps/s $\color{#35bf28}+3.03\%$
test_values_stack_nested 0.1209ms 52.3510μs 19.1018 KOps/s 18.8627 KOps/s $\color{#35bf28}+1.27\%$
test_values_stack_nested_leaf 0.1326ms 59.4907μs 16.8093 KOps/s 16.3866 KOps/s $\color{#35bf28}+2.58\%$
test_values_stack_nested_locked 0.1192ms 55.3284μs 18.0739 KOps/s 18.0534 KOps/s $\color{#35bf28}+0.11\%$
test_membership 13.0368μs 0.8434μs 1.1857 MOps/s 1.1977 MOps/s $\color{#d91a1a}-1.00\%$
test_membership_nested 63.0810μs 2.9432μs 339.7664 KOps/s 333.8217 KOps/s $\color{#35bf28}+1.78\%$
test_membership_nested_leaf 38.1010μs 2.9208μs 342.3763 KOps/s 345.7206 KOps/s $\color{#d91a1a}-0.97\%$
test_membership_stacked_nested 74.8410μs 2.8024μs 356.8348 KOps/s 345.6654 KOps/s $\color{#35bf28}+3.23\%$
test_membership_stacked_nested_leaf 35.6910μs 2.8005μs 357.0745 KOps/s 348.5392 KOps/s $\color{#35bf28}+2.45\%$
test_membership_nested_last 85.3620μs 4.3887μs 227.8593 KOps/s 221.1005 KOps/s $\color{#35bf28}+3.06\%$
test_membership_nested_leaf_last 41.4400μs 4.4080μs 226.8600 KOps/s 223.1022 KOps/s $\color{#35bf28}+1.68\%$
test_membership_stacked_nested_last 70.6920μs 4.2363μs 236.0537 KOps/s 230.4592 KOps/s $\color{#35bf28}+2.43\%$
test_membership_stacked_nested_leaf_last 46.3010μs 4.1990μs 238.1493 KOps/s 240.2021 KOps/s $\color{#d91a1a}-0.85\%$
test_nested_getleaf 89.8920μs 21.1414μs 47.3006 KOps/s 47.0413 KOps/s $\color{#35bf28}+0.55\%$
test_nested_get 54.0610μs 19.9169μs 50.2086 KOps/s 49.8759 KOps/s $\color{#35bf28}+0.67\%$
test_stacked_getleaf 95.5410μs 20.9657μs 47.6969 KOps/s 47.7755 KOps/s $\color{#d91a1a}-0.16\%$
test_stacked_get 58.6920μs 20.0174μs 49.9564 KOps/s 51.7332 KOps/s $\color{#d91a1a}-3.43\%$
test_nested_getitemleaf 0.1120ms 21.9563μs 45.5451 KOps/s 45.2052 KOps/s $\color{#35bf28}+0.75\%$
test_nested_getitem 98.9120μs 20.8050μs 48.0654 KOps/s 47.5930 KOps/s $\color{#35bf28}+0.99\%$
test_stacked_getitemleaf 83.7810μs 21.5641μs 46.3734 KOps/s 46.6338 KOps/s $\color{#d91a1a}-0.56\%$
test_stacked_getitem 60.2010μs 20.2528μs 49.3760 KOps/s 50.1207 KOps/s $\color{#d91a1a}-1.49\%$
test_lock_nested 0.5805ms 0.4968ms 2.0130 KOps/s 2.0134 KOps/s $\color{#d91a1a}-0.02\%$
test_lock_stack_nested 0.6694ms 0.5038ms 1.9850 KOps/s 1.9767 KOps/s $\color{#35bf28}+0.42\%$
test_unlock_nested 0.4890ms 0.3947ms 2.5334 KOps/s 2.5275 KOps/s $\color{#35bf28}+0.23\%$
test_unlock_stack_nested 0.6953ms 0.3976ms 2.5150 KOps/s 2.5029 KOps/s $\color{#35bf28}+0.48\%$
test_flatten_speed 0.3201ms 0.1223ms 8.1750 KOps/s 8.0527 KOps/s $\color{#35bf28}+1.52\%$
test_unflatten_speed 0.6681ms 0.5605ms 1.7842 KOps/s 1.7731 KOps/s $\color{#35bf28}+0.63\%$
test_common_ops 0.8652ms 0.6823ms 1.4657 KOps/s 1.4408 KOps/s $\color{#35bf28}+1.73\%$
test_creation 0.2409ms 3.1695μs 315.5031 KOps/s 324.0727 KOps/s $\color{#d91a1a}-2.64\%$
test_creation_empty 39.1710μs 7.0235μs 142.3791 KOps/s 141.9289 KOps/s $\color{#35bf28}+0.32\%$
test_creation_nested_1 41.7100μs 11.4355μs 87.4469 KOps/s 86.9987 KOps/s $\color{#35bf28}+0.52\%$
test_creation_nested_2 57.7710μs 13.3364μs 74.9828 KOps/s 74.2067 KOps/s $\color{#35bf28}+1.05\%$
test_creation_many_keys[10] 83.8810μs 20.2120μs 49.4757 KOps/s 47.8472 KOps/s $\color{#35bf28}+3.40\%$
test_creation_many_keys[50] 0.1543ms 90.0584μs 11.1039 KOps/s 11.0604 KOps/s $\color{#35bf28}+0.39\%$
test_creation_many_keys[100] 0.2656ms 0.1737ms 5.7569 KOps/s 5.6838 KOps/s $\color{#35bf28}+1.29\%$
test_creation_nested_many_keys[10] 0.1218ms 44.4148μs 22.5150 KOps/s 22.3362 KOps/s $\color{#35bf28}+0.80\%$
test_creation_nested_many_keys[50] 0.2709ms 0.1822ms 5.4882 KOps/s 5.4453 KOps/s $\color{#35bf28}+0.79\%$
test_clone 95.3520μs 12.7737μs 78.2859 KOps/s 78.4013 KOps/s $\color{#d91a1a}-0.15\%$
test_getitem[int] 1.5810ms 15.0232μs 66.5637 KOps/s 65.5337 KOps/s $\color{#35bf28}+1.57\%$
test_getitem[slice_int] 0.1410ms 24.4599μs 40.8833 KOps/s 40.8499 KOps/s $\color{#35bf28}+0.08\%$
test_getitem[range] 0.1887ms 60.9249μs 16.4137 KOps/s 16.1459 KOps/s $\color{#35bf28}+1.66\%$
test_getitem[tuple] 0.1455ms 23.5375μs 42.4854 KOps/s 41.9698 KOps/s $\color{#35bf28}+1.23\%$
test_getitem[list] 0.1826ms 55.5558μs 17.9999 KOps/s 17.7777 KOps/s $\color{#35bf28}+1.25\%$
test_setitem_dim[int] 93.9520μs 25.0108μs 39.9827 KOps/s 39.2699 KOps/s $\color{#35bf28}+1.82\%$
test_setitem_dim[slice_int] 80.3010μs 41.2384μs 24.2493 KOps/s 23.0612 KOps/s $\textbf{\color{#35bf28}+5.15\%}$
test_setitem_dim[range] 0.1634ms 90.3217μs 11.0715 KOps/s 10.5698 KOps/s $\color{#35bf28}+4.75\%$
test_setitem_dim[tuple] 71.6710μs 37.8494μs 26.4205 KOps/s 25.6004 KOps/s $\color{#35bf28}+3.20\%$
test_setitem 49.4510μs 16.7780μs 59.6017 KOps/s 59.0317 KOps/s $\color{#35bf28}+0.97\%$
test_set 96.3520μs 16.2337μs 61.6002 KOps/s 61.5299 KOps/s $\color{#35bf28}+0.11\%$
test_set_shared 0.6252ms 0.2112ms 4.7347 KOps/s 4.7331 KOps/s $\color{#35bf28}+0.03\%$
test_update 0.3658ms 20.9942μs 47.6322 KOps/s 47.5063 KOps/s $\color{#35bf28}+0.27\%$
test_update_nested 0.1234ms 31.8472μs 31.4000 KOps/s 30.2760 KOps/s $\color{#35bf28}+3.71\%$
test_update__nested 0.5202ms 32.2201μs 31.0365 KOps/s 30.3590 KOps/s $\color{#35bf28}+2.23\%$
test_set_nested 46.6900μs 18.1836μs 54.9946 KOps/s 54.9443 KOps/s $\color{#35bf28}+0.09\%$
test_set_nested_new 65.4510μs 22.9290μs 43.6128 KOps/s 44.3146 KOps/s $\color{#d91a1a}-1.58\%$
test_select 76.8420μs 39.5003μs 25.3163 KOps/s 25.4196 KOps/s $\color{#d91a1a}-0.41\%$
test_select_nested 0.1332ms 74.9171μs 13.3481 KOps/s 13.4319 KOps/s $\color{#d91a1a}-0.62\%$
test_exclude_nested 0.1713ms 90.0808μs 11.1011 KOps/s 10.9662 KOps/s $\color{#35bf28}+1.23\%$
test_empty[True] 0.4591ms 0.3916ms 2.5538 KOps/s 2.5060 KOps/s $\color{#35bf28}+1.91\%$
test_empty[False] 11.0078μs 1.3851μs 721.9668 KOps/s 717.7826 KOps/s $\color{#35bf28}+0.58\%$
test_to 0.1174ms 67.7717μs 14.7554 KOps/s 14.5465 KOps/s $\color{#35bf28}+1.44\%$
test_to_nonblocking 0.1223ms 68.0905μs 14.6863 KOps/s 14.5649 KOps/s $\color{#35bf28}+0.83\%$
test_unbind_speed 0.3737ms 0.3336ms 2.9975 KOps/s 2.9656 KOps/s $\color{#35bf28}+1.07\%$
test_unbind_speed_stack0 0.4127ms 0.3285ms 3.0443 KOps/s 3.0287 KOps/s $\color{#35bf28}+0.51\%$
test_unbind_speed_stack1 0.1179s 0.8439ms 1.1849 KOps/s 1.1632 KOps/s $\color{#35bf28}+1.87\%$
test_split 0.1174s 1.4380ms 695.3921 Ops/s 640.7905 Ops/s $\textbf{\color{#35bf28}+8.52\%}$
test_chunk 1.2843ms 1.2231ms 817.6096 Ops/s 807.2963 Ops/s $\color{#35bf28}+1.28\%$
test_to_cpu_blocking 0.1360s 21.8071ms 45.8567 Ops/s 51.2946 Ops/s $\textbf{\color{#d91a1a}-10.60\%}$
test_to_cpu_global_sync 11.6487ms 11.4439ms 87.3828 Ops/s 86.8013 Ops/s $\color{#35bf28}+0.67\%$
test_to_cpu_event_sync 12.7457ms 12.3910ms 80.7040 Ops/s 79.9272 Ops/s $\color{#35bf28}+0.97\%$
test_to_cpu_default 12.7594ms 12.3548ms 80.9403 Ops/s 71.5900 Ops/s $\textbf{\color{#35bf28}+13.06\%}$
test_consolidate[False-None] 4.2168ms 4.1186ms 242.8025 Ops/s 241.1200 Ops/s $\color{#35bf28}+0.70\%$
test_consolidate[default-None] 3.1848ms 2.2010ms 454.3406 Ops/s 445.4293 Ops/s $\color{#35bf28}+2.00\%$
test_consolidate[reduce-overhead-None] 3.2064ms 2.1254ms 470.4987 Ops/s 458.6061 Ops/s $\color{#35bf28}+2.59\%$
test_consolidate_njt[False-None] 5.7599ms 5.0716ms 197.1780 Ops/s 195.8033 Ops/s $\color{#35bf28}+0.70\%$
test_to[False-False-None] 2.2294ms 2.1588ms 463.2121 Ops/s 454.8620 Ops/s $\color{#35bf28}+1.84\%$
test_to[True-False-None] 2.1665ms 1.8975ms 527.0113 Ops/s 268.7321 Ops/s $\textbf{\color{#35bf28}+96.11\%}$
test_to[within-False-None] 6.2489ms 6.1224ms 163.3358 Ops/s 162.9581 Ops/s $\color{#35bf28}+0.23\%$
test_to[True-default-None] 10.6767ms 10.2344ms 97.7097 Ops/s 96.0370 Ops/s $\color{#35bf28}+1.74\%$
test_to_njt[False-False-None] 7.0923ms 6.9349ms 144.1976 Ops/s 144.0152 Ops/s $\color{#35bf28}+0.13\%$
test_to_njt[True-False-None] 4.0552ms 3.8594ms 259.1057 Ops/s 261.6345 Ops/s $\color{#d91a1a}-0.97\%$
test_to_njt[within-False-None] 9.3260ms 8.9454ms 111.7894 Ops/s 111.7770 Ops/s $\color{#35bf28}+0.01\%$
test_creation[device0] 0.3904ms 0.1153ms 8.6721 KOps/s 8.8176 KOps/s $\color{#d91a1a}-1.65\%$
test_creation_from_tensor 0.7050ms 0.1164ms 8.5902 KOps/s 8.8682 KOps/s $\color{#d91a1a}-3.13\%$
test_add_one[memmap_tensor0] 0.3590ms 6.0495μs 165.3032 KOps/s 163.7646 KOps/s $\color{#35bf28}+0.94\%$
test_contiguous[memmap_tensor0] 26.4600μs 0.6846μs 1.4608 MOps/s 2.0157 MOps/s $\textbf{\color{#d91a1a}-27.53\%}$
test_stack[memmap_tensor0] 82.8320μs 4.0477μs 247.0529 KOps/s 242.0952 KOps/s $\color{#35bf28}+2.05\%$
test_memmaptd_index 1.3421ms 0.2725ms 3.6703 KOps/s 3.6909 KOps/s $\color{#d91a1a}-0.56\%$
test_memmaptd_index_astensor 0.5359ms 0.3728ms 2.6825 KOps/s 2.6839 KOps/s $\color{#d91a1a}-0.05\%$
test_memmaptd_index_op 0.9783ms 0.6104ms 1.6384 KOps/s 1.6281 KOps/s $\color{#35bf28}+0.63\%$
test_serialize_model 0.1368s 0.1356s 7.3737 Ops/s 5.9620 Ops/s $\textbf{\color{#35bf28}+23.68\%}$
test_serialize_model_pickle 1.3475s 1.2107s 0.8260 Ops/s 0.8252 Ops/s $\color{#35bf28}+0.09\%$
test_serialize_weights 0.1377s 0.1359s 7.3559 Ops/s 7.3881 Ops/s $\color{#d91a1a}-0.44\%$
test_serialize_weights_returnearly 0.2849s 46.9622ms 21.2937 Ops/s 29.5406 Ops/s $\textbf{\color{#d91a1a}-27.92\%}$
test_serialize_weights_pickle 1.3759s 1.2201s 0.8196 Ops/s 0.8182 Ops/s $\color{#35bf28}+0.17\%$
test_consolidate[few_large-0] 61.0773ms 7.9256ms 126.1730 Ops/s 136.7411 Ops/s $\textbf{\color{#d91a1a}-7.73\%}$
test_consolidate[few_large-8] 60.4500ms 8.3348ms 119.9794 Ops/s 139.0825 Ops/s $\textbf{\color{#d91a1a}-13.74\%}$
test_consolidate[many_small-0] 0.1282s 0.1225s 8.1601 Ops/s 8.2801 Ops/s $\color{#d91a1a}-1.45\%$
test_consolidate[many_small-8] 0.2256s 48.1253ms 20.7791 Ops/s 22.2024 Ops/s $\textbf{\color{#d91a1a}-6.41\%}$
test_consolidate_to_file[few_large-0] 0.1313s 22.0918ms 45.2657 Ops/s 43.1333 Ops/s $\color{#35bf28}+4.94\%$
test_consolidate_to_file[few_large-8] 0.6003s 0.1491s 6.7072 Ops/s 13.9545 Ops/s $\textbf{\color{#d91a1a}-51.94\%}$
test_consolidate_to_file[many_small-0] 1.6170s 1.1005s 0.9087 Ops/s 1.2351 Ops/s $\textbf{\color{#d91a1a}-26.43\%}$
test_consolidate_to_file[many_small-8] 0.8910s 0.7245s 1.3802 Ops/s 1.2746 Ops/s $\textbf{\color{#35bf28}+8.29\%}$
test_reshape_pytree 12.9792ms 33.1743μs 30.1438 KOps/s 30.9443 KOps/s $\color{#d91a1a}-2.59\%$
test_reshape_td 75.0410μs 45.0305μs 22.2072 KOps/s 21.5641 KOps/s $\color{#35bf28}+2.98\%$
test_view_pytree 0.2132ms 31.5464μs 31.6994 KOps/s 31.4966 KOps/s $\color{#35bf28}+0.64\%$
test_view_td 88.7620μs 52.7238μs 18.9668 KOps/s 17.8734 KOps/s $\textbf{\color{#35bf28}+6.12\%}$
test_unbind_pytree 0.2262ms 35.5890μs 28.0986 KOps/s 27.3793 KOps/s $\color{#35bf28}+2.63\%$
test_unbind_td 98.9420μs 48.8847μs 20.4563 KOps/s 19.8705 KOps/s $\color{#35bf28}+2.95\%$
test_split_pytree 0.2420ms 40.4582μs 24.7168 KOps/s 24.0993 KOps/s $\color{#35bf28}+2.56\%$
test_split_td 0.1801ms 69.9463μs 14.2967 KOps/s 13.8351 KOps/s $\color{#35bf28}+3.34\%$
test_add_pytree 0.2324ms 39.1768μs 25.5253 KOps/s 25.0787 KOps/s $\color{#35bf28}+1.78\%$
test_add_td 91.2610μs 51.8315μs 19.2933 KOps/s 18.8648 KOps/s $\color{#35bf28}+2.27\%$
test_compile_add_one_nested[tensordict-compile] 0.3670ms 0.2789ms 3.5849 KOps/s 3.5009 KOps/s $\color{#35bf28}+2.40\%$
test_compile_add_one_nested[tensordict-eager] 0.2665ms 0.2006ms 4.9857 KOps/s 4.8211 KOps/s $\color{#35bf28}+3.41\%$
test_compile_add_one_nested[pytree-compile] 0.2382ms 0.1539ms 6.4995 KOps/s 5.9697 KOps/s $\textbf{\color{#35bf28}+8.87\%}$
test_compile_add_one_nested[pytree-eager] 0.4949ms 0.1756ms 5.6959 KOps/s 5.4779 KOps/s $\color{#35bf28}+3.98\%$
test_compile_copy_nested[tensordict-compile] 75.8420μs 21.5758μs 46.3483 KOps/s 46.3420 KOps/s $\color{#35bf28}+0.01\%$
test_compile_copy_nested[tensordict-eager] 0.2300ms 52.3023μs 19.1196 KOps/s 18.3036 KOps/s $\color{#35bf28}+4.46\%$
test_compile_copy_nested[pytree-compile] 0.1974ms 19.2353μs 51.9878 KOps/s 52.2360 KOps/s $\color{#d91a1a}-0.48\%$
test_compile_copy_nested[pytree-eager] 0.3612ms 67.3184μs 14.8548 KOps/s 14.8438 KOps/s $\color{#35bf28}+0.07\%$
test_compile_add_one_flat[tensordict-compile] 0.3676ms 0.3092ms 3.2346 KOps/s 3.1459 KOps/s $\color{#35bf28}+2.82\%$
test_compile_add_one_flat[tensordict-eager] 0.4412ms 0.2876ms 3.4776 KOps/s 3.4235 KOps/s $\color{#35bf28}+1.58\%$
test_compile_add_one_flat[tensorclass-compile] 0.3671ms 0.2182ms 4.5822 KOps/s 4.5070 KOps/s $\color{#35bf28}+1.67\%$
test_compile_add_one_flat[tensorclass-eager] 0.1354ms 75.4356μs 13.2563 KOps/s 12.6638 KOps/s $\color{#35bf28}+4.68\%$
test_compile_add_one_flat[pytree-compile] 0.3825ms 0.2443ms 4.0925 KOps/s 4.0163 KOps/s $\color{#35bf28}+1.90\%$
test_compile_add_one_flat[pytree-eager] 0.7914ms 0.5268ms 1.8984 KOps/s 1.7836 KOps/s $\textbf{\color{#35bf28}+6.44\%}$
test_compile_add_self_flat[tensordict-eager] 0.5104ms 0.3387ms 2.9526 KOps/s 2.9028 KOps/s $\color{#35bf28}+1.72\%$
test_compile_add_self_flat[tensordict-compile] 0.5061ms 0.3061ms 3.2667 KOps/s 3.1074 KOps/s $\textbf{\color{#35bf28}+5.13\%}$
test_compile_add_self_flat[tensorclass-eager] 0.1376ms 90.4911μs 11.0508 KOps/s 10.6642 KOps/s $\color{#35bf28}+3.63\%$
test_compile_add_self_flat[tensorclass-compile] 0.3592ms 0.2105ms 4.7513 KOps/s 4.4890 KOps/s $\textbf{\color{#35bf28}+5.84\%}$
test_compile_add_self_flat[pytree-eager] 0.6813ms 0.4349ms 2.2996 KOps/s 2.2434 KOps/s $\color{#35bf28}+2.51\%$
test_compile_add_self_flat[pytree-compile] 0.5777ms 0.2435ms 4.1060 KOps/s 4.0463 KOps/s $\color{#35bf28}+1.48\%$
test_compile_copy_flat[tensordict-compile] 81.3910μs 25.8490μs 38.6862 KOps/s 37.0460 KOps/s $\color{#35bf28}+4.43\%$
test_compile_copy_flat[tensordict-eager] 79.8220μs 40.2429μs 24.8491 KOps/s 24.1645 KOps/s $\color{#35bf28}+2.83\%$
test_compile_copy_flat[pytree-compile] 0.1967ms 24.5969μs 40.6556 KOps/s 40.3097 KOps/s $\color{#35bf28}+0.86\%$
test_compile_copy_flat[pytree-eager] 0.3300ms 51.3095μs 19.4896 KOps/s 19.3279 KOps/s $\color{#35bf28}+0.84\%$
test_compile_assign_and_add[tensordict-compile] 2.1123ms 0.2589ms 3.8627 KOps/s 3.5965 KOps/s $\textbf{\color{#35bf28}+7.40\%}$
test_compile_assign_and_add[tensordict-eager] 3.5154ms 3.2844ms 304.4664 Ops/s 298.8378 Ops/s $\color{#35bf28}+1.88\%$
test_compile_assign_and_add[pytree-compile] 1.9511ms 0.1959ms 5.1058 KOps/s 5.0484 KOps/s $\color{#35bf28}+1.14\%$
test_compile_assign_and_add[pytree-eager] 2.9734ms 2.7456ms 364.2143 Ops/s 352.9652 Ops/s $\color{#35bf28}+3.19\%$
test_compile_indexing[tensor-tensordict-compile] 0.2553ms 0.1556ms 6.4267 KOps/s 6.2618 KOps/s $\color{#35bf28}+2.63\%$
test_compile_indexing[tensor-tensordict-eager] 0.3807ms 72.8892μs 13.7195 KOps/s 13.6532 KOps/s $\color{#35bf28}+0.49\%$
test_compile_indexing[tensor-tensorclass-compile] 0.2846ms 0.1348ms 7.4197 KOps/s 7.4754 KOps/s $\color{#d91a1a}-0.74\%$
test_compile_indexing[tensor-tensorclass-eager] 0.3246ms 43.6588μs 22.9049 KOps/s 23.0908 KOps/s $\color{#d91a1a}-0.81\%$
test_compile_indexing[tensor-pytree-compile] 0.1760ms 0.1349ms 7.4115 KOps/s 7.3984 KOps/s $\color{#35bf28}+0.18\%$
test_compile_indexing[tensor-pytree-eager] 0.2576ms 43.5291μs 22.9731 KOps/s 22.5335 KOps/s $\color{#35bf28}+1.95\%$
test_compile_indexing[slice-tensordict-compile] 0.1846ms 86.6691μs 11.5381 KOps/s 11.0233 KOps/s $\color{#35bf28}+4.67\%$
test_compile_indexing[slice-tensordict-eager] 0.2115ms 36.5401μs 27.3672 KOps/s 36.2964 KOps/s $\textbf{\color{#d91a1a}-24.60\%}$
test_compile_indexing[slice-tensorclass-compile] 0.1908ms 70.9061μs 14.1032 KOps/s 14.1954 KOps/s $\color{#d91a1a}-0.65\%$
test_compile_indexing[slice-tensorclass-eager] 0.2591ms 21.9697μs 45.5173 KOps/s 45.6774 KOps/s $\color{#d91a1a}-0.35\%$
test_compile_indexing[slice-pytree-compile] 0.1508ms 73.5818μs 13.5903 KOps/s 13.9533 KOps/s $\color{#d91a1a}-2.60\%$
test_compile_indexing[slice-pytree-eager] 0.2503ms 21.8450μs 45.7771 KOps/s 46.0470 KOps/s $\color{#d91a1a}-0.59\%$
test_compile_indexing[int-tensordict-compile] 0.1741ms 89.9676μs 11.1151 KOps/s 11.0376 KOps/s $\color{#35bf28}+0.70\%$
test_compile_indexing[int-tensordict-eager] 0.2481ms 27.7184μs 36.0771 KOps/s 36.5355 KOps/s $\color{#d91a1a}-1.25\%$
test_compile_indexing[int-tensorclass-compile] 0.1120ms 71.7315μs 13.9409 KOps/s 14.0362 KOps/s $\color{#d91a1a}-0.68\%$
test_compile_indexing[int-tensorclass-eager] 0.2453ms 21.7521μs 45.9725 KOps/s 46.2676 KOps/s $\color{#d91a1a}-0.64\%$
test_compile_indexing[int-pytree-compile] 0.1196ms 71.2795μs 14.0293 KOps/s 14.0652 KOps/s $\color{#d91a1a}-0.26\%$
test_compile_indexing[int-pytree-eager] 0.2604ms 21.9142μs 45.6324 KOps/s 45.8383 KOps/s $\color{#d91a1a}-0.45\%$
test_compile_replace[single-eager] 97.6120μs 46.9808μs 21.2853 KOps/s 20.7436 KOps/s $\color{#35bf28}+2.61\%$
test_compile_replace[single-compile] 0.2216ms 0.1460ms 6.8516 KOps/s 6.7735 KOps/s $\color{#35bf28}+1.15\%$
test_compile_replace[multi-eager] 0.6336ms 0.5645ms 1.7716 KOps/s 1.7862 KOps/s $\color{#d91a1a}-0.81\%$
test_compile_replace[multi-compile] 0.3127ms 0.1618ms 6.1811 KOps/s 6.2929 KOps/s $\color{#d91a1a}-1.78\%$
test_compile_tc_getattr_20[eager] 0.2293ms 0.1693ms 5.9079 KOps/s 5.9365 KOps/s $\color{#d91a1a}-0.48\%$
test_compile_tc_getattr_20[compile] 0.4237ms 0.1777ms 5.6284 KOps/s 5.5402 KOps/s $\color{#35bf28}+1.59\%$
test_compile_clone_shallow[20-eager] 47.8210μs 18.7799μs 53.2485 KOps/s 53.4064 KOps/s $\color{#d91a1a}-0.30\%$
test_compile_clone_shallow[20-compile] 91.1820μs 22.9074μs 43.6540 KOps/s 43.5006 KOps/s $\color{#35bf28}+0.35\%$
test_compile_clone_shallow[40-eager] 96.8010μs 32.6299μs 30.6468 KOps/s 30.1546 KOps/s $\color{#35bf28}+1.63\%$
test_compile_clone_shallow[40-compile] 79.3920μs 25.7188μs 38.8820 KOps/s 40.4358 KOps/s $\color{#d91a1a}-3.84\%$
test_compile_clone_shallow[80-eager] 95.8010μs 61.5201μs 16.2548 KOps/s 15.9085 KOps/s $\color{#35bf28}+2.18\%$
test_compile_clone_shallow[80-compile] 89.2710μs 29.2420μs 34.1973 KOps/s 33.9941 KOps/s $\color{#35bf28}+0.60\%$
test_compile_update_inplace[eager] 0.2148ms 62.4778μs 16.0057 KOps/s 15.8351 KOps/s $\color{#35bf28}+1.08\%$
test_compile_update_inplace[compile] 0.3723ms 0.1209ms 8.2681 KOps/s 8.3138 KOps/s $\color{#d91a1a}-0.55\%$
test_mod_add[eager] 0.2801ms 0.2254ms 4.4363 KOps/s 4.4049 KOps/s $\color{#35bf28}+0.71\%$
test_mod_add[compile] 0.2582ms 0.1200ms 8.3341 KOps/s 8.4167 KOps/s $\color{#d91a1a}-0.98\%$
test_mod_add[compile-overhead] 0.3607ms 0.1972ms 5.0719 KOps/s 5.0914 KOps/s $\color{#d91a1a}-0.38\%$
test_mod_wrap[eager] 0.6209ms 0.4781ms 2.0915 KOps/s 2.1271 KOps/s $\color{#d91a1a}-1.67\%$
test_mod_wrap[compile] 0.4391ms 0.3560ms 2.8086 KOps/s 2.8174 KOps/s $\color{#d91a1a}-0.31\%$
test_mod_wrap[compile-overhead] 7.1833ms 4.0184ms 248.8523 Ops/s 259.2416 Ops/s $\color{#d91a1a}-4.01\%$
test_mod_wrap_and_backward[eager] 1.8944ms 1.7684ms 565.4962 Ops/s 572.2781 Ops/s $\color{#d91a1a}-1.19\%$
test_mod_wrap_and_backward[compile] 1.8836ms 1.4467ms 691.2256 Ops/s 697.9951 Ops/s $\color{#d91a1a}-0.97\%$
test_mod_wrap_and_backward[compile-overhead] 1.8256ms 1.0673ms 936.9193 Ops/s 943.4284 Ops/s $\color{#d91a1a}-0.69\%$
test_seq_add[eager] 0.5714ms 0.5111ms 1.9567 KOps/s 1.9588 KOps/s $\color{#d91a1a}-0.11\%$
test_seq_add[compile] 0.2774ms 0.1364ms 7.3323 KOps/s 6.3287 KOps/s $\textbf{\color{#35bf28}+15.86\%}$
test_seq_add[compile-overhead] 0.2895ms 0.2050ms 4.8786 KOps/s 4.7996 KOps/s $\color{#35bf28}+1.65\%$
test_seq_wrap[eager] 5.4960ms 1.4242ms 702.1377 Ops/s 710.9780 Ops/s $\color{#d91a1a}-1.24\%$
test_seq_wrap[compile] 0.4338ms 0.3720ms 2.6884 KOps/s 2.7431 KOps/s $\color{#d91a1a}-1.99\%$
test_seq_wrap[compile-overhead] 0.4594ms 0.3291ms 3.0386 KOps/s 3.0992 KOps/s $\color{#d91a1a}-1.96\%$
test_func_call_runtime[False-eager] 1.1179ms 0.8490ms 1.1778 KOps/s 1.2454 KOps/s $\textbf{\color{#d91a1a}-5.42\%}$
test_func_call_runtime[False-compile] 0.9029ms 0.7582ms 1.3190 KOps/s 1.3301 KOps/s $\color{#d91a1a}-0.84\%$
test_func_call_runtime[False-compile-overhead] 0.7015ms 0.5107ms 1.9579 KOps/s 1.9585 KOps/s $\color{#d91a1a}-0.03\%$
test_func_call_runtime[True-eager] 1.1831ms 1.0905ms 917.0055 Ops/s 953.4191 Ops/s $\color{#d91a1a}-3.82\%$
test_func_call_runtime[True-compile] 0.8683ms 0.7821ms 1.2786 KOps/s 1.3091 KOps/s $\color{#d91a1a}-2.33\%$
test_func_call_runtime[True-compile-overhead] 0.7011ms 0.5279ms 1.8944 KOps/s 1.8958 KOps/s $\color{#d91a1a}-0.07\%$
test_func_call_cm_runtime[False-eager] 0.8504ms 0.7999ms 1.2502 KOps/s 1.2595 KOps/s $\color{#d91a1a}-0.74\%$
test_func_call_cm_runtime[False-compile] 0.9100ms 0.7422ms 1.3473 KOps/s 1.3527 KOps/s $\color{#d91a1a}-0.40\%$
test_func_call_cm_runtime[False-compile-overhead] 0.5472ms 0.4896ms 2.0424 KOps/s 2.0330 KOps/s $\color{#35bf28}+0.46\%$
test_func_call_cm_runtime[True-eager] 1.3291ms 1.2136ms 824.0087 Ops/s 828.4561 Ops/s $\color{#d91a1a}-0.54\%$
test_func_call_cm_runtime[True-compile] 0.9743ms 0.8524ms 1.1732 KOps/s 1.1843 KOps/s $\color{#d91a1a}-0.94\%$
test_func_call_cm_runtime[True-compile-overhead] 1.1482ms 0.8708ms 1.1484 KOps/s 1.1388 KOps/s $\color{#35bf28}+0.84\%$
test_vmap_func_call_cm_runtime[eager] 3.3436ms 2.6994ms 370.4491 Ops/s 370.2702 Ops/s $\color{#35bf28}+0.05\%$
test_vmap_func_call_cm_runtime[compile] 4.2035ms 3.6981ms 270.4059 Ops/s 273.3073 Ops/s $\color{#d91a1a}-1.06\%$
test_vmap_func_call_cm_runtime[compile-overhead] 4.1376ms 3.6541ms 273.6669 Ops/s 274.0456 Ops/s $\color{#d91a1a}-0.14\%$
test_distributed 0.6974ms 0.1451ms 6.8903 KOps/s 5.7088 KOps/s $\textbf{\color{#35bf28}+20.70\%}$
test_tdmodule 0.5002ms 0.1943ms 5.1473 KOps/s 5.1162 KOps/s $\color{#35bf28}+0.61\%$
test_tdmodule_dispatch 0.4042ms 0.3750ms 2.6668 KOps/s 2.6557 KOps/s $\color{#35bf28}+0.42\%$
test_tdseq 46.7310μs 27.0355μs 36.9883 KOps/s 36.5662 KOps/s $\color{#35bf28}+1.15\%$
test_tdseq_dispatch 0.2782ms 0.2209ms 4.5270 KOps/s 4.5882 KOps/s $\color{#d91a1a}-1.33\%$
test_instantiation_functorch 2.1861ms 2.0388ms 490.4868 Ops/s 494.5090 Ops/s $\color{#d91a1a}-0.81\%$
test_exec_functorch 0.2251ms 0.1735ms 5.7650 KOps/s 5.8261 KOps/s $\color{#d91a1a}-1.05\%$
test_exec_functional_call 0.4328ms 0.1540ms 6.4948 KOps/s 6.6028 KOps/s $\color{#d91a1a}-1.64\%$
test_exec_td_decorator 0.5054ms 0.2410ms 4.1500 KOps/s 4.1511 KOps/s $\color{#d91a1a}-0.03\%$
test_vmap_mlp_speed_decorator[True-True] 1.3047ms 1.1133ms 898.2309 Ops/s 894.3201 Ops/s $\color{#35bf28}+0.44\%$
test_vmap_mlp_speed_decorator[True-False] 1.3410ms 1.1104ms 900.6085 Ops/s 896.3512 Ops/s $\color{#35bf28}+0.47\%$
test_vmap_mlp_speed_decorator[False-True] 1.0516ms 0.8157ms 1.2259 KOps/s 1.2223 KOps/s $\color{#35bf28}+0.29\%$
test_vmap_mlp_speed_decorator[False-False] 1.0079ms 0.8126ms 1.2306 KOps/s 1.2249 KOps/s $\color{#35bf28}+0.47\%$
test_vmap_transformer_speed_decorator[True-True] 27.0929ms 24.9681ms 40.0511 Ops/s 39.6651 Ops/s $\color{#35bf28}+0.97\%$
test_vmap_transformer_speed_decorator[True-False] 25.7606ms 24.8907ms 40.1756 Ops/s 39.9865 Ops/s $\color{#35bf28}+0.47\%$
test_vmap_transformer_speed_decorator[False-True] 24.8969ms 24.5078ms 40.8033 Ops/s 40.4841 Ops/s $\color{#35bf28}+0.79\%$
test_vmap_transformer_speed_decorator[False-False] 24.5948ms 24.4345ms 40.9258 Ops/s 40.5813 Ops/s $\color{#35bf28}+0.85\%$
test_to_module_speed[True] 2.3007ms 1.6462ms 607.4629 Ops/s 606.1550 Ops/s $\color{#35bf28}+0.22\%$
test_to_module_speed[False] 2.2101ms 1.6282ms 614.1889 Ops/s 616.1415 Ops/s $\color{#d91a1a}-0.32\%$
test_to_module_plain_tensor_speed[None] 2.4442ms 1.7300ms 578.0438 Ops/s 573.6071 Ops/s $\color{#35bf28}+0.77\%$
test_to_module_plain_tensor_speed[False] 2.1865ms 1.5664ms 638.4161 Ops/s 637.2506 Ops/s $\color{#35bf28}+0.18\%$
test_to_module_plain_tensor_speed[True] 3.5010ms 3.3840ms 295.5093 Ops/s 295.9209 Ops/s $\color{#d91a1a}-0.14\%$
test_tc_init 76.8520μs 43.7882μs 22.8372 KOps/s 23.1300 KOps/s $\color{#d91a1a}-1.27\%$
test_tc_init_tensor_only 50.8510μs 9.3859μs 106.5432 KOps/s 105.6657 KOps/s $\color{#35bf28}+0.83\%$
test_tc_init_nested 0.1266ms 86.3908μs 11.5753 KOps/s 11.7585 KOps/s $\color{#d91a1a}-1.56\%$
test_tc_init_many_fields 45.7610μs 17.1224μs 58.4031 KOps/s 58.9980 KOps/s $\color{#d91a1a}-1.01\%$
test_tc_first_layer_tensor 29.1910μs 1.7874μs 559.4825 KOps/s 565.9812 KOps/s $\color{#d91a1a}-1.15\%$
test_tc_first_layer_tensor_only 2.1346μs 0.3980μs 2.5124 MOps/s 2.4880 MOps/s $\color{#35bf28}+0.98\%$
test_tc_first_layer_tensor_set 30.7100μs 4.2092μs 237.5738 KOps/s 239.2136 KOps/s $\color{#d91a1a}-0.69\%$
test_tc_first_layer_tensor_only_set 23.3010μs 3.4856μs 286.8908 KOps/s 285.8640 KOps/s $\color{#35bf28}+0.36\%$
test_tc_first_layer_nontensor 33.7010μs 6.0540μs 165.1808 KOps/s 160.7409 KOps/s $\color{#35bf28}+2.76\%$
test_tc_second_layer_tensor 27.3210μs 4.3156μs 231.7176 KOps/s 232.8522 KOps/s $\color{#d91a1a}-0.49\%$
test_tc_second_layer_nontensor 57.4710μs 8.2184μs 121.6783 KOps/s 115.3058 KOps/s $\textbf{\color{#35bf28}+5.53\%}$
test_unbind 0.2892s 13.9075ms 71.9038 Ops/s 64.6867 Ops/s $\textbf{\color{#35bf28}+11.16\%}$
test_full_like 5.0174ms 4.6982ms 212.8456 Ops/s 210.0867 Ops/s $\color{#35bf28}+1.31\%$
test_zeros_like 4.7930ms 4.6158ms 216.6459 Ops/s 215.0717 Ops/s $\color{#35bf28}+0.73\%$
test_ones_like 4.8348ms 4.6437ms 215.3444 Ops/s 212.7947 Ops/s $\color{#35bf28}+1.20\%$
test_clone 7.4277ms 6.9054ms 144.8135 Ops/s 141.8537 Ops/s $\color{#35bf28}+2.09\%$
test_squeeze 0.1828ms 14.0992μs 70.9263 KOps/s 66.7479 KOps/s $\textbf{\color{#35bf28}+6.26\%}$
test_unsqueeze 0.1747ms 0.1150ms 8.6947 KOps/s 8.6945 KOps/s $+0.00\%$
test_split 0.4202ms 0.1906ms 5.2464 KOps/s 5.0678 KOps/s $\color{#35bf28}+3.52\%$
test_permute 0.2593ms 0.1988ms 5.0306 KOps/s 4.9731 KOps/s $\color{#35bf28}+1.16\%$
test_stack 22.7322ms 20.4492ms 48.9017 Ops/s 50.5424 Ops/s $\color{#d91a1a}-3.25\%$
test_cat 20.7068ms 19.8200ms 50.4540 Ops/s 50.9136 Ops/s $\color{#d91a1a}-0.90\%$
test_sequential_tensordict 0.9618ms 0.8744ms 1.1437 KOps/s 1.1364 KOps/s $\color{#35bf28}+0.64\%$
test_sequential_graph_module 0.4608ms 0.4080ms 2.4507 KOps/s 2.4125 KOps/s $\color{#35bf28}+1.58\%$
test_nested_tensordict 2.5319ms 1.2260ms 815.6482 Ops/s 811.6106 Ops/s $\color{#35bf28}+0.50\%$
test_nested_graph_module 0.6283ms 0.5111ms 1.9567 KOps/s 1.8231 KOps/s $\textbf{\color{#35bf28}+7.33\%}$

@vmoens
Update
3800572 
[ghstack-poisoned]
vmoens added a commit that referenced this pull request Jul 9, 2026
@vmoens @claude
[Feature] Support return_early in TensorDict.consolidate
83bd403 
consolidate(num_threads>1, return_early=True) raised
NotImplementedError. Now that the threaded and sequential paths share
the storage-view rebuild step, the result can be constructed eagerly
(it only manipulates storage views and metadata) while the chunked
copies keep running in the worker threads, and a TensorDictFuture is
returned, mirroring memmap/memmap_/memmap_like.

Notes:

- With return_early=True the file-backed fused-copy fallback is
  skipped: the point of returning early is freeing the main thread,
  not raw throughput.
- return_early=True still raises for the two combinations that require
  main-thread work after the copies complete: use_buffer=True (final
  buffer-to-file write) and non_blocking=True (device sync).

Co-Authored-By: Claude Fable 5 <noreply@anthropic.com>
ghstack-source-id: a07c8a5
Pull-Request: #1742
@vmoens vmoens mentioned this pull request Jul 9, 2026
Merged
@vmoens
Update
5957bf3 
[ghstack-poisoned]
vmoens added a commit that referenced this pull request Jul 9, 2026
@vmoens @claude
[Feature] Support return_early in TensorDict.consolidate
2f22fbc 
consolidate(num_threads>1, return_early=True) raised
NotImplementedError. Now that the threaded and sequential paths share
the storage-view rebuild step, the result can be constructed eagerly
(it only manipulates storage views and metadata) while the chunked
copies keep running in the worker threads, and a TensorDictFuture is
returned, mirroring memmap/memmap_/memmap_like.

Notes:

- With return_early=True the file-backed fused-copy fallback is
  skipped: the point of returning early is freeing the main thread,
  not raw throughput.
- return_early=True still raises for the two combinations that require
  main-thread work after the copies complete: use_buffer=True (final
  buffer-to-file write) and non_blocking=True (device sync).

Co-Authored-By: Claude Fable 5 <noreply@anthropic.com>
ghstack-source-id: 67161fd
Pull-Request: #1742
@vmoens
vmoens merged commit 5957bf3 into gh/vmoens/95/base Jul 9, 2026
63 of 66 checks passed
@vmoens
vmoens deleted the gh/vmoens/95/head branch July 9, 2026 21:17
Sign up for free to join this conversation on GitHub. Already have an account? Sign in to comment

Reviewers

No reviews

Assignees

No one assigned

Labels

CLA Signed This label is managed by the Facebook bot. Authors need to sign the CLA before a PR can be reviewed. Feature New feature Test

Projects

None yet

Milestone

No milestone

Development

Successfully merging this pull request may close these issues.

1 participant

@vmoens