[DTensor] Add Strategy B (local-shard transfer + redistribute on receiver) by vmoens · Pull Request #1641 · pytorch/tensordict

vmoens · 2026-03-06T15:02:40Z

Stack from ghstack (oldest at bottom):

Implement _dtensor_send_redistribute and _dtensor_recv_redistribute:

Sender: extracts local shards via to_local(), sends placement metadata
(placements, mesh topology, mesh dim names) + local tensor data
Receiver: receives local shards and placement metadata, stores tensors
for the caller to reconstruct as DTensors via from_local() + redistribute()

This avoids materializing full tensors (no memory spike on sender),
and transfers only the data each sender rank actually holds.

Made-with: Cursor

[ghstack-poisoned]

github-actions · 2026-03-06T15:02:48Z

PR Title Label Error

Unknown or invalid prefix [DTensor].

Current title: [DTensor] Add Strategy B (local-shard transfer + redistribute on receiver)

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 · 2026-03-06T15:02:59Z

PR Title Label Error

Unknown or invalid prefix [DTensor].

Current title: [DTensor] Add Strategy B (local-shard transfer + redistribute on receiver)

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 · 2026-03-06T15:31:02Z

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

Total Benchmarks: 261. Improved: $\large\color{#35bf28}13$. Worsened: $\large\color{#d91a1a}8$.

Expand to view detailed results

Name	Max	Mean	Ops	Ops on Repo `HEAD`	Change
test_plain_set_nested	32.1810μs	15.1468μs	66.0206 KOps/s	66.9536 KOps/s	$\color{#d91a1a}-1.39\%$
test_plain_set_stack_nested	42.3010μs	15.3918μs	64.9696 KOps/s	65.5792 KOps/s	$\color{#d91a1a}-0.93\%$
test_plain_set_nested_inplace	44.7110μs	16.9114μs	59.1316 KOps/s	59.2037 KOps/s	$\color{#d91a1a}-0.12\%$
test_plain_set_stack_nested_inplace	58.8020μs	16.8795μs	59.2433 KOps/s	59.6818 KOps/s	$\color{#d91a1a}-0.73\%$
test_items	39.0500μs	6.1571μs	162.4146 KOps/s	167.5811 KOps/s	$\color{#d91a1a}-3.08\%$
test_items_nested	0.5356ms	0.4720ms	2.1185 KOps/s	2.1215 KOps/s	$\color{#d91a1a}-0.14\%$
test_items_nested_locked	0.5183ms	0.4754ms	2.1036 KOps/s	2.1158 KOps/s	$\color{#d91a1a}-0.58\%$
test_items_nested_leaf	0.1278ms	99.2144μs	10.0792 KOps/s	10.1834 KOps/s	$\color{#d91a1a}-1.02\%$
test_items_stack_nested	0.5347ms	0.4698ms	2.1284 KOps/s	2.1366 KOps/s	$\color{#d91a1a}-0.38\%$
test_items_stack_nested_leaf	0.1346ms	98.7403μs	10.1276 KOps/s	10.1144 KOps/s	$\color{#35bf28}+0.13\%$
test_items_stack_nested_locked	0.5065ms	0.4765ms	2.0986 KOps/s	2.1011 KOps/s	$\color{#d91a1a}-0.12\%$
test_keys	27.2800μs	4.2660μs	234.4097 KOps/s	236.4689 KOps/s	$\color{#d91a1a}-0.87\%$
test_keys_nested	0.1980ms	0.1320ms	7.5765 KOps/s	7.5985 KOps/s	$\color{#d91a1a}-0.29\%$
test_keys_nested_locked	1.9953ms	0.1416ms	7.0636 KOps/s	7.1051 KOps/s	$\color{#d91a1a}-0.58\%$
test_keys_nested_leaf	0.1839ms	0.1221ms	8.1902 KOps/s	8.2239 KOps/s	$\color{#d91a1a}-0.41\%$
test_keys_stack_nested	0.1670ms	0.1312ms	7.6229 KOps/s	7.6003 KOps/s	$\color{#35bf28}+0.30\%$
test_keys_stack_nested_leaf	0.1499ms	0.1221ms	8.1922 KOps/s	8.1470 KOps/s	$\color{#35bf28}+0.55\%$
test_keys_stack_nested_locked	0.1688ms	0.1405ms	7.1153 KOps/s	7.1142 KOps/s	$\color{#35bf28}+0.01\%$
test_values	5.8280μs	1.0374μs	963.9323 KOps/s	970.7805 KOps/s	$\color{#d91a1a}-0.71\%$
test_values_nested	83.9210μs	54.1491μs	18.4675 KOps/s	18.6893 KOps/s	$\color{#d91a1a}-1.19\%$
test_values_nested_locked	80.0620μs	57.2992μs	17.4523 KOps/s	17.6347 KOps/s	$\color{#d91a1a}-1.03\%$
test_values_nested_leaf	85.8010μs	61.1878μs	16.3431 KOps/s	16.4695 KOps/s	$\color{#d91a1a}-0.77\%$
test_values_stack_nested	93.8110μs	53.8462μs	18.5714 KOps/s	18.7211 KOps/s	$\color{#d91a1a}-0.80\%$
test_values_stack_nested_leaf	0.1419ms	61.5948μs	16.2351 KOps/s	16.3494 KOps/s	$\color{#d91a1a}-0.70\%$
test_values_stack_nested_locked	96.5720μs	57.3054μs	17.4504 KOps/s	17.5155 KOps/s	$\color{#d91a1a}-0.37\%$
test_membership	6.3635μs	0.8662μs	1.1544 MOps/s	1.1847 MOps/s	$\color{#d91a1a}-2.56\%$
test_membership_nested	32.4810μs	2.9825μs	335.2856 KOps/s	343.8662 KOps/s	$\color{#d91a1a}-2.50\%$
test_membership_nested_leaf	66.5720μs	2.9871μs	334.7706 KOps/s	343.1503 KOps/s	$\color{#d91a1a}-2.44\%$
test_membership_stacked_nested	34.6400μs	2.9246μs	341.9248 KOps/s	341.6643 KOps/s	$\color{#35bf28}+0.08\%$
test_membership_stacked_nested_leaf	21.3810μs	2.9498μs	339.0114 KOps/s	344.3124 KOps/s	$\color{#d91a1a}-1.54\%$
test_membership_nested_last	40.3510μs	4.4311μs	225.6768 KOps/s	227.0283 KOps/s	$\color{#d91a1a}-0.60\%$
test_membership_nested_leaf_last	35.4610μs	4.4412μs	225.1644 KOps/s	229.0837 KOps/s	$\color{#d91a1a}-1.71\%$
test_membership_stacked_nested_last	43.2110μs	4.4026μs	227.1365 KOps/s	229.2578 KOps/s	$\color{#d91a1a}-0.93\%$
test_membership_stacked_nested_leaf_last	26.4600μs	4.3903μs	227.7729 KOps/s	231.3810 KOps/s	$\color{#d91a1a}-1.56\%$
test_nested_getleaf	56.5110μs	21.7565μs	45.9633 KOps/s	46.9593 KOps/s	$\color{#d91a1a}-2.12\%$
test_nested_get	46.7210μs	20.7477μs	48.1981 KOps/s	48.0656 KOps/s	$\color{#35bf28}+0.28\%$
test_stacked_getleaf	54.2710μs	21.5849μs	46.3287 KOps/s	46.7389 KOps/s	$\color{#d91a1a}-0.88\%$
test_stacked_get	54.4910μs	20.7693μs	48.1480 KOps/s	48.8391 KOps/s	$\color{#d91a1a}-1.42\%$
test_nested_getitemleaf	59.1720μs	22.4359μs	44.5715 KOps/s	45.7240 KOps/s	$\color{#d91a1a}-2.52\%$
test_nested_getitem	49.9310μs	20.9883μs	47.6455 KOps/s	48.2107 KOps/s	$\color{#d91a1a}-1.17\%$
test_stacked_getitemleaf	48.7710μs	22.2896μs	44.8641 KOps/s	45.4545 KOps/s	$\color{#d91a1a}-1.30\%$
test_stacked_getitem	52.4020μs	21.0130μs	47.5895 KOps/s	48.7947 KOps/s	$\color{#d91a1a}-2.47\%$
test_lock_nested	0.5463ms	0.4827ms	2.0716 KOps/s	2.0980 KOps/s	$\color{#d91a1a}-1.26\%$
test_lock_stack_nested	0.5796ms	0.4838ms	2.0669 KOps/s	2.0649 KOps/s	$\color{#35bf28}+0.10\%$
test_unlock_nested	0.4572ms	0.3930ms	2.5443 KOps/s	2.5637 KOps/s	$\color{#d91a1a}-0.75\%$
test_unlock_stack_nested	0.4292ms	0.3917ms	2.5528 KOps/s	2.5376 KOps/s	$\color{#35bf28}+0.60\%$
test_flatten_speed	0.1531ms	0.1232ms	8.1190 KOps/s	8.1090 KOps/s	$\color{#35bf28}+0.12\%$
test_unflatten_speed	0.6326ms	0.5855ms	1.7080 KOps/s	1.7285 KOps/s	$\color{#d91a1a}-1.19\%$
test_common_ops	0.8364ms	0.7028ms	1.4228 KOps/s	1.4201 KOps/s	$\color{#35bf28}+0.19\%$
test_creation	92.7220μs	3.0396μs	328.9925 KOps/s	319.0697 KOps/s	$\color{#35bf28}+3.11\%$
test_creation_empty	27.8100μs	6.9935μs	142.9908 KOps/s	144.0987 KOps/s	$\color{#d91a1a}-0.77\%$
test_creation_nested_1	34.5510μs	11.6727μs	85.6699 KOps/s	86.5848 KOps/s	$\color{#d91a1a}-1.06\%$
test_creation_nested_2	38.8000μs	13.3228μs	75.0594 KOps/s	74.6021 KOps/s	$\color{#35bf28}+0.61\%$
test_creation_many_keys[10]	50.7310μs	21.0395μs	47.5297 KOps/s	48.0968 KOps/s	$\color{#d91a1a}-1.18\%$
test_creation_many_keys[50]	0.1209ms	90.2223μs	11.0837 KOps/s	11.1251 KOps/s	$\color{#d91a1a}-0.37\%$
test_creation_many_keys[100]	0.2061ms	0.1750ms	5.7149 KOps/s	5.6420 KOps/s	$\color{#35bf28}+1.29\%$
test_creation_nested_many_keys[10]	84.0720μs	44.9793μs	22.2325 KOps/s	22.3232 KOps/s	$\color{#d91a1a}-0.41\%$
test_creation_nested_many_keys[50]	0.2261ms	0.1834ms	5.4528 KOps/s	5.4002 KOps/s	$\color{#35bf28}+0.97\%$
test_clone	43.7610μs	13.5151μs	73.9914 KOps/s	74.1943 KOps/s	$\color{#d91a1a}-0.27\%$
test_getitem[int]	1.5612ms	15.0222μs	66.5680 KOps/s	59.3741 KOps/s	$\textbf{\color{#35bf28}+12.12\%}$
test_getitem[slice_int]	0.1327ms	24.1744μs	41.3660 KOps/s	41.1236 KOps/s	$\color{#35bf28}+0.59\%$
test_getitem[range]	0.1753ms	64.0474μs	15.6134 KOps/s	15.6341 KOps/s	$\color{#d91a1a}-0.13\%$
test_getitem[tuple]	0.1383ms	23.7951μs	42.0254 KOps/s	41.4073 KOps/s	$\color{#35bf28}+1.49\%$
test_getitem[list]	0.1849ms	59.9424μs	16.6827 KOps/s	17.0034 KOps/s	$\color{#d91a1a}-1.89\%$
test_setitem_dim[int]	48.2710μs	26.0197μs	38.4324 KOps/s	38.0077 KOps/s	$\color{#35bf28}+1.12\%$
test_setitem_dim[slice_int]	65.5510μs	43.0457μs	23.2311 KOps/s	22.7716 KOps/s	$\color{#35bf28}+2.02\%$
test_setitem_dim[range]	0.1222ms	95.8264μs	10.4355 KOps/s	10.5173 KOps/s	$\color{#d91a1a}-0.78\%$
test_setitem_dim[tuple]	62.9110μs	39.7554μs	25.1538 KOps/s	24.2820 KOps/s	$\color{#35bf28}+3.59\%$
test_setitem	49.3610μs	17.8660μs	55.9723 KOps/s	55.7887 KOps/s	$\color{#35bf28}+0.33\%$
test_set	43.8210μs	17.0892μs	58.5164 KOps/s	58.2053 KOps/s	$\color{#35bf28}+0.53\%$
test_set_shared	0.5002ms	0.2051ms	4.8745 KOps/s	4.9014 KOps/s	$\color{#d91a1a}-0.55\%$
test_update	0.3330ms	21.7580μs	45.9602 KOps/s	45.9180 KOps/s	$\color{#35bf28}+0.09\%$
test_update_nested	66.3510μs	33.5549μs	29.8019 KOps/s	29.6310 KOps/s	$\color{#35bf28}+0.58\%$
test_update__nested	0.4519ms	34.3032μs	29.1518 KOps/s	28.7383 KOps/s	$\color{#35bf28}+1.44\%$
test_set_nested	49.4410μs	18.9229μs	52.8459 KOps/s	52.6048 KOps/s	$\color{#35bf28}+0.46\%$
test_set_nested_new	63.8120μs	24.2971μs	41.1572 KOps/s	41.4499 KOps/s	$\color{#d91a1a}-0.71\%$
test_select	78.9120μs	41.3308μs	24.1951 KOps/s	24.2355 KOps/s	$\color{#d91a1a}-0.17\%$
test_select_nested	0.1113ms	74.4233μs	13.4367 KOps/s	13.2320 KOps/s	$\color{#35bf28}+1.55\%$
test_exclude_nested	0.1283ms	92.7921μs	10.7768 KOps/s	10.7313 KOps/s	$\color{#35bf28}+0.42\%$
test_empty[True]	0.4570ms	0.4023ms	2.4855 KOps/s	2.4852 KOps/s	$\color{#35bf28}+0.02\%$
test_empty[False]	9.5377μs	1.3160μs	759.8712 KOps/s	758.8738 KOps/s	$\color{#35bf28}+0.13\%$
test_to	0.1057ms	71.4464μs	13.9965 KOps/s	13.5615 KOps/s	$\color{#35bf28}+3.21\%$
test_to_nonblocking	0.1039ms	64.8198μs	15.4274 KOps/s	15.3954 KOps/s	$\color{#35bf28}+0.21\%$
test_unbind_speed	0.3708ms	0.3330ms	3.0026 KOps/s	2.9906 KOps/s	$\color{#35bf28}+0.40\%$
test_unbind_speed_stack0	0.4152ms	0.3331ms	3.0024 KOps/s	3.0098 KOps/s	$\color{#d91a1a}-0.25\%$
test_unbind_speed_stack1	0.1038s	0.8458ms	1.1823 KOps/s	1.1741 KOps/s	$\color{#35bf28}+0.70\%$
test_split	0.1033s	1.2698ms	787.5393 Ops/s	782.3303 Ops/s	$\color{#35bf28}+0.67\%$
test_chunk	0.1036s	1.2154ms	822.7732 Ops/s	922.4359 Ops/s	$\textbf{\color{#d91a1a}-10.80\%}$
test_to_cpu_blocking	19.4731ms	19.3896ms	51.5740 Ops/s	35.1867 Ops/s	$\textbf{\color{#35bf28}+46.57\%}$
test_to_cpu_global_sync	11.5389ms	11.4543ms	87.3032 Ops/s	78.9137 Ops/s	$\textbf{\color{#35bf28}+10.63\%}$
test_to_cpu_event_sync	12.7311ms	12.4365ms	80.4082 Ops/s	80.6696 Ops/s	$\color{#d91a1a}-0.32\%$
test_to_cpu_default	0.1155s	13.7389ms	72.7861 Ops/s	80.6910 Ops/s	$\textbf{\color{#d91a1a}-9.80\%}$
test_consolidate[False-None]	4.2299ms	4.1610ms	240.3281 Ops/s	216.8407 Ops/s	$\textbf{\color{#35bf28}+10.83\%}$
test_consolidate[default-None]	2.1527ms	2.0072ms	498.2046 Ops/s	487.1938 Ops/s	$\color{#35bf28}+2.26\%$
test_consolidate[reduce-overhead-None]	2.0092ms	1.9125ms	522.8664 Ops/s	507.9723 Ops/s	$\color{#35bf28}+2.93\%$
test_consolidate_njt[False-None]	8.8629ms	8.6190ms	116.0224 Ops/s	117.6604 Ops/s	$\color{#d91a1a}-1.39\%$
test_to[False-False-None]	2.2964ms	2.1104ms	473.8402 Ops/s	470.6205 Ops/s	$\color{#35bf28}+0.68\%$
test_to[True-False-None]	2.2513ms	1.9376ms	516.1085 Ops/s	511.7222 Ops/s	$\color{#35bf28}+0.86\%$
test_to[within-False-None]	6.2786ms	6.1682ms	162.1229 Ops/s	162.6628 Ops/s	$\color{#d91a1a}-0.33\%$
test_to[True-default-None]	9.3031ms	8.9433ms	111.8152 Ops/s	110.5601 Ops/s	$\color{#35bf28}+1.14\%$
test_to_njt[False-False-None]	8.6230ms	8.5094ms	117.5173 Ops/s	117.8506 Ops/s	$\color{#d91a1a}-0.28\%$
test_to_njt[True-False-None]	7.3353ms	6.9771ms	143.3266 Ops/s	143.6131 Ops/s	$\color{#d91a1a}-0.20\%$
test_to_njt[within-False-None]	15.9085ms	15.6804ms	63.7737 Ops/s	64.1588 Ops/s	$\color{#d91a1a}-0.60\%$
test_creation[device0]	0.3839ms	0.1139ms	8.7815 KOps/s	8.7002 KOps/s	$\color{#35bf28}+0.93\%$
test_creation_from_tensor	0.4129ms	0.1115ms	8.9667 KOps/s	8.5047 KOps/s	$\textbf{\color{#35bf28}+5.43\%}$
test_add_one[memmap_tensor0]	0.2939ms	6.6013μs	151.4864 KOps/s	150.9784 KOps/s	$\color{#35bf28}+0.34\%$
test_contiguous[memmap_tensor0]	26.3200μs	0.6761μs	1.4791 MOps/s	1.9012 MOps/s	$\textbf{\color{#d91a1a}-22.20\%}$
test_stack[memmap_tensor0]	33.3110μs	4.6485μs	215.1223 KOps/s	220.8156 KOps/s	$\color{#d91a1a}-2.58\%$
test_memmaptd_index	1.1468ms	0.2720ms	3.6771 KOps/s	3.7740 KOps/s	$\color{#d91a1a}-2.57\%$
test_memmaptd_index_astensor	0.5318ms	0.3775ms	2.6489 KOps/s	2.7067 KOps/s	$\color{#d91a1a}-2.14\%$
test_memmaptd_index_op	0.8914ms	0.6246ms	1.6009 KOps/s	1.6053 KOps/s	$\color{#d91a1a}-0.27\%$
test_serialize_model	0.3161s	0.1657s	6.0333 Ops/s	7.3102 Ops/s	$\textbf{\color{#d91a1a}-17.47\%}$
test_serialize_model_pickle	2.1095s	1.3626s	0.7339 Ops/s	0.8254 Ops/s	$\textbf{\color{#d91a1a}-11.08\%}$
test_serialize_weights	0.1391s	0.1368s	7.3085 Ops/s	7.3401 Ops/s	$\color{#d91a1a}-0.43\%$
test_serialize_weights_returnearly	0.4557s	93.6088ms	10.6828 Ops/s	10.3950 Ops/s	$\color{#35bf28}+2.77\%$
test_serialize_weights_pickle	1.3775s	1.2180s	0.8210 Ops/s	0.8186 Ops/s	$\color{#35bf28}+0.29\%$
test_reshape_pytree	0.2076ms	33.0853μs	30.2249 KOps/s	30.7492 KOps/s	$\color{#d91a1a}-1.70\%$
test_reshape_td	0.2161ms	45.9692μs	21.7537 KOps/s	22.2743 KOps/s	$\color{#d91a1a}-2.34\%$
test_view_pytree	0.2207ms	32.7049μs	30.5764 KOps/s	30.8383 KOps/s	$\color{#d91a1a}-0.85\%$
test_view_td	93.1220μs	53.6069μs	18.6543 KOps/s	18.0148 KOps/s	$\color{#35bf28}+3.55\%$
test_unbind_pytree	0.2440ms	36.7841μs	27.1856 KOps/s	27.6223 KOps/s	$\color{#d91a1a}-1.58\%$
test_unbind_td	0.1026ms	50.3707μs	19.8528 KOps/s	19.8542 KOps/s	$-0.01\%$
test_split_pytree	0.2494ms	42.7075μs	23.4151 KOps/s	23.3278 KOps/s	$\color{#35bf28}+0.37\%$
test_split_td	0.1764ms	64.2355μs	15.5677 KOps/s	15.3542 KOps/s	$\color{#35bf28}+1.39\%$
test_add_pytree	0.2346ms	44.7264μs	22.3582 KOps/s	23.4353 KOps/s	$\color{#d91a1a}-4.60\%$
test_add_td	97.5120μs	59.8405μs	16.7111 KOps/s	17.6940 KOps/s	$\textbf{\color{#d91a1a}-5.55\%}$
test_compile_add_one_nested[tensordict-compile]	0.2298ms	0.1456ms	6.8683 KOps/s	6.6940 KOps/s	$\color{#35bf28}+2.60\%$
test_compile_add_one_nested[tensordict-eager]	0.3070ms	0.2037ms	4.9100 KOps/s	5.0094 KOps/s	$\color{#d91a1a}-1.98\%$
test_compile_add_one_nested[pytree-compile]	0.1809ms	0.1085ms	9.2195 KOps/s	9.1408 KOps/s	$\color{#35bf28}+0.86\%$
test_compile_add_one_nested[pytree-eager]	0.4350ms	0.1806ms	5.5378 KOps/s	5.5728 KOps/s	$\color{#d91a1a}-0.63\%$
test_compile_copy_nested[tensordict-compile]	0.5438ms	10.7770μs	92.7906 KOps/s	97.0721 KOps/s	$\color{#d91a1a}-4.41\%$
test_compile_copy_nested[tensordict-eager]	0.1389ms	54.8017μs	18.2476 KOps/s	18.5604 KOps/s	$\color{#d91a1a}-1.69\%$
test_compile_copy_nested[pytree-compile]	0.1468ms	9.7741μs	102.3110 KOps/s	102.5887 KOps/s	$\color{#d91a1a}-0.27\%$
test_compile_copy_nested[pytree-eager]	0.4648ms	70.2915μs	14.2265 KOps/s	14.6184 KOps/s	$\color{#d91a1a}-2.68\%$
test_compile_add_one_flat[tensordict-compile]	0.2396ms	0.1748ms	5.7217 KOps/s	5.4352 KOps/s	$\textbf{\color{#35bf28}+5.27\%}$
test_compile_add_one_flat[tensordict-eager]	0.4062ms	0.2825ms	3.5401 KOps/s	3.5559 KOps/s	$\color{#d91a1a}-0.45\%$
test_compile_add_one_flat[tensorclass-compile]	0.2572ms	0.1163ms	8.5982 KOps/s	8.3570 KOps/s	$\color{#35bf28}+2.89\%$
test_compile_add_one_flat[tensorclass-eager]	0.1186ms	73.0184μs	13.6952 KOps/s	13.3137 KOps/s	$\color{#35bf28}+2.87\%$
test_compile_add_one_flat[pytree-compile]	0.2058ms	0.1570ms	6.3705 KOps/s	6.1541 KOps/s	$\color{#35bf28}+3.52\%$
test_compile_add_one_flat[pytree-eager]	0.8704ms	0.5222ms	1.9149 KOps/s	1.8938 KOps/s	$\color{#35bf28}+1.12\%$
test_compile_add_self_flat[tensordict-eager]	0.5103ms	0.3364ms	2.9724 KOps/s	2.9605 KOps/s	$\color{#35bf28}+0.40\%$
test_compile_add_self_flat[tensordict-compile]	0.3392ms	0.1784ms	5.6057 KOps/s	5.2271 KOps/s	$\textbf{\color{#35bf28}+7.24\%}$
test_compile_add_self_flat[tensorclass-eager]	0.5512ms	91.3722μs	10.9442 KOps/s	11.0511 KOps/s	$\color{#d91a1a}-0.97\%$
test_compile_add_self_flat[tensorclass-compile]	0.6031ms	0.1180ms	8.4780 KOps/s	8.0530 KOps/s	$\textbf{\color{#35bf28}+5.28\%}$
test_compile_add_self_flat[pytree-eager]	0.9203ms	0.4292ms	2.3300 KOps/s	2.2828 KOps/s	$\color{#35bf28}+2.07\%$
test_compile_add_self_flat[pytree-compile]	0.2153ms	0.1637ms	6.1095 KOps/s	6.1582 KOps/s	$\color{#d91a1a}-0.79\%$
test_compile_copy_flat[tensordict-compile]	76.2110μs	13.6038μs	73.5087 KOps/s	75.0175 KOps/s	$\color{#d91a1a}-2.01\%$
test_compile_copy_flat[tensordict-eager]	79.3510μs	42.2465μs	23.6706 KOps/s	23.6349 KOps/s	$\color{#35bf28}+0.15\%$
test_compile_copy_flat[pytree-compile]	89.9720μs	10.8265μs	92.3661 KOps/s	91.2390 KOps/s	$\color{#35bf28}+1.24\%$
test_compile_copy_flat[pytree-eager]	0.4120ms	52.5714μs	19.0217 KOps/s	18.9061 KOps/s	$\color{#35bf28}+0.61\%$
test_compile_assign_and_add[tensordict-compile]	2.0444ms	0.1744ms	5.7353 KOps/s	5.4847 KOps/s	$\color{#35bf28}+4.57\%$
test_compile_assign_and_add[tensordict-eager]	3.7626ms	3.3156ms	301.6008 Ops/s	298.7962 Ops/s	$\color{#35bf28}+0.94\%$
test_compile_assign_and_add[pytree-compile]	2.0568ms	0.1637ms	6.1105 KOps/s	6.0612 KOps/s	$\color{#35bf28}+0.81\%$
test_compile_assign_and_add[pytree-eager]	3.0112ms	2.7957ms	357.6930 Ops/s	358.2589 Ops/s	$\color{#d91a1a}-0.16\%$
test_compile_indexing[tensor-tensordict-compile]	0.2316ms	0.1083ms	9.2368 KOps/s	8.7653 KOps/s	$\textbf{\color{#35bf28}+5.38\%}$
test_compile_indexing[tensor-tensordict-eager]	0.5292ms	73.8997μs	13.5318 KOps/s	13.4443 KOps/s	$\color{#35bf28}+0.65\%$
test_compile_indexing[tensor-tensorclass-compile]	0.1357ms	95.3919μs	10.4831 KOps/s	10.2594 KOps/s	$\color{#35bf28}+2.18\%$
test_compile_indexing[tensor-tensorclass-eager]	0.2564ms	44.5166μs	22.4635 KOps/s	22.4871 KOps/s	$\color{#d91a1a}-0.10\%$
test_compile_indexing[tensor-pytree-compile]	0.2646ms	96.0234μs	10.4141 KOps/s	10.2069 KOps/s	$\color{#35bf28}+2.03\%$
test_compile_indexing[tensor-pytree-eager]	0.2468ms	44.2655μs	22.5909 KOps/s	22.3824 KOps/s	$\color{#35bf28}+0.93\%$
test_compile_indexing[slice-tensordict-compile]	0.2152ms	56.6530μs	17.6513 KOps/s	17.3836 KOps/s	$\color{#35bf28}+1.54\%$
test_compile_indexing[slice-tensordict-eager]	0.2217ms	27.5752μs	36.2645 KOps/s	35.3967 KOps/s	$\color{#35bf28}+2.45\%$
test_compile_indexing[slice-tensorclass-compile]	0.1529ms	44.4304μs	22.5071 KOps/s	22.0552 KOps/s	$\color{#35bf28}+2.05\%$
test_compile_indexing[slice-tensorclass-eager]	0.2389ms	22.6116μs	44.2251 KOps/s	44.3341 KOps/s	$\color{#d91a1a}-0.25\%$
test_compile_indexing[slice-pytree-compile]	83.3710μs	45.6824μs	21.8903 KOps/s	21.7985 KOps/s	$\color{#35bf28}+0.42\%$
test_compile_indexing[slice-pytree-eager]	0.2678ms	22.5748μs	44.2971 KOps/s	44.3017 KOps/s	$\color{#d91a1a}-0.01\%$
test_compile_indexing[int-tensordict-compile]	96.6020μs	57.6537μs	17.3450 KOps/s	17.2775 KOps/s	$\color{#35bf28}+0.39\%$
test_compile_indexing[int-tensordict-eager]	0.2404ms	27.5627μs	36.2809 KOps/s	36.0228 KOps/s	$\color{#35bf28}+0.72\%$
test_compile_indexing[int-tensorclass-compile]	87.0420μs	45.3031μs	22.0736 KOps/s	22.0020 KOps/s	$\color{#35bf28}+0.32\%$
test_compile_indexing[int-tensorclass-eager]	0.2612ms	22.7525μs	43.9513 KOps/s	44.4379 KOps/s	$\color{#d91a1a}-1.10\%$
test_compile_indexing[int-pytree-compile]	76.8710μs	45.5995μs	21.9301 KOps/s	21.8727 KOps/s	$\color{#35bf28}+0.26\%$
test_compile_indexing[int-pytree-eager]	0.1950ms	22.4864μs	44.4713 KOps/s	44.4217 KOps/s	$\color{#35bf28}+0.11\%$
test_compile_replace[single-eager]	89.6120μs	47.0910μs	21.2355 KOps/s	20.9408 KOps/s	$\color{#35bf28}+1.41\%$
test_compile_replace[single-compile]	0.1845ms	0.1045ms	9.5686 KOps/s	9.3573 KOps/s	$\color{#35bf28}+2.26\%$
test_compile_replace[multi-eager]	0.6523ms	0.5659ms	1.7670 KOps/s	1.7610 KOps/s	$\color{#35bf28}+0.34\%$
test_compile_replace[multi-compile]	0.2462ms	0.1150ms	8.6924 KOps/s	8.7640 KOps/s	$\color{#d91a1a}-0.82\%$
test_compile_tc_getattr_20[eager]	0.2114ms	0.1658ms	6.0321 KOps/s	6.0138 KOps/s	$\color{#35bf28}+0.31\%$
test_compile_tc_getattr_20[compile]	0.4606ms	0.1183ms	8.4502 KOps/s	8.3246 KOps/s	$\color{#35bf28}+1.51\%$
test_compile_clone_shallow[20-eager]	40.8600μs	19.9150μs	50.2134 KOps/s	51.2542 KOps/s	$\color{#d91a1a}-2.03\%$
test_compile_clone_shallow[20-compile]	0.1131ms	11.5656μs	86.4629 KOps/s	88.0743 KOps/s	$\color{#d91a1a}-1.83\%$
test_compile_clone_shallow[40-eager]	55.3110μs	34.9478μs	28.6141 KOps/s	29.1445 KOps/s	$\color{#d91a1a}-1.82\%$
test_compile_clone_shallow[40-compile]	67.4310μs	12.9507μs	77.2158 KOps/s	65.5660 KOps/s	$\textbf{\color{#35bf28}+17.77\%}$
test_compile_clone_shallow[80-eager]	93.1620μs	64.7994μs	15.4322 KOps/s	15.8322 KOps/s	$\color{#d91a1a}-2.53\%$
test_compile_clone_shallow[80-compile]	0.1338ms	15.1225μs	66.1264 KOps/s	67.2489 KOps/s	$\color{#d91a1a}-1.67\%$
test_compile_update_inplace[eager]	95.8420μs	60.4709μs	16.5369 KOps/s	16.6286 KOps/s	$\color{#d91a1a}-0.55\%$
test_compile_update_inplace[compile]	0.2629ms	0.1387ms	7.2076 KOps/s	7.0666 KOps/s	$\color{#35bf28}+1.99\%$
test_mod_add[eager]	91.3710μs	48.7276μs	20.5222 KOps/s	20.3061 KOps/s	$\color{#35bf28}+1.06\%$
test_mod_add[compile]	0.2022ms	0.1033ms	9.6842 KOps/s	9.2906 KOps/s	$\color{#35bf28}+4.24\%$
test_mod_add[compile-overhead]	0.3413ms	0.1493ms	6.6989 KOps/s	6.5525 KOps/s	$\color{#35bf28}+2.24\%$
test_mod_wrap[eager]	0.3880ms	0.2872ms	3.4825 KOps/s	3.4268 KOps/s	$\color{#35bf28}+1.62\%$
test_mod_wrap[compile]	0.4396ms	0.3537ms	2.8274 KOps/s	2.7086 KOps/s	$\color{#35bf28}+4.39\%$
test_mod_wrap[compile-overhead]	7.5076ms	4.0493ms	246.9581 Ops/s	250.2105 Ops/s	$\color{#d91a1a}-1.30\%$
test_mod_wrap_and_backward[eager]	1.6183ms	1.4827ms	674.4373 Ops/s	669.6368 Ops/s	$\color{#35bf28}+0.72\%$
test_mod_wrap_and_backward[compile]	1.6568ms	1.4443ms	692.3622 Ops/s	688.1573 Ops/s	$\color{#35bf28}+0.61\%$
test_mod_wrap_and_backward[compile-overhead]	1.2718ms	0.8830ms	1.1326 KOps/s	1.1043 KOps/s	$\color{#35bf28}+2.56\%$
test_seq_add[eager]	0.2408ms	0.1602ms	6.2418 KOps/s	6.4741 KOps/s	$\color{#d91a1a}-3.59\%$
test_seq_add[compile]	0.5721ms	0.1135ms	8.8126 KOps/s	8.4787 KOps/s	$\color{#35bf28}+3.94\%$
test_seq_add[compile-overhead]	0.2295ms	0.1593ms	6.2758 KOps/s	6.3060 KOps/s	$\color{#d91a1a}-0.48\%$
test_seq_wrap[eager]	0.6669ms	0.5364ms	1.8644 KOps/s	1.9247 KOps/s	$\color{#d91a1a}-3.13\%$
test_seq_wrap[compile]	0.4691ms	0.3648ms	2.7409 KOps/s	2.7263 KOps/s	$\color{#35bf28}+0.53\%$
test_seq_wrap[compile-overhead]	0.3259ms	0.2657ms	3.7630 KOps/s	3.6363 KOps/s	$\color{#35bf28}+3.48\%$
test_func_call_runtime[False-eager]	0.9708ms	0.8361ms	1.1961 KOps/s	1.1656 KOps/s	$\color{#35bf28}+2.61\%$
test_func_call_runtime[False-compile]	1.1015ms	0.9067ms	1.1029 KOps/s	1.0932 KOps/s	$\color{#35bf28}+0.88\%$
test_func_call_runtime[False-compile-overhead]	0.5274ms	0.4650ms	2.1505 KOps/s	2.1470 KOps/s	$\color{#35bf28}+0.16\%$
test_func_call_runtime[True-eager]	1.2911ms	1.0692ms	935.2462 Ops/s	932.8066 Ops/s	$\color{#35bf28}+0.26\%$
test_func_call_runtime[True-compile]	1.0133ms	0.9228ms	1.0837 KOps/s	1.0442 KOps/s	$\color{#35bf28}+3.78\%$
test_func_call_runtime[True-compile-overhead]	0.5513ms	0.4758ms	2.1016 KOps/s	2.0688 KOps/s	$\color{#35bf28}+1.58\%$
test_func_call_cm_runtime[False-eager]	1.0179ms	0.8808ms	1.1354 KOps/s	1.1948 KOps/s	$\color{#d91a1a}-4.97\%$
test_func_call_cm_runtime[False-compile]	1.0342ms	0.9261ms	1.0798 KOps/s	1.0873 KOps/s	$\color{#d91a1a}-0.69\%$
test_func_call_cm_runtime[False-compile-overhead]	0.5199ms	0.4644ms	2.1532 KOps/s	2.1385 KOps/s	$\color{#35bf28}+0.68\%$
test_func_call_cm_runtime[True-eager]	1.3037ms	1.2114ms	825.5106 Ops/s	811.2490 Ops/s	$\color{#35bf28}+1.76\%$
test_func_call_cm_runtime[True-compile]	1.3048ms	0.9904ms	1.0097 KOps/s	1.0356 KOps/s	$\color{#d91a1a}-2.50\%$
test_func_call_cm_runtime[True-compile-overhead]	0.5713ms	0.5107ms	1.9581 KOps/s	1.9458 KOps/s	$\color{#35bf28}+0.63\%$
test_vmap_func_call_cm_runtime[eager]	2.8511ms	2.3599ms	423.7469 Ops/s	421.5673 Ops/s	$\color{#35bf28}+0.52\%$
test_vmap_func_call_cm_runtime[compile]	1.1068ms	0.9785ms	1.0220 KOps/s	1.0099 KOps/s	$\color{#35bf28}+1.20\%$
test_vmap_func_call_cm_runtime[compile-overhead]	0.6429ms	0.5161ms	1.9376 KOps/s	1.9212 KOps/s	$\color{#35bf28}+0.85\%$
test_distributed	0.6426ms	0.1524ms	6.5618 KOps/s	6.4797 KOps/s	$\color{#35bf28}+1.27\%$
test_tdmodule	0.5126ms	29.5166μs	33.8792 KOps/s	35.6232 KOps/s	$\color{#d91a1a}-4.90\%$
test_tdmodule_dispatch	83.2020μs	48.2500μs	20.7254 KOps/s	21.7412 KOps/s	$\color{#d91a1a}-4.67\%$
test_tdseq	42.0910μs	29.0220μs	34.4567 KOps/s	36.9620 KOps/s	$\textbf{\color{#d91a1a}-6.78\%}$
test_tdseq_dispatch	71.2420μs	50.8562μs	19.6633 KOps/s	20.6630 KOps/s	$\color{#d91a1a}-4.84\%$
test_instantiation_functorch	2.1724ms	2.0915ms	478.1368 Ops/s	474.1385 Ops/s	$\color{#35bf28}+0.84\%$
test_exec_functorch	0.3045ms	0.1793ms	5.5758 KOps/s	5.5400 KOps/s	$\color{#35bf28}+0.65\%$
test_exec_functional_call	0.6110ms	0.1617ms	6.1851 KOps/s	6.2255 KOps/s	$\color{#d91a1a}-0.65\%$
test_exec_td_decorator	0.4468ms	0.2398ms	4.1704 KOps/s	4.2321 KOps/s	$\color{#d91a1a}-1.46\%$
test_vmap_mlp_speed_decorator[True-True]	1.2508ms	0.8288ms	1.2065 KOps/s	1.2046 KOps/s	$\color{#35bf28}+0.16\%$
test_vmap_mlp_speed_decorator[True-False]	1.2691ms	0.8228ms	1.2154 KOps/s	1.2000 KOps/s	$\color{#35bf28}+1.29\%$
test_vmap_mlp_speed_decorator[False-True]	1.1922ms	0.7082ms	1.4120 KOps/s	1.3905 KOps/s	$\color{#35bf28}+1.55\%$
test_vmap_mlp_speed_decorator[False-False]	1.1442ms	0.7079ms	1.4126 KOps/s	1.3877 KOps/s	$\color{#35bf28}+1.80\%$
test_vmap_transformer_speed_decorator[True-True]	21.0638ms	20.3833ms	49.0597 Ops/s	48.6559 Ops/s	$\color{#35bf28}+0.83\%$
test_vmap_transformer_speed_decorator[True-False]	21.0182ms	20.4122ms	48.9904 Ops/s	48.7575 Ops/s	$\color{#35bf28}+0.48\%$
test_vmap_transformer_speed_decorator[False-True]	20.5852ms	20.1750ms	49.5663 Ops/s	49.1742 Ops/s	$\color{#35bf28}+0.80\%$
test_vmap_transformer_speed_decorator[False-False]	20.6188ms	20.2062ms	49.4897 Ops/s	49.3036 Ops/s	$\color{#35bf28}+0.38\%$
test_to_module_speed[True]	1.9559ms	1.4829ms	674.3381 Ops/s	666.6685 Ops/s	$\color{#35bf28}+1.15\%$
test_to_module_speed[False]	1.9000ms	1.4656ms	682.3000 Ops/s	677.0204 Ops/s	$\color{#35bf28}+0.78\%$
test_tc_init	78.4720μs	44.1731μs	22.6382 KOps/s	21.6726 KOps/s	$\color{#35bf28}+4.46\%$
test_tc_init_tensor_only	37.6610μs	9.8583μs	101.4375 KOps/s	101.1959 KOps/s	$\color{#35bf28}+0.24\%$
test_tc_init_nested	0.1322ms	88.9981μs	11.2362 KOps/s	11.0806 KOps/s	$\color{#35bf28}+1.40\%$
test_tc_init_many_fields	44.3810μs	16.5169μs	60.5440 KOps/s	60.6681 KOps/s	$\color{#d91a1a}-0.20\%$
test_tc_first_layer_tensor	26.2910μs	1.7887μs	559.0527 KOps/s	547.0724 KOps/s	$\color{#35bf28}+2.19\%$
test_tc_first_layer_tensor_only	1.6551μs	0.3980μs	2.5123 MOps/s	2.5551 MOps/s	$\color{#d91a1a}-1.68\%$
test_tc_first_layer_tensor_set	38.2400μs	3.9215μs	255.0037 KOps/s	256.8133 KOps/s	$\color{#d91a1a}-0.70\%$
test_tc_first_layer_tensor_only_set	29.4000μs	3.2961μs	303.3904 KOps/s	308.6595 KOps/s	$\color{#d91a1a}-1.71\%$
test_tc_first_layer_nontensor	70.2310μs	6.1509μs	162.5782 KOps/s	162.6540 KOps/s	$\color{#d91a1a}-0.05\%$
test_tc_second_layer_tensor	39.1210μs	4.3450μs	230.1505 KOps/s	229.2418 KOps/s	$\color{#35bf28}+0.40\%$
test_tc_second_layer_nontensor	36.3710μs	8.6864μs	115.1230 KOps/s	115.0659 KOps/s	$\color{#35bf28}+0.05\%$
test_unbind	0.2489s	14.2096ms	70.3751 Ops/s	65.2251 Ops/s	$\textbf{\color{#35bf28}+7.90\%}$
test_full_like	7.4904ms	4.4217ms	226.1598 Ops/s	135.7072 Ops/s	$\textbf{\color{#35bf28}+66.65\%}$
test_zeros_like	4.9139ms	4.3789ms	228.3656 Ops/s	227.7485 Ops/s	$\color{#35bf28}+0.27\%$
test_ones_like	5.0021ms	4.3989ms	227.3313 Ops/s	227.4915 Ops/s	$\color{#d91a1a}-0.07\%$
test_clone	6.8466ms	6.5937ms	151.6608 Ops/s	151.8710 Ops/s	$\color{#d91a1a}-0.14\%$
test_squeeze	0.1593ms	14.5028μs	68.9521 KOps/s	66.9965 KOps/s	$\color{#35bf28}+2.92\%$
test_unsqueeze	0.2695ms	0.1120ms	8.9259 KOps/s	8.5408 KOps/s	$\color{#35bf28}+4.51\%$
test_split	0.2919ms	0.1861ms	5.3748 KOps/s	5.0122 KOps/s	$\textbf{\color{#35bf28}+7.24\%}$
test_permute	0.6461ms	0.2048ms	4.8820 KOps/s	4.7573 KOps/s	$\color{#35bf28}+2.62\%$
test_stack	52.5094ms	51.6754ms	19.3516 Ops/s	19.3453 Ops/s	$\color{#35bf28}+0.03\%$
test_cat	52.1261ms	51.5872ms	19.3846 Ops/s	19.3896 Ops/s	$\color{#d91a1a}-0.03\%$
test_sequential_tensordict	0.2946ms	0.2239ms	4.4658 KOps/s	4.5153 KOps/s	$\color{#d91a1a}-1.10\%$
test_sequential_graph_module	0.2683ms	0.1208ms	8.2798 KOps/s	8.1561 KOps/s	$\color{#35bf28}+1.52\%$
test_nested_tensordict	0.4667ms	0.3001ms	3.3321 KOps/s	3.4455 KOps/s	$\color{#d91a1a}-3.29\%$
test_nested_graph_module	0.1942ms	0.1361ms	7.3455 KOps/s	7.7342 KOps/s	$\textbf{\color{#d91a1a}-5.03\%}$

github-actions · 2026-03-06T15:31:15Z

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

Total Benchmarks: 261. Improved: $\large\color{#35bf28}19$. Worsened: $\large\color{#d91a1a}13$.

Expand to view detailed results

Name	Max	Mean	Ops	Ops on Repo `HEAD`	Change
test_plain_set_nested	70.0210μs	14.9025μs	67.1029 KOps/s	67.4927 KOps/s	$\color{#d91a1a}-0.58\%$
test_plain_set_stack_nested	35.7600μs	15.3172μs	65.2860 KOps/s	65.5820 KOps/s	$\color{#d91a1a}-0.45\%$
test_plain_set_nested_inplace	42.9400μs	16.7802μs	59.5942 KOps/s	59.9319 KOps/s	$\color{#d91a1a}-0.56\%$
test_plain_set_stack_nested_inplace	49.2200μs	16.6127μs	60.1948 KOps/s	59.7499 KOps/s	$\color{#35bf28}+0.74\%$
test_items	36.6400μs	5.9482μs	168.1174 KOps/s	164.9846 KOps/s	$\color{#35bf28}+1.90\%$
test_items_nested	0.5273ms	0.4653ms	2.1492 KOps/s	2.1365 KOps/s	$\color{#35bf28}+0.59\%$
test_items_nested_locked	0.5309ms	0.4690ms	2.1324 KOps/s	2.1475 KOps/s	$\color{#d91a1a}-0.70\%$
test_items_nested_leaf	0.1198ms	97.6912μs	10.2363 KOps/s	10.1974 KOps/s	$\color{#35bf28}+0.38\%$
test_items_stack_nested	0.5011ms	0.4633ms	2.1585 KOps/s	2.1546 KOps/s	$\color{#35bf28}+0.18\%$
test_items_stack_nested_leaf	0.1353ms	99.2045μs	10.0802 KOps/s	10.1370 KOps/s	$\color{#d91a1a}-0.56\%$
test_items_stack_nested_locked	0.5471ms	0.4699ms	2.1283 KOps/s	2.1427 KOps/s	$\color{#d91a1a}-0.67\%$
test_keys	29.4000μs	4.2232μs	236.7860 KOps/s	236.3948 KOps/s	$\color{#35bf28}+0.17\%$
test_keys_nested	0.1645ms	0.1299ms	7.6958 KOps/s	7.7237 KOps/s	$\color{#d91a1a}-0.36\%$
test_keys_nested_locked	1.9065ms	0.1405ms	7.1157 KOps/s	7.2193 KOps/s	$\color{#d91a1a}-1.44\%$
test_keys_nested_leaf	0.1595ms	0.1214ms	8.2347 KOps/s	8.3253 KOps/s	$\color{#d91a1a}-1.09\%$
test_keys_stack_nested	0.1953ms	0.1306ms	7.6550 KOps/s	7.6775 KOps/s	$\color{#d91a1a}-0.29\%$
test_keys_stack_nested_leaf	0.1549ms	0.1211ms	8.2568 KOps/s	8.3460 KOps/s	$\color{#d91a1a}-1.07\%$
test_keys_stack_nested_locked	0.1754ms	0.1384ms	7.2253 KOps/s	7.2502 KOps/s	$\color{#d91a1a}-0.34\%$
test_values	3.7371μs	0.9972μs	1.0028 MOps/s	976.3076 KOps/s	$\color{#35bf28}+2.72\%$
test_values_nested	88.9110μs	52.8237μs	18.9309 KOps/s	19.1762 KOps/s	$\color{#d91a1a}-1.28\%$
test_values_nested_locked	82.3510μs	56.4141μs	17.7261 KOps/s	17.9614 KOps/s	$\color{#d91a1a}-1.31\%$
test_values_nested_leaf	87.0210μs	60.5944μs	16.5032 KOps/s	16.7238 KOps/s	$\color{#d91a1a}-1.32\%$
test_values_stack_nested	84.9110μs	53.1956μs	18.7985 KOps/s	19.0615 KOps/s	$\color{#d91a1a}-1.38\%$
test_values_stack_nested_leaf	92.9910μs	60.4435μs	16.5444 KOps/s	16.7065 KOps/s	$\color{#d91a1a}-0.97\%$
test_values_stack_nested_locked	0.1267ms	56.4845μs	17.7040 KOps/s	18.0458 KOps/s	$\color{#d91a1a}-1.89\%$
test_membership	5.3500μs	0.8638μs	1.1577 MOps/s	1.1816 MOps/s	$\color{#d91a1a}-2.02\%$
test_membership_nested	32.9900μs	2.9360μs	340.6020 KOps/s	347.6101 KOps/s	$\color{#d91a1a}-2.02\%$
test_membership_nested_leaf	30.2800μs	2.9432μs	339.7703 KOps/s	350.1143 KOps/s	$\color{#d91a1a}-2.95\%$
test_membership_stacked_nested	39.6500μs	2.9388μs	340.2754 KOps/s	346.6530 KOps/s	$\color{#d91a1a}-1.84\%$
test_membership_stacked_nested_leaf	31.5010μs	2.9112μs	343.4991 KOps/s	347.2106 KOps/s	$\color{#d91a1a}-1.07\%$
test_membership_nested_last	34.7700μs	4.4077μs	226.8756 KOps/s	231.3285 KOps/s	$\color{#d91a1a}-1.92\%$
test_membership_nested_leaf_last	39.6810μs	4.3454μs	230.1261 KOps/s	229.0440 KOps/s	$\color{#35bf28}+0.47\%$
test_membership_stacked_nested_last	35.4610μs	4.4229μs	226.0983 KOps/s	229.5250 KOps/s	$\color{#d91a1a}-1.49\%$
test_membership_stacked_nested_leaf_last	40.9700μs	4.3812μs	228.2475 KOps/s	229.9962 KOps/s	$\color{#d91a1a}-0.76\%$
test_nested_getleaf	48.5410μs	21.9044μs	45.6530 KOps/s	46.1097 KOps/s	$\color{#d91a1a}-0.99\%$
test_nested_get	71.6010μs	20.6680μs	48.3839 KOps/s	48.9018 KOps/s	$\color{#d91a1a}-1.06\%$
test_stacked_getleaf	51.0000μs	21.6846μs	46.1157 KOps/s	46.2680 KOps/s	$\color{#d91a1a}-0.33\%$
test_stacked_get	51.6800μs	20.4729μs	48.8451 KOps/s	48.7964 KOps/s	$\color{#35bf28}+0.10\%$
test_nested_getitemleaf	53.3500μs	22.1884μs	45.0685 KOps/s	45.5723 KOps/s	$\color{#d91a1a}-1.11\%$
test_nested_getitem	48.5310μs	20.9507μs	47.7310 KOps/s	47.9075 KOps/s	$\color{#d91a1a}-0.37\%$
test_stacked_getitemleaf	41.2310μs	22.1553μs	45.1359 KOps/s	45.6772 KOps/s	$\color{#d91a1a}-1.19\%$
test_stacked_getitem	62.3610μs	20.9663μs	47.6957 KOps/s	47.9379 KOps/s	$\color{#d91a1a}-0.51\%$
test_lock_nested	0.5826ms	0.4784ms	2.0904 KOps/s	2.0770 KOps/s	$\color{#35bf28}+0.64\%$
test_lock_stack_nested	0.5827ms	0.4833ms	2.0693 KOps/s	2.0499 KOps/s	$\color{#35bf28}+0.94\%$
test_unlock_nested	0.5241ms	0.3898ms	2.5654 KOps/s	2.5701 KOps/s	$\color{#d91a1a}-0.18\%$
test_unlock_stack_nested	0.4621ms	0.3908ms	2.5590 KOps/s	2.5341 KOps/s	$\color{#35bf28}+0.98\%$
test_flatten_speed	0.1616ms	0.1213ms	8.2447 KOps/s	8.1240 KOps/s	$\color{#35bf28}+1.49\%$
test_unflatten_speed	0.6181ms	0.5737ms	1.7430 KOps/s	1.7592 KOps/s	$\color{#d91a1a}-0.92\%$
test_common_ops	0.8883ms	0.7084ms	1.4116 KOps/s	1.4271 KOps/s	$\color{#d91a1a}-1.09\%$
test_creation	0.1058ms	3.1745μs	315.0125 KOps/s	317.0173 KOps/s	$\color{#d91a1a}-0.63\%$
test_creation_empty	32.8300μs	6.9840μs	143.1842 KOps/s	141.7945 KOps/s	$\color{#35bf28}+0.98\%$
test_creation_nested_1	40.4100μs	11.5269μs	86.7536 KOps/s	86.4803 KOps/s	$\color{#35bf28}+0.32\%$
test_creation_nested_2	74.1310μs	13.2103μs	75.6983 KOps/s	74.3462 KOps/s	$\color{#35bf28}+1.82\%$
test_creation_many_keys[10]	56.0700μs	21.1531μs	47.2744 KOps/s	47.4640 KOps/s	$\color{#d91a1a}-0.40\%$
test_creation_many_keys[50]	0.1241ms	89.8037μs	11.1354 KOps/s	11.0934 KOps/s	$\color{#35bf28}+0.38\%$
test_creation_many_keys[100]	0.2303ms	0.1768ms	5.6566 KOps/s	5.6100 KOps/s	$\color{#35bf28}+0.83\%$
test_creation_nested_many_keys[10]	89.4110μs	45.0526μs	22.1963 KOps/s	22.2149 KOps/s	$\color{#d91a1a}-0.08\%$
test_creation_nested_many_keys[50]	0.2543ms	0.1840ms	5.4334 KOps/s	5.4032 KOps/s	$\color{#35bf28}+0.56\%$
test_clone	51.9510μs	13.0777μs	76.4659 KOps/s	76.0663 KOps/s	$\color{#35bf28}+0.53\%$
test_getitem[int]	1.4897ms	15.1676μs	65.9300 KOps/s	60.5236 KOps/s	$\textbf{\color{#35bf28}+8.93\%}$
test_getitem[slice_int]	0.1473ms	24.0143μs	41.6419 KOps/s	41.5276 KOps/s	$\color{#35bf28}+0.28\%$
test_getitem[range]	0.1873ms	64.3114μs	15.5493 KOps/s	15.8168 KOps/s	$\color{#d91a1a}-1.69\%$
test_getitem[tuple]	0.1416ms	23.6764μs	42.2362 KOps/s	42.4855 KOps/s	$\color{#d91a1a}-0.59\%$
test_getitem[list]	0.2104ms	57.2951μs	17.4535 KOps/s	17.1695 KOps/s	$\color{#35bf28}+1.65\%$
test_setitem_dim[int]	39.7810μs	24.8517μs	40.2387 KOps/s	38.0571 KOps/s	$\textbf{\color{#35bf28}+5.73\%}$
test_setitem_dim[slice_int]	65.0310μs	42.8803μs	23.3207 KOps/s	22.9625 KOps/s	$\color{#35bf28}+1.56\%$
test_setitem_dim[range]	0.1376ms	95.9416μs	10.4230 KOps/s	10.4985 KOps/s	$\color{#d91a1a}-0.72\%$
test_setitem_dim[tuple]	79.1010μs	39.8204μs	25.1128 KOps/s	24.4569 KOps/s	$\color{#35bf28}+2.68\%$
test_setitem	51.8710μs	17.5950μs	56.8344 KOps/s	56.1191 KOps/s	$\color{#35bf28}+1.27\%$
test_set	46.9900μs	16.9739μs	58.9138 KOps/s	59.3895 KOps/s	$\color{#d91a1a}-0.80\%$
test_set_shared	0.5571ms	0.2102ms	4.7578 KOps/s	4.9094 KOps/s	$\color{#d91a1a}-3.09\%$
test_update	0.4067ms	21.3782μs	46.7767 KOps/s	46.5369 KOps/s	$\color{#35bf28}+0.52\%$
test_update_nested	76.3210μs	32.3924μs	30.8714 KOps/s	30.0669 KOps/s	$\color{#35bf28}+2.68\%$
test_update__nested	0.5057ms	33.7980μs	29.5876 KOps/s	29.0195 KOps/s	$\color{#35bf28}+1.96\%$
test_set_nested	48.5610μs	18.7228μs	53.4107 KOps/s	53.5866 KOps/s	$\color{#d91a1a}-0.33\%$
test_set_nested_new	64.4310μs	23.5900μs	42.3908 KOps/s	41.5842 KOps/s	$\color{#35bf28}+1.94\%$
test_select	78.1110μs	40.2547μs	24.8418 KOps/s	24.4331 KOps/s	$\color{#35bf28}+1.67\%$
test_select_nested	0.1314ms	75.6211μs	13.2238 KOps/s	13.3544 KOps/s	$\color{#d91a1a}-0.98\%$
test_exclude_nested	0.1245ms	93.2843μs	10.7199 KOps/s	10.8564 KOps/s	$\color{#d91a1a}-1.26\%$
test_empty[True]	0.4726ms	0.4005ms	2.4969 KOps/s	2.4986 KOps/s	$\color{#d91a1a}-0.07\%$
test_empty[False]	10.9367μs	1.3441μs	743.9899 KOps/s	758.7407 KOps/s	$\color{#d91a1a}-1.94\%$
test_to	0.1045ms	74.7460μs	13.3786 KOps/s	13.4407 KOps/s	$\color{#d91a1a}-0.46\%$
test_to_nonblocking	0.1105ms	64.7264μs	15.4497 KOps/s	15.4278 KOps/s	$\color{#35bf28}+0.14\%$
test_unbind_speed	0.3883ms	0.3352ms	2.9832 KOps/s	3.0104 KOps/s	$\color{#d91a1a}-0.90\%$
test_unbind_speed_stack0	0.3800ms	0.3326ms	3.0063 KOps/s	3.0418 KOps/s	$\color{#d91a1a}-1.17\%$
test_unbind_speed_stack1	0.1035s	1.0576ms	945.5808 Ops/s	1.1912 KOps/s	$\textbf{\color{#d91a1a}-20.62\%}$
test_split	1.2514ms	1.1486ms	870.6000 Ops/s	786.3578 Ops/s	$\textbf{\color{#35bf28}+10.71\%}$
test_chunk	0.1038s	1.2097ms	826.6805 Ops/s	916.7735 Ops/s	$\textbf{\color{#d91a1a}-9.83\%}$
test_to_cpu_blocking	19.3758ms	19.2917ms	51.8357 Ops/s	51.3026 Ops/s	$\color{#35bf28}+1.04\%$
test_to_cpu_global_sync	11.3359ms	11.2012ms	89.2765 Ops/s	78.9689 Ops/s	$\textbf{\color{#35bf28}+13.05\%}$
test_to_cpu_event_sync	0.1152s	13.4113ms	74.5641 Ops/s	80.2244 Ops/s	$\textbf{\color{#d91a1a}-7.06\%}$
test_to_cpu_default	12.4567ms	12.1935ms	82.0109 Ops/s	80.6764 Ops/s	$\color{#35bf28}+1.65\%$
test_consolidate[False-None]	4.2616ms	4.1606ms	240.3515 Ops/s	216.9879 Ops/s	$\textbf{\color{#35bf28}+10.77\%}$
test_consolidate[default-None]	2.1458ms	2.0100ms	497.5119 Ops/s	487.1495 Ops/s	$\color{#35bf28}+2.13\%$
test_consolidate[reduce-overhead-None]	2.0355ms	1.9435ms	514.5298 Ops/s	508.6437 Ops/s	$\color{#35bf28}+1.16\%$
test_consolidate_njt[False-None]	8.6913ms	8.4440ms	118.4276 Ops/s	117.7129 Ops/s	$\color{#35bf28}+0.61\%$
test_to[False-False-None]	2.3227ms	2.0744ms	482.0637 Ops/s	473.9723 Ops/s	$\color{#35bf28}+1.71\%$
test_to[True-False-None]	2.1310ms	1.8945ms	527.8518 Ops/s	520.0717 Ops/s	$\color{#35bf28}+1.50\%$
test_to[within-False-None]	6.3861ms	6.1350ms	162.9991 Ops/s	163.1155 Ops/s	$\color{#d91a1a}-0.07\%$
test_to[True-default-None]	8.9004ms	8.6410ms	115.7279 Ops/s	111.2757 Ops/s	$\color{#35bf28}+4.00\%$
test_to_njt[False-False-None]	8.5305ms	8.4515ms	118.3218 Ops/s	117.1108 Ops/s	$\color{#35bf28}+1.03\%$
test_to_njt[True-False-None]	7.0806ms	6.8804ms	145.3407 Ops/s	142.1915 Ops/s	$\color{#35bf28}+2.21\%$
test_to_njt[within-False-None]	15.6386ms	15.3588ms	65.1092 Ops/s	63.7603 Ops/s	$\color{#35bf28}+2.12\%$
test_creation[device0]	0.4141ms	0.1144ms	8.7447 KOps/s	8.6670 KOps/s	$\color{#35bf28}+0.90\%$
test_creation_from_tensor	0.3957ms	0.1118ms	8.9442 KOps/s	8.8637 KOps/s	$\color{#35bf28}+0.91\%$
test_add_one[memmap_tensor0]	0.3490ms	6.4542μs	154.9385 KOps/s	155.9302 KOps/s	$\color{#d91a1a}-0.64\%$
test_contiguous[memmap_tensor0]	14.2600μs	0.6840μs	1.4620 MOps/s	2.1148 MOps/s	$\textbf{\color{#d91a1a}-30.87\%}$
test_stack[memmap_tensor0]	46.9600μs	4.7189μs	211.9120 KOps/s	212.4636 KOps/s	$\color{#d91a1a}-0.26\%$
test_memmaptd_index	0.9804ms	0.2749ms	3.6372 KOps/s	3.7386 KOps/s	$\color{#d91a1a}-2.71\%$
test_memmaptd_index_astensor	0.5560ms	0.3765ms	2.6561 KOps/s	2.7436 KOps/s	$\color{#d91a1a}-3.19\%$
test_memmaptd_index_op	0.7807ms	0.6222ms	1.6071 KOps/s	1.6353 KOps/s	$\color{#d91a1a}-1.72\%$
test_serialize_model	0.1389s	0.1364s	7.3322 Ops/s	7.2803 Ops/s	$\color{#35bf28}+0.71\%$
test_serialize_model_pickle	1.3680s	1.1894s	0.8407 Ops/s	0.8261 Ops/s	$\color{#35bf28}+1.78\%$
test_serialize_weights	0.1386s	0.1360s	7.3520 Ops/s	7.3839 Ops/s	$\color{#d91a1a}-0.43\%$
test_serialize_weights_returnearly	0.4390s	93.1447ms	10.7360 Ops/s	11.2691 Ops/s	$\color{#d91a1a}-4.73\%$
test_serialize_weights_pickle	1.3506s	1.2143s	0.8235 Ops/s	0.8231 Ops/s	$\color{#35bf28}+0.05\%$
test_reshape_pytree	0.2052ms	34.0810μs	29.3419 KOps/s	30.7136 KOps/s	$\color{#d91a1a}-4.47\%$
test_reshape_td	83.1210μs	47.2263μs	21.1746 KOps/s	22.4973 KOps/s	$\textbf{\color{#d91a1a}-5.88\%}$
test_view_pytree	0.2160ms	33.8747μs	29.5206 KOps/s	30.9972 KOps/s	$\color{#d91a1a}-4.76\%$
test_view_td	87.9710μs	53.2281μs	18.7871 KOps/s	18.9527 KOps/s	$\color{#d91a1a}-0.87\%$
test_unbind_pytree	0.2512ms	36.8672μs	27.1244 KOps/s	27.7626 KOps/s	$\color{#d91a1a}-2.30\%$
test_unbind_td	82.3310μs	50.3740μs	19.8515 KOps/s	20.2738 KOps/s	$\color{#d91a1a}-2.08\%$
test_split_pytree	0.2184ms	44.6286μs	22.4072 KOps/s	23.8758 KOps/s	$\textbf{\color{#d91a1a}-6.15\%}$
test_split_td	0.1164ms	68.2950μs	14.6424 KOps/s	15.5342 KOps/s	$\textbf{\color{#d91a1a}-5.74\%}$
test_add_pytree	0.2286ms	44.3322μs	22.5570 KOps/s	23.9562 KOps/s	$\textbf{\color{#d91a1a}-5.84\%}$
test_add_td	91.0720μs	57.5619μs	17.3726 KOps/s	18.3801 KOps/s	$\textbf{\color{#d91a1a}-5.48\%}$
test_compile_add_one_nested[tensordict-compile]	0.3247ms	0.1445ms	6.9205 KOps/s	6.8813 KOps/s	$\color{#35bf28}+0.57\%$
test_compile_add_one_nested[tensordict-eager]	0.2902ms	0.2056ms	4.8646 KOps/s	4.9825 KOps/s	$\color{#d91a1a}-2.37\%$
test_compile_add_one_nested[pytree-compile]	0.1966ms	0.1080ms	9.2579 KOps/s	9.0051 KOps/s	$\color{#35bf28}+2.81\%$
test_compile_add_one_nested[pytree-eager]	0.4271ms	0.1812ms	5.5190 KOps/s	5.5632 KOps/s	$\color{#d91a1a}-0.79\%$
test_compile_copy_nested[tensordict-compile]	0.3082ms	10.5469μs	94.8150 KOps/s	96.9610 KOps/s	$\color{#d91a1a}-2.21\%$
test_compile_copy_nested[tensordict-eager]	82.1910μs	53.8457μs	18.5716 KOps/s	18.2623 KOps/s	$\color{#35bf28}+1.69\%$
test_compile_copy_nested[pytree-compile]	45.5700μs	9.6465μs	103.6646 KOps/s	99.7691 KOps/s	$\color{#35bf28}+3.90\%$
test_compile_copy_nested[pytree-eager]	0.4265ms	69.4631μs	14.3961 KOps/s	14.6397 KOps/s	$\color{#d91a1a}-1.66\%$
test_compile_add_one_flat[tensordict-compile]	0.2278ms	0.1776ms	5.6313 KOps/s	5.2680 KOps/s	$\textbf{\color{#35bf28}+6.89\%}$
test_compile_add_one_flat[tensordict-eager]	0.3283ms	0.2777ms	3.6005 KOps/s	3.4991 KOps/s	$\color{#35bf28}+2.90\%$
test_compile_add_one_flat[tensorclass-compile]	0.1684ms	0.1176ms	8.5047 KOps/s	8.0644 KOps/s	$\textbf{\color{#35bf28}+5.46\%}$
test_compile_add_one_flat[tensorclass-eager]	0.1190ms	73.5225μs	13.6013 KOps/s	13.4587 KOps/s	$\color{#35bf28}+1.06\%$
test_compile_add_one_flat[pytree-compile]	0.1936ms	0.1587ms	6.2996 KOps/s	6.0994 KOps/s	$\color{#35bf28}+3.28\%$
test_compile_add_one_flat[pytree-eager]	0.8014ms	0.5285ms	1.8920 KOps/s	1.7904 KOps/s	$\textbf{\color{#35bf28}+5.68\%}$
test_compile_add_self_flat[tensordict-eager]	0.5086ms	0.3324ms	3.0083 KOps/s	2.9570 KOps/s	$\color{#35bf28}+1.74\%$
test_compile_add_self_flat[tensordict-compile]	0.2306ms	0.1795ms	5.5720 KOps/s	5.2719 KOps/s	$\textbf{\color{#35bf28}+5.69\%}$
test_compile_add_self_flat[tensorclass-eager]	0.1357ms	90.1068μs	11.0979 KOps/s	11.1199 KOps/s	$\color{#d91a1a}-0.20\%$
test_compile_add_self_flat[tensorclass-compile]	0.1620ms	0.1198ms	8.3468 KOps/s	7.9888 KOps/s	$\color{#35bf28}+4.48\%$
test_compile_add_self_flat[pytree-eager]	0.6583ms	0.4368ms	2.2893 KOps/s	2.2565 KOps/s	$\color{#35bf28}+1.45\%$
test_compile_add_self_flat[pytree-compile]	0.2022ms	0.1586ms	6.3037 KOps/s	6.1288 KOps/s	$\color{#35bf28}+2.85\%$
test_compile_copy_flat[tensordict-compile]	42.8700μs	13.3910μs	74.6770 KOps/s	70.5391 KOps/s	$\textbf{\color{#35bf28}+5.87\%}$
test_compile_copy_flat[tensordict-eager]	76.6210μs	41.9615μs	23.8314 KOps/s	24.2200 KOps/s	$\color{#d91a1a}-1.60\%$
test_compile_copy_flat[pytree-compile]	48.4910μs	10.9713μs	91.1468 KOps/s	91.6971 KOps/s	$\color{#d91a1a}-0.60\%$
test_compile_copy_flat[pytree-eager]	0.4054ms	52.1982μs	19.1578 KOps/s	19.0813 KOps/s	$\color{#35bf28}+0.40\%$
test_compile_assign_and_add[tensordict-compile]	2.0150ms	0.1744ms	5.7334 KOps/s	5.5068 KOps/s	$\color{#35bf28}+4.11\%$
test_compile_assign_and_add[tensordict-eager]	3.3722ms	3.2867ms	304.2594 Ops/s	299.9706 Ops/s	$\color{#35bf28}+1.43\%$
test_compile_assign_and_add[pytree-compile]	2.0316ms	0.1622ms	6.1651 KOps/s	6.0482 KOps/s	$\color{#35bf28}+1.93\%$
test_compile_assign_and_add[pytree-eager]	2.9207ms	2.7932ms	358.0111 Ops/s	356.5978 Ops/s	$\color{#35bf28}+0.40\%$
test_compile_indexing[tensor-tensordict-compile]	0.1720ms	0.1103ms	9.0649 KOps/s	8.8330 KOps/s	$\color{#35bf28}+2.63\%$
test_compile_indexing[tensor-tensordict-eager]	0.3163ms	73.2233μs	13.6569 KOps/s	13.3764 KOps/s	$\color{#35bf28}+2.10\%$
test_compile_indexing[tensor-tensorclass-compile]	0.1519ms	96.5276μs	10.3597 KOps/s	10.0810 KOps/s	$\color{#35bf28}+2.76\%$
test_compile_indexing[tensor-tensorclass-eager]	0.2491ms	44.4916μs	22.4761 KOps/s	22.2730 KOps/s	$\color{#35bf28}+0.91\%$
test_compile_indexing[tensor-pytree-compile]	0.1393ms	97.6363μs	10.2421 KOps/s	10.1081 KOps/s	$\color{#35bf28}+1.33\%$
test_compile_indexing[tensor-pytree-eager]	0.2769ms	44.4980μs	22.4729 KOps/s	22.1346 KOps/s	$\color{#35bf28}+1.53\%$
test_compile_indexing[slice-tensordict-compile]	0.1009ms	56.8707μs	17.5837 KOps/s	17.4205 KOps/s	$\color{#35bf28}+0.94\%$
test_compile_indexing[slice-tensordict-eager]	0.2186ms	26.7467μs	37.3877 KOps/s	36.2341 KOps/s	$\color{#35bf28}+3.18\%$
test_compile_indexing[slice-tensorclass-compile]	81.4210μs	43.9204μs	22.7684 KOps/s	22.2723 KOps/s	$\color{#35bf28}+2.23\%$
test_compile_indexing[slice-tensorclass-eager]	0.2514ms	22.3896μs	44.6635 KOps/s	44.4624 KOps/s	$\color{#35bf28}+0.45\%$
test_compile_indexing[slice-pytree-compile]	79.3010μs	45.6584μs	21.9018 KOps/s	22.0406 KOps/s	$\color{#d91a1a}-0.63\%$
test_compile_indexing[slice-pytree-eager]	0.2863ms	22.5522μs	44.3415 KOps/s	44.9063 KOps/s	$\color{#d91a1a}-1.26\%$
test_compile_indexing[int-tensordict-compile]	0.1157ms	57.1069μs	17.5110 KOps/s	17.0801 KOps/s	$\color{#35bf28}+2.52\%$
test_compile_indexing[int-tensordict-eager]	0.2773ms	26.9246μs	37.1408 KOps/s	36.3254 KOps/s	$\color{#35bf28}+2.24\%$
test_compile_indexing[int-tensorclass-compile]	89.2210μs	44.3599μs	22.5429 KOps/s	22.4137 KOps/s	$\color{#35bf28}+0.58\%$
test_compile_indexing[int-tensorclass-eager]	0.2589ms	22.4277μs	44.5878 KOps/s	44.9856 KOps/s	$\color{#d91a1a}-0.88\%$
test_compile_indexing[int-pytree-compile]	75.8210μs	44.1036μs	22.6739 KOps/s	22.3387 KOps/s	$\color{#35bf28}+1.50\%$
test_compile_indexing[int-pytree-eager]	0.2519ms	22.2362μs	44.9717 KOps/s	45.1486 KOps/s	$\color{#d91a1a}-0.39\%$
test_compile_replace[single-eager]	84.0110μs	46.5068μs	21.5022 KOps/s	21.0591 KOps/s	$\color{#35bf28}+2.10\%$
test_compile_replace[single-compile]	0.1851ms	0.1048ms	9.5437 KOps/s	9.2366 KOps/s	$\color{#35bf28}+3.32\%$
test_compile_replace[multi-eager]	0.6060ms	0.5601ms	1.7855 KOps/s	1.7670 KOps/s	$\color{#35bf28}+1.05\%$
test_compile_replace[multi-compile]	0.1772ms	0.1114ms	8.9776 KOps/s	8.4086 KOps/s	$\textbf{\color{#35bf28}+6.77\%}$
test_compile_tc_getattr_20[eager]	0.2417ms	0.1783ms	5.6086 KOps/s	5.9433 KOps/s	$\textbf{\color{#d91a1a}-5.63\%}$
test_compile_tc_getattr_20[compile]	0.1789ms	0.1190ms	8.4046 KOps/s	8.2019 KOps/s	$\color{#35bf28}+2.47\%$
test_compile_clone_shallow[20-eager]	44.8400μs	19.3473μs	51.6869 KOps/s	51.1154 KOps/s	$\color{#35bf28}+1.12\%$
test_compile_clone_shallow[20-compile]	81.5410μs	11.2698μs	88.7328 KOps/s	88.0581 KOps/s	$\color{#35bf28}+0.77\%$
test_compile_clone_shallow[40-eager]	66.3810μs	34.3775μs	29.0888 KOps/s	29.1795 KOps/s	$\color{#d91a1a}-0.31\%$
test_compile_clone_shallow[40-compile]	44.8300μs	12.5829μs	79.4726 KOps/s	78.3976 KOps/s	$\color{#35bf28}+1.37\%$
test_compile_clone_shallow[80-eager]	0.1018ms	62.9000μs	15.8982 KOps/s	15.6393 KOps/s	$\color{#35bf28}+1.66\%$
test_compile_clone_shallow[80-compile]	50.5600μs	14.5533μs	68.7128 KOps/s	65.9108 KOps/s	$\color{#35bf28}+4.25\%$
test_compile_update_inplace[eager]	97.3220μs	59.7122μs	16.7470 KOps/s	16.7158 KOps/s	$\color{#35bf28}+0.19\%$
test_compile_update_inplace[compile]	0.2106ms	0.1394ms	7.1734 KOps/s	6.7007 KOps/s	$\textbf{\color{#35bf28}+7.05\%}$
test_mod_add[eager]	99.2610μs	50.3944μs	19.8435 KOps/s	19.6742 KOps/s	$\color{#35bf28}+0.86\%$
test_mod_add[compile]	0.1551ms	0.1050ms	9.5281 KOps/s	9.3658 KOps/s	$\color{#35bf28}+1.73\%$
test_mod_add[compile-overhead]	0.2324ms	0.1470ms	6.8043 KOps/s	6.5235 KOps/s	$\color{#35bf28}+4.30\%$
test_mod_wrap[eager]	0.3800ms	0.3026ms	3.3043 KOps/s	3.3755 KOps/s	$\color{#d91a1a}-2.11\%$
test_mod_wrap[compile]	0.3991ms	0.3460ms	2.8902 KOps/s	2.8114 KOps/s	$\color{#35bf28}+2.80\%$
test_mod_wrap[compile-overhead]	7.3809ms	4.0795ms	245.1299 Ops/s	253.4924 Ops/s	$\color{#d91a1a}-3.30\%$
test_mod_wrap_and_backward[eager]	1.6236ms	1.5002ms	666.5711 Ops/s	658.1628 Ops/s	$\color{#35bf28}+1.28\%$
test_mod_wrap_and_backward[compile]	1.6361ms	1.5553ms	642.9525 Ops/s	683.0620 Ops/s	$\textbf{\color{#d91a1a}-5.87\%}$
test_mod_wrap_and_backward[compile-overhead]	1.4647ms	1.0000ms	1.0000 KOps/s	1.0961 KOps/s	$\textbf{\color{#d91a1a}-8.77\%}$
test_seq_add[eager]	0.2233ms	0.1516ms	6.5957 KOps/s	6.2680 KOps/s	$\textbf{\color{#35bf28}+5.23\%}$
test_seq_add[compile]	0.1725ms	0.1119ms	8.9332 KOps/s	8.4925 KOps/s	$\textbf{\color{#35bf28}+5.19\%}$
test_seq_add[compile-overhead]	0.2260ms	0.1658ms	6.0324 KOps/s	6.1969 KOps/s	$\color{#d91a1a}-2.66\%$
test_seq_wrap[eager]	0.6172ms	0.5523ms	1.8108 KOps/s	1.8757 KOps/s	$\color{#d91a1a}-3.46\%$
test_seq_wrap[compile]	0.4690ms	0.3869ms	2.5847 KOps/s	2.6826 KOps/s	$\color{#d91a1a}-3.65\%$
test_seq_wrap[compile-overhead]	0.3371ms	0.2662ms	3.7564 KOps/s	3.7167 KOps/s	$\color{#35bf28}+1.07\%$
test_func_call_runtime[False-eager]	0.9022ms	0.8331ms	1.2003 KOps/s	1.1648 KOps/s	$\color{#35bf28}+3.05\%$
test_func_call_runtime[False-compile]	0.9706ms	0.9072ms	1.1023 KOps/s	1.0866 KOps/s	$\color{#35bf28}+1.45\%$
test_func_call_runtime[False-compile-overhead]	0.5613ms	0.4642ms	2.1541 KOps/s	2.1384 KOps/s	$\color{#35bf28}+0.74\%$
test_func_call_runtime[True-eager]	1.1363ms	1.0781ms	927.5463 Ops/s	920.0990 Ops/s	$\color{#35bf28}+0.81\%$
test_func_call_runtime[True-compile]	0.9818ms	0.9222ms	1.0844 KOps/s	1.0648 KOps/s	$\color{#35bf28}+1.84\%$
test_func_call_runtime[True-compile-overhead]	0.5368ms	0.4772ms	2.0954 KOps/s	2.0762 KOps/s	$\color{#35bf28}+0.92\%$
test_func_call_cm_runtime[False-eager]	0.9468ms	0.8861ms	1.1286 KOps/s	1.1690 KOps/s	$\color{#d91a1a}-3.46\%$
test_func_call_cm_runtime[False-compile]	1.0012ms	0.9159ms	1.0918 KOps/s	1.0788 KOps/s	$\color{#35bf28}+1.21\%$
test_func_call_cm_runtime[False-compile-overhead]	0.5330ms	0.4672ms	2.1404 KOps/s	2.1303 KOps/s	$\color{#35bf28}+0.48\%$
test_func_call_cm_runtime[True-eager]	1.3060ms	1.2272ms	814.8518 Ops/s	804.3518 Ops/s	$\color{#35bf28}+1.31\%$
test_func_call_cm_runtime[True-compile]	1.0109ms	0.9572ms	1.0447 KOps/s	1.0276 KOps/s	$\color{#35bf28}+1.66\%$
test_func_call_cm_runtime[True-compile-overhead]	0.5798ms	0.5107ms	1.9580 KOps/s	1.9258 KOps/s	$\color{#35bf28}+1.67\%$
test_vmap_func_call_cm_runtime[eager]	2.8496ms	2.3629ms	423.2044 Ops/s	416.3101 Ops/s	$\color{#35bf28}+1.66\%$
test_vmap_func_call_cm_runtime[compile]	1.0564ms	0.9777ms	1.0228 KOps/s	1.0082 KOps/s	$\color{#35bf28}+1.45\%$
test_vmap_func_call_cm_runtime[compile-overhead]	0.5682ms	0.5167ms	1.9355 KOps/s	1.9150 KOps/s	$\color{#35bf28}+1.07\%$
test_distributed	2.7916ms	0.1749ms	5.7161 KOps/s	6.5135 KOps/s	$\textbf{\color{#d91a1a}-12.24\%}$
test_tdmodule	0.1779ms	28.3807μs	35.2352 KOps/s	35.8465 KOps/s	$\color{#d91a1a}-1.71\%$
test_tdmodule_dispatch	72.7610μs	45.2216μs	22.1133 KOps/s	21.8889 KOps/s	$\color{#35bf28}+1.03\%$
test_tdseq	66.3010μs	27.1554μs	36.8251 KOps/s	36.8426 KOps/s	$\color{#d91a1a}-0.05\%$
test_tdseq_dispatch	75.4710μs	47.5791μs	21.0176 KOps/s	20.9630 KOps/s	$\color{#35bf28}+0.26\%$
test_instantiation_functorch	2.1943ms	2.0794ms	480.9151 Ops/s	480.1015 Ops/s	$\color{#35bf28}+0.17\%$
test_exec_functorch	0.2304ms	0.1803ms	5.5463 KOps/s	5.5594 KOps/s	$\color{#d91a1a}-0.24\%$
test_exec_functional_call	0.2231ms	0.1637ms	6.1091 KOps/s	6.2130 KOps/s	$\color{#d91a1a}-1.67\%$
test_exec_td_decorator	0.4400ms	0.2390ms	4.1849 KOps/s	4.2503 KOps/s	$\color{#d91a1a}-1.54\%$
test_vmap_mlp_speed_decorator[True-True]	0.9979ms	0.8215ms	1.2173 KOps/s	1.2050 KOps/s	$\color{#35bf28}+1.03\%$
test_vmap_mlp_speed_decorator[True-False]	1.0183ms	0.8413ms	1.1887 KOps/s	1.1695 KOps/s	$\color{#35bf28}+1.64\%$
test_vmap_mlp_speed_decorator[False-True]	0.9378ms	0.7199ms	1.3891 KOps/s	1.3472 KOps/s	$\color{#35bf28}+3.11\%$
test_vmap_mlp_speed_decorator[False-False]	0.9199ms	0.7166ms	1.3954 KOps/s	1.3562 KOps/s	$\color{#35bf28}+2.89\%$
test_vmap_transformer_speed_decorator[True-True]	20.9544ms	20.4908ms	48.8025 Ops/s	47.1333 Ops/s	$\color{#35bf28}+3.54\%$
test_vmap_transformer_speed_decorator[True-False]	21.0121ms	20.5128ms	48.7500 Ops/s	48.1325 Ops/s	$\color{#35bf28}+1.28\%$
test_vmap_transformer_speed_decorator[False-True]	20.8740ms	20.2702ms	49.3336 Ops/s	48.9000 Ops/s	$\color{#35bf28}+0.89\%$
test_vmap_transformer_speed_decorator[False-False]	20.8591ms	20.2718ms	49.3295 Ops/s	48.9145 Ops/s	$\color{#35bf28}+0.85\%$
test_to_module_speed[True]	2.0504ms	1.4758ms	677.5965 Ops/s	673.4486 Ops/s	$\color{#35bf28}+0.62\%$
test_to_module_speed[False]	1.9318ms	1.4536ms	687.9563 Ops/s	680.8005 Ops/s	$\color{#35bf28}+1.05\%$
test_tc_init	81.0210μs	45.1692μs	22.1390 KOps/s	22.1994 KOps/s	$\color{#d91a1a}-0.27\%$
test_tc_init_tensor_only	37.4010μs	9.6840μs	103.2635 KOps/s	101.9391 KOps/s	$\color{#35bf28}+1.30\%$
test_tc_init_nested	0.1191ms	87.8365μs	11.3848 KOps/s	11.3506 KOps/s	$\color{#35bf28}+0.30\%$
test_tc_init_many_fields	48.1200μs	16.3521μs	61.1542 KOps/s	60.8158 KOps/s	$\color{#35bf28}+0.56\%$
test_tc_first_layer_tensor	28.5110μs	1.8326μs	545.6834 KOps/s	540.4270 KOps/s	$\color{#35bf28}+0.97\%$
test_tc_first_layer_tensor_only	3.5200μs	0.4102μs	2.4380 MOps/s	2.4594 MOps/s	$\color{#d91a1a}-0.87\%$
test_tc_first_layer_tensor_set	28.4000μs	3.9700μs	251.8920 KOps/s	252.9086 KOps/s	$\color{#d91a1a}-0.40\%$
test_tc_first_layer_tensor_only_set	23.2000μs	3.2942μs	303.5661 KOps/s	305.0497 KOps/s	$\color{#d91a1a}-0.49\%$
test_tc_first_layer_nontensor	48.2000μs	6.1494μs	162.6178 KOps/s	160.6697 KOps/s	$\color{#35bf28}+1.21\%$
test_tc_second_layer_tensor	33.8600μs	4.4201μs	226.2390 KOps/s	227.4721 KOps/s	$\color{#d91a1a}-0.54\%$
test_tc_second_layer_nontensor	49.2210μs	8.6439μs	115.6881 KOps/s	115.0925 KOps/s	$\color{#35bf28}+0.52\%$
test_unbind	0.2628s	16.5450ms	60.4412 Ops/s	53.0614 Ops/s	$\textbf{\color{#35bf28}+13.91\%}$
test_full_like	5.4464ms	4.3942ms	227.5740 Ops/s	58.3779 Ops/s	$\textbf{\color{#35bf28}+289.83\%}$
test_zeros_like	4.9649ms	4.3873ms	227.9285 Ops/s	59.6096 Ops/s	$\textbf{\color{#35bf28}+282.37\%}$
test_ones_like	4.6448ms	4.4049ms	227.0183 Ops/s	59.5299 Ops/s	$\textbf{\color{#35bf28}+281.35\%}$
test_clone	6.9395ms	6.5257ms	153.2391 Ops/s	56.4190 Ops/s	$\textbf{\color{#35bf28}+171.61\%}$
test_squeeze	60.4210μs	14.5311μs	68.8180 KOps/s	70.9669 KOps/s	$\color{#d91a1a}-3.03\%$
test_unsqueeze	0.1594ms	0.1107ms	9.0347 KOps/s	8.9854 KOps/s	$\color{#35bf28}+0.55\%$
test_split	0.2438ms	0.1847ms	5.4140 KOps/s	5.3905 KOps/s	$\color{#35bf28}+0.44\%$
test_permute	0.2746ms	0.2108ms	4.7446 KOps/s	4.8579 KOps/s	$\color{#d91a1a}-2.33\%$
test_stack	53.3136ms	51.9025ms	19.2669 Ops/s	19.3997 Ops/s	$\color{#d91a1a}-0.68\%$
test_cat	52.4766ms	51.6763ms	19.3512 Ops/s	19.4082 Ops/s	$\color{#d91a1a}-0.29\%$
test_sequential_tensordict	0.3276ms	0.2262ms	4.4207 KOps/s	4.5361 KOps/s	$\color{#d91a1a}-2.54\%$
test_sequential_graph_module	0.5419ms	0.1218ms	8.2132 KOps/s	8.2197 KOps/s	$\color{#d91a1a}-0.08\%$
test_nested_tensordict	0.3323ms	0.2888ms	3.4629 KOps/s	3.4701 KOps/s	$\color{#d91a1a}-0.21\%$
test_nested_graph_module	0.2501ms	0.1336ms	7.4847 KOps/s	7.7406 KOps/s	$\color{#d91a1a}-3.31\%$

Update

c37b533

[ghstack-poisoned]

vmoens mentioned this pull request Mar 6, 2026

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

Open

This was referenced Mar 6, 2026

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

Open

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

Open

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

Provide feedback

Saved searches

Use saved searches to filter your results more quickly

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

[DTensor] Add Strategy B (local-shard transfer + redistribute on receiver)#1641
vmoens wants to merge 1 commit intogh/vmoens/82/basefrom
gh/vmoens/82/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: $\large\color{#35bf28}13$. Worsened: $\large\color{#d91a1a}8$.

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: $\large\color{#35bf28}19$. Worsened: $\large\color{#d91a1a}13$.

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