Skip to content

[Distributed] Add distributed benchmark script - #1592

Merged
vmoens merged 4 commits into
gh/vmoens/75/basefrom
gh/vmoens/75/head
Feb 19, 2026
Merged

[Distributed] Add distributed benchmark script#1592
vmoens merged 4 commits into
gh/vmoens/75/basefrom
gh/vmoens/75/head

Conversation

@vmoens

@vmoens vmoens commented Feb 18, 2026 •
edited
Loading

Copy link
Copy Markdown
Collaborator

Stack from ghstack (oldest at bottom):

Benchmark comparing leaf-by-leaf vs consolidated send/recv, broadcast,
all_reduce, and init_remote across various TD sizes. Reports latency
and speedup for each configuration.

Co-authored-by: Cursor cursoragent@cursor.com

@vmoens
Update
efc31ab 
[ghstack-poisoned]
vmoens added a commit that referenced this pull request Feb 18, 2026
@vmoens @cursoragent
[Distributed] Add distributed benchmark script
9fbaabe 
Benchmark comparing leaf-by-leaf vs consolidated send/recv, broadcast,
all_reduce, and init_remote across various TD sizes. Reports latency
and speedup for each configuration.

Co-authored-by: Cursor <cursoragent@cursor.com>
ghstack-source-id: 99241bf
Pull-Request: #1592
This was referenced Feb 18, 2026
Merged
Merged
Merged
Merged
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 Feb 18, 2026
@github-actions

github-actions Bot commented Feb 18, 2026 •
edited
Loading

Copy link
Copy Markdown
Contributor

$\color{#D29922}\textsf{\Large&amp;#x26A0;\kern{0.2cm}\normalsize Warning}$ Result of CPU Benchmark Tests

Total Benchmarks: 243. Improved: $\large\color{#35bf28}17$. Worsened: $\large\color{#d91a1a}9$.

Expand to view detailed results
Name Max Mean Ops Ops on Repo HEAD Change
test_plain_set_nested 34.9110μs 14.0664μs 71.0914 KOps/s 70.4387 KOps/s $\color{#35bf28}+0.93\%$
test_plain_set_stack_nested 35.2810μs 14.3011μs 69.9249 KOps/s 69.6554 KOps/s $\color{#35bf28}+0.39\%$
test_plain_set_nested_inplace 48.1810μs 15.7327μs 63.5618 KOps/s 63.2926 KOps/s $\color{#35bf28}+0.43\%$
test_plain_set_stack_nested_inplace 48.1410μs 15.7099μs 63.6541 KOps/s 63.6430 KOps/s $\color{#35bf28}+0.02\%$
test_items 32.0600μs 5.4901μs 182.1475 KOps/s 181.2353 KOps/s $\color{#35bf28}+0.50\%$
test_items_nested 0.5120ms 0.4378ms 2.2840 KOps/s 2.3123 KOps/s $\color{#d91a1a}-1.22\%$
test_items_nested_locked 0.7225ms 0.4432ms 2.2561 KOps/s 2.2949 KOps/s $\color{#d91a1a}-1.69\%$
test_items_nested_leaf 0.1476ms 91.0848μs 10.9788 KOps/s 10.8047 KOps/s $\color{#35bf28}+1.61\%$
test_items_stack_nested 0.5195ms 0.4400ms 2.2730 KOps/s 2.2891 KOps/s $\color{#d91a1a}-0.71\%$
test_items_stack_nested_leaf 0.1319ms 90.7057μs 11.0247 KOps/s 10.8070 KOps/s $\color{#35bf28}+2.01\%$
test_items_stack_nested_locked 0.5024ms 0.4412ms 2.2667 KOps/s 2.2767 KOps/s $\color{#d91a1a}-0.44\%$
test_keys 28.2400μs 4.3439μs 230.2078 KOps/s 229.7243 KOps/s $\color{#35bf28}+0.21\%$
test_keys_nested 0.1698ms 0.1162ms 8.6041 KOps/s 8.5537 KOps/s $\color{#35bf28}+0.59\%$
test_keys_nested_locked 88.1587ms 0.1377ms 7.2648 KOps/s 7.9320 KOps/s $\textbf{\color{#d91a1a}-8.41\%}$
test_keys_nested_leaf 0.1463ms 0.1072ms 9.3317 KOps/s 9.2829 KOps/s $\color{#35bf28}+0.53\%$
test_keys_stack_nested 0.2199ms 0.1173ms 8.5283 KOps/s 8.5553 KOps/s $\color{#d91a1a}-0.32\%$
test_keys_stack_nested_leaf 0.1726ms 0.1083ms 9.2360 KOps/s 9.2762 KOps/s $\color{#d91a1a}-0.43\%$
test_keys_stack_nested_locked 0.1918ms 0.1258ms 7.9482 KOps/s 8.0633 KOps/s $\color{#d91a1a}-1.43\%$
test_values 7.7522μs 0.9881μs 1.0120 MOps/s 1.0039 MOps/s $\color{#35bf28}+0.81\%$
test_values_nested 70.2110μs 46.2134μs 21.6387 KOps/s 21.7567 KOps/s $\color{#d91a1a}-0.54\%$
test_values_nested_locked 77.4620μs 49.1175μs 20.3593 KOps/s 20.3505 KOps/s $\color{#35bf28}+0.04\%$
test_values_nested_leaf 83.6620μs 52.5329μs 19.0357 KOps/s 19.1656 KOps/s $\color{#d91a1a}-0.68\%$
test_values_stack_nested 72.1010μs 46.3887μs 21.5570 KOps/s 21.6621 KOps/s $\color{#d91a1a}-0.49\%$
test_values_stack_nested_leaf 76.2410μs 52.5327μs 19.0358 KOps/s 19.0891 KOps/s $\color{#d91a1a}-0.28\%$
test_values_stack_nested_locked 75.2110μs 49.2288μs 20.3133 KOps/s 20.2599 KOps/s $\color{#35bf28}+0.26\%$
test_membership 5.8002μs 0.8036μs 1.2444 MOps/s 1.2763 MOps/s $\color{#d91a1a}-2.50\%$
test_membership_nested 26.4810μs 2.7541μs 363.0994 KOps/s 367.3861 KOps/s $\color{#d91a1a}-1.17\%$
test_membership_nested_leaf 29.5600μs 2.7675μs 361.3391 KOps/s 366.1913 KOps/s $\color{#d91a1a}-1.33\%$
test_membership_stacked_nested 31.7810μs 2.7560μs 362.8457 KOps/s 369.7980 KOps/s $\color{#d91a1a}-1.88\%$
test_membership_stacked_nested_leaf 27.3900μs 2.7945μs 357.8482 KOps/s 366.8566 KOps/s $\color{#d91a1a}-2.46\%$
test_membership_nested_last 28.3500μs 4.1141μs 243.0687 KOps/s 247.3362 KOps/s $\color{#d91a1a}-1.73\%$
test_membership_nested_leaf_last 44.7900μs 4.0533μs 246.7145 KOps/s 244.1224 KOps/s $\color{#35bf28}+1.06\%$
test_membership_stacked_nested_last 23.8000μs 4.1247μs 242.4407 KOps/s 243.2659 KOps/s $\color{#d91a1a}-0.34\%$
test_membership_stacked_nested_leaf_last 32.9700μs 4.1192μs 242.7638 KOps/s 243.8334 KOps/s $\color{#d91a1a}-0.44\%$
test_nested_getleaf 98.0310μs 20.4053μs 49.0069 KOps/s 49.0428 KOps/s $\color{#d91a1a}-0.07\%$
test_nested_get 44.9810μs 19.1123μs 52.3223 KOps/s 52.5504 KOps/s $\color{#d91a1a}-0.43\%$
test_stacked_getleaf 52.2710μs 20.2834μs 49.3014 KOps/s 49.4903 KOps/s $\color{#d91a1a}-0.38\%$
test_stacked_get 43.9400μs 19.3484μs 51.6837 KOps/s 51.0089 KOps/s $\color{#35bf28}+1.32\%$
test_nested_getitemleaf 42.0800μs 20.9274μs 47.7843 KOps/s 48.2857 KOps/s $\color{#d91a1a}-1.04\%$
test_nested_getitem 92.9210μs 19.8312μs 50.4257 KOps/s 50.1819 KOps/s $\color{#35bf28}+0.49\%$
test_stacked_getitemleaf 62.5710μs 20.2003μs 49.5042 KOps/s 47.5454 KOps/s $\color{#35bf28}+4.12\%$
test_stacked_getitem 46.7010μs 19.8214μs 50.4506 KOps/s 51.5609 KOps/s $\color{#d91a1a}-2.15\%$
test_lock_nested 8.2411ms 0.4557ms 2.1946 KOps/s 2.2016 KOps/s $\color{#d91a1a}-0.32\%$
test_lock_stack_nested 0.5186ms 0.4545ms 2.2004 KOps/s 2.1775 KOps/s $\color{#35bf28}+1.05\%$
test_unlock_nested 0.4901ms 0.3646ms 2.7428 KOps/s 2.7433 KOps/s $\color{#d91a1a}-0.02\%$
test_unlock_stack_nested 0.4751ms 0.3652ms 2.7379 KOps/s 2.6999 KOps/s $\color{#35bf28}+1.41\%$
test_flatten_speed 0.1566ms 0.1165ms 8.5874 KOps/s 8.6312 KOps/s $\color{#d91a1a}-0.51\%$
test_unflatten_speed 0.6147ms 0.5456ms 1.8328 KOps/s 1.8434 KOps/s $\color{#d91a1a}-0.58\%$
test_common_ops 0.7957ms 0.6649ms 1.5041 KOps/s 1.4875 KOps/s $\color{#35bf28}+1.11\%$
test_creation 0.1077ms 2.7515μs 363.4386 KOps/s 362.9052 KOps/s $\color{#35bf28}+0.15\%$
test_creation_empty 35.2910μs 5.7508μs 173.8887 KOps/s 174.6662 KOps/s $\color{#d91a1a}-0.45\%$
test_creation_nested_1 32.9710μs 10.0638μs 99.3665 KOps/s 101.1235 KOps/s $\color{#d91a1a}-1.74\%$
test_creation_nested_2 35.1910μs 11.4133μs 87.6170 KOps/s 90.1234 KOps/s $\color{#d91a1a}-2.78\%$
test_creation_many_keys[10] 56.5110μs 17.0149μs 58.7722 KOps/s 59.8045 KOps/s $\color{#d91a1a}-1.73\%$
test_creation_many_keys[50] 99.6520μs 73.0324μs 13.6926 KOps/s 14.0005 KOps/s $\color{#d91a1a}-2.20\%$
test_creation_many_keys[100] 0.2113ms 0.1427ms 7.0072 KOps/s 7.1465 KOps/s $\color{#d91a1a}-1.95\%$
test_creation_nested_many_keys[10] 65.3020μs 36.7316μs 27.2245 KOps/s 27.6528 KOps/s $\color{#d91a1a}-1.55\%$
test_creation_nested_many_keys[50] 0.1872ms 0.1484ms 6.7379 KOps/s 6.8247 KOps/s $\color{#d91a1a}-1.27\%$
test_clone 55.5610μs 12.7985μs 78.1341 KOps/s 76.7749 KOps/s $\color{#35bf28}+1.77\%$
test_getitem[int] 1.6750ms 13.9168μs 71.8554 KOps/s 58.4838 KOps/s $\textbf{\color{#35bf28}+22.86\%}$
test_getitem[slice_int] 0.1368ms 23.8058μs 42.0066 KOps/s 42.7288 KOps/s $\color{#d91a1a}-1.69\%$
test_getitem[range] 0.1780ms 59.8438μs 16.7102 KOps/s 16.3283 KOps/s $\color{#35bf28}+2.34\%$
test_getitem[tuple] 0.1449ms 22.7355μs 43.9840 KOps/s 43.9361 KOps/s $\color{#35bf28}+0.11\%$
test_getitem[list] 0.1817ms 55.8440μs 17.9070 KOps/s 17.7535 KOps/s $\color{#35bf28}+0.86\%$
test_setitem_dim[int] 50.4010μs 27.5192μs 36.3382 KOps/s 39.1250 KOps/s $\textbf{\color{#d91a1a}-7.12\%}$
test_setitem_dim[slice_int] 67.1510μs 44.8204μs 22.3113 KOps/s 23.6008 KOps/s $\textbf{\color{#d91a1a}-5.46\%}$
test_setitem_dim[range] 0.1146ms 91.0952μs 10.9775 KOps/s 10.9144 KOps/s $\color{#35bf28}+0.58\%$
test_setitem_dim[tuple] 73.3020μs 38.3434μs 26.0801 KOps/s 25.5212 KOps/s $\color{#35bf28}+2.19\%$
test_setitem 53.5910μs 16.9948μs 58.8416 KOps/s 57.1009 KOps/s $\color{#35bf28}+3.05\%$
test_set 48.2310μs 16.2729μs 61.4520 KOps/s 60.3090 KOps/s $\color{#35bf28}+1.90\%$
test_set_shared 0.5712ms 0.2119ms 4.7183 KOps/s 4.8919 KOps/s $\color{#d91a1a}-3.55\%$
test_update 0.2115ms 22.0879μs 45.2736 KOps/s 47.3409 KOps/s $\color{#d91a1a}-4.37\%$
test_update_nested 67.1410μs 33.2775μs 30.0504 KOps/s 31.0611 KOps/s $\color{#d91a1a}-3.25\%$
test_update__nested 0.5056ms 34.7669μs 28.7630 KOps/s 29.6169 KOps/s $\color{#d91a1a}-2.88\%$
test_set_nested 56.1110μs 19.6064μs 51.0038 KOps/s 54.7286 KOps/s $\textbf{\color{#d91a1a}-6.81\%}$
test_set_nested_new 56.1510μs 22.7604μs 43.9359 KOps/s 42.1259 KOps/s $\color{#35bf28}+4.30\%$
test_select 69.1510μs 39.2282μs 25.4919 KOps/s 25.1768 KOps/s $\color{#35bf28}+1.25\%$
test_select_nested 97.4020μs 69.8597μs 14.3144 KOps/s 14.2828 KOps/s $\color{#35bf28}+0.22\%$
test_exclude_nested 0.1197ms 86.3642μs 11.5789 KOps/s 11.5603 KOps/s $\color{#35bf28}+0.16\%$
test_empty[True] 0.4283ms 0.3764ms 2.6570 KOps/s 2.7169 KOps/s $\color{#d91a1a}-2.21\%$
test_empty[False] 19.3628μs 1.2537μs 797.6570 KOps/s 790.1337 KOps/s $\color{#35bf28}+0.95\%$
test_to 0.1004ms 69.5998μs 14.3679 KOps/s 13.5582 KOps/s $\textbf{\color{#35bf28}+5.97\%}$
test_to_nonblocking 0.1065ms 65.3769μs 15.2959 KOps/s 15.9020 KOps/s $\color{#d91a1a}-3.81\%$
test_unbind_speed 0.3651ms 0.3100ms 3.2257 KOps/s 3.2317 KOps/s $\color{#d91a1a}-0.19\%$
test_unbind_speed_stack0 0.4393ms 0.2999ms 3.3346 KOps/s 3.2841 KOps/s $\color{#35bf28}+1.54\%$
test_unbind_speed_stack1 0.1025s 0.8781ms 1.1388 KOps/s 1.0031 KOps/s $\textbf{\color{#35bf28}+13.53\%}$
test_split 1.1670ms 1.0943ms 913.8061 Ops/s 918.5199 Ops/s $\color{#d91a1a}-0.51\%$
test_chunk 0.1024s 1.1555ms 865.4167 Ops/s 778.1075 Ops/s $\textbf{\color{#35bf28}+11.22\%}$
test_to_cpu_blocking 19.4402ms 19.3534ms 51.6706 Ops/s 47.7691 Ops/s $\textbf{\color{#35bf28}+8.17\%}$
test_to_cpu_global_sync 11.3335ms 11.0678ms 90.3521 Ops/s 90.5345 Ops/s $\color{#d91a1a}-0.20\%$
test_to_cpu_event_sync 0.1144s 13.2603ms 75.4131 Ops/s 83.1264 Ops/s $\textbf{\color{#d91a1a}-9.28\%}$
test_to_cpu_default 12.1399ms 12.0041ms 83.3052 Ops/s 74.9951 Ops/s $\textbf{\color{#35bf28}+11.08\%}$
test_consolidate[False-None] 4.1938ms 3.9924ms 250.4767 Ops/s 251.8804 Ops/s $\color{#d91a1a}-0.56\%$
test_consolidate[default-None] 2.0321ms 1.9292ms 518.3484 Ops/s 499.8568 Ops/s $\color{#35bf28}+3.70\%$
test_consolidate[reduce-overhead-None] 1.9621ms 1.8675ms 535.4882 Ops/s 521.6729 Ops/s $\color{#35bf28}+2.65\%$
test_consolidate_njt[False-None] 8.5243ms 8.1233ms 123.1026 Ops/s 121.4173 Ops/s $\color{#35bf28}+1.39\%$
test_to[False-False-None] 2.1763ms 2.0335ms 491.7727 Ops/s 490.2943 Ops/s $\color{#35bf28}+0.30\%$
test_to[True-False-None] 2.0897ms 1.9151ms 522.1658 Ops/s 528.2890 Ops/s $\color{#d91a1a}-1.16\%$
test_to[within-False-None] 6.2166ms 5.9029ms 169.4087 Ops/s 169.3064 Ops/s $\color{#35bf28}+0.06\%$
test_to[True-default-None] 0.1763s 8.5091ms 117.5211 Ops/s 130.5589 Ops/s $\textbf{\color{#d91a1a}-9.99\%}$
test_to_njt[False-False-None] 8.4034ms 8.2524ms 121.1771 Ops/s 119.6660 Ops/s $\color{#35bf28}+1.26\%$
test_to_njt[True-False-None] 7.3130ms 6.8303ms 146.4059 Ops/s 146.4987 Ops/s $\color{#d91a1a}-0.06\%$
test_to_njt[within-False-None] 16.2792ms 15.5464ms 64.3237 Ops/s 65.9486 Ops/s $\color{#d91a1a}-2.46\%$
test_creation[device0] 0.3355ms 0.1145ms 8.7325 KOps/s 8.6603 KOps/s $\color{#35bf28}+0.83\%$
test_creation_from_tensor 0.4144ms 0.1170ms 8.5491 KOps/s 8.7161 KOps/s $\color{#d91a1a}-1.92\%$
test_add_one[memmap_tensor0] 0.2398ms 6.4183μs 155.8040 KOps/s 152.0378 KOps/s $\color{#35bf28}+2.48\%$
test_contiguous[memmap_tensor0] 20.7910μs 0.6313μs 1.5839 MOps/s 2.1654 MOps/s $\textbf{\color{#d91a1a}-26.85\%}$
test_stack[memmap_tensor0] 59.8210μs 4.5099μs 221.7363 KOps/s 219.1814 KOps/s $\color{#35bf28}+1.17\%$
test_memmaptd_index 1.1049ms 0.2528ms 3.9551 KOps/s 3.9835 KOps/s $\color{#d91a1a}-0.71\%$
test_memmaptd_index_astensor 0.4899ms 0.3388ms 2.9515 KOps/s 2.9394 KOps/s $\color{#35bf28}+0.41\%$
test_memmaptd_index_op 0.8199ms 0.5747ms 1.7401 KOps/s 1.6999 KOps/s $\color{#35bf28}+2.36\%$
test_serialize_model 0.1392s 0.1369s 7.3023 Ops/s 7.3107 Ops/s $\color{#d91a1a}-0.12\%$
test_serialize_model_pickle 1.3721s 1.1978s 0.8348 Ops/s 0.8260 Ops/s $\color{#35bf28}+1.07\%$
test_serialize_weights 0.1363s 0.1348s 7.4204 Ops/s 7.3788 Ops/s $\color{#35bf28}+0.56\%$
test_serialize_weights_returnearly 0.4117s 90.6697ms 11.0290 Ops/s 14.5091 Ops/s $\textbf{\color{#d91a1a}-23.99\%}$
test_serialize_weights_pickle 1.3626s 1.2131s 0.8243 Ops/s 0.8187 Ops/s $\color{#35bf28}+0.69\%$
test_reshape_pytree 0.2010ms 32.5392μs 30.7322 KOps/s 30.4717 KOps/s $\color{#35bf28}+0.85\%$
test_reshape_td 0.4481ms 42.4952μs 23.5321 KOps/s 23.3679 KOps/s $\color{#35bf28}+0.70\%$
test_view_pytree 0.2323ms 31.7891μs 31.4573 KOps/s 30.6128 KOps/s $\color{#35bf28}+2.76\%$
test_view_td 0.4611ms 48.9759μs 20.4182 KOps/s 19.5864 KOps/s $\color{#35bf28}+4.25\%$
test_unbind_pytree 0.2446ms 35.0158μs 28.5585 KOps/s 28.0545 KOps/s $\color{#35bf28}+1.80\%$
test_unbind_td 0.4698ms 46.1826μs 21.6532 KOps/s 21.4427 KOps/s $\color{#35bf28}+0.98\%$
test_split_pytree 0.2520ms 41.8581μs 23.8902 KOps/s 24.3142 KOps/s $\color{#d91a1a}-1.74\%$
test_split_td 0.4873ms 62.9087μs 15.8960 KOps/s 16.0765 KOps/s $\color{#d91a1a}-1.12\%$
test_add_pytree 0.2348ms 40.8122μs 24.5025 KOps/s 24.2224 KOps/s $\color{#35bf28}+1.16\%$
test_add_td 0.4756ms 51.2926μs 19.4960 KOps/s 19.4106 KOps/s $\color{#35bf28}+0.44\%$
test_compile_add_one_nested[tensordict-compile] 0.5838ms 0.1353ms 7.3929 KOps/s 6.8787 KOps/s $\textbf{\color{#35bf28}+7.48\%}$
test_compile_add_one_nested[tensordict-eager] 0.5089ms 0.1879ms 5.3233 KOps/s 5.3710 KOps/s $\color{#d91a1a}-0.89\%$
test_compile_add_one_nested[pytree-compile] 0.4493ms 0.1067ms 9.3689 KOps/s 9.2549 KOps/s $\color{#35bf28}+1.23\%$
test_compile_add_one_nested[pytree-eager] 0.4332ms 0.1762ms 5.6741 KOps/s 5.6035 KOps/s $\color{#35bf28}+1.26\%$
test_compile_copy_nested[tensordict-compile] 0.2041ms 28.1867μs 35.4777 KOps/s 33.1164 KOps/s $\textbf{\color{#35bf28}+7.13\%}$
test_compile_copy_nested[tensordict-eager] 87.7410μs 49.6018μs 20.1606 KOps/s 20.3268 KOps/s $\color{#d91a1a}-0.82\%$
test_compile_copy_nested[pytree-compile] 0.1105ms 9.3765μs 106.6497 KOps/s 103.9187 KOps/s $\color{#35bf28}+2.63\%$
test_compile_copy_nested[pytree-eager] 0.4702ms 65.7319μs 15.2133 KOps/s 15.0670 KOps/s $\color{#35bf28}+0.97\%$
test_compile_add_one_flat[tensordict-compile] 0.4720ms 0.1735ms 5.7637 KOps/s 5.7263 KOps/s $\color{#35bf28}+0.65\%$
test_compile_add_one_flat[tensordict-eager] 0.3322ms 0.2499ms 4.0016 KOps/s 3.9470 KOps/s $\color{#35bf28}+1.38\%$
test_compile_add_one_flat[tensorclass-compile] 0.2954ms 0.1131ms 8.8429 KOps/s 8.6968 KOps/s $\color{#35bf28}+1.68\%$
test_compile_add_one_flat[tensorclass-eager] 0.1138ms 68.2087μs 14.6609 KOps/s 14.6812 KOps/s $\color{#d91a1a}-0.14\%$
test_compile_add_one_flat[pytree-compile] 0.2108ms 0.1552ms 6.4427 KOps/s 6.2976 KOps/s $\color{#35bf28}+2.30\%$
test_compile_add_one_flat[pytree-eager] 0.8090ms 0.5286ms 1.8918 KOps/s 1.8984 KOps/s $\color{#d91a1a}-0.34\%$
test_compile_add_self_flat[tensordict-eager] 0.4637ms 0.3027ms 3.3039 KOps/s 3.2622 KOps/s $\color{#35bf28}+1.28\%$
test_compile_add_self_flat[tensordict-compile] 0.2378ms 0.1749ms 5.7174 KOps/s 5.2679 KOps/s $\textbf{\color{#35bf28}+8.53\%}$
test_compile_add_self_flat[tensorclass-eager] 0.1264ms 83.7085μs 11.9462 KOps/s 11.9649 KOps/s $\color{#d91a1a}-0.16\%$
test_compile_add_self_flat[tensorclass-compile] 0.2033ms 0.1154ms 8.6665 KOps/s 8.2159 KOps/s $\textbf{\color{#35bf28}+5.48\%}$
test_compile_add_self_flat[pytree-eager] 0.1751s 0.5240ms 1.9083 KOps/s 2.2443 KOps/s $\textbf{\color{#d91a1a}-14.97\%}$
test_compile_add_self_flat[pytree-compile] 0.3004ms 0.1569ms 6.3753 KOps/s 6.2190 KOps/s $\color{#35bf28}+2.51\%$
test_compile_copy_flat[tensordict-compile] 95.2010μs 22.9282μs 43.6144 KOps/s 37.8296 KOps/s $\textbf{\color{#35bf28}+15.29\%}$
test_compile_copy_flat[tensordict-eager] 83.4320μs 40.3992μs 24.7530 KOps/s 25.0457 KOps/s $\color{#d91a1a}-1.17\%$
test_compile_copy_flat[pytree-compile] 0.1800ms 10.3851μs 96.2921 KOps/s 93.3166 KOps/s $\color{#35bf28}+3.19\%$
test_compile_copy_flat[pytree-eager] 0.4179ms 51.0695μs 19.5811 KOps/s 19.6183 KOps/s $\color{#d91a1a}-0.19\%$
test_compile_assign_and_add[tensordict-compile] 1.9342ms 0.1716ms 5.8263 KOps/s 5.6210 KOps/s $\color{#35bf28}+3.65\%$
test_compile_assign_and_add[tensordict-eager] 3.3855ms 3.2580ms 306.9378 Ops/s 309.6704 Ops/s $\color{#d91a1a}-0.88\%$
test_compile_assign_and_add[pytree-compile] 1.8957ms 0.1581ms 6.3246 KOps/s 6.1434 KOps/s $\color{#35bf28}+2.95\%$
test_compile_assign_and_add[pytree-eager] 2.9259ms 2.7859ms 358.9472 Ops/s 360.1698 Ops/s $\color{#d91a1a}-0.34\%$
test_compile_indexing[tensor-tensordict-compile] 0.1708ms 0.1060ms 9.4326 KOps/s 9.2385 KOps/s $\color{#35bf28}+2.10\%$
test_compile_indexing[tensor-tensordict-eager] 0.3100ms 70.3789μs 14.2088 KOps/s 13.8600 KOps/s $\color{#35bf28}+2.52\%$
test_compile_indexing[tensor-tensorclass-compile] 0.2197ms 95.8165μs 10.4366 KOps/s 10.6072 KOps/s $\color{#d91a1a}-1.61\%$
test_compile_indexing[tensor-tensorclass-eager] 0.2530ms 44.1427μs 22.6538 KOps/s 22.6348 KOps/s $\color{#35bf28}+0.08\%$
test_compile_indexing[tensor-pytree-compile] 0.1421ms 94.7831μs 10.5504 KOps/s 10.4732 KOps/s $\color{#35bf28}+0.74\%$
test_compile_indexing[tensor-pytree-eager] 0.2580ms 44.0727μs 22.6898 KOps/s 22.6874 KOps/s $\color{#35bf28}+0.01\%$
test_compile_indexing[slice-tensordict-compile] 0.2041ms 54.0627μs 18.4971 KOps/s 17.5344 KOps/s $\textbf{\color{#35bf28}+5.49\%}$
test_compile_indexing[slice-tensordict-eager] 0.2273ms 26.4611μs 37.7913 KOps/s 36.9429 KOps/s $\color{#35bf28}+2.30\%$
test_compile_indexing[slice-tensorclass-compile] 0.1725ms 43.6275μs 22.9213 KOps/s 22.2423 KOps/s $\color{#35bf28}+3.05\%$
test_compile_indexing[slice-tensorclass-eager] 0.2553ms 21.6911μs 46.1019 KOps/s 45.7723 KOps/s $\color{#35bf28}+0.72\%$
test_compile_indexing[slice-pytree-compile] 0.1048ms 43.7205μs 22.8725 KOps/s 21.8786 KOps/s $\color{#35bf28}+4.54\%$
test_compile_indexing[slice-pytree-eager] 0.2921ms 21.4699μs 46.5769 KOps/s 45.8204 KOps/s $\color{#35bf28}+1.65\%$
test_compile_indexing[int-tensordict-compile] 0.1085ms 55.0307μs 18.1717 KOps/s 17.2195 KOps/s $\textbf{\color{#35bf28}+5.53\%}$
test_compile_indexing[int-tensordict-eager] 0.2758ms 26.4278μs 37.8390 KOps/s 37.9979 KOps/s $\color{#d91a1a}-0.42\%$
test_compile_indexing[int-tensorclass-compile] 81.9510μs 43.7854μs 22.8386 KOps/s 21.7385 KOps/s $\textbf{\color{#35bf28}+5.06\%}$
test_compile_indexing[int-tensorclass-eager] 0.2594ms 21.5888μs 46.3203 KOps/s 44.8693 KOps/s $\color{#35bf28}+3.23\%$
test_compile_indexing[int-pytree-compile] 83.3010μs 43.9962μs 22.7292 KOps/s 21.9089 KOps/s $\color{#35bf28}+3.74\%$
test_compile_indexing[int-pytree-eager] 0.2632ms 21.3079μs 46.9310 KOps/s 45.6379 KOps/s $\color{#35bf28}+2.83\%$
test_mod_add[eager] 0.1056ms 48.3484μs 20.6832 KOps/s 20.7050 KOps/s $\color{#d91a1a}-0.11\%$
test_mod_add[compile] 0.2340ms 0.1016ms 9.8456 KOps/s 9.5101 KOps/s $\color{#35bf28}+3.53\%$
test_mod_add[compile-overhead] 0.2267ms 0.1434ms 6.9743 KOps/s 6.6853 KOps/s $\color{#35bf28}+4.32\%$
test_mod_wrap[eager] 0.3912ms 0.2847ms 3.5126 KOps/s 3.5255 KOps/s $\color{#d91a1a}-0.37\%$
test_mod_wrap[compile] 0.4308ms 0.3391ms 2.9492 KOps/s 2.8835 KOps/s $\color{#35bf28}+2.28\%$
test_mod_wrap[compile-overhead] 7.4529ms 4.0876ms 244.6432 Ops/s 247.4820 Ops/s $\color{#d91a1a}-1.15\%$
test_mod_wrap_and_backward[eager] 1.5681ms 1.4743ms 678.2827 Ops/s 674.7862 Ops/s $\color{#35bf28}+0.52\%$
test_mod_wrap_and_backward[compile] 1.5188ms 1.4167ms 705.8801 Ops/s 701.8515 Ops/s $\color{#35bf28}+0.57\%$
test_mod_wrap_and_backward[compile-overhead] 1.2442ms 0.8620ms 1.1601 KOps/s 1.1323 KOps/s $\color{#35bf28}+2.45\%$
test_seq_add[eager] 0.2298ms 0.1529ms 6.5414 KOps/s 6.5254 KOps/s $\color{#35bf28}+0.25\%$
test_seq_add[compile] 0.1499ms 0.1113ms 8.9844 KOps/s 8.3155 KOps/s $\textbf{\color{#35bf28}+8.04\%}$
test_seq_add[compile-overhead] 0.2232ms 0.1495ms 6.6908 KOps/s 6.3533 KOps/s $\textbf{\color{#35bf28}+5.31\%}$
test_seq_wrap[eager] 0.5903ms 0.5119ms 1.9536 KOps/s 1.9298 KOps/s $\color{#35bf28}+1.24\%$
test_seq_wrap[compile] 0.4144ms 0.3567ms 2.8036 KOps/s 2.7342 KOps/s $\color{#35bf28}+2.54\%$
test_seq_wrap[compile-overhead] 0.3268ms 0.2574ms 3.8846 KOps/s 3.7486 KOps/s $\color{#35bf28}+3.63\%$
test_func_call_runtime[False-eager] 0.9148ms 0.8279ms 1.2078 KOps/s 1.2150 KOps/s $\color{#d91a1a}-0.59\%$
test_func_call_runtime[False-compile] 1.0028ms 0.8860ms 1.1287 KOps/s 1.1282 KOps/s $\color{#35bf28}+0.04\%$
test_func_call_runtime[False-compile-overhead] 0.5123ms 0.4431ms 2.2566 KOps/s 2.2189 KOps/s $\color{#35bf28}+1.70\%$
test_func_call_runtime[True-eager] 1.1511ms 1.0587ms 944.5766 Ops/s 942.4404 Ops/s $\color{#35bf28}+0.23\%$
test_func_call_runtime[True-compile] 1.3396ms 0.8941ms 1.1185 KOps/s 1.0983 KOps/s $\color{#35bf28}+1.84\%$
test_func_call_runtime[True-compile-overhead] 0.5385ms 0.4553ms 2.1964 KOps/s 2.1368 KOps/s $\color{#35bf28}+2.79\%$
test_func_call_cm_runtime[False-eager] 0.9248ms 0.8264ms 1.2101 KOps/s 1.1782 KOps/s $\color{#35bf28}+2.71\%$
test_func_call_cm_runtime[False-compile] 0.9675ms 0.8752ms 1.1425 KOps/s 1.0960 KOps/s $\color{#35bf28}+4.25\%$
test_func_call_cm_runtime[False-compile-overhead] 0.5048ms 0.4453ms 2.2459 KOps/s 2.2311 KOps/s $\color{#35bf28}+0.67\%$
test_func_call_cm_runtime[True-eager] 1.3098ms 1.1986ms 834.2730 Ops/s 833.9750 Ops/s $\color{#35bf28}+0.04\%$
test_func_call_cm_runtime[True-compile] 1.0489ms 0.9235ms 1.0828 KOps/s 1.0652 KOps/s $\color{#35bf28}+1.65\%$
test_func_call_cm_runtime[True-compile-overhead] 0.8546ms 0.4910ms 2.0367 KOps/s 2.0139 KOps/s $\color{#35bf28}+1.13\%$
test_vmap_func_call_cm_runtime[eager] 2.8145ms 2.3326ms 428.7012 Ops/s 427.6713 Ops/s $\color{#35bf28}+0.24\%$
test_vmap_func_call_cm_runtime[compile] 1.0858ms 0.9421ms 1.0615 KOps/s 1.0463 KOps/s $\color{#35bf28}+1.45\%$
test_vmap_func_call_cm_runtime[compile-overhead] 0.5606ms 0.4899ms 2.0411 KOps/s 1.9816 KOps/s $\color{#35bf28}+3.00\%$
test_distributed 2.5974ms 0.1616ms 6.1887 KOps/s 6.4727 KOps/s $\color{#d91a1a}-4.39\%$
test_tdmodule 0.4197ms 27.5744μs 36.2656 KOps/s 37.4568 KOps/s $\color{#d91a1a}-3.18\%$
test_tdmodule_dispatch 76.3210μs 44.4879μs 22.4780 KOps/s 22.5014 KOps/s $\color{#d91a1a}-0.10\%$
test_tdseq 60.4410μs 26.5407μs 37.6780 KOps/s 37.4326 KOps/s $\color{#35bf28}+0.66\%$
test_tdseq_dispatch 66.0910μs 46.2299μs 21.6310 KOps/s 21.4044 KOps/s $\color{#35bf28}+1.06\%$
test_instantiation_functorch 2.0774ms 1.9829ms 504.3086 Ops/s 503.9477 Ops/s $\color{#35bf28}+0.07\%$
test_exec_functorch 0.2315ms 0.1735ms 5.7653 KOps/s 5.7379 KOps/s $\color{#35bf28}+0.48\%$
test_exec_functional_call 0.2233ms 0.1547ms 6.4624 KOps/s 6.4316 KOps/s $\color{#35bf28}+0.48\%$
test_exec_td_decorator 0.4423ms 0.2332ms 4.2877 KOps/s 4.3708 KOps/s $\color{#d91a1a}-1.90\%$
test_vmap_mlp_speed_decorator[True-True] 0.9795ms 0.8074ms 1.2385 KOps/s 1.2337 KOps/s $\color{#35bf28}+0.39\%$
test_vmap_mlp_speed_decorator[True-False] 1.0286ms 0.8081ms 1.2375 KOps/s 1.2423 KOps/s $\color{#d91a1a}-0.38\%$
test_vmap_mlp_speed_decorator[False-True] 0.9000ms 0.7011ms 1.4264 KOps/s 1.4250 KOps/s $\color{#35bf28}+0.10\%$
test_vmap_mlp_speed_decorator[False-False] 0.9103ms 0.7031ms 1.4223 KOps/s 1.4295 KOps/s $\color{#d91a1a}-0.50\%$
test_vmap_transformer_speed_decorator[True-True] 21.2250ms 20.3290ms 49.1909 Ops/s 49.2557 Ops/s $\color{#d91a1a}-0.13\%$
test_vmap_transformer_speed_decorator[True-False] 21.1238ms 20.3399ms 49.1646 Ops/s 49.3539 Ops/s $\color{#d91a1a}-0.38\%$
test_vmap_transformer_speed_decorator[False-True] 20.2415ms 20.0765ms 49.8094 Ops/s 49.7289 Ops/s $\color{#35bf28}+0.16\%$
test_vmap_transformer_speed_decorator[False-False] 20.6262ms 20.1644ms 49.5924 Ops/s 49.6717 Ops/s $\color{#d91a1a}-0.16\%$
test_to_module_speed[True] 1.4887ms 1.4089ms 709.7925 Ops/s 708.6220 Ops/s $\color{#35bf28}+0.17\%$
test_to_module_speed[False] 1.4824ms 1.3889ms 719.9937 Ops/s 725.7452 Ops/s $\color{#d91a1a}-0.79\%$
test_tc_init 77.8720μs 43.6742μs 22.8968 KOps/s 22.7790 KOps/s $\color{#35bf28}+0.52\%$
test_tc_init_tensor_only 42.7610μs 9.3623μs 106.8113 KOps/s 107.5690 KOps/s $\color{#d91a1a}-0.70\%$
test_tc_init_nested 0.1260ms 86.9406μs 11.5021 KOps/s 11.3615 KOps/s $\color{#35bf28}+1.24\%$
test_tc_init_many_fields 53.6110μs 15.7940μs 63.3152 KOps/s 64.1484 KOps/s $\color{#d91a1a}-1.30\%$
test_tc_first_layer_tensor 36.0300μs 1.7532μs 570.3964 KOps/s 586.4208 KOps/s $\color{#d91a1a}-2.73\%$
test_tc_first_layer_tensor_only 4.6730μs 0.7156μs 1.3973 MOps/s 1.4002 MOps/s $\color{#d91a1a}-0.20\%$
test_tc_first_layer_tensor_set 44.6200μs 3.7456μs 266.9781 KOps/s 265.6453 KOps/s $\color{#35bf28}+0.50\%$
test_tc_first_layer_tensor_only_set 22.1800μs 2.9832μs 335.2068 KOps/s 337.1415 KOps/s $\color{#d91a1a}-0.57\%$
test_tc_first_layer_nontensor 37.8000μs 5.8474μs 171.0166 KOps/s 171.2278 KOps/s $\color{#d91a1a}-0.12\%$
test_tc_second_layer_tensor 25.0200μs 4.1323μs 241.9956 KOps/s 239.7148 KOps/s $\color{#35bf28}+0.95\%$
test_tc_second_layer_nontensor 85.4810μs 8.1866μs 122.1510 KOps/s 120.7470 KOps/s $\color{#35bf28}+1.16\%$
test_unbind 0.2419s 14.2061ms 70.3923 Ops/s 56.5558 Ops/s $\textbf{\color{#35bf28}+24.47\%}$
test_full_like 5.3433ms 4.3929ms 227.6378 Ops/s 228.8184 Ops/s $\color{#d91a1a}-0.52\%$
test_zeros_like 4.9471ms 4.3668ms 229.0029 Ops/s 229.3778 Ops/s $\color{#d91a1a}-0.16\%$
test_ones_like 4.4659ms 4.3698ms 228.8423 Ops/s 228.5930 Ops/s $\color{#35bf28}+0.11\%$
test_clone 6.6619ms 6.4840ms 154.2246 Ops/s 153.4678 Ops/s $\color{#35bf28}+0.49\%$
test_squeeze 0.2190ms 13.8078μs 72.4230 KOps/s 72.9423 KOps/s $\color{#d91a1a}-0.71\%$
test_unsqueeze 0.1600ms 0.1126ms 8.8782 KOps/s 9.2380 KOps/s $\color{#d91a1a}-3.90\%$
test_split 0.2581ms 0.1835ms 5.4502 KOps/s 5.5069 KOps/s $\color{#d91a1a}-1.03\%$
test_permute 0.2862ms 0.2126ms 4.7033 KOps/s 4.8875 KOps/s $\color{#d91a1a}-3.77\%$
test_stack 51.9632ms 51.6299ms 19.3686 Ops/s 19.4004 Ops/s $\color{#d91a1a}-0.16\%$
test_cat 51.9518ms 51.5876ms 19.3845 Ops/s 19.4478 Ops/s $\color{#d91a1a}-0.33\%$

@github-actions

github-actions Bot commented Feb 18, 2026 •
edited
Loading

Copy link
Copy Markdown
Contributor

$\color{#D29922}\textsf{\Large&amp;#x26A0;\kern{0.2cm}\normalsize Warning}$ Result of GPU Benchmark Tests

Total Benchmarks: 243. Improved: $\large\color{#35bf28}11$. Worsened: $\large\color{#d91a1a}11$.

Expand to view detailed results
Name Max Mean Ops Ops on Repo HEAD Change
test_plain_set_nested 57.1810μs 14.5481μs 68.7374 KOps/s 68.2881 KOps/s $\color{#35bf28}+0.66\%$
test_plain_set_stack_nested 38.5110μs 14.7930μs 67.5994 KOps/s 67.1345 KOps/s $\color{#35bf28}+0.69\%$
test_plain_set_nested_inplace 43.4310μs 16.2824μs 61.4158 KOps/s 61.4332 KOps/s $\color{#d91a1a}-0.03\%$
test_plain_set_stack_nested_inplace 58.7910μs 16.2538μs 61.5242 KOps/s 61.1762 KOps/s $\color{#35bf28}+0.57\%$
test_items 40.2510μs 5.7714μs 173.2667 KOps/s 168.5438 KOps/s $\color{#35bf28}+2.80\%$
test_items_nested 0.4981ms 0.4524ms 2.2106 KOps/s 2.1832 KOps/s $\color{#35bf28}+1.25\%$
test_items_nested_locked 0.6128ms 0.4512ms 2.2165 KOps/s 2.1599 KOps/s $\color{#35bf28}+2.62\%$
test_items_nested_leaf 0.1295ms 94.0997μs 10.6270 KOps/s 10.4220 KOps/s $\color{#35bf28}+1.97\%$
test_items_stack_nested 0.5154ms 0.4480ms 2.2323 KOps/s 2.1884 KOps/s $\color{#35bf28}+2.00\%$
test_items_stack_nested_leaf 0.1360ms 96.9772μs 10.3117 KOps/s 10.3976 KOps/s $\color{#d91a1a}-0.83\%$
test_items_stack_nested_locked 0.5221ms 0.4535ms 2.2049 KOps/s 2.1378 KOps/s $\color{#35bf28}+3.14\%$
test_keys 41.1410μs 4.2217μs 236.8707 KOps/s 237.8805 KOps/s $\color{#d91a1a}-0.42\%$
test_keys_nested 0.1650ms 0.1195ms 8.3711 KOps/s 8.3648 KOps/s $\color{#35bf28}+0.08\%$
test_keys_nested_locked 92.7930ms 0.1418ms 7.0502 KOps/s 7.8081 KOps/s $\textbf{\color{#d91a1a}-9.71\%}$
test_keys_nested_leaf 0.1615ms 0.1100ms 9.0876 KOps/s 9.0451 KOps/s $\color{#35bf28}+0.47\%$
test_keys_stack_nested 0.1824ms 0.1205ms 8.2966 KOps/s 8.3723 KOps/s $\color{#d91a1a}-0.90\%$
test_keys_stack_nested_leaf 0.1508ms 0.1097ms 9.1194 KOps/s 9.0395 KOps/s $\color{#35bf28}+0.88\%$
test_keys_stack_nested_locked 0.3008ms 0.1285ms 7.7807 KOps/s 7.8312 KOps/s $\color{#d91a1a}-0.64\%$
test_values 12.9362μs 1.0223μs 978.2070 KOps/s 980.3603 KOps/s $\color{#d91a1a}-0.22\%$
test_values_nested 84.2010μs 47.4492μs 21.0752 KOps/s 20.8487 KOps/s $\color{#35bf28}+1.09\%$
test_values_nested_locked 80.9010μs 50.9299μs 19.6348 KOps/s 19.6194 KOps/s $\color{#35bf28}+0.08\%$
test_values_nested_leaf 96.6210μs 54.1434μs 18.4695 KOps/s 18.4550 KOps/s $\color{#35bf28}+0.08\%$
test_values_stack_nested 98.5510μs 47.0449μs 21.2563 KOps/s 21.0928 KOps/s $\color{#35bf28}+0.77\%$
test_values_stack_nested_leaf 95.5710μs 54.1221μs 18.4767 KOps/s 18.4602 KOps/s $\color{#35bf28}+0.09\%$
test_values_stack_nested_locked 88.3110μs 51.1807μs 19.5386 KOps/s 19.6838 KOps/s $\color{#d91a1a}-0.74\%$
test_membership 10.7402μs 0.8357μs 1.1967 MOps/s 982.8943 KOps/s $\textbf{\color{#35bf28}+21.75\%}$
test_membership_nested 30.8210μs 2.9178μs 342.7200 KOps/s 337.4040 KOps/s $\color{#35bf28}+1.58\%$
test_membership_nested_leaf 32.8710μs 2.9383μs 340.3287 KOps/s 341.1086 KOps/s $\color{#d91a1a}-0.23\%$
test_membership_stacked_nested 32.9500μs 2.9151μs 343.0395 KOps/s 338.0490 KOps/s $\color{#35bf28}+1.48\%$
test_membership_stacked_nested_leaf 58.5300μs 2.9202μs 342.4443 KOps/s 341.3736 KOps/s $\color{#35bf28}+0.31\%$
test_membership_nested_last 34.8510μs 4.3343μs 230.7175 KOps/s 226.1165 KOps/s $\color{#35bf28}+2.03\%$
test_membership_nested_leaf_last 29.8600μs 4.3285μs 231.0244 KOps/s 228.5739 KOps/s $\color{#35bf28}+1.07\%$
test_membership_stacked_nested_last 48.0200μs 4.3017μs 232.4645 KOps/s 227.5529 KOps/s $\color{#35bf28}+2.16\%$
test_membership_stacked_nested_leaf_last 39.6910μs 4.3180μs 231.5876 KOps/s 227.1297 KOps/s $\color{#35bf28}+1.96\%$
test_nested_getleaf 46.7910μs 21.4623μs 46.5934 KOps/s 46.1760 KOps/s $\color{#35bf28}+0.90\%$
test_nested_get 71.7910μs 20.1930μs 49.5222 KOps/s 48.6173 KOps/s $\color{#35bf28}+1.86\%$
test_stacked_getleaf 48.8400μs 21.4292μs 46.6652 KOps/s 46.5650 KOps/s $\color{#35bf28}+0.22\%$
test_stacked_get 43.8510μs 20.5689μs 48.6172 KOps/s 48.7398 KOps/s $\color{#d91a1a}-0.25\%$
test_nested_getitemleaf 47.6400μs 21.9043μs 45.6531 KOps/s 45.5851 KOps/s $\color{#35bf28}+0.15\%$
test_nested_getitem 50.2510μs 20.6374μs 48.4558 KOps/s 48.1907 KOps/s $\color{#35bf28}+0.55\%$
test_stacked_getitemleaf 50.7200μs 21.8636μs 45.7381 KOps/s 45.3543 KOps/s $\color{#35bf28}+0.85\%$
test_stacked_getitem 55.0410μs 21.0131μs 47.5893 KOps/s 47.3535 KOps/s $\color{#35bf28}+0.50\%$
test_lock_nested 8.0103ms 0.4785ms 2.0899 KOps/s 2.0763 KOps/s $\color{#35bf28}+0.65\%$
test_lock_stack_nested 0.5219ms 0.4774ms 2.0948 KOps/s 2.0521 KOps/s $\color{#35bf28}+2.08\%$
test_unlock_nested 0.4587ms 0.3813ms 2.6224 KOps/s 2.5996 KOps/s $\color{#35bf28}+0.88\%$
test_unlock_stack_nested 0.4167ms 0.3820ms 2.6179 KOps/s 2.5476 KOps/s $\color{#35bf28}+2.76\%$
test_flatten_speed 0.1636ms 0.1208ms 8.2797 KOps/s 8.1963 KOps/s $\color{#35bf28}+1.02\%$
test_unflatten_speed 0.6228ms 0.5679ms 1.7607 KOps/s 1.7491 KOps/s $\color{#35bf28}+0.66\%$
test_common_ops 0.8050ms 0.6812ms 1.4679 KOps/s 1.4573 KOps/s $\color{#35bf28}+0.73\%$
test_creation 79.1810μs 2.9174μs 342.7689 KOps/s 341.1657 KOps/s $\color{#35bf28}+0.47\%$
test_creation_empty 29.8500μs 6.1401μs 162.8645 KOps/s 161.1477 KOps/s $\color{#35bf28}+1.07\%$
test_creation_nested_1 35.3410μs 10.4973μs 95.2627 KOps/s 94.9725 KOps/s $\color{#35bf28}+0.31\%$
test_creation_nested_2 36.3810μs 11.9082μs 83.9756 KOps/s 83.4443 KOps/s $\color{#35bf28}+0.64\%$
test_creation_many_keys[10] 45.6710μs 18.4061μs 54.3298 KOps/s 54.7932 KOps/s $\color{#d91a1a}-0.85\%$
test_creation_many_keys[50] 0.1127ms 78.0061μs 12.8195 KOps/s 12.8602 KOps/s $\color{#d91a1a}-0.32\%$
test_creation_many_keys[100] 0.1920ms 0.1525ms 6.5566 KOps/s 6.5683 KOps/s $\color{#d91a1a}-0.18\%$
test_creation_nested_many_keys[10] 95.9720μs 37.9176μs 26.3730 KOps/s 25.5944 KOps/s $\color{#35bf28}+3.04\%$
test_creation_nested_many_keys[50] 0.1899ms 0.1587ms 6.3005 KOps/s 6.3093 KOps/s $\color{#d91a1a}-0.14\%$
test_clone 43.3300μs 13.2382μs 75.5388 KOps/s 72.8683 KOps/s $\color{#35bf28}+3.66\%$
test_getitem[int] 1.7300ms 14.5624μs 68.6698 KOps/s 56.4278 KOps/s $\textbf{\color{#35bf28}+21.70\%}$
test_getitem[slice_int] 0.1467ms 24.4960μs 40.8230 KOps/s 40.4058 KOps/s $\color{#35bf28}+1.03\%$
test_getitem[range] 0.1686ms 62.2168μs 16.0728 KOps/s 14.8164 KOps/s $\textbf{\color{#35bf28}+8.48\%}$
test_getitem[tuple] 0.1393ms 23.9380μs 41.7746 KOps/s 41.3861 KOps/s $\color{#35bf28}+0.94\%$
test_getitem[list] 0.1846ms 57.9607μs 17.2531 KOps/s 16.0813 KOps/s $\textbf{\color{#35bf28}+7.29\%}$
test_setitem_dim[int] 59.9510μs 25.8060μs 38.7507 KOps/s 37.5010 KOps/s $\color{#35bf28}+3.33\%$
test_setitem_dim[slice_int] 65.0110μs 42.3235μs 23.6275 KOps/s 21.8940 KOps/s $\textbf{\color{#35bf28}+7.92\%}$
test_setitem_dim[range] 0.1380ms 93.4982μs 10.6954 KOps/s 10.5771 KOps/s $\color{#35bf28}+1.12\%$
test_setitem_dim[tuple] 59.3910μs 39.7009μs 25.1883 KOps/s 24.7399 KOps/s $\color{#35bf28}+1.81\%$
test_setitem 47.6600μs 17.8488μs 56.0262 KOps/s 55.0308 KOps/s $\color{#35bf28}+1.81\%$
test_set 49.0800μs 16.9769μs 58.9035 KOps/s 58.4045 KOps/s $\color{#35bf28}+0.85\%$
test_set_shared 0.5056ms 0.2076ms 4.8173 KOps/s 4.7893 KOps/s $\color{#35bf28}+0.58\%$
test_update 0.3568ms 21.9784μs 45.4993 KOps/s 46.2367 KOps/s $\color{#d91a1a}-1.59\%$
test_update_nested 73.8810μs 33.5106μs 29.8413 KOps/s 29.7613 KOps/s $\color{#35bf28}+0.27\%$
test_update__nested 0.5548ms 35.0052μs 28.5672 KOps/s 28.5228 KOps/s $\color{#35bf28}+0.16\%$
test_set_nested 49.6110μs 18.7886μs 53.2237 KOps/s 50.4864 KOps/s $\textbf{\color{#35bf28}+5.42\%}$
test_set_nested_new 67.5510μs 23.4684μs 42.6105 KOps/s 41.2887 KOps/s $\color{#35bf28}+3.20\%$
test_select 80.6010μs 42.7160μs 23.4104 KOps/s 24.0454 KOps/s $\color{#d91a1a}-2.64\%$
test_select_nested 0.1087ms 73.8431μs 13.5422 KOps/s 13.4276 KOps/s $\color{#35bf28}+0.85\%$
test_exclude_nested 0.1309ms 90.6873μs 11.0269 KOps/s 10.9540 KOps/s $\color{#35bf28}+0.67\%$
test_empty[True] 0.4643ms 0.3815ms 2.6213 KOps/s 2.5826 KOps/s $\color{#35bf28}+1.50\%$
test_empty[False] 7.4750μs 1.3239μs 755.3611 KOps/s 764.2013 KOps/s $\color{#d91a1a}-1.16\%$
test_to 0.1084ms 72.2498μs 13.8409 KOps/s 13.4449 KOps/s $\color{#35bf28}+2.95\%$
test_to_nonblocking 0.1077ms 65.1125μs 15.3580 KOps/s 15.3386 KOps/s $\color{#35bf28}+0.13\%$
test_unbind_speed 0.3596ms 0.3285ms 3.0446 KOps/s 3.0087 KOps/s $\color{#35bf28}+1.19\%$
test_unbind_speed_stack0 0.3662ms 0.3250ms 3.0765 KOps/s 3.0543 KOps/s $\color{#35bf28}+0.73\%$
test_unbind_speed_stack1 0.1027s 0.8391ms 1.1918 KOps/s 1.1768 KOps/s $\color{#35bf28}+1.27\%$
test_split 0.1029s 1.2724ms 785.9358 Ops/s 780.5690 Ops/s $\color{#35bf28}+0.69\%$
test_chunk 0.1035s 1.2132ms 824.2404 Ops/s 920.7283 Ops/s $\textbf{\color{#d91a1a}-10.48\%}$
test_to_cpu_blocking 28.6803ms 28.3096ms 35.3237 Ops/s 45.3643 Ops/s $\textbf{\color{#d91a1a}-22.13\%}$
test_to_cpu_global_sync 11.2211ms 11.0916ms 90.1585 Ops/s 79.8557 Ops/s $\textbf{\color{#35bf28}+12.90\%}$
test_to_cpu_event_sync 12.2795ms 11.9272ms 83.8418 Ops/s 82.3619 Ops/s $\color{#35bf28}+1.80\%$
test_to_cpu_default 0.1192s 13.2602ms 75.4139 Ops/s 82.0791 Ops/s $\textbf{\color{#d91a1a}-8.12\%}$
test_consolidate[False-None] 4.1644ms 4.1129ms 243.1399 Ops/s 219.6862 Ops/s $\textbf{\color{#35bf28}+10.68\%}$
test_consolidate[default-None] 2.4365ms 2.0204ms 494.9524 Ops/s 480.7218 Ops/s $\color{#35bf28}+2.96\%$
test_consolidate[reduce-overhead-None] 2.0901ms 1.9461ms 513.8522 Ops/s 502.0567 Ops/s $\color{#35bf28}+2.35\%$
test_consolidate_njt[False-None] 8.6054ms 8.4334ms 118.5763 Ops/s 117.0782 Ops/s $\color{#35bf28}+1.28\%$
test_to[False-False-None] 2.2500ms 2.0754ms 481.8454 Ops/s 474.9566 Ops/s $\color{#35bf28}+1.45\%$
test_to[True-False-None] 2.1534ms 1.9167ms 521.7329 Ops/s 519.9147 Ops/s $\color{#35bf28}+0.35\%$
test_to[within-False-None] 6.3749ms 6.0947ms 164.0776 Ops/s 162.6744 Ops/s $\color{#35bf28}+0.86\%$
test_to[True-default-None] 7.7019ms 7.5421ms 132.5895 Ops/s 130.4438 Ops/s $\color{#35bf28}+1.64\%$
test_to_njt[False-False-None] 8.6068ms 8.4822ms 117.8940 Ops/s 116.9972 Ops/s $\color{#35bf28}+0.77\%$
test_to_njt[True-False-None] 7.0598ms 6.9339ms 144.2198 Ops/s 143.6744 Ops/s $\color{#35bf28}+0.38\%$
test_to_njt[within-False-None] 15.5947ms 15.4805ms 64.5974 Ops/s 64.1983 Ops/s $\color{#35bf28}+0.62\%$
test_creation[device0] 0.4490ms 0.1162ms 8.6077 KOps/s 8.3266 KOps/s $\color{#35bf28}+3.38\%$
test_creation_from_tensor 0.4421ms 0.1147ms 8.7188 KOps/s 8.5567 KOps/s $\color{#35bf28}+1.89\%$
test_add_one[memmap_tensor0] 0.1797ms 6.4679μs 154.6106 KOps/s 152.2048 KOps/s $\color{#35bf28}+1.58\%$
test_contiguous[memmap_tensor0] 22.5100μs 0.6812μs 1.4679 MOps/s 2.1678 MOps/s $\textbf{\color{#d91a1a}-32.29\%}$
test_stack[memmap_tensor0] 32.6310μs 4.6994μs 212.7926 KOps/s 215.1892 KOps/s $\color{#d91a1a}-1.11\%$
test_memmaptd_index 1.0279ms 0.2621ms 3.8153 KOps/s 3.7995 KOps/s $\color{#35bf28}+0.42\%$
test_memmaptd_index_astensor 0.5083ms 0.3578ms 2.7945 KOps/s 2.8046 KOps/s $\color{#d91a1a}-0.36\%$
test_memmaptd_index_op 0.7586ms 0.5954ms 1.6795 KOps/s 1.6594 KOps/s $\color{#35bf28}+1.21\%$
test_serialize_model 0.1400s 0.1376s 7.2681 Ops/s 7.2972 Ops/s $\color{#d91a1a}-0.40\%$
test_serialize_model_pickle 1.3502s 1.2100s 0.8264 Ops/s 0.8225 Ops/s $\color{#35bf28}+0.47\%$
test_serialize_weights 0.1403s 0.1359s 7.3596 Ops/s 7.3416 Ops/s $\color{#35bf28}+0.24\%$
test_serialize_weights_returnearly 0.3970s 90.4382ms 11.0573 Ops/s 10.5470 Ops/s $\color{#35bf28}+4.84\%$
test_serialize_weights_pickle 1.3637s 1.2131s 0.8244 Ops/s 0.8235 Ops/s $\color{#35bf28}+0.11\%$
test_reshape_pytree 0.2168ms 33.3251μs 30.0074 KOps/s 30.1701 KOps/s $\color{#d91a1a}-0.54\%$
test_reshape_td 81.8110μs 44.7086μs 22.3670 KOps/s 23.2374 KOps/s $\color{#d91a1a}-3.75\%$
test_view_pytree 0.2281ms 32.8125μs 30.4761 KOps/s 30.4103 KOps/s $\color{#35bf28}+0.22\%$
test_view_td 84.0310μs 52.1442μs 19.1776 KOps/s 19.2591 KOps/s $\color{#d91a1a}-0.42\%$
test_unbind_pytree 0.2394ms 36.3936μs 27.4774 KOps/s 27.0616 KOps/s $\color{#35bf28}+1.54\%$
test_unbind_td 0.1214ms 48.2909μs 20.7078 KOps/s 19.9965 KOps/s $\color{#35bf28}+3.56\%$
test_split_pytree 0.2554ms 42.1773μs 23.7094 KOps/s 23.4516 KOps/s $\color{#35bf28}+1.10\%$
test_split_td 0.1515ms 64.7416μs 15.4460 KOps/s 15.1964 KOps/s $\color{#35bf28}+1.64\%$
test_add_pytree 0.2551ms 41.4557μs 24.1222 KOps/s 24.0027 KOps/s $\color{#35bf28}+0.50\%$
test_add_td 89.7010μs 52.7210μs 18.9678 KOps/s 18.4183 KOps/s $\color{#35bf28}+2.98\%$
test_compile_add_one_nested[tensordict-compile] 0.2021ms 0.1401ms 7.1401 KOps/s 6.7959 KOps/s $\textbf{\color{#35bf28}+5.06\%}$
test_compile_add_one_nested[tensordict-eager] 0.4889ms 0.1920ms 5.2088 KOps/s 5.2989 KOps/s $\color{#d91a1a}-1.70\%$
test_compile_add_one_nested[pytree-compile] 0.1653ms 0.1092ms 9.1536 KOps/s 9.0035 KOps/s $\color{#35bf28}+1.67\%$
test_compile_add_one_nested[pytree-eager] 0.4396ms 0.1785ms 5.6008 KOps/s 5.5179 KOps/s $\color{#35bf28}+1.50\%$
test_compile_copy_nested[tensordict-compile] 0.2190ms 29.5172μs 33.8785 KOps/s 30.8949 KOps/s $\textbf{\color{#35bf28}+9.66\%}$
test_compile_copy_nested[tensordict-eager] 90.9410μs 52.5926μs 19.0141 KOps/s 18.9861 KOps/s $\color{#35bf28}+0.15\%$
test_compile_copy_nested[pytree-compile] 0.1474ms 10.1050μs 98.9609 KOps/s 102.6486 KOps/s $\color{#d91a1a}-3.59\%$
test_compile_copy_nested[pytree-eager] 0.4720ms 68.3792μs 14.6243 KOps/s 14.5822 KOps/s $\color{#35bf28}+0.29\%$
test_compile_add_one_flat[tensordict-compile] 0.2307ms 0.1779ms 5.6201 KOps/s 5.4270 KOps/s $\color{#35bf28}+3.56\%$
test_compile_add_one_flat[tensordict-eager] 0.3253ms 0.2557ms 3.9111 KOps/s 3.9474 KOps/s $\color{#d91a1a}-0.92\%$
test_compile_add_one_flat[tensorclass-compile] 0.2468ms 0.1172ms 8.5316 KOps/s 8.1833 KOps/s $\color{#35bf28}+4.26\%$
test_compile_add_one_flat[tensorclass-eager] 0.1257ms 69.3284μs 14.4241 KOps/s 13.8418 KOps/s $\color{#35bf28}+4.21\%$
test_compile_add_one_flat[pytree-compile] 0.2017ms 0.1602ms 6.2427 KOps/s 6.1095 KOps/s $\color{#35bf28}+2.18\%$
test_compile_add_one_flat[pytree-eager] 0.8026ms 0.5243ms 1.9071 KOps/s 1.8623 KOps/s $\color{#35bf28}+2.41\%$
test_compile_add_self_flat[tensordict-eager] 0.3844ms 0.3073ms 3.2545 KOps/s 3.2523 KOps/s $\color{#35bf28}+0.07\%$
test_compile_add_self_flat[tensordict-compile] 0.4277ms 0.2285ms 4.3756 KOps/s 5.3088 KOps/s $\textbf{\color{#d91a1a}-17.58\%}$
test_compile_add_self_flat[tensorclass-eager] 0.1575ms 85.6908μs 11.6699 KOps/s 11.4211 KOps/s $\color{#35bf28}+2.18\%$
test_compile_add_self_flat[tensorclass-compile] 0.2955ms 0.1204ms 8.3089 KOps/s 8.1090 KOps/s $\color{#35bf28}+2.47\%$
test_compile_add_self_flat[pytree-eager] 0.7139ms 0.4357ms 2.2950 KOps/s 2.2758 KOps/s $\color{#35bf28}+0.84\%$
test_compile_add_self_flat[pytree-compile] 0.2400ms 0.1613ms 6.1981 KOps/s 6.1100 KOps/s $\color{#35bf28}+1.44\%$
test_compile_copy_flat[tensordict-compile] 92.3310μs 24.2066μs 41.3110 KOps/s 40.5272 KOps/s $\color{#35bf28}+1.93\%$
test_compile_copy_flat[tensordict-eager] 75.5610μs 41.8269μs 23.9081 KOps/s 23.7159 KOps/s $\color{#35bf28}+0.81\%$
test_compile_copy_flat[pytree-compile] 0.1002ms 10.8361μs 92.2845 KOps/s 90.3947 KOps/s $\color{#35bf28}+2.09\%$
test_compile_copy_flat[pytree-eager] 0.4081ms 52.1616μs 19.1712 KOps/s 19.1902 KOps/s $\color{#d91a1a}-0.10\%$
test_compile_assign_and_add[tensordict-compile] 2.0108ms 0.1745ms 5.7299 KOps/s 5.5181 KOps/s $\color{#35bf28}+3.84\%$
test_compile_assign_and_add[tensordict-eager] 3.4187ms 3.2839ms 304.5137 Ops/s 302.0961 Ops/s $\color{#35bf28}+0.80\%$
test_compile_assign_and_add[pytree-compile] 1.9985ms 0.1609ms 6.2144 KOps/s 6.0308 KOps/s $\color{#35bf28}+3.04\%$
test_compile_assign_and_add[pytree-eager] 2.8921ms 2.7656ms 361.5799 Ops/s 355.8009 Ops/s $\color{#35bf28}+1.62\%$
test_compile_indexing[tensor-tensordict-compile] 0.1821ms 0.1106ms 9.0401 KOps/s 8.9709 KOps/s $\color{#35bf28}+0.77\%$
test_compile_indexing[tensor-tensordict-eager] 0.3233ms 71.6624μs 13.9543 KOps/s 13.5918 KOps/s $\color{#35bf28}+2.67\%$
test_compile_indexing[tensor-tensorclass-compile] 0.1989ms 0.1020ms 9.8083 KOps/s 10.1460 KOps/s $\color{#d91a1a}-3.33\%$
test_compile_indexing[tensor-tensorclass-eager] 0.2655ms 47.7077μs 20.9610 KOps/s 22.1545 KOps/s $\textbf{\color{#d91a1a}-5.39\%}$
test_compile_indexing[tensor-pytree-compile] 0.1729ms 0.1024ms 9.7623 KOps/s 10.0360 KOps/s $\color{#d91a1a}-2.73\%$
test_compile_indexing[tensor-pytree-eager] 0.2628ms 47.7825μs 20.9282 KOps/s 22.1461 KOps/s $\textbf{\color{#d91a1a}-5.50\%}$
test_compile_indexing[slice-tensordict-compile] 0.1174ms 56.3077μs 17.7596 KOps/s 17.6125 KOps/s $\color{#35bf28}+0.83\%$
test_compile_indexing[slice-tensordict-eager] 0.2301ms 27.7479μs 36.0388 KOps/s 35.4845 KOps/s $\color{#35bf28}+1.56\%$
test_compile_indexing[slice-tensorclass-compile] 0.1086ms 45.6100μs 21.9250 KOps/s 22.0577 KOps/s $\color{#d91a1a}-0.60\%$
test_compile_indexing[slice-tensorclass-eager] 0.2632ms 22.6003μs 44.2473 KOps/s 44.2866 KOps/s $\color{#d91a1a}-0.09\%$
test_compile_indexing[slice-pytree-compile] 84.8110μs 45.7640μs 21.8512 KOps/s 21.4648 KOps/s $\color{#35bf28}+1.80\%$
test_compile_indexing[slice-pytree-eager] 0.3368ms 22.5890μs 44.2693 KOps/s 44.3284 KOps/s $\color{#d91a1a}-0.13\%$
test_compile_indexing[int-tensordict-compile] 92.6510μs 56.1228μs 17.8181 KOps/s 17.3779 KOps/s $\color{#35bf28}+2.53\%$
test_compile_indexing[int-tensordict-eager] 0.2408ms 28.3241μs 35.3057 KOps/s 35.7571 KOps/s $\color{#d91a1a}-1.26\%$
test_compile_indexing[int-tensorclass-compile] 81.9510μs 45.9451μs 21.7651 KOps/s 21.2680 KOps/s $\color{#35bf28}+2.34\%$
test_compile_indexing[int-tensorclass-eager] 0.2609ms 22.3857μs 44.6713 KOps/s 44.3077 KOps/s $\color{#35bf28}+0.82\%$
test_compile_indexing[int-pytree-compile] 0.1158ms 45.3906μs 22.0310 KOps/s 21.2862 KOps/s $\color{#35bf28}+3.50\%$
test_compile_indexing[int-pytree-eager] 0.2786ms 22.5831μs 44.2809 KOps/s 44.5003 KOps/s $\color{#d91a1a}-0.49\%$
test_mod_add[eager] 0.1113ms 51.0000μs 19.6079 KOps/s 20.0205 KOps/s $\color{#d91a1a}-2.06\%$
test_mod_add[compile] 0.2069ms 0.1037ms 9.6473 KOps/s 9.3963 KOps/s $\color{#35bf28}+2.67\%$
test_mod_add[compile-overhead] 0.2308ms 0.1478ms 6.7662 KOps/s 6.4479 KOps/s $\color{#35bf28}+4.94\%$
test_mod_wrap[eager] 0.3609ms 0.2865ms 3.4903 KOps/s 3.4429 KOps/s $\color{#35bf28}+1.38\%$
test_mod_wrap[compile] 0.4113ms 0.3465ms 2.8857 KOps/s 2.8415 KOps/s $\color{#35bf28}+1.55\%$
test_mod_wrap[compile-overhead] 7.3607ms 4.0550ms 246.6090 Ops/s 247.1324 Ops/s $\color{#d91a1a}-0.21\%$
test_mod_wrap_and_backward[eager] 1.6144ms 1.4842ms 673.7816 Ops/s 671.0764 Ops/s $\color{#35bf28}+0.40\%$
test_mod_wrap_and_backward[compile] 2.2406ms 1.4841ms 673.7999 Ops/s 679.8854 Ops/s $\color{#d91a1a}-0.90\%$
test_mod_wrap_and_backward[compile-overhead] 1.4440ms 0.9908ms 1.0093 KOps/s 1.1100 KOps/s $\textbf{\color{#d91a1a}-9.07\%}$
test_seq_add[eager] 0.2270ms 0.1576ms 6.3471 KOps/s 6.4688 KOps/s $\color{#d91a1a}-1.88\%$
test_seq_add[compile] 0.1861ms 0.1176ms 8.5048 KOps/s 8.5169 KOps/s $\color{#d91a1a}-0.14\%$
test_seq_add[compile-overhead] 0.3324ms 0.1572ms 6.3614 KOps/s 6.2972 KOps/s $\color{#35bf28}+1.02\%$
test_seq_wrap[eager] 0.6242ms 0.5227ms 1.9133 KOps/s 1.9309 KOps/s $\color{#d91a1a}-0.91\%$
test_seq_wrap[compile] 0.4260ms 0.3664ms 2.7290 KOps/s 2.6648 KOps/s $\color{#35bf28}+2.41\%$
test_seq_wrap[compile-overhead] 0.3532ms 0.2686ms 3.7224 KOps/s 3.7113 KOps/s $\color{#35bf28}+0.30\%$
test_func_call_runtime[False-eager] 0.9672ms 0.8878ms 1.1264 KOps/s 1.1922 KOps/s $\textbf{\color{#d91a1a}-5.52\%}$
test_func_call_runtime[False-compile] 0.9866ms 0.9100ms 1.0989 KOps/s 1.0948 KOps/s $\color{#35bf28}+0.38\%$
test_func_call_runtime[False-compile-overhead] 0.5388ms 0.4631ms 2.1593 KOps/s 2.1381 KOps/s $\color{#35bf28}+0.99\%$
test_func_call_runtime[True-eager] 1.2889ms 1.0814ms 924.7238 Ops/s 918.1853 Ops/s $\color{#35bf28}+0.71\%$
test_func_call_runtime[True-compile] 0.9959ms 0.9203ms 1.0867 KOps/s 1.0766 KOps/s $\color{#35bf28}+0.93\%$
test_func_call_runtime[True-compile-overhead] 0.5739ms 0.4819ms 2.0752 KOps/s 2.0782 KOps/s $\color{#d91a1a}-0.14\%$
test_func_call_cm_runtime[False-eager] 0.9569ms 0.8710ms 1.1481 KOps/s 1.1935 KOps/s $\color{#d91a1a}-3.80\%$
test_func_call_cm_runtime[False-compile] 1.1069ms 0.9345ms 1.0701 KOps/s 1.0911 KOps/s $\color{#d91a1a}-1.92\%$
test_func_call_cm_runtime[False-compile-overhead] 0.5615ms 0.4671ms 2.1411 KOps/s 2.1324 KOps/s $\color{#35bf28}+0.41\%$
test_func_call_cm_runtime[True-eager] 1.3540ms 1.2530ms 798.0941 Ops/s 807.8485 Ops/s $\color{#d91a1a}-1.21\%$
test_func_call_cm_runtime[True-compile] 1.0461ms 0.9542ms 1.0481 KOps/s 1.0389 KOps/s $\color{#35bf28}+0.89\%$
test_func_call_cm_runtime[True-compile-overhead] 0.5673ms 0.5067ms 1.9736 KOps/s 1.9465 KOps/s $\color{#35bf28}+1.39\%$
test_vmap_func_call_cm_runtime[eager] 2.8291ms 2.3471ms 426.0552 Ops/s 419.4404 Ops/s $\color{#35bf28}+1.58\%$
test_vmap_func_call_cm_runtime[compile] 1.0864ms 1.0164ms 983.8624 Ops/s 1.0177 KOps/s $\color{#d91a1a}-3.32\%$
test_vmap_func_call_cm_runtime[compile-overhead] 0.5677ms 0.5095ms 1.9628 KOps/s 1.9265 KOps/s $\color{#35bf28}+1.88\%$
test_distributed 0.5481ms 0.1524ms 6.5601 KOps/s 6.2583 KOps/s $\color{#35bf28}+4.82\%$
test_tdmodule 80.9910μs 27.7037μs 36.0963 KOps/s 35.6873 KOps/s $\color{#35bf28}+1.15\%$
test_tdmodule_dispatch 75.3610μs 44.7910μs 22.3259 KOps/s 21.9177 KOps/s $\color{#35bf28}+1.86\%$
test_tdseq 45.0210μs 26.8589μs 37.2316 KOps/s 36.9914 KOps/s $\color{#35bf28}+0.65\%$
test_tdseq_dispatch 66.9710μs 46.9570μs 21.2961 KOps/s 21.2480 KOps/s $\color{#35bf28}+0.23\%$
test_instantiation_functorch 2.1363ms 2.0677ms 483.6262 Ops/s 481.9214 Ops/s $\color{#35bf28}+0.35\%$
test_exec_functorch 0.2275ms 0.1792ms 5.5791 KOps/s 5.4682 KOps/s $\color{#35bf28}+2.03\%$
test_exec_functional_call 0.1970ms 0.1595ms 6.2686 KOps/s 6.1211 KOps/s $\color{#35bf28}+2.41\%$
test_exec_td_decorator 0.4349ms 0.2349ms 4.2570 KOps/s 4.1814 KOps/s $\color{#35bf28}+1.81\%$
test_vmap_mlp_speed_decorator[True-True] 1.0030ms 0.8167ms 1.2245 KOps/s 1.2116 KOps/s $\color{#35bf28}+1.06\%$
test_vmap_mlp_speed_decorator[True-False] 0.9899ms 0.8131ms 1.2298 KOps/s 1.2141 KOps/s $\color{#35bf28}+1.30\%$
test_vmap_mlp_speed_decorator[False-True] 0.8772ms 0.7048ms 1.4188 KOps/s 1.4018 KOps/s $\color{#35bf28}+1.21\%$
test_vmap_mlp_speed_decorator[False-False] 0.8750ms 0.7053ms 1.4178 KOps/s 1.4001 KOps/s $\color{#35bf28}+1.26\%$
test_vmap_transformer_speed_decorator[True-True] 21.2019ms 20.4666ms 48.8602 Ops/s 48.4233 Ops/s $\color{#35bf28}+0.90\%$
test_vmap_transformer_speed_decorator[True-False] 21.1625ms 20.4698ms 48.8525 Ops/s 48.4171 Ops/s $\color{#35bf28}+0.90\%$
test_vmap_transformer_speed_decorator[False-True] 20.4097ms 20.2520ms 49.3777 Ops/s 48.8937 Ops/s $\color{#35bf28}+0.99\%$
test_vmap_transformer_speed_decorator[False-False] 20.4158ms 20.2748ms 49.3223 Ops/s 48.8769 Ops/s $\color{#35bf28}+0.91\%$
test_to_module_speed[True] 1.5653ms 1.4824ms 674.5989 Ops/s 668.3374 Ops/s $\color{#35bf28}+0.94\%$
test_to_module_speed[False] 1.5780ms 1.4434ms 692.8321 Ops/s 686.1020 Ops/s $\color{#35bf28}+0.98\%$
test_tc_init 77.4310μs 45.0138μs 22.2154 KOps/s 21.5021 KOps/s $\color{#35bf28}+3.32\%$
test_tc_init_tensor_only 36.7400μs 9.8323μs 101.7059 KOps/s 100.3924 KOps/s $\color{#35bf28}+1.31\%$
test_tc_init_nested 0.1426ms 89.7859μs 11.1376 KOps/s 10.7872 KOps/s $\color{#35bf28}+3.25\%$
test_tc_init_many_fields 43.4310μs 16.6697μs 59.9891 KOps/s 60.1053 KOps/s $\color{#d91a1a}-0.19\%$
test_tc_first_layer_tensor 20.1400μs 1.8474μs 541.3132 KOps/s 530.6089 KOps/s $\color{#35bf28}+2.02\%$
test_tc_first_layer_tensor_only 6.1001μs 0.7484μs 1.3362 MOps/s 1.3153 MOps/s $\color{#35bf28}+1.59\%$
test_tc_first_layer_tensor_set 33.0110μs 3.9467μs 253.3775 KOps/s 253.8627 KOps/s $\color{#d91a1a}-0.19\%$
test_tc_first_layer_tensor_only_set 26.4800μs 3.1364μs 318.8344 KOps/s 316.1648 KOps/s $\color{#35bf28}+0.84\%$
test_tc_first_layer_nontensor 34.9000μs 6.1463μs 162.6993 KOps/s 161.1125 KOps/s $\color{#35bf28}+0.98\%$
test_tc_second_layer_tensor 25.2800μs 4.3658μs 229.0521 KOps/s 223.6983 KOps/s $\color{#35bf28}+2.39\%$
test_tc_second_layer_nontensor 56.0200μs 8.6857μs 115.1318 KOps/s 114.8125 KOps/s $\color{#35bf28}+0.28\%$
test_unbind 0.2416s 16.1508ms 61.9165 Ops/s 55.7825 Ops/s $\textbf{\color{#35bf28}+11.00\%}$
test_full_like 7.6214ms 4.5355ms 220.4810 Ops/s 224.5851 Ops/s $\color{#d91a1a}-1.83\%$
test_zeros_like 9.4544ms 7.3738ms 135.6146 Ops/s 226.2308 Ops/s $\textbf{\color{#d91a1a}-40.05\%}$
test_ones_like 5.2091ms 4.4439ms 225.0259 Ops/s 224.6749 Ops/s $\color{#35bf28}+0.16\%$
test_clone 7.5424ms 6.9005ms 144.9177 Ops/s 146.5495 Ops/s $\color{#d91a1a}-1.11\%$
test_squeeze 70.5710μs 14.3668μs 69.6049 KOps/s 70.9958 KOps/s $\color{#d91a1a}-1.96\%$
test_unsqueeze 0.2677ms 0.1120ms 8.9260 KOps/s 9.1191 KOps/s $\color{#d91a1a}-2.12\%$
test_split 0.2297ms 0.1826ms 5.4769 KOps/s 5.4145 KOps/s $\color{#35bf28}+1.15\%$
test_permute 0.2664ms 0.2098ms 4.7658 KOps/s 4.7918 KOps/s $\color{#d91a1a}-0.54\%$
test_stack 53.3208ms 51.9916ms 19.2339 Ops/s 19.1857 Ops/s $\color{#35bf28}+0.25\%$
test_cat 53.0114ms 52.1032ms 19.1927 Ops/s 19.1408 Ops/s $\color{#35bf28}+0.27\%$

@vmoens
Update
0d9503c 
[ghstack-poisoned]
vmoens added a commit that referenced this pull request Feb 18, 2026
@vmoens @cursoragent
[Distributed] Add distributed benchmark script
4ac7d27 
Benchmark comparing leaf-by-leaf vs consolidated send/recv, broadcast,
all_reduce, and init_remote across various TD sizes. Reports latency
and speedup for each configuration.

Co-authored-by: Cursor <cursoragent@cursor.com>
ghstack-source-id: 00be8c4
Pull-Request: #1592
Co-authored-by: Cursor <cursoragent@cursor.com>
@vmoens vmoens mentioned this pull request Feb 18, 2026
Merged
@vmoens
Update
49efd19 
[ghstack-poisoned]
This was referenced Feb 19, 2026
Closed
Merged
@github-actions

github-actions Bot commented Feb 19, 2026

Copy link
Copy Markdown
Contributor

PR Title Label Error

Unknown or invalid prefix [Distributed].

Current title: [Distributed] Add distributed benchmark script

Supported Prefixes

Your PR title must start with exactly one of these prefixes:

Prefix Label Applied Example
[BugFix] BugFix [BugFix] Fix memory leak in TensorDict
[Feature] Feature [Feature] Add new storage backend
[Doc] or [Docs] Documentation [Doc] Update installation guide
[Refactor] Refactor [Refactor] Clean up module imports
[CI] CI [CI] Fix workflow permissions
[Test] or [Tests] Test [Test] Add unit tests for nn module
[Compile] Compile [Compile] Fix torch.compile issue
[Performance] or [Perf] Perf [Perf] Optimize tensor operations
[Deprecation] Deprecation [Deprecation] Mark old function
[Setup] Setup [Setup] Update build configuration

Note: Common variations like singular/plural are supported (e.g., [Doc] or [Docs]).

@vmoens
Update
41955e5 
[ghstack-poisoned]
@github-actions

github-actions Bot commented Feb 19, 2026

Copy link
Copy Markdown
Contributor

PR Title Label Error

Unknown or invalid prefix [Distributed].

Current title: [Distributed] Add distributed benchmark script

Supported Prefixes

Your PR title must start with exactly one of these prefixes:

Prefix Label Applied Example
[BugFix] BugFix [BugFix] Fix memory leak in TensorDict
[Feature] Feature [Feature] Add new storage backend
[Doc] or [Docs] Documentation [Doc] Update installation guide
[Refactor] Refactor [Refactor] Clean up module imports
[CI] CI [CI] Fix workflow permissions
[Test] or [Tests] Test [Test] Add unit tests for nn module
[Compile] Compile [Compile] Fix torch.compile issue
[Performance] or [Perf] Perf [Perf] Optimize tensor operations
[Deprecation] Deprecation [Deprecation] Mark old function
[Setup] Setup [Setup] Update build configuration

Note: Common variations like singular/plural are supported (e.g., [Doc] or [Docs]).

vmoens added a commit that referenced this pull request Feb 19, 2026
@vmoens @cursoragent
[Distributed] Add distributed benchmark script
b4dcdb7 
Benchmark comparing leaf-by-leaf vs consolidated send/recv, broadcast,
all_reduce, and init_remote across various TD sizes. Reports latency
and speedup for each configuration.

Co-authored-by: Cursor <cursoragent@cursor.com>
ghstack-source-id: 45de218
Pull-Request: #1592
Co-authored-by: Cursor <cursoragent@cursor.com>
@vmoens
vmoens merged commit 41955e5 into gh/vmoens/75/base Feb 19, 2026
35 of 52 checks passed
@vmoens
vmoens deleted the gh/vmoens/75/head branch February 19, 2026 10:14
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

Benchmarks CLA Signed This label is managed by the Facebook bot. Authors need to sign the CLA before a PR can be reviewed. Distributed

Projects

None yet

Milestone

No milestone

Development

Successfully merging this pull request may close these issues.

1 participant

@vmoens