[DTensor] Add transfer plan computation for cross-mesh DTensor redistribution by vmoens · Pull Request #1639 · pytorch/tensordict

vmoens · 2026-03-06T15:02:34Z

Stack from ghstack (oldest at bottom):

Add tensordict/_dtensor.py with:

Shard algebra: compute per-rank local slices from mesh shape + placements
Transfer plan computation: given src/dst mesh+placements, compute the
minimal set of P2P transfers (which byte ranges go from which src rank
to which dst rank)
Transport abstraction: _TransportBackend protocol with
_TorchDistributedBackend and _UCXXBackend implementations

The transfer plan is pure computation (no GPU, no distributed runtime)
and can be tested in isolation. It supports Shard, Replicate, and Partial
placements on arbitrary n-D meshes, uneven sharding, and custom rank maps.

All new classes are private (underscore-prefixed).

Made-with: Cursor

[ghstack-poisoned]

github-actions · 2026-03-06T15:02:44Z

PR Title Label Error

Unknown or invalid prefix [DTensor].

Current title: [DTensor] Add transfer plan computation for cross-mesh DTensor redistribution

Supported Prefixes

Your PR title must start with exactly one of these prefixes (case-insensitive):

Prefix	Label Applied	Example
`[BugFix]` or `[Fix]`	bug	`[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]`	Performance	`[Perf] Optimize tensor operations`
`[Deprecation]`	Deprecation	`[Deprecation] Mark old function`
`[Setup]`	setup	`[Setup] Update build configuration`
`[Distributed]` or `[Dist]`	Distributed	`[Distributed] Add scatter collective`
`[Benchmark]` or `[Bench]`	Benchmarks	`[Benchmark] Add compile benchmark`
`[Typing]` or `[Type]`	Typing	`[Typing] Add type stubs`
`[BC-breaking]` or `[BC]`	BC-breaking	`[BC-breaking] Remove deprecated API`
`[Formatting]` or `[Format]`	Formatting	`[Format] Fix code style`
`[Quality]`	Quality	`[Quality] Improve error messages`

Note: Matching is case-insensitive. Common variations (singular/plural) are supported.

github-actions · 2026-03-06T15:02:55Z

PR Title Label Error

Unknown or invalid prefix [DTensor].

Current title: [DTensor] Add transfer plan computation for cross-mesh DTensor redistribution

Supported Prefixes

Your PR title must start with exactly one of these prefixes (case-insensitive):

Prefix	Label Applied	Example
`[BugFix]` or `[Fix]`	bug	`[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]`	Performance	`[Perf] Optimize tensor operations`
`[Deprecation]`	Deprecation	`[Deprecation] Mark old function`
`[Setup]`	setup	`[Setup] Update build configuration`
`[Distributed]` or `[Dist]`	Distributed	`[Distributed] Add scatter collective`
`[Benchmark]` or `[Bench]`	Benchmarks	`[Benchmark] Add compile benchmark`
`[Typing]` or `[Type]`	Typing	`[Typing] Add type stubs`
`[BC-breaking]` or `[BC]`	BC-breaking	`[BC-breaking] Remove deprecated API`
`[Formatting]` or `[Format]`	Formatting	`[Format] Fix code style`
`[Quality]`	Quality	`[Quality] Improve error messages`

Note: Matching is case-insensitive. Common variations (singular/plural) are supported.

github-actions · 2026-03-06T15:20:44Z

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

Total Benchmarks: 261. Improved: $\large\color{#35bf28}20$. Worsened: $\large\color{#d91a1a}9$.

Expand to view detailed results

Name	Max	Mean	Ops	Ops on Repo `HEAD`	Change
test_plain_set_nested	40.5100μs	15.3135μs	65.3018 KOps/s	66.6704 KOps/s	$\color{#d91a1a}-2.05\%$
test_plain_set_stack_nested	37.0400μs	15.1823μs	65.8661 KOps/s	66.1368 KOps/s	$\color{#d91a1a}-0.41\%$
test_plain_set_nested_inplace	44.0410μs	16.6865μs	59.9287 KOps/s	59.9007 KOps/s	$\color{#35bf28}+0.05\%$
test_plain_set_stack_nested_inplace	42.1510μs	16.5489μs	60.4270 KOps/s	59.7537 KOps/s	$\color{#35bf28}+1.13\%$
test_items	38.4900μs	6.1109μs	163.6408 KOps/s	165.9487 KOps/s	$\color{#d91a1a}-1.39\%$
test_items_nested	0.5131ms	0.4636ms	2.1568 KOps/s	2.1567 KOps/s	$+0.00\%$
test_items_nested_locked	0.5155ms	0.4675ms	2.1393 KOps/s	2.1354 KOps/s	$\color{#35bf28}+0.18\%$
test_items_nested_leaf	0.1661ms	96.7166μs	10.3395 KOps/s	10.1746 KOps/s	$\color{#35bf28}+1.62\%$
test_items_stack_nested	0.5120ms	0.4674ms	2.1397 KOps/s	2.1545 KOps/s	$\color{#d91a1a}-0.69\%$
test_items_stack_nested_leaf	0.1355ms	98.6486μs	10.1370 KOps/s	10.0290 KOps/s	$\color{#35bf28}+1.08\%$
test_items_stack_nested_locked	0.5117ms	0.4687ms	2.1335 KOps/s	2.1468 KOps/s	$\color{#d91a1a}-0.62\%$
test_keys	36.4610μs	4.2331μs	236.2341 KOps/s	234.7756 KOps/s	$\color{#35bf28}+0.62\%$
test_keys_nested	0.1964ms	0.1296ms	7.7185 KOps/s	7.7226 KOps/s	$\color{#d91a1a}-0.05\%$
test_keys_nested_locked	0.7704ms	0.1397ms	7.1581 KOps/s	7.2127 KOps/s	$\color{#d91a1a}-0.76\%$
test_keys_nested_leaf	0.1654ms	0.1217ms	8.2157 KOps/s	8.2975 KOps/s	$\color{#d91a1a}-0.99\%$
test_keys_stack_nested	0.1769ms	0.1309ms	7.6422 KOps/s	7.6467 KOps/s	$\color{#d91a1a}-0.06\%$
test_keys_stack_nested_leaf	0.1556ms	0.1213ms	8.2449 KOps/s	8.2653 KOps/s	$\color{#d91a1a}-0.25\%$
test_keys_stack_nested_locked	0.1834ms	0.1399ms	7.1489 KOps/s	7.2519 KOps/s	$\color{#d91a1a}-1.42\%$
test_values	11.3022μs	1.0240μs	976.5860 KOps/s	975.2531 KOps/s	$\color{#35bf28}+0.14\%$
test_values_nested	83.0610μs	52.7437μs	18.9596 KOps/s	19.0916 KOps/s	$\color{#d91a1a}-0.69\%$
test_values_nested_locked	94.1610μs	56.3215μs	17.7552 KOps/s	17.8145 KOps/s	$\color{#d91a1a}-0.33\%$
test_values_nested_leaf	96.5820μs	60.6813μs	16.4795 KOps/s	17.0871 KOps/s	$\color{#d91a1a}-3.56\%$
test_values_stack_nested	88.0610μs	52.9292μs	18.8932 KOps/s	18.7569 KOps/s	$\color{#35bf28}+0.73\%$
test_values_stack_nested_leaf	93.4920μs	60.6546μs	16.4868 KOps/s	16.5513 KOps/s	$\color{#d91a1a}-0.39\%$
test_values_stack_nested_locked	87.2910μs	56.4233μs	17.7232 KOps/s	17.8506 KOps/s	$\color{#d91a1a}-0.71\%$
test_membership	11.4202μs	0.8372μs	1.1945 MOps/s	1.1639 MOps/s	$\color{#35bf28}+2.62\%$
test_membership_nested	30.2100μs	2.8838μs	346.7629 KOps/s	344.6520 KOps/s	$\color{#35bf28}+0.61\%$
test_membership_nested_leaf	25.4910μs	2.8968μs	345.2087 KOps/s	341.1969 KOps/s	$\color{#35bf28}+1.18\%$
test_membership_stacked_nested	36.3910μs	2.8713μs	348.2736 KOps/s	347.8237 KOps/s	$\color{#35bf28}+0.13\%$
test_membership_stacked_nested_leaf	22.2300μs	2.8790μs	347.3448 KOps/s	348.6078 KOps/s	$\color{#d91a1a}-0.36\%$
test_membership_nested_last	29.6910μs	4.3460μs	230.0944 KOps/s	228.8973 KOps/s	$\color{#35bf28}+0.52\%$
test_membership_nested_leaf_last	30.6700μs	4.3463μs	230.0814 KOps/s	227.5173 KOps/s	$\color{#35bf28}+1.13\%$
test_membership_stacked_nested_last	71.0810μs	4.3516μs	229.8011 KOps/s	227.8177 KOps/s	$\color{#35bf28}+0.87\%$
test_membership_stacked_nested_leaf_last	31.7900μs	4.3232μs	231.3092 KOps/s	230.6605 KOps/s	$\color{#35bf28}+0.28\%$
test_nested_getleaf	51.1410μs	21.6296μs	46.2330 KOps/s	45.8269 KOps/s	$\color{#35bf28}+0.89\%$
test_nested_get	58.2610μs	20.7229μs	48.2557 KOps/s	49.1516 KOps/s	$\color{#d91a1a}-1.82\%$
test_stacked_getleaf	47.4010μs	21.5452μs	46.4141 KOps/s	45.5110 KOps/s	$\color{#35bf28}+1.98\%$
test_stacked_get	57.5610μs	20.5948μs	48.5560 KOps/s	47.8800 KOps/s	$\color{#35bf28}+1.41\%$
test_nested_getitemleaf	89.6220μs	22.0759μs	45.2983 KOps/s	45.1702 KOps/s	$\color{#35bf28}+0.28\%$
test_nested_getitem	45.3810μs	21.0461μs	47.5147 KOps/s	47.0940 KOps/s	$\color{#35bf28}+0.89\%$
test_stacked_getitemleaf	47.8200μs	22.0222μs	45.4087 KOps/s	44.4461 KOps/s	$\color{#35bf28}+2.17\%$
test_stacked_getitem	48.8010μs	21.0484μs	47.5096 KOps/s	46.5242 KOps/s	$\color{#35bf28}+2.12\%$
test_lock_nested	7.8906ms	0.4856ms	2.0595 KOps/s	2.0952 KOps/s	$\color{#d91a1a}-1.71\%$
test_lock_stack_nested	0.5357ms	0.4798ms	2.0842 KOps/s	2.0583 KOps/s	$\color{#35bf28}+1.26\%$
test_unlock_nested	0.4606ms	0.3885ms	2.5739 KOps/s	2.5737 KOps/s	$+0.01\%$
test_unlock_stack_nested	0.4593ms	0.3900ms	2.5638 KOps/s	2.5179 KOps/s	$\color{#35bf28}+1.83\%$
test_flatten_speed	0.1729ms	0.1213ms	8.2414 KOps/s	8.1155 KOps/s	$\color{#35bf28}+1.55\%$
test_unflatten_speed	0.6417ms	0.5774ms	1.7319 KOps/s	1.7350 KOps/s	$\color{#d91a1a}-0.18\%$
test_common_ops	0.9198ms	0.7015ms	1.4255 KOps/s	1.4382 KOps/s	$\color{#d91a1a}-0.88\%$
test_creation	90.0210μs	3.1642μs	316.0374 KOps/s	318.5614 KOps/s	$\color{#d91a1a}-0.79\%$
test_creation_empty	51.4610μs	6.9460μs	143.9673 KOps/s	141.6286 KOps/s	$\color{#35bf28}+1.65\%$
test_creation_nested_1	49.0510μs	11.5697μs	86.4325 KOps/s	86.5407 KOps/s	$\color{#d91a1a}-0.12\%$
test_creation_nested_2	46.6510μs	13.3625μs	74.8364 KOps/s	74.7257 KOps/s	$\color{#35bf28}+0.15\%$
test_creation_many_keys[10]	54.0910μs	20.9429μs	47.7490 KOps/s	47.0981 KOps/s	$\color{#35bf28}+1.38\%$
test_creation_many_keys[50]	0.1253ms	89.2396μs	11.2058 KOps/s	10.9006 KOps/s	$\color{#35bf28}+2.80\%$
test_creation_many_keys[100]	0.2831ms	0.1763ms	5.6725 KOps/s	5.5082 KOps/s	$\color{#35bf28}+2.98\%$
test_creation_nested_many_keys[10]	73.2210μs	45.3257μs	22.0625 KOps/s	21.7019 KOps/s	$\color{#35bf28}+1.66\%$
test_creation_nested_many_keys[50]	0.2274ms	0.1831ms	5.4623 KOps/s	5.4804 KOps/s	$\color{#d91a1a}-0.33\%$
test_clone	95.7920μs	13.5147μs	73.9933 KOps/s	73.9038 KOps/s	$\color{#35bf28}+0.12\%$
test_getitem[int]	1.5267ms	15.1437μs	66.0342 KOps/s	59.4539 KOps/s	$\textbf{\color{#35bf28}+11.07\%}$
test_getitem[slice_int]	0.1332ms	24.1946μs	41.3316 KOps/s	41.3898 KOps/s	$\color{#d91a1a}-0.14\%$
test_getitem[range]	0.1765ms	63.8715μs	15.6564 KOps/s	15.6007 KOps/s	$\color{#35bf28}+0.36\%$
test_getitem[tuple]	0.1398ms	24.1541μs	41.4008 KOps/s	42.0665 KOps/s	$\color{#d91a1a}-1.58\%$
test_getitem[list]	0.1786ms	58.6637μs	17.0463 KOps/s	16.6128 KOps/s	$\color{#35bf28}+2.61\%$
test_setitem_dim[int]	46.8710μs	26.4326μs	37.8320 KOps/s	37.0551 KOps/s	$\color{#35bf28}+2.10\%$
test_setitem_dim[slice_int]	65.5210μs	43.1571μs	23.1711 KOps/s	22.1555 KOps/s	$\color{#35bf28}+4.58\%$
test_setitem_dim[range]	0.1196ms	95.6660μs	10.4530 KOps/s	9.9936 KOps/s	$\color{#35bf28}+4.60\%$
test_setitem_dim[tuple]	69.8410μs	40.5963μs	24.6328 KOps/s	24.2068 KOps/s	$\color{#35bf28}+1.76\%$
test_setitem	54.7410μs	18.0037μs	55.5442 KOps/s	55.2244 KOps/s	$\color{#35bf28}+0.58\%$
test_set	51.8710μs	17.2381μs	58.0110 KOps/s	58.5751 KOps/s	$\color{#d91a1a}-0.96\%$
test_set_shared	0.6216ms	0.2130ms	4.6959 KOps/s	4.8817 KOps/s	$\color{#d91a1a}-3.81\%$
test_update	0.2031ms	21.7312μs	46.0168 KOps/s	44.9664 KOps/s	$\color{#35bf28}+2.34\%$
test_update_nested	68.6210μs	33.1134μs	30.1993 KOps/s	30.0251 KOps/s	$\color{#35bf28}+0.58\%$
test_update__nested	0.4477ms	34.1817μs	29.2554 KOps/s	28.9103 KOps/s	$\color{#35bf28}+1.19\%$
test_set_nested	60.7110μs	20.3960μs	49.0291 KOps/s	51.5897 KOps/s	$\color{#d91a1a}-4.96\%$
test_set_nested_new	58.2110μs	26.0347μs	38.4103 KOps/s	41.6554 KOps/s	$\textbf{\color{#d91a1a}-7.79\%}$
test_select	75.1310μs	43.2440μs	23.1246 KOps/s	24.4331 KOps/s	$\textbf{\color{#d91a1a}-5.36\%}$
test_select_nested	0.1050ms	74.9398μs	13.3441 KOps/s	13.5116 KOps/s	$\color{#d91a1a}-1.24\%$
test_exclude_nested	0.1314ms	91.9688μs	10.8733 KOps/s	10.9017 KOps/s	$\color{#d91a1a}-0.26\%$
test_empty[True]	0.4634ms	0.3995ms	2.5029 KOps/s	2.5096 KOps/s	$\color{#d91a1a}-0.27\%$
test_empty[False]	7.8475μs	1.3270μs	753.5974 KOps/s	768.0517 KOps/s	$\color{#d91a1a}-1.88\%$
test_to	0.1024ms	71.7361μs	13.9400 KOps/s	13.4951 KOps/s	$\color{#35bf28}+3.30\%$
test_to_nonblocking	0.1222ms	65.4454μs	15.2799 KOps/s	15.2487 KOps/s	$\color{#35bf28}+0.20\%$
test_unbind_speed	0.3731ms	0.3338ms	2.9956 KOps/s	2.9876 KOps/s	$\color{#35bf28}+0.27\%$
test_unbind_speed_stack0	0.4247ms	0.3290ms	3.0398 KOps/s	3.0080 KOps/s	$\color{#35bf28}+1.06\%$
test_unbind_speed_stack1	0.1035s	0.8384ms	1.1927 KOps/s	1.1673 KOps/s	$\color{#35bf28}+2.17\%$
test_split	0.1035s	1.2668ms	789.4121 Ops/s	782.4156 Ops/s	$\color{#35bf28}+0.89\%$
test_chunk	0.1032s	1.2125ms	824.7189 Ops/s	918.6403 Ops/s	$\textbf{\color{#d91a1a}-10.22\%}$
test_to_cpu_blocking	19.8678ms	19.7660ms	50.5919 Ops/s	34.1241 Ops/s	$\textbf{\color{#35bf28}+48.26\%}$
test_to_cpu_global_sync	11.7691ms	11.6839ms	85.5879 Ops/s	76.8206 Ops/s	$\textbf{\color{#35bf28}+11.41\%}$
test_to_cpu_event_sync	13.4181ms	12.6749ms	78.8960 Ops/s	78.7987 Ops/s	$\color{#35bf28}+0.12\%$
test_to_cpu_default	0.1161s	14.0025ms	71.4159 Ops/s	78.5336 Ops/s	$\textbf{\color{#d91a1a}-9.06\%}$
test_consolidate[False-None]	4.2335ms	4.1546ms	240.6945 Ops/s	214.6916 Ops/s	$\textbf{\color{#35bf28}+12.11\%}$
test_consolidate[default-None]	2.6717ms	2.0491ms	488.0300 Ops/s	479.1963 Ops/s	$\color{#35bf28}+1.84\%$
test_consolidate[reduce-overhead-None]	2.0520ms	1.9743ms	506.5149 Ops/s	504.8202 Ops/s	$\color{#35bf28}+0.34\%$
test_consolidate_njt[False-None]	0.1913s	10.1671ms	98.3560 Ops/s	116.7773 Ops/s	$\textbf{\color{#d91a1a}-15.77\%}$
test_to[False-False-None]	2.2287ms	2.1288ms	469.7501 Ops/s	465.5847 Ops/s	$\color{#35bf28}+0.89\%$
test_to[True-False-None]	2.1483ms	1.9124ms	522.9123 Ops/s	519.6354 Ops/s	$\color{#35bf28}+0.63\%$
test_to[within-False-None]	6.3868ms	6.1781ms	161.8616 Ops/s	162.6492 Ops/s	$\color{#d91a1a}-0.48\%$
test_to[True-default-None]	9.0233ms	8.8430ms	113.0842 Ops/s	109.1644 Ops/s	$\color{#35bf28}+3.59\%$
test_to_njt[False-False-None]	8.8336ms	8.5131ms	117.4664 Ops/s	116.1406 Ops/s	$\color{#35bf28}+1.14\%$
test_to_njt[True-False-None]	7.1127ms	6.9275ms	144.3515 Ops/s	141.6555 Ops/s	$\color{#35bf28}+1.90\%$
test_to_njt[within-False-None]	15.8162ms	15.6220ms	64.0124 Ops/s	62.8856 Ops/s	$\color{#35bf28}+1.79\%$
test_creation[device0]	0.4486ms	0.1167ms	8.5680 KOps/s	8.6259 KOps/s	$\color{#d91a1a}-0.67\%$
test_creation_from_tensor	0.5865ms	0.1142ms	8.7602 KOps/s	8.6896 KOps/s	$\color{#35bf28}+0.81\%$
test_add_one[memmap_tensor0]	0.3086ms	6.6452μs	150.4854 KOps/s	146.8191 KOps/s	$\color{#35bf28}+2.50\%$
test_contiguous[memmap_tensor0]	26.3900μs	0.6691μs	1.4945 MOps/s	2.1624 MOps/s	$\textbf{\color{#d91a1a}-30.89\%}$
test_stack[memmap_tensor0]	32.7000μs	4.6024μs	217.2789 KOps/s	219.1664 KOps/s	$\color{#d91a1a}-0.86\%$
test_memmaptd_index	1.0726ms	0.2649ms	3.7755 KOps/s	3.7607 KOps/s	$\color{#35bf28}+0.39\%$
test_memmaptd_index_astensor	0.5351ms	0.3686ms	2.7128 KOps/s	2.7020 KOps/s	$\color{#35bf28}+0.40\%$
test_memmaptd_index_op	0.1639s	0.7315ms	1.3671 KOps/s	1.5870 KOps/s	$\textbf{\color{#d91a1a}-13.86\%}$
test_serialize_model	0.1394s	0.1371s	7.2957 Ops/s	7.2560 Ops/s	$\color{#35bf28}+0.55\%$
test_serialize_model_pickle	1.3973s	1.2125s	0.8247 Ops/s	0.8249 Ops/s	$\color{#d91a1a}-0.02\%$
test_serialize_weights	0.1375s	0.1359s	7.3585 Ops/s	7.3159 Ops/s	$\color{#35bf28}+0.58\%$
test_serialize_weights_returnearly	0.2825s	94.1937ms	10.6164 Ops/s	14.4578 Ops/s	$\textbf{\color{#d91a1a}-26.57\%}$
test_serialize_weights_pickle	1.3687s	1.2137s	0.8239 Ops/s	0.8236 Ops/s	$\color{#35bf28}+0.04\%$
test_reshape_pytree	0.2005ms	32.9318μs	30.3658 KOps/s	30.1674 KOps/s	$\color{#35bf28}+0.66\%$
test_reshape_td	89.1320μs	45.2927μs	22.0786 KOps/s	22.0490 KOps/s	$\color{#35bf28}+0.13\%$
test_view_pytree	0.2194ms	32.5197μs	30.7506 KOps/s	30.3671 KOps/s	$\color{#35bf28}+1.26\%$
test_view_td	0.1118ms	53.1348μs	18.8200 KOps/s	18.3961 KOps/s	$\color{#35bf28}+2.30\%$
test_unbind_pytree	0.2346ms	36.5464μs	27.3625 KOps/s	26.8394 KOps/s	$\color{#35bf28}+1.95\%$
test_unbind_td	0.1909ms	49.7230μs	20.1114 KOps/s	19.9336 KOps/s	$\color{#35bf28}+0.89\%$
test_split_pytree	0.2483ms	42.7336μs	23.4008 KOps/s	23.2711 KOps/s	$\color{#35bf28}+0.56\%$
test_split_td	0.2188ms	63.6834μs	15.7027 KOps/s	15.5271 KOps/s	$\color{#35bf28}+1.13\%$
test_add_pytree	0.2045ms	42.5387μs	23.5080 KOps/s	23.1369 KOps/s	$\color{#35bf28}+1.60\%$
test_add_td	0.1176ms	55.3437μs	18.0689 KOps/s	17.8144 KOps/s	$\color{#35bf28}+1.43\%$
test_compile_add_one_nested[tensordict-compile]	0.1924ms	0.1401ms	7.1368 KOps/s	6.8622 KOps/s	$\color{#35bf28}+4.00\%$
test_compile_add_one_nested[tensordict-eager]	0.2816ms	0.2025ms	4.9384 KOps/s	4.9537 KOps/s	$\color{#d91a1a}-0.31\%$
test_compile_add_one_nested[pytree-compile]	0.1421ms	0.1074ms	9.3133 KOps/s	8.8475 KOps/s	$\textbf{\color{#35bf28}+5.26\%}$
test_compile_add_one_nested[pytree-eager]	0.4575ms	0.1841ms	5.4311 KOps/s	5.3750 KOps/s	$\color{#35bf28}+1.05\%$
test_compile_copy_nested[tensordict-compile]	0.3024ms	9.8977μs	101.0339 KOps/s	94.6967 KOps/s	$\textbf{\color{#35bf28}+6.69\%}$
test_compile_copy_nested[tensordict-eager]	0.5811ms	53.3188μs	18.7551 KOps/s	18.3925 KOps/s	$\color{#35bf28}+1.97\%$
test_compile_copy_nested[pytree-compile]	0.1393ms	9.8219μs	101.8129 KOps/s	102.0449 KOps/s	$\color{#d91a1a}-0.23\%$
test_compile_copy_nested[pytree-eager]	0.4914ms	69.3723μs	14.4150 KOps/s	14.0529 KOps/s	$\color{#35bf28}+2.58\%$
test_compile_add_one_flat[tensordict-compile]	0.2376ms	0.1770ms	5.6507 KOps/s	3.3181 KOps/s	$\textbf{\color{#35bf28}+70.30\%}$
test_compile_add_one_flat[tensordict-eager]	0.3988ms	0.2820ms	3.5464 KOps/s	3.4817 KOps/s	$\color{#35bf28}+1.86\%$
test_compile_add_one_flat[tensorclass-compile]	0.2138ms	0.1174ms	8.5144 KOps/s	7.9935 KOps/s	$\textbf{\color{#35bf28}+6.52\%}$
test_compile_add_one_flat[tensorclass-eager]	0.1159ms	73.5535μs	13.5955 KOps/s	13.7048 KOps/s	$\color{#d91a1a}-0.80\%$
test_compile_add_one_flat[pytree-compile]	0.2050ms	0.1570ms	6.3685 KOps/s	6.0588 KOps/s	$\textbf{\color{#35bf28}+5.11\%}$
test_compile_add_one_flat[pytree-eager]	0.8087ms	0.5387ms	1.8563 KOps/s	1.8497 KOps/s	$\color{#35bf28}+0.35\%$
test_compile_add_self_flat[tensordict-eager]	0.4379ms	0.3344ms	2.9904 KOps/s	2.9294 KOps/s	$\color{#35bf28}+2.08\%$
test_compile_add_self_flat[tensordict-compile]	0.2322ms	0.1786ms	5.5994 KOps/s	5.0613 KOps/s	$\textbf{\color{#35bf28}+10.63\%}$
test_compile_add_self_flat[tensorclass-eager]	0.1805ms	90.2627μs	11.0788 KOps/s	11.1964 KOps/s	$\color{#d91a1a}-1.05\%$
test_compile_add_self_flat[tensorclass-compile]	0.6740ms	0.1211ms	8.2573 KOps/s	7.7812 KOps/s	$\textbf{\color{#35bf28}+6.12\%}$
test_compile_add_self_flat[pytree-eager]	0.8842ms	0.4464ms	2.2401 KOps/s	2.1966 KOps/s	$\color{#35bf28}+1.98\%$
test_compile_add_self_flat[pytree-compile]	0.1906ms	0.1573ms	6.3591 KOps/s	6.0389 KOps/s	$\textbf{\color{#35bf28}+5.30\%}$
test_compile_copy_flat[tensordict-compile]	0.1201ms	13.6590μs	73.2117 KOps/s	73.3661 KOps/s	$\color{#d91a1a}-0.21\%$
test_compile_copy_flat[tensordict-eager]	0.5035ms	42.2612μs	23.6624 KOps/s	23.7952 KOps/s	$\color{#d91a1a}-0.56\%$
test_compile_copy_flat[pytree-compile]	0.4519ms	10.8009μs	92.5848 KOps/s	91.8209 KOps/s	$\color{#35bf28}+0.83\%$
test_compile_copy_flat[pytree-eager]	0.5001ms	53.0541μs	18.8487 KOps/s	18.7992 KOps/s	$\color{#35bf28}+0.26\%$
test_compile_assign_and_add[tensordict-compile]	2.0126ms	0.1778ms	5.6231 KOps/s	5.4066 KOps/s	$\color{#35bf28}+4.00\%$
test_compile_assign_and_add[tensordict-eager]	3.8599ms	3.3187ms	301.3217 Ops/s	297.6538 Ops/s	$\color{#35bf28}+1.23\%$
test_compile_assign_and_add[pytree-compile]	1.9557ms	0.1609ms	6.2162 KOps/s	5.9751 KOps/s	$\color{#35bf28}+4.03\%$
test_compile_assign_and_add[pytree-eager]	3.2885ms	2.8536ms	350.4325 Ops/s	350.5013 Ops/s	$\color{#d91a1a}-0.02\%$
test_compile_indexing[tensor-tensordict-compile]	0.2543ms	0.1096ms	9.1201 KOps/s	8.7245 KOps/s	$\color{#35bf28}+4.53\%$
test_compile_indexing[tensor-tensordict-eager]	0.5133ms	75.0424μs	13.3258 KOps/s	13.2396 KOps/s	$\color{#35bf28}+0.65\%$
test_compile_indexing[tensor-tensorclass-compile]	0.6187ms	95.5601μs	10.4646 KOps/s	10.1551 KOps/s	$\color{#35bf28}+3.05\%$
test_compile_indexing[tensor-tensorclass-eager]	0.5210ms	44.9058μs	22.2688 KOps/s	20.7297 KOps/s	$\textbf{\color{#35bf28}+7.42\%}$
test_compile_indexing[tensor-pytree-compile]	0.5621ms	96.0202μs	10.4145 KOps/s	10.0847 KOps/s	$\color{#35bf28}+3.27\%$
test_compile_indexing[tensor-pytree-eager]	0.2492ms	44.4537μs	22.4953 KOps/s	21.7287 KOps/s	$\color{#35bf28}+3.53\%$
test_compile_indexing[slice-tensordict-compile]	0.5253ms	56.5248μs	17.6913 KOps/s	17.0063 KOps/s	$\color{#35bf28}+4.03\%$
test_compile_indexing[slice-tensordict-eager]	0.2172ms	27.4556μs	36.4224 KOps/s	35.8536 KOps/s	$\color{#35bf28}+1.59\%$
test_compile_indexing[slice-tensorclass-compile]	0.1304ms	44.4844μs	22.4798 KOps/s	22.2755 KOps/s	$\color{#35bf28}+0.92\%$
test_compile_indexing[slice-tensorclass-eager]	0.4814ms	22.6257μs	44.1975 KOps/s	44.1138 KOps/s	$\color{#35bf28}+0.19\%$
test_compile_indexing[slice-pytree-compile]	0.4725ms	45.3822μs	22.0351 KOps/s	21.5143 KOps/s	$\color{#35bf28}+2.42\%$
test_compile_indexing[slice-pytree-eager]	0.4622ms	22.6451μs	44.1596 KOps/s	43.6134 KOps/s	$\color{#35bf28}+1.25\%$
test_compile_indexing[int-tensordict-compile]	0.4997ms	57.0164μs	17.5388 KOps/s	16.7974 KOps/s	$\color{#35bf28}+4.41\%$
test_compile_indexing[int-tensordict-eager]	0.5809ms	27.5512μs	36.2961 KOps/s	36.0663 KOps/s	$\color{#35bf28}+0.64\%$
test_compile_indexing[int-tensorclass-compile]	84.4320μs	44.9726μs	22.2357 KOps/s	21.8972 KOps/s	$\color{#35bf28}+1.55\%$
test_compile_indexing[int-tensorclass-eager]	0.4635ms	22.5697μs	44.3072 KOps/s	43.6325 KOps/s	$\color{#35bf28}+1.55\%$
test_compile_indexing[int-pytree-compile]	0.4873ms	45.1276μs	22.1594 KOps/s	22.0899 KOps/s	$\color{#35bf28}+0.31\%$
test_compile_indexing[int-pytree-eager]	0.4532ms	22.4719μs	44.5001 KOps/s	43.7609 KOps/s	$\color{#35bf28}+1.69\%$
test_compile_replace[single-eager]	0.4921ms	47.7454μs	20.9444 KOps/s	21.0288 KOps/s	$\color{#d91a1a}-0.40\%$
test_compile_replace[single-compile]	0.1804ms	0.1060ms	9.4366 KOps/s	9.1537 KOps/s	$\color{#35bf28}+3.09\%$
test_compile_replace[multi-eager]	1.0405ms	0.5729ms	1.7456 KOps/s	1.7741 KOps/s	$\color{#d91a1a}-1.60\%$
test_compile_replace[multi-compile]	0.1438ms	0.1127ms	8.8754 KOps/s	8.6665 KOps/s	$\color{#35bf28}+2.41\%$
test_compile_tc_getattr_20[eager]	0.6056ms	0.1731ms	5.7768 KOps/s	5.7892 KOps/s	$\color{#d91a1a}-0.21\%$
test_compile_tc_getattr_20[compile]	0.2723ms	0.1196ms	8.3583 KOps/s	8.1217 KOps/s	$\color{#35bf28}+2.91\%$
test_compile_clone_shallow[20-eager]	0.4468ms	19.8725μs	50.3207 KOps/s	52.5398 KOps/s	$\color{#d91a1a}-4.22\%$
test_compile_clone_shallow[20-compile]	0.4872ms	11.2274μs	89.0676 KOps/s	85.1435 KOps/s	$\color{#35bf28}+4.61\%$
test_compile_clone_shallow[40-eager]	0.4535ms	34.5778μs	28.9203 KOps/s	29.2583 KOps/s	$\color{#d91a1a}-1.16\%$
test_compile_clone_shallow[40-compile]	67.6610μs	12.7920μs	78.1740 KOps/s	75.8066 KOps/s	$\color{#35bf28}+3.12\%$
test_compile_clone_shallow[80-eager]	0.5073ms	63.0851μs	15.8516 KOps/s	15.5398 KOps/s	$\color{#35bf28}+2.01\%$
test_compile_clone_shallow[80-compile]	0.4475ms	15.0108μs	66.6186 KOps/s	64.0099 KOps/s	$\color{#35bf28}+4.08\%$
test_compile_update_inplace[eager]	0.5093ms	59.5563μs	16.7908 KOps/s	17.0130 KOps/s	$\color{#d91a1a}-1.31\%$
test_compile_update_inplace[compile]	0.5708ms	0.1393ms	7.1775 KOps/s	6.9047 KOps/s	$\color{#35bf28}+3.95\%$
test_mod_add[eager]	0.4843ms	49.8419μs	20.0635 KOps/s	20.2420 KOps/s	$\color{#d91a1a}-0.88\%$
test_mod_add[compile]	0.3205ms	0.1043ms	9.5910 KOps/s	9.4114 KOps/s	$\color{#35bf28}+1.91\%$
test_mod_add[compile-overhead]	0.2357ms	0.1495ms	6.6886 KOps/s	6.5218 KOps/s	$\color{#35bf28}+2.56\%$
test_mod_wrap[eager]	0.7431ms	0.2907ms	3.4397 KOps/s	3.4074 KOps/s	$\color{#35bf28}+0.95\%$
test_mod_wrap[compile]	0.4420ms	0.3607ms	2.7724 KOps/s	2.8072 KOps/s	$\color{#d91a1a}-1.24\%$
test_mod_wrap[compile-overhead]	7.0034ms	3.7894ms	263.8942 Ops/s	250.1506 Ops/s	$\textbf{\color{#35bf28}+5.49\%}$
test_mod_wrap_and_backward[eager]	1.6084ms	1.4942ms	669.2355 Ops/s	661.8537 Ops/s	$\color{#35bf28}+1.12\%$
test_mod_wrap_and_backward[compile]	1.5289ms	1.4413ms	693.8331 Ops/s	678.4257 Ops/s	$\color{#35bf28}+2.27\%$
test_mod_wrap_and_backward[compile-overhead]	1.3034ms	0.8939ms	1.1187 KOps/s	1.0945 KOps/s	$\color{#35bf28}+2.21\%$
test_seq_add[eager]	0.2372ms	0.1610ms	6.2105 KOps/s	6.4532 KOps/s	$\color{#d91a1a}-3.76\%$
test_seq_add[compile]	0.5209ms	0.1187ms	8.4236 KOps/s	7.9138 KOps/s	$\textbf{\color{#35bf28}+6.44\%}$
test_seq_add[compile-overhead]	0.2302ms	0.1614ms	6.1952 KOps/s	6.2071 KOps/s	$\color{#d91a1a}-0.19\%$
test_seq_wrap[eager]	0.6240ms	0.5464ms	1.8302 KOps/s	1.9115 KOps/s	$\color{#d91a1a}-4.25\%$
test_seq_wrap[compile]	0.4235ms	0.3667ms	2.7273 KOps/s	2.6553 KOps/s	$\color{#35bf28}+2.71\%$
test_seq_wrap[compile-overhead]	0.3346ms	0.2640ms	3.7884 KOps/s	3.6608 KOps/s	$\color{#35bf28}+3.49\%$
test_func_call_runtime[False-eager]	0.9486ms	0.8753ms	1.1424 KOps/s	1.1775 KOps/s	$\color{#d91a1a}-2.98\%$
test_func_call_runtime[False-compile]	1.1083ms	0.9222ms	1.0843 KOps/s	1.0664 KOps/s	$\color{#35bf28}+1.68\%$
test_func_call_runtime[False-compile-overhead]	0.5218ms	0.4641ms	2.1546 KOps/s	2.1392 KOps/s	$\color{#35bf28}+0.72\%$
test_func_call_runtime[True-eager]	1.2159ms	1.0936ms	914.4053 Ops/s	910.2943 Ops/s	$\color{#35bf28}+0.45\%$
test_func_call_runtime[True-compile]	1.0116ms	0.9388ms	1.0652 KOps/s	1.0605 KOps/s	$\color{#35bf28}+0.44\%$
test_func_call_runtime[True-compile-overhead]	0.5583ms	0.4766ms	2.0982 KOps/s	2.0642 KOps/s	$\color{#35bf28}+1.64\%$
test_func_call_cm_runtime[False-eager]	1.0179ms	0.8406ms	1.1897 KOps/s	1.1708 KOps/s	$\color{#35bf28}+1.61\%$
test_func_call_cm_runtime[False-compile]	1.0422ms	0.9502ms	1.0524 KOps/s	1.0668 KOps/s	$\color{#d91a1a}-1.35\%$
test_func_call_cm_runtime[False-compile-overhead]	0.6329ms	0.4631ms	2.1593 KOps/s	2.1140 KOps/s	$\color{#35bf28}+2.15\%$
test_func_call_cm_runtime[True-eager]	1.3953ms	1.2321ms	811.5934 Ops/s	800.7785 Ops/s	$\color{#35bf28}+1.35\%$
test_func_call_cm_runtime[True-compile]	1.1294ms	0.9556ms	1.0465 KOps/s	1.0183 KOps/s	$\color{#35bf28}+2.77\%$
test_func_call_cm_runtime[True-compile-overhead]	0.5987ms	0.5091ms	1.9642 KOps/s	1.9209 KOps/s	$\color{#35bf28}+2.25\%$
test_vmap_func_call_cm_runtime[eager]	2.8791ms	2.3773ms	420.6493 Ops/s	415.8002 Ops/s	$\color{#35bf28}+1.17\%$
test_vmap_func_call_cm_runtime[compile]	1.0982ms	0.9811ms	1.0193 KOps/s	991.3028 Ops/s	$\color{#35bf28}+2.82\%$
test_vmap_func_call_cm_runtime[compile-overhead]	0.5636ms	0.5146ms	1.9434 KOps/s	1.9079 KOps/s	$\color{#35bf28}+1.86\%$
test_distributed	0.6722ms	0.1534ms	6.5190 KOps/s	6.3736 KOps/s	$\color{#35bf28}+2.28\%$
test_tdmodule	0.3947ms	27.6611μs	36.1519 KOps/s	35.5631 KOps/s	$\color{#35bf28}+1.66\%$
test_tdmodule_dispatch	78.6320μs	45.1207μs	22.1628 KOps/s	21.9093 KOps/s	$\color{#35bf28}+1.16\%$
test_tdseq	77.5110μs	26.7269μs	37.4154 KOps/s	37.3301 KOps/s	$\color{#35bf28}+0.23\%$
test_tdseq_dispatch	69.4310μs	46.7983μs	21.3683 KOps/s	20.9876 KOps/s	$\color{#35bf28}+1.81\%$
test_instantiation_functorch	2.1454ms	2.0844ms	479.7445 Ops/s	476.9101 Ops/s	$\color{#35bf28}+0.59\%$
test_exec_functorch	0.2475ms	0.1777ms	5.6268 KOps/s	5.4605 KOps/s	$\color{#35bf28}+3.05\%$
test_exec_functional_call	0.2105ms	0.1615ms	6.1918 KOps/s	6.0892 KOps/s	$\color{#35bf28}+1.69\%$
test_exec_td_decorator	0.4380ms	0.2362ms	4.2330 KOps/s	4.0755 KOps/s	$\color{#35bf28}+3.86\%$
test_vmap_mlp_speed_decorator[True-True]	1.0419ms	0.8321ms	1.2018 KOps/s	1.1580 KOps/s	$\color{#35bf28}+3.78\%$
test_vmap_mlp_speed_decorator[True-False]	1.0400ms	0.8300ms	1.2048 KOps/s	1.1590 KOps/s	$\color{#35bf28}+3.95\%$
test_vmap_mlp_speed_decorator[False-True]	1.0039ms	0.7186ms	1.3916 KOps/s	1.3467 KOps/s	$\color{#35bf28}+3.34\%$
test_vmap_mlp_speed_decorator[False-False]	0.8929ms	0.7156ms	1.3974 KOps/s	1.3345 KOps/s	$\color{#35bf28}+4.71\%$
test_vmap_transformer_speed_decorator[True-True]	21.2028ms	20.6326ms	48.4671 Ops/s	47.9364 Ops/s	$\color{#35bf28}+1.11\%$
test_vmap_transformer_speed_decorator[True-False]	21.3554ms	20.6374ms	48.4558 Ops/s	47.9084 Ops/s	$\color{#35bf28}+1.14\%$
test_vmap_transformer_speed_decorator[False-True]	21.1218ms	20.4469ms	48.9072 Ops/s	48.3888 Ops/s	$\color{#35bf28}+1.07\%$
test_vmap_transformer_speed_decorator[False-False]	20.6631ms	20.4102ms	48.9951 Ops/s	48.3589 Ops/s	$\color{#35bf28}+1.32\%$
test_to_module_speed[True]	1.5686ms	1.4869ms	672.5447 Ops/s	675.5950 Ops/s	$\color{#d91a1a}-0.45\%$
test_to_module_speed[False]	1.5552ms	1.4690ms	680.7153 Ops/s	687.2723 Ops/s	$\color{#d91a1a}-0.95\%$
test_tc_init	73.7910μs	44.3123μs	22.5671 KOps/s	22.3258 KOps/s	$\color{#35bf28}+1.08\%$
test_tc_init_tensor_only	33.5810μs	9.7679μs	102.3764 KOps/s	103.3791 KOps/s	$\color{#d91a1a}-0.97\%$
test_tc_init_nested	0.1233ms	88.1582μs	11.3432 KOps/s	11.1883 KOps/s	$\color{#35bf28}+1.39\%$
test_tc_init_many_fields	39.4500μs	16.3144μs	61.2957 KOps/s	60.8914 KOps/s	$\color{#35bf28}+0.66\%$
test_tc_first_layer_tensor	31.5700μs	1.8372μs	544.3064 KOps/s	551.8987 KOps/s	$\color{#d91a1a}-1.38\%$
test_tc_first_layer_tensor_only	1.5495μs	0.3963μs	2.5231 MOps/s	2.5109 MOps/s	$\color{#35bf28}+0.49\%$
test_tc_first_layer_tensor_set	27.0400μs	3.9351μs	254.1210 KOps/s	254.2313 KOps/s	$\color{#d91a1a}-0.04\%$
test_tc_first_layer_tensor_only_set	24.5600μs	3.2652μs	306.2606 KOps/s	300.6727 KOps/s	$\color{#35bf28}+1.86\%$
test_tc_first_layer_nontensor	26.6900μs	6.2300μs	160.5142 KOps/s	160.7150 KOps/s	$\color{#d91a1a}-0.12\%$
test_tc_second_layer_tensor	34.4210μs	4.4901μs	222.7108 KOps/s	223.7364 KOps/s	$\color{#d91a1a}-0.46\%$
test_tc_second_layer_nontensor	42.8510μs	8.8142μs	113.4534 KOps/s	115.1649 KOps/s	$\color{#d91a1a}-1.49\%$
test_unbind	0.2649s	17.9828ms	55.6088 Ops/s	66.5562 Ops/s	$\textbf{\color{#d91a1a}-16.45\%}$
test_full_like	7.5270ms	4.3436ms	230.2237 Ops/s	59.2888 Ops/s	$\textbf{\color{#35bf28}+288.31\%}$
test_zeros_like	5.0215ms	4.3789ms	228.3676 Ops/s	59.3357 Ops/s	$\textbf{\color{#35bf28}+284.87\%}$
test_ones_like	4.5911ms	4.3751ms	228.5672 Ops/s	59.5702 Ops/s	$\textbf{\color{#35bf28}+283.69\%}$
test_clone	6.5930ms	6.4375ms	155.3397 Ops/s	56.2241 Ops/s	$\textbf{\color{#35bf28}+176.29\%}$
test_squeeze	79.2910μs	13.9319μs	71.7776 KOps/s	69.9429 KOps/s	$\color{#35bf28}+2.62\%$
test_unsqueeze	0.1745ms	0.1116ms	8.9595 KOps/s	8.9747 KOps/s	$\color{#d91a1a}-0.17\%$
test_split	0.2935ms	0.1828ms	5.4715 KOps/s	5.3034 KOps/s	$\color{#35bf28}+3.17\%$
test_permute	0.3204ms	0.2027ms	4.9329 KOps/s	4.6784 KOps/s	$\textbf{\color{#35bf28}+5.44\%}$
test_stack	55.0164ms	54.2037ms	18.4489 Ops/s	19.3471 Ops/s	$\color{#d91a1a}-4.64\%$
test_cat	54.1943ms	51.6324ms	19.3677 Ops/s	19.4073 Ops/s	$\color{#d91a1a}-0.20\%$
test_sequential_tensordict	0.6281ms	0.2215ms	4.5150 KOps/s	4.5973 KOps/s	$\color{#d91a1a}-1.79\%$
test_sequential_graph_module	0.1588ms	0.1263ms	7.9204 KOps/s	8.1941 KOps/s	$\color{#d91a1a}-3.34\%$
test_nested_tensordict	0.3384ms	0.2841ms	3.5201 KOps/s	3.5288 KOps/s	$\color{#d91a1a}-0.25\%$
test_nested_graph_module	0.5495ms	0.1359ms	7.3568 KOps/s	7.3948 KOps/s	$\color{#d91a1a}-0.51\%$

github-actions · 2026-03-06T15:20:44Z

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

Total Benchmarks: 261. Improved: $\large\color{#35bf28}24$. Worsened: $\large\color{#d91a1a}5$.

Expand to view detailed results

Name	Max	Mean	Ops	Ops on Repo `HEAD`	Change
test_plain_set_nested	29.9510μs	14.4746μs	69.0863 KOps/s	70.3572 KOps/s	$\color{#d91a1a}-1.81\%$
test_plain_set_stack_nested	38.4310μs	14.8071μs	67.5354 KOps/s	68.4167 KOps/s	$\color{#d91a1a}-1.29\%$
test_plain_set_nested_inplace	37.3010μs	16.2131μs	61.6785 KOps/s	61.8554 KOps/s	$\color{#d91a1a}-0.29\%$
test_plain_set_stack_nested_inplace	40.6910μs	16.2810μs	61.4213 KOps/s	62.6643 KOps/s	$\color{#d91a1a}-1.98\%$
test_items	33.8000μs	5.5689μs	179.5691 KOps/s	179.9665 KOps/s	$\color{#d91a1a}-0.22\%$
test_items_nested	0.4881ms	0.4490ms	2.2271 KOps/s	2.2359 KOps/s	$\color{#d91a1a}-0.39\%$
test_items_nested_locked	0.5027ms	0.4532ms	2.2064 KOps/s	2.2107 KOps/s	$\color{#d91a1a}-0.19\%$
test_items_nested_leaf	0.1512ms	91.9447μs	10.8761 KOps/s	10.8080 KOps/s	$\color{#35bf28}+0.63\%$
test_items_stack_nested	0.4797ms	0.4505ms	2.2195 KOps/s	2.2440 KOps/s	$\color{#d91a1a}-1.09\%$
test_items_stack_nested_leaf	0.1541ms	94.4595μs	10.5865 KOps/s	10.6554 KOps/s	$\color{#d91a1a}-0.65\%$
test_items_stack_nested_locked	0.5449ms	0.4538ms	2.2038 KOps/s	2.2231 KOps/s	$\color{#d91a1a}-0.87\%$
test_keys	28.0310μs	4.1360μs	241.7796 KOps/s	240.9997 KOps/s	$\color{#35bf28}+0.32\%$
test_keys_nested	0.1816ms	0.1283ms	7.7942 KOps/s	7.8401 KOps/s	$\color{#d91a1a}-0.58\%$
test_keys_nested_locked	0.7720ms	0.1357ms	7.3682 KOps/s	7.3580 KOps/s	$\color{#35bf28}+0.14\%$
test_keys_nested_leaf	0.1639ms	0.1186ms	8.4295 KOps/s	8.5272 KOps/s	$\color{#d91a1a}-1.15\%$
test_keys_stack_nested	0.1691ms	0.1290ms	7.7523 KOps/s	7.8091 KOps/s	$\color{#d91a1a}-0.73\%$
test_keys_stack_nested_leaf	0.1495ms	0.1183ms	8.4512 KOps/s	8.4457 KOps/s	$\color{#35bf28}+0.07\%$
test_keys_stack_nested_locked	0.1639ms	0.1363ms	7.3342 KOps/s	7.3446 KOps/s	$\color{#d91a1a}-0.14\%$
test_values	6.1502μs	1.0115μs	988.5902 KOps/s	998.6009 KOps/s	$\color{#d91a1a}-1.00\%$
test_values_nested	75.9710μs	51.5790μs	19.3877 KOps/s	19.6293 KOps/s	$\color{#d91a1a}-1.23\%$
test_values_nested_locked	90.5910μs	55.1383μs	18.1362 KOps/s	18.4124 KOps/s	$\color{#d91a1a}-1.50\%$
test_values_nested_leaf	0.1064ms	59.1945μs	16.8935 KOps/s	17.0319 KOps/s	$\color{#d91a1a}-0.81\%$
test_values_stack_nested	85.0210μs	51.7311μs	19.3307 KOps/s	19.5398 KOps/s	$\color{#d91a1a}-1.07\%$
test_values_stack_nested_leaf	93.0620μs	59.2799μs	16.8691 KOps/s	17.1487 KOps/s	$\color{#d91a1a}-1.63\%$
test_values_stack_nested_locked	0.1002ms	55.2764μs	18.0909 KOps/s	18.5954 KOps/s	$\color{#d91a1a}-2.71\%$
test_membership	4.8418μs	0.8146μs	1.2275 MOps/s	1.2204 MOps/s	$\color{#35bf28}+0.59\%$
test_membership_nested	32.6310μs	2.7648μs	361.6950 KOps/s	362.3219 KOps/s	$\color{#d91a1a}-0.17\%$
test_membership_nested_leaf	19.3455μs	2.6404μs	378.7293 KOps/s	360.8127 KOps/s	$\color{#35bf28}+4.97\%$
test_membership_stacked_nested	31.9000μs	2.7177μs	367.9635 KOps/s	362.2735 KOps/s	$\color{#35bf28}+1.57\%$
test_membership_stacked_nested_leaf	28.9600μs	2.7408μs	364.8513 KOps/s	366.4649 KOps/s	$\color{#d91a1a}-0.44\%$
test_membership_nested_last	32.3000μs	4.1465μs	241.1668 KOps/s	241.7224 KOps/s	$\color{#d91a1a}-0.23\%$
test_membership_nested_leaf_last	34.4110μs	4.1278μs	242.2603 KOps/s	241.9704 KOps/s	$\color{#35bf28}+0.12\%$
test_membership_stacked_nested_last	28.0000μs	4.1215μs	242.6305 KOps/s	244.2299 KOps/s	$\color{#d91a1a}-0.65\%$
test_membership_stacked_nested_leaf_last	39.4810μs	4.1062μs	243.5338 KOps/s	243.0951 KOps/s	$\color{#35bf28}+0.18\%$
test_nested_getleaf	50.1810μs	21.0310μs	47.5489 KOps/s	49.6247 KOps/s	$\color{#d91a1a}-4.18\%$
test_nested_get	53.8210μs	19.4881μs	51.3133 KOps/s	52.6842 KOps/s	$\color{#d91a1a}-2.60\%$
test_stacked_getleaf	80.2220μs	20.3114μs	49.2334 KOps/s	50.1236 KOps/s	$\color{#d91a1a}-1.78\%$
test_stacked_get	42.6910μs	19.5531μs	51.1427 KOps/s	52.5698 KOps/s	$\color{#d91a1a}-2.71\%$
test_nested_getitemleaf	65.5010μs	20.8844μs	47.8826 KOps/s	49.0730 KOps/s	$\color{#d91a1a}-2.43\%$
test_nested_getitem	46.5310μs	19.7303μs	50.6834 KOps/s	51.4573 KOps/s	$\color{#d91a1a}-1.50\%$
test_stacked_getitemleaf	45.9010μs	20.8309μs	48.0056 KOps/s	48.9079 KOps/s	$\color{#d91a1a}-1.84\%$
test_stacked_getitem	62.2820μs	20.0357μs	49.9110 KOps/s	51.1194 KOps/s	$\color{#d91a1a}-2.36\%$
test_lock_nested	8.0686ms	0.4650ms	2.1506 KOps/s	2.1924 KOps/s	$\color{#d91a1a}-1.91\%$
test_lock_stack_nested	0.4932ms	0.4570ms	2.1883 KOps/s	2.1611 KOps/s	$\color{#35bf28}+1.26\%$
test_unlock_nested	0.4447ms	0.3723ms	2.6862 KOps/s	2.6995 KOps/s	$\color{#d91a1a}-0.49\%$
test_unlock_stack_nested	0.4151ms	0.3709ms	2.6964 KOps/s	2.6729 KOps/s	$\color{#35bf28}+0.88\%$
test_flatten_speed	0.1597ms	0.1174ms	8.5172 KOps/s	8.5876 KOps/s	$\color{#d91a1a}-0.82\%$
test_unflatten_speed	0.5794ms	0.5477ms	1.8257 KOps/s	1.8296 KOps/s	$\color{#d91a1a}-0.21\%$
test_common_ops	0.8562ms	0.6781ms	1.4747 KOps/s	1.4584 KOps/s	$\color{#35bf28}+1.11\%$
test_creation	70.7510μs	2.9560μs	338.2925 KOps/s	337.5566 KOps/s	$\color{#35bf28}+0.22\%$
test_creation_empty	45.1310μs	6.6304μs	150.8195 KOps/s	152.5791 KOps/s	$\color{#d91a1a}-1.15\%$
test_creation_nested_1	44.8910μs	11.0896μs	90.1742 KOps/s	91.5865 KOps/s	$\color{#d91a1a}-1.54\%$
test_creation_nested_2	38.1510μs	12.5533μs	79.6606 KOps/s	79.5836 KOps/s	$\color{#35bf28}+0.10\%$
test_creation_many_keys[10]	40.3900μs	19.7750μs	50.5690 KOps/s	50.4786 KOps/s	$\color{#35bf28}+0.18\%$
test_creation_many_keys[50]	0.1159ms	85.7428μs	11.6628 KOps/s	11.9174 KOps/s	$\color{#d91a1a}-2.14\%$
test_creation_many_keys[100]	0.2247ms	0.1703ms	5.8727 KOps/s	5.9365 KOps/s	$\color{#d91a1a}-1.08\%$
test_creation_nested_many_keys[10]	72.4320μs	42.7515μs	23.3910 KOps/s	23.4469 KOps/s	$\color{#d91a1a}-0.24\%$
test_creation_nested_many_keys[50]	0.2127ms	0.1751ms	5.7115 KOps/s	5.7755 KOps/s	$\color{#d91a1a}-1.11\%$
test_clone	39.7000μs	12.4730μs	80.1731 KOps/s	75.7557 KOps/s	$\textbf{\color{#35bf28}+5.83\%}$
test_getitem[int]	1.6508ms	14.5103μs	68.9166 KOps/s	62.1008 KOps/s	$\textbf{\color{#35bf28}+10.98\%}$
test_getitem[slice_int]	0.1313ms	23.1169μs	43.2583 KOps/s	43.0199 KOps/s	$\color{#35bf28}+0.55\%$
test_getitem[range]	0.1716ms	60.3854μs	16.5603 KOps/s	16.3056 KOps/s	$\color{#35bf28}+1.56\%$
test_getitem[tuple]	0.1389ms	22.6100μs	44.2281 KOps/s	43.6674 KOps/s	$\color{#35bf28}+1.28\%$
test_getitem[list]	0.1736ms	57.1289μs	17.5043 KOps/s	17.6820 KOps/s	$\color{#d91a1a}-1.01\%$
test_setitem_dim[int]	45.2610μs	25.3471μs	39.4522 KOps/s	39.5695 KOps/s	$\color{#d91a1a}-0.30\%$
test_setitem_dim[slice_int]	74.6610μs	41.7705μs	23.9403 KOps/s	23.8929 KOps/s	$\color{#35bf28}+0.20\%$
test_setitem_dim[range]	0.1276ms	91.3920μs	10.9419 KOps/s	10.4145 KOps/s	$\textbf{\color{#35bf28}+5.06\%}$
test_setitem_dim[tuple]	59.9810μs	38.7164μs	25.8289 KOps/s	24.4909 KOps/s	$\textbf{\color{#35bf28}+5.46\%}$
test_setitem	54.5310μs	16.9992μs	58.8265 KOps/s	55.2165 KOps/s	$\textbf{\color{#35bf28}+6.54\%}$
test_set	0.5076ms	16.0838μs	62.1744 KOps/s	59.7936 KOps/s	$\color{#35bf28}+3.98\%$
test_set_shared	0.6236ms	0.2054ms	4.8674 KOps/s	4.7702 KOps/s	$\color{#35bf28}+2.04\%$
test_update	0.1869ms	21.0581μs	47.4876 KOps/s	43.9517 KOps/s	$\textbf{\color{#35bf28}+8.05\%}$
test_update_nested	64.7120μs	31.6188μs	31.6268 KOps/s	29.4382 KOps/s	$\textbf{\color{#35bf28}+7.43\%}$
test_update__nested	0.5234ms	32.8994μs	30.3957 KOps/s	27.7926 KOps/s	$\textbf{\color{#35bf28}+9.37\%}$
test_set_nested	54.6810μs	19.1531μs	52.2109 KOps/s	49.6973 KOps/s	$\textbf{\color{#35bf28}+5.06\%}$
test_set_nested_new	59.6210μs	23.0374μs	43.4077 KOps/s	40.0664 KOps/s	$\textbf{\color{#35bf28}+8.34\%}$
test_select	68.8410μs	38.4854μs	25.9839 KOps/s	24.8204 KOps/s	$\color{#35bf28}+4.69\%$
test_select_nested	99.9020μs	71.1734μs	14.0502 KOps/s	14.0214 KOps/s	$\color{#35bf28}+0.21\%$
test_exclude_nested	0.1148ms	87.2090μs	11.4667 KOps/s	11.3100 KOps/s	$\color{#35bf28}+1.39\%$
test_empty[True]	0.4556ms	0.3842ms	2.6025 KOps/s	2.5994 KOps/s	$\color{#35bf28}+0.12\%$
test_empty[False]	6.3427μs	1.2565μs	795.8755 KOps/s	793.4727 KOps/s	$\color{#35bf28}+0.30\%$
test_to	0.1027ms	71.6340μs	13.9599 KOps/s	14.0452 KOps/s	$\color{#d91a1a}-0.61\%$
test_to_nonblocking	0.1110ms	67.7509μs	14.7599 KOps/s	15.6195 KOps/s	$\textbf{\color{#d91a1a}-5.50\%}$
test_unbind_speed	0.3511ms	0.3199ms	3.1262 KOps/s	3.1617 KOps/s	$\color{#d91a1a}-1.12\%$
test_unbind_speed_stack0	0.3829ms	0.3173ms	3.1515 KOps/s	3.1851 KOps/s	$\color{#d91a1a}-1.05\%$
test_unbind_speed_stack1	0.1040s	0.8081ms	1.2375 KOps/s	1.2369 KOps/s	$\color{#35bf28}+0.05\%$
test_split	0.1040s	1.2188ms	820.4528 Ops/s	827.3493 Ops/s	$\color{#d91a1a}-0.83\%$
test_chunk	0.1043s	1.1565ms	864.6781 Ops/s	968.2886 Ops/s	$\textbf{\color{#d91a1a}-10.70\%}$
test_to_cpu_blocking	19.5850ms	19.5258ms	51.2142 Ops/s	46.2068 Ops/s	$\textbf{\color{#35bf28}+10.84\%}$
test_to_cpu_global_sync	11.3958ms	11.3046ms	88.4599 Ops/s	88.5779 Ops/s	$\color{#d91a1a}-0.13\%$
test_to_cpu_event_sync	12.6012ms	12.3321ms	81.0890 Ops/s	81.2191 Ops/s	$\color{#d91a1a}-0.16\%$
test_to_cpu_default	12.6064ms	12.3354ms	81.0675 Ops/s	81.1779 Ops/s	$\color{#d91a1a}-0.14\%$
test_consolidate[False-None]	4.0282ms	3.9598ms	252.5374 Ops/s	222.9650 Ops/s	$\textbf{\color{#35bf28}+13.26\%}$
test_consolidate[default-None]	2.0184ms	1.9238ms	519.8022 Ops/s	502.6801 Ops/s	$\color{#35bf28}+3.41\%$
test_consolidate[reduce-overhead-None]	1.9170ms	1.8431ms	542.5663 Ops/s	520.1939 Ops/s	$\color{#35bf28}+4.30\%$
test_consolidate_njt[False-None]	8.4179ms	8.1817ms	122.2241 Ops/s	120.7256 Ops/s	$\color{#35bf28}+1.24\%$
test_to[False-False-None]	2.1522ms	2.0648ms	484.2979 Ops/s	479.8149 Ops/s	$\color{#35bf28}+0.93\%$
test_to[True-False-None]	2.1716ms	1.8522ms	539.8931 Ops/s	525.5844 Ops/s	$\color{#35bf28}+2.72\%$
test_to[within-False-None]	6.2031ms	5.8644ms	170.5191 Ops/s	165.6574 Ops/s	$\color{#35bf28}+2.93\%$
test_to[True-default-None]	8.8218ms	8.6810ms	115.1945 Ops/s	113.4335 Ops/s	$\color{#35bf28}+1.55\%$
test_to_njt[False-False-None]	8.4674ms	8.2635ms	121.0147 Ops/s	120.3563 Ops/s	$\color{#35bf28}+0.55\%$
test_to_njt[True-False-None]	6.8743ms	6.7520ms	148.1041 Ops/s	148.1472 Ops/s	$\color{#d91a1a}-0.03\%$
test_to_njt[within-False-None]	15.2683ms	15.1747ms	65.8990 Ops/s	66.3607 Ops/s	$\color{#d91a1a}-0.70\%$
test_creation[device0]	0.3525ms	0.1117ms	8.9509 KOps/s	8.4630 KOps/s	$\textbf{\color{#35bf28}+5.76\%}$
test_creation_from_tensor	0.3916ms	0.1091ms	9.1635 KOps/s	8.6482 KOps/s	$\textbf{\color{#35bf28}+5.96\%}$
test_add_one[memmap_tensor0]	0.2254ms	6.1031μs	163.8518 KOps/s	156.8092 KOps/s	$\color{#35bf28}+4.49\%$
test_contiguous[memmap_tensor0]	17.3100μs	0.6025μs	1.6596 MOps/s	2.2522 MOps/s	$\textbf{\color{#d91a1a}-26.31\%}$
test_stack[memmap_tensor0]	29.8100μs	4.3685μs	228.9096 KOps/s	215.2068 KOps/s	$\textbf{\color{#35bf28}+6.37\%}$
test_memmaptd_index	1.0549ms	0.2620ms	3.8165 KOps/s	3.8182 KOps/s	$\color{#d91a1a}-0.04\%$
test_memmaptd_index_astensor	0.5168ms	0.3636ms	2.7501 KOps/s	2.7762 KOps/s	$\color{#d91a1a}-0.94\%$
test_memmaptd_index_op	0.8357ms	0.5952ms	1.6800 KOps/s	1.6445 KOps/s	$\color{#35bf28}+2.16\%$
test_serialize_model	0.1401s	0.1364s	7.3334 Ops/s	5.8654 Ops/s	$\textbf{\color{#35bf28}+25.03\%}$
test_serialize_model_pickle	1.3480s	1.2102s	0.8263 Ops/s	0.8384 Ops/s	$\color{#d91a1a}-1.44\%$
test_serialize_weights	0.1374s	0.1355s	7.3786 Ops/s	7.3286 Ops/s	$\color{#35bf28}+0.68\%$
test_serialize_weights_returnearly	0.4333s	91.4515ms	10.9348 Ops/s	10.5539 Ops/s	$\color{#35bf28}+3.61\%$
test_serialize_weights_pickle	1.3507s	1.2109s	0.8258 Ops/s	0.8227 Ops/s	$\color{#35bf28}+0.38\%$
test_reshape_pytree	0.2225ms	30.8765μs	32.3871 KOps/s	31.2228 KOps/s	$\color{#35bf28}+3.73\%$
test_reshape_td	68.3210μs	43.9953μs	22.7297 KOps/s	22.8735 KOps/s	$\color{#d91a1a}-0.63\%$
test_view_pytree	0.2150ms	30.5850μs	32.6958 KOps/s	31.6981 KOps/s	$\color{#35bf28}+3.15\%$
test_view_td	76.1610μs	50.3693μs	19.8534 KOps/s	20.0181 KOps/s	$\color{#d91a1a}-0.82\%$
test_unbind_pytree	0.2325ms	34.4356μs	29.0397 KOps/s	28.1408 KOps/s	$\color{#35bf28}+3.19\%$
test_unbind_td	0.1075ms	47.5965μs	21.0099 KOps/s	21.1483 KOps/s	$\color{#d91a1a}-0.65\%$
test_split_pytree	0.2554ms	39.7371μs	25.1654 KOps/s	24.4558 KOps/s	$\color{#35bf28}+2.90\%$
test_split_td	0.1485ms	58.7680μs	17.0161 KOps/s	16.3856 KOps/s	$\color{#35bf28}+3.85\%$
test_add_pytree	0.2255ms	39.7526μs	25.1556 KOps/s	24.4978 KOps/s	$\color{#35bf28}+2.69\%$
test_add_td	0.1045ms	52.0055μs	19.2287 KOps/s	18.8809 KOps/s	$\color{#35bf28}+1.84\%$
test_compile_add_one_nested[tensordict-compile]	0.2125ms	0.1397ms	7.1602 KOps/s	7.0295 KOps/s	$\color{#35bf28}+1.86\%$
test_compile_add_one_nested[tensordict-eager]	0.4245ms	0.2070ms	4.8309 KOps/s	5.1782 KOps/s	$\textbf{\color{#d91a1a}-6.71\%}$
test_compile_add_one_nested[pytree-compile]	0.6752ms	0.1082ms	9.2421 KOps/s	8.7872 KOps/s	$\textbf{\color{#35bf28}+5.18\%}$
test_compile_add_one_nested[pytree-eager]	0.6522ms	0.1865ms	5.3630 KOps/s	5.6410 KOps/s	$\color{#d91a1a}-4.93\%$
test_compile_copy_nested[tensordict-compile]	0.2503ms	10.3265μs	96.8384 KOps/s	103.5455 KOps/s	$\textbf{\color{#d91a1a}-6.48\%}$
test_compile_copy_nested[tensordict-eager]	77.2220μs	50.9487μs	19.6276 KOps/s	19.4398 KOps/s	$\color{#35bf28}+0.97\%$
test_compile_copy_nested[pytree-compile]	46.2610μs	9.5292μs	104.9408 KOps/s	104.7371 KOps/s	$\color{#35bf28}+0.19\%$
test_compile_copy_nested[pytree-eager]	0.4473ms	64.2487μs	15.5645 KOps/s	15.2384 KOps/s	$\color{#35bf28}+2.14\%$
test_compile_add_one_flat[tensordict-compile]	0.2376ms	0.1742ms	5.7404 KOps/s	5.5242 KOps/s	$\color{#35bf28}+3.91\%$
test_compile_add_one_flat[tensordict-eager]	0.4061ms	0.2757ms	3.6270 KOps/s	3.6383 KOps/s	$\color{#d91a1a}-0.31\%$
test_compile_add_one_flat[tensorclass-compile]	0.1889ms	0.1167ms	8.5696 KOps/s	8.4864 KOps/s	$\color{#35bf28}+0.98\%$
test_compile_add_one_flat[tensorclass-eager]	0.1186ms	72.5933μs	13.7754 KOps/s	13.5592 KOps/s	$\color{#35bf28}+1.59\%$
test_compile_add_one_flat[pytree-compile]	0.2202ms	0.1565ms	6.3899 KOps/s	6.2737 KOps/s	$\color{#35bf28}+1.85\%$
test_compile_add_one_flat[pytree-eager]	0.9155ms	0.5268ms	1.8981 KOps/s	1.8948 KOps/s	$\color{#35bf28}+0.18\%$
test_compile_add_self_flat[tensordict-eager]	0.3842ms	0.3282ms	3.0469 KOps/s	3.0455 KOps/s	$\color{#35bf28}+0.05\%$
test_compile_add_self_flat[tensordict-compile]	0.2272ms	0.1773ms	5.6417 KOps/s	3.3697 KOps/s	$\textbf{\color{#35bf28}+67.43\%}$
test_compile_add_self_flat[tensorclass-eager]	0.1349ms	91.3949μs	10.9415 KOps/s	11.0864 KOps/s	$\color{#d91a1a}-1.31\%$
test_compile_add_self_flat[tensorclass-compile]	0.2070ms	0.1201ms	8.3282 KOps/s	7.8547 KOps/s	$\textbf{\color{#35bf28}+6.03\%}$
test_compile_add_self_flat[pytree-eager]	0.6685ms	0.4339ms	2.3048 KOps/s	2.2176 KOps/s	$\color{#35bf28}+3.93\%$
test_compile_add_self_flat[pytree-compile]	0.1944ms	0.1554ms	6.4341 KOps/s	6.2204 KOps/s	$\color{#35bf28}+3.43\%$
test_compile_copy_flat[tensordict-compile]	43.3010μs	13.1670μs	75.9477 KOps/s	74.8361 KOps/s	$\color{#35bf28}+1.49\%$
test_compile_copy_flat[tensordict-eager]	74.3010μs	40.6168μs	24.6204 KOps/s	24.9185 KOps/s	$\color{#d91a1a}-1.20\%$
test_compile_copy_flat[pytree-compile]	66.2010μs	10.3497μs	96.6216 KOps/s	94.5111 KOps/s	$\color{#35bf28}+2.23\%$
test_compile_copy_flat[pytree-eager]	0.4259ms	50.5396μs	19.7865 KOps/s	19.5230 KOps/s	$\color{#35bf28}+1.35\%$
test_compile_assign_and_add[tensordict-compile]	1.9381ms	0.1700ms	5.8830 KOps/s	5.4184 KOps/s	$\textbf{\color{#35bf28}+8.57\%}$
test_compile_assign_and_add[tensordict-eager]	3.3482ms	3.2374ms	308.8885 Ops/s	302.0445 Ops/s	$\color{#35bf28}+2.27\%$
test_compile_assign_and_add[pytree-compile]	1.9173ms	0.1567ms	6.3829 KOps/s	6.2232 KOps/s	$\color{#35bf28}+2.57\%$
test_compile_assign_and_add[pytree-eager]	2.8839ms	2.7564ms	362.7860 Ops/s	356.4926 Ops/s	$\color{#35bf28}+1.77\%$
test_compile_indexing[tensor-tensordict-compile]	0.1446ms	0.1065ms	9.3867 KOps/s	8.8786 KOps/s	$\textbf{\color{#35bf28}+5.72\%}$
test_compile_indexing[tensor-tensordict-eager]	0.3095ms	73.8427μs	13.5423 KOps/s	13.9160 KOps/s	$\color{#d91a1a}-2.69\%$
test_compile_indexing[tensor-tensorclass-compile]	0.2027ms	96.2995μs	10.3843 KOps/s	10.4544 KOps/s	$\color{#d91a1a}-0.67\%$
test_compile_indexing[tensor-tensorclass-eager]	0.2503ms	43.2035μs	23.1463 KOps/s	21.6090 KOps/s	$\textbf{\color{#35bf28}+7.11\%}$
test_compile_indexing[tensor-pytree-compile]	0.1444ms	98.7164μs	10.1300 KOps/s	10.3920 KOps/s	$\color{#d91a1a}-2.52\%$
test_compile_indexing[tensor-pytree-eager]	0.2698ms	45.1328μs	22.1568 KOps/s	22.9686 KOps/s	$\color{#d91a1a}-3.53\%$
test_compile_indexing[slice-tensordict-compile]	0.1099ms	57.2495μs	17.4674 KOps/s	17.1955 KOps/s	$\color{#35bf28}+1.58\%$
test_compile_indexing[slice-tensordict-eager]	0.2254ms	27.0528μs	36.9647 KOps/s	38.0890 KOps/s	$\color{#d91a1a}-2.95\%$
test_compile_indexing[slice-tensorclass-compile]	96.9520μs	42.8970μs	23.3116 KOps/s	22.4267 KOps/s	$\color{#35bf28}+3.95\%$
test_compile_indexing[slice-tensorclass-eager]	0.2668ms	21.3862μs	46.7591 KOps/s	46.2092 KOps/s	$\color{#35bf28}+1.19\%$
test_compile_indexing[slice-pytree-compile]	75.8610μs	44.5869μs	22.4281 KOps/s	22.2481 KOps/s	$\color{#35bf28}+0.81\%$
test_compile_indexing[slice-pytree-eager]	0.2604ms	21.5860μs	46.3264 KOps/s	46.5913 KOps/s	$\color{#d91a1a}-0.57\%$
test_compile_indexing[int-tensordict-compile]	0.1069ms	57.7712μs	17.3097 KOps/s	17.1301 KOps/s	$\color{#35bf28}+1.05\%$
test_compile_indexing[int-tensordict-eager]	0.2233ms	26.7918μs	37.3248 KOps/s	37.8390 KOps/s	$\color{#d91a1a}-1.36\%$
test_compile_indexing[int-tensorclass-compile]	82.7920μs	44.1563μs	22.6468 KOps/s	22.3488 KOps/s	$\color{#35bf28}+1.33\%$
test_compile_indexing[int-tensorclass-eager]	0.2647ms	21.3704μs	46.7936 KOps/s	46.3613 KOps/s	$\color{#35bf28}+0.93\%$
test_compile_indexing[int-pytree-compile]	80.3020μs	44.3870μs	22.5291 KOps/s	22.3233 KOps/s	$\color{#35bf28}+0.92\%$
test_compile_indexing[int-pytree-eager]	0.2498ms	21.5562μs	46.3904 KOps/s	46.7378 KOps/s	$\color{#d91a1a}-0.74\%$
test_compile_replace[single-eager]	0.1005ms	48.2132μs	20.7412 KOps/s	21.3571 KOps/s	$\color{#d91a1a}-2.88\%$
test_compile_replace[single-compile]	0.1780ms	0.1026ms	9.7509 KOps/s	9.4054 KOps/s	$\color{#35bf28}+3.67\%$
test_compile_replace[multi-eager]	0.6414ms	0.5654ms	1.7686 KOps/s	1.8008 KOps/s	$\color{#d91a1a}-1.79\%$
test_compile_replace[multi-compile]	0.1669ms	0.1126ms	8.8821 KOps/s	8.9205 KOps/s	$\color{#d91a1a}-0.43\%$
test_compile_tc_getattr_20[eager]	0.2085ms	0.1683ms	5.9420 KOps/s	5.8394 KOps/s	$\color{#35bf28}+1.76\%$
test_compile_tc_getattr_20[compile]	0.1738ms	0.1192ms	8.3917 KOps/s	8.4679 KOps/s	$\color{#d91a1a}-0.90\%$
test_compile_clone_shallow[20-eager]	58.8520μs	18.6122μs	53.7282 KOps/s	54.3742 KOps/s	$\color{#d91a1a}-1.19\%$
test_compile_clone_shallow[20-compile]	39.9200μs	10.8501μs	92.1649 KOps/s	89.7936 KOps/s	$\color{#35bf28}+2.64\%$
test_compile_clone_shallow[40-eager]	69.0320μs	32.8805μs	30.4131 KOps/s	30.6981 KOps/s	$\color{#d91a1a}-0.93\%$
test_compile_clone_shallow[40-compile]	40.7010μs	12.1000μs	82.6446 KOps/s	82.2381 KOps/s	$\color{#35bf28}+0.49\%$
test_compile_clone_shallow[80-eager]	96.6420μs	61.9429μs	16.1439 KOps/s	16.4497 KOps/s	$\color{#d91a1a}-1.86\%$
test_compile_clone_shallow[80-compile]	50.1710μs	14.6803μs	68.1184 KOps/s	67.4496 KOps/s	$\color{#35bf28}+0.99\%$
test_compile_update_inplace[eager]	0.1045ms	57.8300μs	17.2921 KOps/s	17.3509 KOps/s	$\color{#d91a1a}-0.34\%$
test_compile_update_inplace[compile]	0.1896ms	0.1354ms	7.3872 KOps/s	7.1918 KOps/s	$\color{#35bf28}+2.72\%$
test_mod_add[eager]	93.5420μs	47.4959μs	21.0544 KOps/s	20.7287 KOps/s	$\color{#35bf28}+1.57\%$
test_mod_add[compile]	0.1569ms	0.1020ms	9.8038 KOps/s	9.1179 KOps/s	$\textbf{\color{#35bf28}+7.52\%}$
test_mod_add[compile-overhead]	0.2541ms	0.1460ms	6.8513 KOps/s	6.6825 KOps/s	$\color{#35bf28}+2.53\%$
test_mod_wrap[eager]	0.3559ms	0.2838ms	3.5240 KOps/s	3.3969 KOps/s	$\color{#35bf28}+3.74\%$
test_mod_wrap[compile]	0.4017ms	0.3399ms	2.9417 KOps/s	2.9073 KOps/s	$\color{#35bf28}+1.18\%$
test_mod_wrap[compile-overhead]	7.2408ms	3.9803ms	251.2402 Ops/s	248.7134 Ops/s	$\color{#35bf28}+1.02\%$
test_mod_wrap_and_backward[eager]	1.6520ms	1.4635ms	683.2720 Ops/s	659.7152 Ops/s	$\color{#35bf28}+3.57\%$
test_mod_wrap_and_backward[compile]	1.5053ms	1.4172ms	705.6383 Ops/s	702.0643 Ops/s	$\color{#35bf28}+0.51\%$
test_mod_wrap_and_backward[compile-overhead]	1.2463ms	0.8654ms	1.1555 KOps/s	1.1289 KOps/s	$\color{#35bf28}+2.36\%$
test_seq_add[eager]	0.2142ms	0.1560ms	6.4091 KOps/s	6.4102 KOps/s	$\color{#d91a1a}-0.02\%$
test_seq_add[compile]	0.5438ms	0.1115ms	8.9684 KOps/s	8.6403 KOps/s	$\color{#35bf28}+3.80\%$
test_seq_add[compile-overhead]	0.1953ms	0.1509ms	6.6279 KOps/s	6.4133 KOps/s	$\color{#35bf28}+3.35\%$
test_seq_wrap[eager]	0.5923ms	0.5235ms	1.9102 KOps/s	1.9489 KOps/s	$\color{#d91a1a}-1.99\%$
test_seq_wrap[compile]	0.4722ms	0.3733ms	2.6785 KOps/s	2.7610 KOps/s	$\color{#d91a1a}-2.99\%$
test_seq_wrap[compile-overhead]	0.3763ms	0.2570ms	3.8913 KOps/s	3.8327 KOps/s	$\color{#35bf28}+1.53\%$
test_func_call_runtime[False-eager]	0.9040ms	0.8082ms	1.2374 KOps/s	1.2070 KOps/s	$\color{#35bf28}+2.52\%$
test_func_call_runtime[False-compile]	0.9300ms	0.8836ms	1.1317 KOps/s	1.1039 KOps/s	$\color{#35bf28}+2.52\%$
test_func_call_runtime[False-compile-overhead]	0.4870ms	0.4431ms	2.2567 KOps/s	2.2417 KOps/s	$\color{#35bf28}+0.67\%$
test_func_call_runtime[True-eager]	1.1410ms	1.0417ms	959.9346 Ops/s	940.9236 Ops/s	$\color{#35bf28}+2.02\%$
test_func_call_runtime[True-compile]	0.9705ms	0.8895ms	1.1243 KOps/s	1.1109 KOps/s	$\color{#35bf28}+1.20\%$
test_func_call_runtime[True-compile-overhead]	0.5042ms	0.4565ms	2.1907 KOps/s	2.1678 KOps/s	$\color{#35bf28}+1.06\%$
test_func_call_cm_runtime[False-eager]	1.5462ms	0.8113ms	1.2327 KOps/s	1.2001 KOps/s	$\color{#35bf28}+2.71\%$
test_func_call_cm_runtime[False-compile]	1.1085ms	0.8823ms	1.1334 KOps/s	1.1226 KOps/s	$\color{#35bf28}+0.96\%$
test_func_call_cm_runtime[False-compile-overhead]	0.4966ms	0.4449ms	2.2478 KOps/s	2.2250 KOps/s	$\color{#35bf28}+1.03\%$
test_func_call_cm_runtime[True-eager]	1.2856ms	1.1809ms	846.7858 Ops/s	830.1097 Ops/s	$\color{#35bf28}+2.01\%$
test_func_call_cm_runtime[True-compile]	0.9874ms	0.9282ms	1.0773 KOps/s	1.0536 KOps/s	$\color{#35bf28}+2.26\%$
test_func_call_cm_runtime[True-compile-overhead]	0.5276ms	0.4893ms	2.0436 KOps/s	2.0194 KOps/s	$\color{#35bf28}+1.20\%$
test_vmap_func_call_cm_runtime[eager]	2.8286ms	2.3211ms	430.8296 Ops/s	428.5711 Ops/s	$\color{#35bf28}+0.53\%$
test_vmap_func_call_cm_runtime[compile]	1.0199ms	0.9534ms	1.0489 KOps/s	1.0426 KOps/s	$\color{#35bf28}+0.60\%$
test_vmap_func_call_cm_runtime[compile-overhead]	0.5469ms	0.4985ms	2.0059 KOps/s	2.0041 KOps/s	$\color{#35bf28}+0.09\%$
test_distributed	0.5721ms	0.1511ms	6.6186 KOps/s	5.7106 KOps/s	$\textbf{\color{#35bf28}+15.90\%}$
test_tdmodule	0.4250ms	27.2310μs	36.7229 KOps/s	37.5028 KOps/s	$\color{#d91a1a}-2.08\%$
test_tdmodule_dispatch	73.5220μs	44.2255μs	22.6114 KOps/s	22.8906 KOps/s	$\color{#d91a1a}-1.22\%$
test_tdseq	46.2900μs	26.2316μs	38.1219 KOps/s	38.1126 KOps/s	$\color{#35bf28}+0.02\%$
test_tdseq_dispatch	66.2820μs	46.3619μs	21.5695 KOps/s	21.7397 KOps/s	$\color{#d91a1a}-0.78\%$
test_instantiation_functorch	2.0758ms	1.9834ms	504.1743 Ops/s	503.7004 Ops/s	$\color{#35bf28}+0.09\%$
test_exec_functorch	0.2286ms	0.1733ms	5.7690 KOps/s	5.7484 KOps/s	$\color{#35bf28}+0.36\%$
test_exec_functional_call	0.2239ms	0.1550ms	6.4528 KOps/s	6.4715 KOps/s	$\color{#d91a1a}-0.29\%$
test_exec_td_decorator	0.4360ms	0.2247ms	4.4495 KOps/s	4.3638 KOps/s	$\color{#35bf28}+1.96\%$
test_vmap_mlp_speed_decorator[True-True]	0.9777ms	0.8018ms	1.2471 KOps/s	1.2273 KOps/s	$\color{#35bf28}+1.62\%$
test_vmap_mlp_speed_decorator[True-False]	0.9848ms	0.7996ms	1.2507 KOps/s	1.2277 KOps/s	$\color{#35bf28}+1.87\%$
test_vmap_mlp_speed_decorator[False-True]	0.8945ms	0.6921ms	1.4449 KOps/s	1.4310 KOps/s	$\color{#35bf28}+0.97\%$
test_vmap_mlp_speed_decorator[False-False]	0.8729ms	0.6909ms	1.4475 KOps/s	1.4257 KOps/s	$\color{#35bf28}+1.53\%$
test_vmap_transformer_speed_decorator[True-True]	20.2735ms	20.1165ms	49.7105 Ops/s	49.5519 Ops/s	$\color{#35bf28}+0.32\%$
test_vmap_transformer_speed_decorator[True-False]	20.7495ms	20.1821ms	49.5489 Ops/s	49.4994 Ops/s	$\color{#35bf28}+0.10\%$
test_vmap_transformer_speed_decorator[False-True]	20.6017ms	19.9651ms	50.0874 Ops/s	50.0068 Ops/s	$\color{#35bf28}+0.16\%$
test_vmap_transformer_speed_decorator[False-False]	20.6242ms	20.0127ms	49.9682 Ops/s	49.9142 Ops/s	$\color{#35bf28}+0.11\%$
test_to_module_speed[True]	1.9627ms	1.3813ms	723.9311 Ops/s	714.4193 Ops/s	$\color{#35bf28}+1.33\%$
test_to_module_speed[False]	2.1921ms	1.3663ms	731.9232 Ops/s	722.7401 Ops/s	$\color{#35bf28}+1.27\%$
test_tc_init	87.2420μs	43.1205μs	23.1908 KOps/s	22.5554 KOps/s	$\color{#35bf28}+2.82\%$
test_tc_init_tensor_only	38.1310μs	9.2560μs	108.0383 KOps/s	107.7774 KOps/s	$\color{#35bf28}+0.24\%$
test_tc_init_nested	0.3794ms	85.8077μs	11.6540 KOps/s	11.6720 KOps/s	$\color{#d91a1a}-0.15\%$
test_tc_init_many_fields	41.2910μs	15.5260μs	64.4082 KOps/s	64.4895 KOps/s	$\color{#d91a1a}-0.13\%$
test_tc_first_layer_tensor	24.3910μs	1.7087μs	585.2442 KOps/s	587.5501 KOps/s	$\color{#d91a1a}-0.39\%$
test_tc_first_layer_tensor_only	1.6695μs	0.3842μs	2.6025 MOps/s	2.6049 MOps/s	$\color{#d91a1a}-0.09\%$
test_tc_first_layer_tensor_set	28.0500μs	3.6653μs	272.8292 KOps/s	273.0340 KOps/s	$\color{#d91a1a}-0.07\%$
test_tc_first_layer_tensor_only_set	23.7100μs	3.1102μs	321.5250 KOps/s	318.2200 KOps/s	$\color{#35bf28}+1.04\%$
test_tc_first_layer_nontensor	29.3500μs	5.8894μs	169.7978 KOps/s	166.7028 KOps/s	$\color{#35bf28}+1.86\%$
test_tc_second_layer_tensor	27.2110μs	4.1721μs	239.6897 KOps/s	246.1866 KOps/s	$\color{#d91a1a}-2.64\%$
test_tc_second_layer_nontensor	34.5010μs	8.2797μs	120.7772 KOps/s	119.9295 KOps/s	$\color{#35bf28}+0.71\%$
test_unbind	0.2633s	17.1216ms	58.4057 Ops/s	57.7874 Ops/s	$\color{#35bf28}+1.07\%$
test_full_like	5.0753ms	4.3787ms	228.3776 Ops/s	227.0288 Ops/s	$\color{#35bf28}+0.59\%$
test_zeros_like	4.9252ms	4.3642ms	229.1378 Ops/s	228.9288 Ops/s	$\color{#35bf28}+0.09\%$
test_ones_like	5.0137ms	4.3782ms	228.4049 Ops/s	229.1025 Ops/s	$\color{#d91a1a}-0.30\%$
test_clone	6.9092ms	6.5122ms	153.5571 Ops/s	153.4009 Ops/s	$\color{#35bf28}+0.10\%$
test_squeeze	0.1803ms	13.5758μs	73.6604 KOps/s	73.6824 KOps/s	$\color{#d91a1a}-0.03\%$
test_unsqueeze	0.2799ms	0.1092ms	9.1575 KOps/s	9.1813 KOps/s	$\color{#d91a1a}-0.26\%$
test_split	0.2498ms	0.1791ms	5.5830 KOps/s	5.5412 KOps/s	$\color{#35bf28}+0.76\%$
test_permute	0.2597ms	0.2067ms	4.8387 KOps/s	4.9896 KOps/s	$\color{#d91a1a}-3.02\%$
test_stack	44.3271ms	43.1766ms	23.1607 Ops/s	23.1353 Ops/s	$\color{#35bf28}+0.11\%$
test_cat	43.4310ms	43.1665ms	23.1661 Ops/s	23.1944 Ops/s	$\color{#d91a1a}-0.12\%$
test_sequential_tensordict	0.3209ms	0.2127ms	4.7010 KOps/s	4.7183 KOps/s	$\color{#d91a1a}-0.37\%$
test_sequential_graph_module	0.1664ms	0.1166ms	8.5776 KOps/s	8.7058 KOps/s	$\color{#d91a1a}-1.47\%$
test_nested_tensordict	0.3800ms	0.2735ms	3.6569 KOps/s	3.6183 KOps/s	$\color{#35bf28}+1.07\%$
test_nested_graph_module	0.1727ms	0.1257ms	7.9527 KOps/s	8.0026 KOps/s	$\color{#d91a1a}-0.62\%$

Update

898b8ba

[ghstack-poisoned]

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 Mar 6, 2026

github-actions bot added the Test label Mar 6, 2026

This was referenced Mar 6, 2026

[DTensor] Add unified dtensor_send/dtensor_recv API with Strategy A #1640

Open

[DTensor] Add Strategy B (local-shard transfer + redistribute on receiver) #1641

Open

[DTensor] Add Strategy C (optimal P2P transfer using transfer plan) #1642

Open

Provide feedback

Saved searches

Use saved searches to filter your results more quickly

[DTensor] Add transfer plan computation for cross-mesh DTensor redistribution#1639

[DTensor] Add transfer plan computation for cross-mesh DTensor redistribution#1639
vmoens wants to merge 1 commit intogh/vmoens/80/basefrom
gh/vmoens/80/head

vmoens commented Mar 6, 2026 •

edited

Loading

Uh oh!

github-actions bot commented Mar 6, 2026

Uh oh!

github-actions bot commented Mar 6, 2026

Uh oh!

github-actions bot commented Mar 6, 2026

Uh oh!

github-actions bot commented Mar 6, 2026

Uh oh!

Reviewers

Assignees

Labels

Projects

Milestone

Development

Uh oh!

1 participant

Conversation

vmoens commented Mar 6, 2026 • edited Loading Uh oh! There was an error while loading. Please reload this page.

Uh oh!

Uh oh!

github-actions bot commented Mar 6, 2026

PR Title Label Error

Supported Prefixes

Uh oh!

github-actions bot commented Mar 6, 2026

PR Title Label Error

Supported Prefixes

Uh oh!

github-actions bot commented Mar 6, 2026

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

Total Benchmarks: 261. Improved: $\large\color{#35bf28}20$. Worsened: $\large\color{#d91a1a}9$.

Uh oh!

github-actions bot commented Mar 6, 2026

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

Total Benchmarks: 261. Improved: $\large\color{#35bf28}24$. Worsened: $\large\color{#d91a1a}5$.

Uh oh!

Reviewers

Assignees

Labels

Projects

Milestone

Development

Uh oh!

1 participant

vmoens commented Mar 6, 2026 •

edited

Loading

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

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