Skip to content

[DTensor] Add ModelTransferPlan for multi-parameter precomputed transfers - #1650

Open
vmoens wants to merge 5 commits into
gh/vmoens/90/basefrom
gh/vmoens/90/head
Open

[DTensor] Add ModelTransferPlan for multi-parameter precomputed transfers#1650
vmoens wants to merge 5 commits into
gh/vmoens/90/basefrom
gh/vmoens/90/head

Conversation

@vmoens

@vmoens vmoens commented Mar 10, 2026 •
edited
Loading

Copy link
Copy Markdown
Collaborator

Stack from ghstack (oldest at bottom):

Implement ParameterPlan and ModelTransferPlan classes for precomputed
model-level weight transfers between differently-sharded meshes:

  • ModelTransferPlan.build() computes per-parameter transfer plans from
    ShardingDescriptors, with optional name mapping, per-parameter
    transforms, and automatic strategy selection (optimal/materialize/
    direct_copy).
  • ModelTransferPlan.execute() runs all parameter transfers via P2P.
  • Batching groups small parameters within a buffer_size limit.
  • summary(), total_bytes, per_rank_bytes for debugging and profiling.

Also re-exports ModelTransferPlan and ParameterPlan from the public
tensordict.dtensor module.

Made-with: Cursor

@vmoens
Update
fe0869e 
[ghstack-poisoned]
@github-actions

github-actions Bot commented Mar 10, 2026

Copy link
Copy Markdown
Contributor

PR Title Label Error

Unknown or invalid prefix [DTensor].

Current title: [DTensor] Add ModelTransferPlan for multi-parameter precomputed transfers

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.

@vmoens vmoens mentioned this pull request Mar 10, 2026
Open
@github-actions github-actions Bot added the Test label Mar 10, 2026
@vmoens vmoens mentioned this pull request Mar 10, 2026
Open
@meta-cla meta-cla Bot added the CLA Signed This label is managed by the Facebook bot. Authors need to sign the CLA before a PR can be reviewed. label Mar 10, 2026
This was referenced Mar 10, 2026
Open
Open
Open
Open
Open
@github-actions

github-actions Bot commented Mar 10, 2026

Copy link
Copy Markdown
Contributor

PR Title Label Error

Unknown or invalid prefix [DTensor].

Current title: [DTensor] Add ModelTransferPlan for multi-parameter precomputed transfers

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

github-actions Bot commented Mar 10, 2026 •
edited
Loading

Copy link
Copy Markdown
Contributor

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

Total Benchmarks: 261. Improved: $\large\color{#35bf28}26$. Worsened: $\large\color{#d91a1a}16$.

Expand to view detailed results
Name Max Mean Ops Ops on Repo HEAD Change
test_plain_set_nested 35.7600μs 15.0153μs 66.5988 KOps/s 68.1563 KOps/s $\color{#d91a1a}-2.29\%$
test_plain_set_stack_nested 37.0000μs 15.2807μs 65.4420 KOps/s 66.1648 KOps/s $\color{#d91a1a}-1.09\%$
test_plain_set_nested_inplace 52.6500μs 16.8233μs 59.4415 KOps/s 59.1668 KOps/s $\color{#35bf28}+0.46\%$
test_plain_set_stack_nested_inplace 49.5200μs 16.7189μs 59.8124 KOps/s 60.3917 KOps/s $\color{#d91a1a}-0.96\%$
test_items 42.2410μs 6.0591μs 165.0421 KOps/s 165.6190 KOps/s $\color{#d91a1a}-0.35\%$
test_items_nested 0.4946ms 0.4635ms 2.1573 KOps/s 2.0964 KOps/s $\color{#35bf28}+2.91\%$
test_items_nested_locked 0.5387ms 0.4673ms 2.1399 KOps/s 2.1002 KOps/s $\color{#35bf28}+1.89\%$
test_items_nested_leaf 0.1356ms 98.5716μs 10.1449 KOps/s 10.1426 KOps/s $\color{#35bf28}+0.02\%$
test_items_stack_nested 0.7006ms 0.4694ms 2.1303 KOps/s 2.1023 KOps/s $\color{#35bf28}+1.33\%$
test_items_stack_nested_leaf 0.1347ms 98.8420μs 10.1172 KOps/s 10.1826 KOps/s $\color{#d91a1a}-0.64\%$
test_items_stack_nested_locked 0.5054ms 0.4711ms 2.1225 KOps/s 2.1043 KOps/s $\color{#35bf28}+0.87\%$
test_keys 53.2910μs 4.2032μs 237.9131 KOps/s 227.0156 KOps/s $\color{#35bf28}+4.80\%$
test_keys_nested 0.1824ms 0.1301ms 7.6873 KOps/s 7.6316 KOps/s $\color{#35bf28}+0.73\%$
test_keys_nested_locked 0.7905ms 0.1383ms 7.2320 KOps/s 7.1544 KOps/s $\color{#35bf28}+1.08\%$
test_keys_nested_leaf 0.1871ms 0.1198ms 8.3490 KOps/s 8.2489 KOps/s $\color{#35bf28}+1.21\%$
test_keys_stack_nested 0.2020ms 0.1295ms 7.7236 KOps/s 7.5892 KOps/s $\color{#35bf28}+1.77\%$
test_keys_stack_nested_leaf 0.1719ms 0.1202ms 8.3165 KOps/s 8.2230 KOps/s $\color{#35bf28}+1.14\%$
test_keys_stack_nested_locked 0.1844ms 0.1368ms 7.3080 KOps/s 7.2305 KOps/s $\color{#35bf28}+1.07\%$
test_values 15.7542μs 1.0163μs 983.9416 KOps/s 983.4529 KOps/s $\color{#35bf28}+0.05\%$
test_values_nested 83.2810μs 52.3497μs 19.1023 KOps/s 18.8465 KOps/s $\color{#35bf28}+1.36\%$
test_values_nested_locked 88.9210μs 55.1967μs 18.1170 KOps/s 18.0412 KOps/s $\color{#35bf28}+0.42\%$
test_values_nested_leaf 86.0810μs 60.3267μs 16.5764 KOps/s 16.2905 KOps/s $\color{#35bf28}+1.76\%$
test_values_stack_nested 82.7310μs 52.3592μs 19.0988 KOps/s 18.6989 KOps/s $\color{#35bf28}+2.14\%$
test_values_stack_nested_leaf 84.0410μs 60.2943μs 16.5853 KOps/s 16.3536 KOps/s $\color{#35bf28}+1.42\%$
test_values_stack_nested_locked 87.5310μs 55.5239μs 18.0103 KOps/s 17.5752 KOps/s $\color{#35bf28}+2.48\%$
test_membership 5.0133μs 0.8533μs 1.1719 MOps/s 1.1587 MOps/s $\color{#35bf28}+1.14\%$
test_membership_nested 20.9810μs 2.9490μs 339.1021 KOps/s 342.8521 KOps/s $\color{#d91a1a}-1.09\%$
test_membership_nested_leaf 36.8100μs 2.9355μs 340.6617 KOps/s 339.5254 KOps/s $\color{#35bf28}+0.33\%$
test_membership_stacked_nested 22.9700μs 2.9729μs 336.3707 KOps/s 340.4422 KOps/s $\color{#d91a1a}-1.20\%$
test_membership_stacked_nested_leaf 36.1200μs 2.9454μs 339.5136 KOps/s 341.5827 KOps/s $\color{#d91a1a}-0.61\%$
test_membership_nested_last 40.2700μs 4.4448μs 224.9837 KOps/s 226.6329 KOps/s $\color{#d91a1a}-0.73\%$
test_membership_nested_leaf_last 33.6510μs 4.4191μs 226.2891 KOps/s 226.1309 KOps/s $\color{#35bf28}+0.07\%$
test_membership_stacked_nested_last 26.2810μs 4.3965μs 227.4550 KOps/s 227.9241 KOps/s $\color{#d91a1a}-0.21\%$
test_membership_stacked_nested_leaf_last 34.2700μs 4.4243μs 226.0241 KOps/s 225.3349 KOps/s $\color{#35bf28}+0.31\%$
test_nested_getleaf 56.2410μs 21.5365μs 46.4328 KOps/s 45.5524 KOps/s $\color{#35bf28}+1.93\%$
test_nested_get 48.3810μs 20.0506μs 49.8738 KOps/s 48.2346 KOps/s $\color{#35bf28}+3.40\%$
test_stacked_getleaf 49.1310μs 21.4388μs 46.6444 KOps/s 45.6897 KOps/s $\color{#35bf28}+2.09\%$
test_stacked_get 47.0310μs 19.9771μs 50.0573 KOps/s 48.6882 KOps/s $\color{#35bf28}+2.81\%$
test_nested_getitemleaf 43.9000μs 21.4827μs 46.5491 KOps/s 45.0811 KOps/s $\color{#35bf28}+3.26\%$
test_nested_getitem 50.3410μs 20.3133μs 49.2289 KOps/s 47.4261 KOps/s $\color{#35bf28}+3.80\%$
test_stacked_getitemleaf 51.6710μs 21.4921μs 46.5288 KOps/s 44.9811 KOps/s $\color{#35bf28}+3.44\%$
test_stacked_getitem 54.4300μs 20.3191μs 49.2147 KOps/s 47.5772 KOps/s $\color{#35bf28}+3.44\%$
test_lock_nested 8.1448ms 0.4871ms 2.0529 KOps/s 2.0786 KOps/s $\color{#d91a1a}-1.24\%$
test_lock_stack_nested 0.5465ms 0.4784ms 2.0901 KOps/s 2.0449 KOps/s $\color{#35bf28}+2.21\%$
test_unlock_nested 0.4623ms 0.3902ms 2.5626 KOps/s 2.5611 KOps/s $\color{#35bf28}+0.06\%$
test_unlock_stack_nested 0.4356ms 0.3865ms 2.5874 KOps/s 2.5380 KOps/s $\color{#35bf28}+1.95\%$
test_flatten_speed 0.1697ms 0.1230ms 8.1280 KOps/s 8.0417 KOps/s $\color{#35bf28}+1.07\%$
test_unflatten_speed 0.6650ms 0.5653ms 1.7691 KOps/s 1.7293 KOps/s $\color{#35bf28}+2.30\%$
test_common_ops 0.8522ms 0.7010ms 1.4264 KOps/s 1.4218 KOps/s $\color{#35bf28}+0.32\%$
test_creation 0.1268ms 3.1443μs 318.0338 KOps/s 315.7167 KOps/s $\color{#35bf28}+0.73\%$
test_creation_empty 28.4600μs 6.9808μs 143.2494 KOps/s 142.5427 KOps/s $\color{#35bf28}+0.50\%$
test_creation_nested_1 44.1810μs 11.5135μs 86.8547 KOps/s 86.7963 KOps/s $\color{#35bf28}+0.07\%$
test_creation_nested_2 46.0410μs 13.2377μs 75.5416 KOps/s 74.6080 KOps/s $\color{#35bf28}+1.25\%$
test_creation_many_keys[10] 54.3710μs 20.9984μs 47.6227 KOps/s 47.3797 KOps/s $\color{#35bf28}+0.51\%$
test_creation_many_keys[50] 0.1292ms 91.3063μs 10.9521 KOps/s 11.0223 KOps/s $\color{#d91a1a}-0.64\%$
test_creation_many_keys[100] 0.2302ms 0.1784ms 5.6064 KOps/s 5.6075 KOps/s $\color{#d91a1a}-0.02\%$
test_creation_nested_many_keys[10] 73.9710μs 45.2201μs 22.1141 KOps/s 22.0872 KOps/s $\color{#35bf28}+0.12\%$
test_creation_nested_many_keys[50] 0.2637ms 0.1862ms 5.3706 KOps/s 5.4040 KOps/s $\color{#d91a1a}-0.62\%$
test_clone 56.0710μs 13.2648μs 75.3872 KOps/s 75.0128 KOps/s $\color{#35bf28}+0.50\%$
test_getitem[int] 1.6415ms 15.2853μs 65.4221 KOps/s 58.9221 KOps/s $\textbf{\color{#35bf28}+11.03\%}$
test_getitem[slice_int] 0.1352ms 24.3530μs 41.0626 KOps/s 40.5856 KOps/s $\color{#35bf28}+1.18\%$
test_getitem[range] 0.1903ms 64.0826μs 15.6049 KOps/s 15.6231 KOps/s $\color{#d91a1a}-0.12\%$
test_getitem[tuple] 0.1430ms 23.9769μs 41.7068 KOps/s 41.1089 KOps/s $\color{#35bf28}+1.45\%$
test_getitem[list] 0.1812ms 59.0654μs 16.9304 KOps/s 17.0402 KOps/s $\color{#d91a1a}-0.64\%$
test_setitem_dim[int] 44.7910μs 25.2374μs 39.6237 KOps/s 38.2822 KOps/s $\color{#35bf28}+3.50\%$
test_setitem_dim[slice_int] 67.9610μs 44.1887μs 22.6302 KOps/s 22.6890 KOps/s $\color{#d91a1a}-0.26\%$
test_setitem_dim[range] 0.1197ms 95.8392μs 10.4341 KOps/s 10.6378 KOps/s $\color{#d91a1a}-1.91\%$
test_setitem_dim[tuple] 80.5710μs 40.6545μs 24.5975 KOps/s 25.1967 KOps/s $\color{#d91a1a}-2.38\%$
test_setitem 57.5100μs 18.0838μs 55.2980 KOps/s 54.9544 KOps/s $\color{#35bf28}+0.63\%$
test_set 59.4310μs 17.2344μs 58.0233 KOps/s 58.2144 KOps/s $\color{#d91a1a}-0.33\%$
test_set_shared 0.5275ms 0.2078ms 4.8128 KOps/s 4.9501 KOps/s $\color{#d91a1a}-2.77\%$
test_update 0.3362ms 22.1101μs 45.2282 KOps/s 45.5044 KOps/s $\color{#d91a1a}-0.61\%$
test_update_nested 75.7310μs 33.2373μs 30.0867 KOps/s 30.6226 KOps/s $\color{#d91a1a}-1.75\%$
test_update__nested 0.5331ms 34.3216μs 29.1361 KOps/s 28.6145 KOps/s $\color{#35bf28}+1.82\%$
test_set_nested 49.3610μs 19.4058μs 51.5309 KOps/s 51.6158 KOps/s $\color{#d91a1a}-0.16\%$
test_set_nested_new 53.1410μs 23.8187μs 41.9839 KOps/s 41.7686 KOps/s $\color{#35bf28}+0.52\%$
test_select 81.4910μs 40.2866μs 24.8221 KOps/s 24.5885 KOps/s $\color{#35bf28}+0.95\%$
test_select_nested 0.1044ms 73.9680μs 13.5194 KOps/s 13.4616 KOps/s $\color{#35bf28}+0.43\%$
test_exclude_nested 0.1262ms 91.6881μs 10.9065 KOps/s 10.7930 KOps/s $\color{#35bf28}+1.05\%$
test_empty[True] 0.4340ms 0.3986ms 2.5090 KOps/s 2.5069 KOps/s $\color{#35bf28}+0.08\%$
test_empty[False] 8.2175μs 1.3073μs 764.9230 KOps/s 762.9016 KOps/s $\color{#35bf28}+0.26\%$
test_to 0.1105ms 73.4271μs 13.6189 KOps/s 13.4474 KOps/s $\color{#35bf28}+1.28\%$
test_to_nonblocking 0.1779ms 65.4881μs 15.2700 KOps/s 14.6823 KOps/s $\color{#35bf28}+4.00\%$
test_unbind_speed 0.3686ms 0.3302ms 3.0286 KOps/s 2.9823 KOps/s $\color{#35bf28}+1.55\%$
test_unbind_speed_stack0 0.3730ms 0.3322ms 3.0103 KOps/s 2.9881 KOps/s $\color{#35bf28}+0.74\%$
test_unbind_speed_stack1 0.1038s 0.8362ms 1.1959 KOps/s 1.1691 KOps/s $\color{#35bf28}+2.29\%$
test_split 0.1054s 1.2685ms 788.3069 Ops/s 775.3816 Ops/s $\color{#35bf28}+1.67\%$
test_chunk 0.1037s 1.2088ms 827.2999 Ops/s 921.8374 Ops/s $\textbf{\color{#d91a1a}-10.26\%}$
test_to_cpu_blocking 19.8797ms 19.7116ms 50.7316 Ops/s 50.1914 Ops/s $\color{#35bf28}+1.08\%$
test_to_cpu_global_sync 11.4894ms 11.3829ms 87.8507 Ops/s 77.2766 Ops/s $\textbf{\color{#35bf28}+13.68\%}$
test_to_cpu_event_sync 12.6800ms 12.3399ms 81.0377 Ops/s 79.5454 Ops/s $\color{#35bf28}+1.88\%$
test_to_cpu_default 0.1158s 13.6650ms 73.1798 Ops/s 79.1496 Ops/s $\textbf{\color{#d91a1a}-7.54\%}$
test_consolidate[False-None] 4.3436ms 4.1816ms 239.1435 Ops/s 212.6772 Ops/s $\textbf{\color{#35bf28}+12.44\%}$
test_consolidate[default-None] 2.1940ms 2.0797ms 480.8442 Ops/s 467.7396 Ops/s $\color{#35bf28}+2.80\%$
test_consolidate[reduce-overhead-None] 2.6297ms 2.0146ms 496.3733 Ops/s 492.8317 Ops/s $\color{#35bf28}+0.72\%$
test_consolidate_njt[False-None] 0.1982s 10.3158ms 96.9390 Ops/s 115.1302 Ops/s $\textbf{\color{#d91a1a}-15.80\%}$
test_to[False-False-None] 2.3188ms 2.1414ms 466.9937 Ops/s 462.1942 Ops/s $\color{#35bf28}+1.04\%$
test_to[True-False-None] 2.1505ms 1.9366ms 516.3683 Ops/s 508.5795 Ops/s $\color{#35bf28}+1.53\%$
test_to[within-False-None] 6.3686ms 6.2135ms 160.9395 Ops/s 160.5910 Ops/s $\color{#35bf28}+0.22\%$
test_to[True-default-None] 9.3160ms 8.7995ms 113.6433 Ops/s 110.4077 Ops/s $\color{#35bf28}+2.93\%$
test_to_njt[False-False-None] 8.9027ms 8.5454ms 117.0214 Ops/s 116.7409 Ops/s $\color{#35bf28}+0.24\%$
test_to_njt[True-False-None] 7.4650ms 7.0734ms 141.3750 Ops/s 143.0060 Ops/s $\color{#d91a1a}-1.14\%$
test_to_njt[within-False-None] 16.5992ms 15.8038ms 63.2759 Ops/s 63.0818 Ops/s $\color{#35bf28}+0.31\%$
test_creation[device0] 0.4660ms 0.1156ms 8.6471 KOps/s 8.7607 KOps/s $\color{#d91a1a}-1.30\%$
test_creation_from_tensor 0.4115ms 0.1150ms 8.6933 KOps/s 8.9448 KOps/s $\color{#d91a1a}-2.81\%$
test_add_one[memmap_tensor0] 0.2595ms 6.6277μs 150.8809 KOps/s 151.4702 KOps/s $\color{#d91a1a}-0.39\%$
test_contiguous[memmap_tensor0] 34.7410μs 0.6806μs 1.4692 MOps/s 2.2086 MOps/s $\textbf{\color{#d91a1a}-33.48\%}$
test_stack[memmap_tensor0] 29.0210μs 4.8050μs 208.1154 KOps/s 212.3690 KOps/s $\color{#d91a1a}-2.00\%$
test_memmaptd_index 1.0433ms 0.2691ms 3.7162 KOps/s 3.6916 KOps/s $\color{#35bf28}+0.67\%$
test_memmaptd_index_astensor 0.5311ms 0.3706ms 2.6985 KOps/s 2.6694 KOps/s $\color{#35bf28}+1.09\%$
test_memmaptd_index_op 0.1738s 0.7513ms 1.3311 KOps/s 1.5804 KOps/s $\textbf{\color{#d91a1a}-15.77\%}$
test_serialize_model 0.1384s 0.1361s 7.3463 Ops/s 7.4517 Ops/s $\color{#d91a1a}-1.41\%$
test_serialize_model_pickle 1.3502s 1.2117s 0.8253 Ops/s 0.8267 Ops/s $\color{#d91a1a}-0.17\%$
test_serialize_weights 0.1373s 0.1345s 7.4325 Ops/s 7.4203 Ops/s $\color{#35bf28}+0.17\%$
test_serialize_weights_returnearly 0.4338s 87.7285ms 11.3988 Ops/s 11.4135 Ops/s $\color{#d91a1a}-0.13\%$
test_serialize_weights_pickle 1.3710s 1.1976s 0.8350 Ops/s 0.8107 Ops/s $\color{#35bf28}+3.00\%$
test_reshape_pytree 0.2061ms 33.0008μs 30.3023 KOps/s 30.2641 KOps/s $\color{#35bf28}+0.13\%$
test_reshape_td 72.9910μs 47.2364μs 21.1701 KOps/s 20.5747 KOps/s $\color{#35bf28}+2.89\%$
test_view_pytree 0.2134ms 32.6117μs 30.6639 KOps/s 30.1915 KOps/s $\color{#35bf28}+1.56\%$
test_view_td 83.9410μs 53.7847μs 18.5926 KOps/s 18.3182 KOps/s $\color{#35bf28}+1.50\%$
test_unbind_pytree 0.2480ms 35.9663μs 27.8038 KOps/s 27.5057 KOps/s $\color{#35bf28}+1.08\%$
test_unbind_td 0.1619ms 49.9496μs 20.0202 KOps/s 19.6621 KOps/s $\color{#35bf28}+1.82\%$
test_split_pytree 0.1819ms 42.1058μs 23.7497 KOps/s 23.0937 KOps/s $\color{#35bf28}+2.84\%$
test_split_td 0.1683ms 65.4325μs 15.2829 KOps/s 15.1776 KOps/s $\color{#35bf28}+0.69\%$
test_add_pytree 0.1875ms 41.8425μs 23.8992 KOps/s 23.7698 KOps/s $\color{#35bf28}+0.54\%$
test_add_td 92.3910μs 55.3359μs 18.0715 KOps/s 17.8272 KOps/s $\color{#35bf28}+1.37\%$
test_compile_add_one_nested[tensordict-compile] 0.1971ms 0.1408ms 7.1014 KOps/s 6.8675 KOps/s $\color{#35bf28}+3.40\%$
test_compile_add_one_nested[tensordict-eager] 0.3427ms 0.2030ms 4.9250 KOps/s 4.9817 KOps/s $\color{#d91a1a}-1.14\%$
test_compile_add_one_nested[pytree-compile] 0.1643ms 0.1084ms 9.2288 KOps/s 8.4448 KOps/s $\textbf{\color{#35bf28}+9.28\%}$
test_compile_add_one_nested[pytree-eager] 0.4304ms 0.1823ms 5.4851 KOps/s 5.4889 KOps/s $\color{#d91a1a}-0.07\%$
test_compile_copy_nested[tensordict-compile] 0.3347ms 11.3346μs 88.2252 KOps/s 95.0912 KOps/s $\textbf{\color{#d91a1a}-7.22\%}$
test_compile_copy_nested[tensordict-eager] 82.3810μs 54.7268μs 18.2726 KOps/s 18.6051 KOps/s $\color{#d91a1a}-1.79\%$
test_compile_copy_nested[pytree-compile] 0.1566ms 9.9553μs 100.4494 KOps/s 101.5541 KOps/s $\color{#d91a1a}-1.09\%$
test_compile_copy_nested[pytree-eager] 0.4327ms 68.3152μs 14.6380 KOps/s 14.7379 KOps/s $\color{#d91a1a}-0.68\%$
test_compile_add_one_flat[tensordict-compile] 0.2786ms 0.1761ms 5.6774 KOps/s 3.6353 KOps/s $\textbf{\color{#35bf28}+56.17\%}$
test_compile_add_one_flat[tensordict-eager] 0.4201ms 0.2789ms 3.5851 KOps/s 3.4765 KOps/s $\color{#35bf28}+3.12\%$
test_compile_add_one_flat[tensorclass-compile] 0.2704ms 0.1164ms 8.5880 KOps/s 8.0168 KOps/s $\textbf{\color{#35bf28}+7.12\%}$
test_compile_add_one_flat[tensorclass-eager] 0.2107ms 72.4013μs 13.8119 KOps/s 13.2647 KOps/s $\color{#35bf28}+4.13\%$
test_compile_add_one_flat[pytree-compile] 0.1965ms 0.1574ms 6.3521 KOps/s 6.0362 KOps/s $\textbf{\color{#35bf28}+5.23\%}$
test_compile_add_one_flat[pytree-eager] 0.7942ms 0.5383ms 1.8575 KOps/s 1.8435 KOps/s $\color{#35bf28}+0.76\%$
test_compile_add_self_flat[tensordict-eager] 0.4621ms 0.3333ms 3.0001 KOps/s 2.9233 KOps/s $\color{#35bf28}+2.63\%$
test_compile_add_self_flat[tensordict-compile] 0.2370ms 0.1785ms 5.6018 KOps/s 5.1300 KOps/s $\textbf{\color{#35bf28}+9.20\%}$
test_compile_add_self_flat[tensorclass-eager] 0.1415ms 89.1280μs 11.2198 KOps/s 10.8008 KOps/s $\color{#35bf28}+3.88\%$
test_compile_add_self_flat[tensorclass-compile] 0.1709ms 0.1191ms 8.3943 KOps/s 8.0111 KOps/s $\color{#35bf28}+4.78\%$
test_compile_add_self_flat[pytree-eager] 0.6507ms 0.4401ms 2.2723 KOps/s 2.2297 KOps/s $\color{#35bf28}+1.91\%$
test_compile_add_self_flat[pytree-compile] 0.2014ms 0.1571ms 6.3641 KOps/s 5.9168 KOps/s $\textbf{\color{#35bf28}+7.56\%}$
test_compile_copy_flat[tensordict-compile] 0.1231ms 13.0984μs 76.3453 KOps/s 73.3938 KOps/s $\color{#35bf28}+4.02\%$
test_compile_copy_flat[tensordict-eager] 77.5910μs 42.1770μs 23.7096 KOps/s 23.7264 KOps/s $\color{#d91a1a}-0.07\%$
test_compile_copy_flat[pytree-compile] 60.5910μs 10.8001μs 92.5918 KOps/s 90.4019 KOps/s $\color{#35bf28}+2.42\%$
test_compile_copy_flat[pytree-eager] 0.4108ms 52.3365μs 19.1071 KOps/s 18.9136 KOps/s $\color{#35bf28}+1.02\%$
test_compile_assign_and_add[tensordict-compile] 2.0312ms 0.1741ms 5.7425 KOps/s 5.3981 KOps/s $\textbf{\color{#35bf28}+6.38\%}$
test_compile_assign_and_add[tensordict-eager] 3.4852ms 3.3190ms 301.2943 Ops/s 289.0529 Ops/s $\color{#35bf28}+4.23\%$
test_compile_assign_and_add[pytree-compile] 1.9883ms 0.1615ms 6.1911 KOps/s 6.0950 KOps/s $\color{#35bf28}+1.58\%$
test_compile_assign_and_add[pytree-eager] 2.9663ms 2.8260ms 353.8582 Ops/s 349.9667 Ops/s $\color{#35bf28}+1.11\%$
test_compile_indexing[tensor-tensordict-compile] 0.2032ms 0.1089ms 9.1786 KOps/s 8.7162 KOps/s $\textbf{\color{#35bf28}+5.31\%}$
test_compile_indexing[tensor-tensordict-eager] 0.3158ms 76.2380μs 13.1168 KOps/s 13.5391 KOps/s $\color{#d91a1a}-3.12\%$
test_compile_indexing[tensor-tensorclass-compile] 0.1427ms 95.7544μs 10.4434 KOps/s 9.9878 KOps/s $\color{#35bf28}+4.56\%$
test_compile_indexing[tensor-tensorclass-eager] 0.2501ms 44.7072μs 22.3678 KOps/s 21.0339 KOps/s $\textbf{\color{#35bf28}+6.34\%}$
test_compile_indexing[tensor-pytree-compile] 0.2460ms 96.0774μs 10.4083 KOps/s 9.5602 KOps/s $\textbf{\color{#35bf28}+8.87\%}$
test_compile_indexing[tensor-pytree-eager] 0.2460ms 44.0970μs 22.6773 KOps/s 20.6533 KOps/s $\textbf{\color{#35bf28}+9.80\%}$
test_compile_indexing[slice-tensordict-compile] 0.1592ms 58.3640μs 17.1338 KOps/s 16.2935 KOps/s $\textbf{\color{#35bf28}+5.16\%}$
test_compile_indexing[slice-tensordict-eager] 0.2291ms 27.6300μs 36.1926 KOps/s 33.9929 KOps/s $\textbf{\color{#35bf28}+6.47\%}$
test_compile_indexing[slice-tensorclass-compile] 0.1534ms 45.2536μs 22.0977 KOps/s 20.4230 KOps/s $\textbf{\color{#35bf28}+8.20\%}$
test_compile_indexing[slice-tensorclass-eager] 0.2594ms 22.5521μs 44.3417 KOps/s 43.8000 KOps/s $\color{#35bf28}+1.24\%$
test_compile_indexing[slice-pytree-compile] 90.7010μs 45.5592μs 21.9495 KOps/s 20.1845 KOps/s $\textbf{\color{#35bf28}+8.74\%}$
test_compile_indexing[slice-pytree-eager] 0.2777ms 22.5135μs 44.4178 KOps/s 43.8191 KOps/s $\color{#35bf28}+1.37\%$
test_compile_indexing[int-tensordict-compile] 95.5610μs 57.9195μs 17.2653 KOps/s 16.6626 KOps/s $\color{#35bf28}+3.62\%$
test_compile_indexing[int-tensordict-eager] 0.2515ms 27.7165μs 36.0796 KOps/s 33.9821 KOps/s $\textbf{\color{#35bf28}+6.17\%}$
test_compile_indexing[int-tensorclass-compile] 84.1510μs 45.4259μs 22.0139 KOps/s 22.0935 KOps/s $\color{#d91a1a}-0.36\%$
test_compile_indexing[int-tensorclass-eager] 0.2547ms 22.2567μs 44.9303 KOps/s 44.3944 KOps/s $\color{#35bf28}+1.21\%$
test_compile_indexing[int-pytree-compile] 79.0510μs 45.1221μs 22.1621 KOps/s 20.1238 KOps/s $\textbf{\color{#35bf28}+10.13\%}$
test_compile_indexing[int-pytree-eager] 0.2575ms 22.2879μs 44.8674 KOps/s 43.9936 KOps/s $\color{#35bf28}+1.99\%$
test_compile_replace[single-eager] 0.1999ms 48.0939μs 20.7927 KOps/s 21.0616 KOps/s $\color{#d91a1a}-1.28\%$
test_compile_replace[single-compile] 0.2418ms 0.1051ms 9.5143 KOps/s 9.3503 KOps/s $\color{#35bf28}+1.75\%$
test_compile_replace[multi-eager] 0.6360ms 0.5721ms 1.7480 KOps/s 1.7697 KOps/s $\color{#d91a1a}-1.23\%$
test_compile_replace[multi-compile] 0.1616ms 0.1110ms 9.0120 KOps/s 8.7822 KOps/s $\color{#35bf28}+2.62\%$
test_compile_tc_getattr_20[eager] 0.2164ms 0.1720ms 5.8135 KOps/s 5.6798 KOps/s $\color{#35bf28}+2.36\%$
test_compile_tc_getattr_20[compile] 0.2560ms 0.1190ms 8.4056 KOps/s 8.2458 KOps/s $\color{#35bf28}+1.94\%$
test_compile_clone_shallow[20-eager] 54.5800μs 19.4429μs 51.4327 KOps/s 51.8646 KOps/s $\color{#d91a1a}-0.83\%$
test_compile_clone_shallow[20-compile] 52.5400μs 11.1569μs 89.6305 KOps/s 87.0219 KOps/s $\color{#35bf28}+3.00\%$
test_compile_clone_shallow[40-eager] 65.8100μs 34.0773μs 29.3450 KOps/s 29.5782 KOps/s $\color{#d91a1a}-0.79\%$
test_compile_clone_shallow[40-compile] 72.6310μs 12.6534μs 79.0304 KOps/s 76.5346 KOps/s $\color{#35bf28}+3.26\%$
test_compile_clone_shallow[80-eager] 91.1910μs 63.6185μs 15.7187 KOps/s 16.0940 KOps/s $\color{#d91a1a}-2.33\%$
test_compile_clone_shallow[80-compile] 65.8510μs 14.9585μs 66.8518 KOps/s 63.7286 KOps/s $\color{#35bf28}+4.90\%$
test_compile_update_inplace[eager] 95.8020μs 59.2275μs 16.8840 KOps/s 15.8562 KOps/s $\textbf{\color{#35bf28}+6.48\%}$
test_compile_update_inplace[compile] 0.3255ms 0.1397ms 7.1591 KOps/s 6.9841 KOps/s $\color{#35bf28}+2.51\%$
test_mod_add[eager] 92.5710μs 50.6309μs 19.7508 KOps/s 19.9134 KOps/s $\color{#d91a1a}-0.82\%$
test_mod_add[compile] 0.5108ms 0.1038ms 9.6353 KOps/s 9.2274 KOps/s $\color{#35bf28}+4.42\%$
test_mod_add[compile-overhead] 0.2326ms 0.1489ms 6.7160 KOps/s 6.5032 KOps/s $\color{#35bf28}+3.27\%$
test_mod_wrap[eager] 0.3794ms 0.2894ms 3.4555 KOps/s 3.3345 KOps/s $\color{#35bf28}+3.63\%$
test_mod_wrap[compile] 0.5007ms 0.3576ms 2.7960 KOps/s 2.8313 KOps/s $\color{#d91a1a}-1.25\%$
test_mod_wrap[compile-overhead] 6.9263ms 3.8701ms 258.3928 Ops/s 248.4914 Ops/s $\color{#35bf28}+3.98\%$
test_mod_wrap_and_backward[eager] 1.6336ms 1.5027ms 665.4757 Ops/s 622.7963 Ops/s $\textbf{\color{#35bf28}+6.85\%}$
test_mod_wrap_and_backward[compile] 1.6561ms 1.4764ms 677.3268 Ops/s 630.8154 Ops/s $\textbf{\color{#35bf28}+7.37\%}$
test_mod_wrap_and_backward[compile-overhead] 1.2435ms 0.8938ms 1.1188 KOps/s 982.9996 Ops/s $\textbf{\color{#35bf28}+13.82\%}$
test_seq_add[eager] 0.2171ms 0.1594ms 6.2722 KOps/s 6.4929 KOps/s $\color{#d91a1a}-3.40\%$
test_seq_add[compile] 0.1964ms 0.1208ms 8.2798 KOps/s 8.4904 KOps/s $\color{#d91a1a}-2.48\%$
test_seq_add[compile-overhead] 0.2207ms 0.1538ms 6.5015 KOps/s 6.2186 KOps/s $\color{#35bf28}+4.55\%$
test_seq_wrap[eager] 0.5865ms 0.5189ms 1.9272 KOps/s 1.9229 KOps/s $\color{#35bf28}+0.23\%$
test_seq_wrap[compile] 0.4871ms 0.3665ms 2.7282 KOps/s 2.6937 KOps/s $\color{#35bf28}+1.28\%$
test_seq_wrap[compile-overhead] 0.3724ms 0.2817ms 3.5499 KOps/s 3.7123 KOps/s $\color{#d91a1a}-4.37\%$
test_func_call_runtime[False-eager] 0.9423ms 0.8411ms 1.1889 KOps/s 1.1838 KOps/s $\color{#35bf28}+0.43\%$
test_func_call_runtime[False-compile] 1.0955ms 0.9330ms 1.0718 KOps/s 1.0836 KOps/s $\color{#d91a1a}-1.09\%$
test_func_call_runtime[False-compile-overhead] 0.5244ms 0.4657ms 2.1474 KOps/s 2.1279 KOps/s $\color{#35bf28}+0.92\%$
test_func_call_runtime[True-eager] 1.2119ms 1.0876ms 919.4512 Ops/s 925.4516 Ops/s $\color{#d91a1a}-0.65\%$
test_func_call_runtime[True-compile] 1.0523ms 0.9737ms 1.0270 KOps/s 1.0704 KOps/s $\color{#d91a1a}-4.06\%$
test_func_call_runtime[True-compile-overhead] 0.5546ms 0.4808ms 2.0798 KOps/s 2.0658 KOps/s $\color{#35bf28}+0.68\%$
test_func_call_cm_runtime[False-eager] 0.9301ms 0.8314ms 1.2028 KOps/s 1.1897 KOps/s $\color{#35bf28}+1.10\%$
test_func_call_cm_runtime[False-compile] 1.0863ms 0.9732ms 1.0275 KOps/s 1.0794 KOps/s $\color{#d91a1a}-4.81\%$
test_func_call_cm_runtime[False-compile-overhead] 0.6436ms 0.4655ms 2.1484 KOps/s 2.1249 KOps/s $\color{#35bf28}+1.11\%$
test_func_call_cm_runtime[True-eager] 1.3377ms 1.2199ms 819.7350 Ops/s 815.6167 Ops/s $\color{#35bf28}+0.50\%$
test_func_call_cm_runtime[True-compile] 1.0113ms 0.9617ms 1.0398 KOps/s 1.0344 KOps/s $\color{#35bf28}+0.52\%$
test_func_call_cm_runtime[True-compile-overhead] 0.6090ms 0.5093ms 1.9634 KOps/s 1.9365 KOps/s $\color{#35bf28}+1.39\%$
test_vmap_func_call_cm_runtime[eager] 2.8693ms 2.3876ms 418.8317 Ops/s 420.2438 Ops/s $\color{#d91a1a}-0.34\%$
test_vmap_func_call_cm_runtime[compile] 1.1230ms 1.0366ms 964.6468 Ops/s 1.0030 KOps/s $\color{#d91a1a}-3.82\%$
test_vmap_func_call_cm_runtime[compile-overhead] 0.6500ms 0.5286ms 1.8918 KOps/s 1.9159 KOps/s $\color{#d91a1a}-1.26\%$
test_distributed 2.5617ms 0.1629ms 6.1392 KOps/s 6.5611 KOps/s $\textbf{\color{#d91a1a}-6.43\%}$
test_tdmodule 0.4935ms 28.3092μs 35.3242 KOps/s 36.2917 KOps/s $\color{#d91a1a}-2.67\%$
test_tdmodule_dispatch 0.1774ms 47.7851μs 20.9270 KOps/s 21.8728 KOps/s $\color{#d91a1a}-4.32\%$
test_tdseq 48.2110μs 28.8963μs 34.6065 KOps/s 37.3508 KOps/s $\textbf{\color{#d91a1a}-7.35\%}$
test_tdseq_dispatch 72.1610μs 50.7146μs 19.7182 KOps/s 20.7971 KOps/s $\textbf{\color{#d91a1a}-5.19\%}$
test_instantiation_functorch 2.2440ms 2.1042ms 475.2290 Ops/s 474.1644 Ops/s $\color{#35bf28}+0.22\%$
test_exec_functorch 0.2684ms 0.1834ms 5.4519 KOps/s 5.5480 KOps/s $\color{#d91a1a}-1.73\%$
test_exec_functional_call 0.2175ms 0.1693ms 5.9059 KOps/s 6.2091 KOps/s $\color{#d91a1a}-4.88\%$
test_exec_td_decorator 0.4600ms 0.2450ms 4.0808 KOps/s 4.2096 KOps/s $\color{#d91a1a}-3.06\%$
test_vmap_mlp_speed_decorator[True-True] 1.3543ms 0.8622ms 1.1598 KOps/s 1.2058 KOps/s $\color{#d91a1a}-3.82\%$
test_vmap_mlp_speed_decorator[True-False] 1.0714ms 0.8612ms 1.1612 KOps/s 1.2073 KOps/s $\color{#d91a1a}-3.82\%$
test_vmap_mlp_speed_decorator[False-True] 0.9225ms 0.7404ms 1.3506 KOps/s 1.4032 KOps/s $\color{#d91a1a}-3.74\%$
test_vmap_mlp_speed_decorator[False-False] 0.9209ms 0.7437ms 1.3446 KOps/s 1.4049 KOps/s $\color{#d91a1a}-4.29\%$
test_vmap_transformer_speed_decorator[True-True] 21.3592ms 20.6251ms 48.4847 Ops/s 48.5821 Ops/s $\color{#d91a1a}-0.20\%$
test_vmap_transformer_speed_decorator[True-False] 21.2033ms 20.5244ms 48.7225 Ops/s 48.6721 Ops/s $\color{#35bf28}+0.10\%$
test_vmap_transformer_speed_decorator[False-True] 21.3212ms 20.4857ms 48.8146 Ops/s 49.1884 Ops/s $\color{#d91a1a}-0.76\%$
test_vmap_transformer_speed_decorator[False-False] 21.4031ms 20.5063ms 48.7656 Ops/s 49.0819 Ops/s $\color{#d91a1a}-0.64\%$
test_to_module_speed[True] 1.6007ms 1.4733ms 678.7674 Ops/s 671.7834 Ops/s $\color{#35bf28}+1.04\%$
test_to_module_speed[False] 1.5606ms 1.4446ms 692.2486 Ops/s 676.6792 Ops/s $\color{#35bf28}+2.30\%$
test_tc_init 69.0510μs 45.4464μs 22.0039 KOps/s 22.1693 KOps/s $\color{#d91a1a}-0.75\%$
test_tc_init_tensor_only 49.8810μs 9.7827μs 102.2216 KOps/s 102.8888 KOps/s $\color{#d91a1a}-0.65\%$
test_tc_init_nested 0.1311ms 93.3763μs 10.7094 KOps/s 11.3012 KOps/s $\textbf{\color{#d91a1a}-5.24\%}$
test_tc_init_many_fields 44.9510μs 16.5235μs 60.5197 KOps/s 60.7302 KOps/s $\color{#d91a1a}-0.35\%$
test_tc_first_layer_tensor 24.6600μs 1.8211μs 549.1065 KOps/s 541.5578 KOps/s $\color{#35bf28}+1.39\%$
test_tc_first_layer_tensor_only 3.4876μs 0.3974μs 2.5161 MOps/s 2.5347 MOps/s $\color{#d91a1a}-0.73\%$
test_tc_first_layer_tensor_set 41.7000μs 3.9268μs 254.6598 KOps/s 255.2723 KOps/s $\color{#d91a1a}-0.24\%$
test_tc_first_layer_tensor_only_set 35.3800μs 3.2964μs 303.3658 KOps/s 307.3726 KOps/s $\color{#d91a1a}-1.30\%$
test_tc_first_layer_nontensor 41.8700μs 6.2816μs 159.1943 KOps/s 150.4863 KOps/s $\textbf{\color{#35bf28}+5.79\%}$
test_tc_second_layer_tensor 42.1000μs 4.4337μs 225.5428 KOps/s 224.1483 KOps/s $\color{#35bf28}+0.62\%$
test_tc_second_layer_nontensor 30.3900μs 8.7772μs 113.9315 KOps/s 108.0982 KOps/s $\textbf{\color{#35bf28}+5.40\%}$
test_unbind 0.2530s 16.9562ms 58.9756 Ops/s 65.8451 Ops/s $\textbf{\color{#d91a1a}-10.43\%}$
test_full_like 17.4938ms 16.7118ms 59.8381 Ops/s 227.7876 Ops/s $\textbf{\color{#d91a1a}-73.73\%}$
test_zeros_like 17.2609ms 16.4951ms 60.6241 Ops/s 228.0322 Ops/s $\textbf{\color{#d91a1a}-73.41\%}$
test_ones_like 16.8505ms 16.5418ms 60.4528 Ops/s 223.7719 Ops/s $\textbf{\color{#d91a1a}-72.98\%}$
test_clone 17.7363ms 17.5058ms 57.1239 Ops/s 143.8456 Ops/s $\textbf{\color{#d91a1a}-60.29\%}$
test_squeeze 0.1180ms 14.4095μs 69.3984 KOps/s 68.6382 KOps/s $\color{#35bf28}+1.11\%$
test_unsqueeze 0.1669ms 0.1172ms 8.5321 KOps/s 8.8585 KOps/s $\color{#d91a1a}-3.68\%$
test_split 0.4003ms 0.1930ms 5.1809 KOps/s 5.3929 KOps/s $\color{#d91a1a}-3.93\%$
test_permute 0.2991ms 0.2176ms 4.5947 KOps/s 4.6720 KOps/s $\color{#d91a1a}-1.65\%$
test_stack 51.2593ms 50.9464ms 19.6285 Ops/s 19.2722 Ops/s $\color{#35bf28}+1.85\%$
test_cat 51.1097ms 50.8387ms 19.6700 Ops/s 19.4813 Ops/s $\color{#35bf28}+0.97\%$
test_sequential_tensordict 0.3065ms 0.2323ms 4.3042 KOps/s 4.5406 KOps/s $\textbf{\color{#d91a1a}-5.21\%}$
test_sequential_graph_module 0.1811ms 0.1213ms 8.2423 KOps/s 8.0679 KOps/s $\color{#35bf28}+2.16\%$
test_nested_tensordict 0.3609ms 0.2994ms 3.3401 KOps/s 3.3587 KOps/s $\color{#d91a1a}-0.55\%$
test_nested_graph_module 0.2036ms 0.1335ms 7.4933 KOps/s 7.3978 KOps/s $\color{#35bf28}+1.29\%$

@github-actions

github-actions Bot commented Mar 10, 2026 •
edited
Loading

Copy link
Copy Markdown
Contributor

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

Total Benchmarks: 261. Improved: $\large\color{#35bf28}26$. Worsened: $\large\color{#d91a1a}5$.

Expand to view detailed results
Name Max Mean Ops Ops on Repo HEAD Change
test_plain_set_nested 41.4210μs 14.7920μs 67.6042 KOps/s 68.2457 KOps/s $\color{#d91a1a}-0.94\%$
test_plain_set_stack_nested 43.9210μs 15.3645μs 65.0852 KOps/s 66.0833 KOps/s $\color{#d91a1a}-1.51\%$
test_plain_set_nested_inplace 42.9510μs 16.6093μs 60.2072 KOps/s 60.1441 KOps/s $\color{#35bf28}+0.10\%$
test_plain_set_stack_nested_inplace 47.9310μs 16.6454μs 60.0768 KOps/s 60.1522 KOps/s $\color{#d91a1a}-0.13\%$
test_items 28.6110μs 5.9738μs 167.3983 KOps/s 166.5742 KOps/s $\color{#35bf28}+0.49\%$
test_items_nested 0.5476ms 0.4649ms 2.1512 KOps/s 2.1351 KOps/s $\color{#35bf28}+0.75\%$
test_items_nested_locked 0.5223ms 0.4677ms 2.1383 KOps/s 2.1166 KOps/s $\color{#35bf28}+1.03\%$
test_items_nested_leaf 0.1346ms 97.1936μs 10.2887 KOps/s 10.1021 KOps/s $\color{#35bf28}+1.85\%$
test_items_stack_nested 0.5260ms 0.4672ms 2.1406 KOps/s 2.1348 KOps/s $\color{#35bf28}+0.28\%$
test_items_stack_nested_leaf 0.1693ms 97.3344μs 10.2739 KOps/s 10.1791 KOps/s $\color{#35bf28}+0.93\%$
test_items_stack_nested_locked 0.5245ms 0.4670ms 2.1413 KOps/s 2.1229 KOps/s $\color{#35bf28}+0.87\%$
test_keys 27.9700μs 4.2142μs 237.2914 KOps/s 236.6504 KOps/s $\color{#35bf28}+0.27\%$
test_keys_nested 0.1771ms 0.1287ms 7.7672 KOps/s 7.6152 KOps/s $\color{#35bf28}+2.00\%$
test_keys_nested_locked 0.7488ms 0.1352ms 7.3950 KOps/s 7.1490 KOps/s $\color{#35bf28}+3.44\%$
test_keys_nested_leaf 0.1785ms 0.1205ms 8.3018 KOps/s 8.2413 KOps/s $\color{#35bf28}+0.73\%$
test_keys_stack_nested 0.1747ms 0.1297ms 7.7109 KOps/s 7.6185 KOps/s $\color{#35bf28}+1.21\%$
test_keys_stack_nested_leaf 0.1740ms 0.1198ms 8.3447 KOps/s 8.2107 KOps/s $\color{#35bf28}+1.63\%$
test_keys_stack_nested_locked 0.2023ms 0.1365ms 7.3271 KOps/s 7.1752 KOps/s $\color{#35bf28}+2.12\%$
test_values 4.4980μs 1.0188μs 981.5402 KOps/s 989.6168 KOps/s $\color{#d91a1a}-0.82\%$
test_values_nested 76.6310μs 52.7142μs 18.9702 KOps/s 18.9842 KOps/s $\color{#d91a1a}-0.07\%$
test_values_nested_locked 92.9520μs 56.0917μs 17.8279 KOps/s 17.7769 KOps/s $\color{#35bf28}+0.29\%$
test_values_nested_leaf 97.9220μs 60.3464μs 16.5710 KOps/s 16.5497 KOps/s $\color{#35bf28}+0.13\%$
test_values_stack_nested 86.1520μs 52.9025μs 18.9027 KOps/s 18.8418 KOps/s $\color{#35bf28}+0.32\%$
test_values_stack_nested_leaf 0.1001ms 60.7062μs 16.4728 KOps/s 16.3750 KOps/s $\color{#35bf28}+0.60\%$
test_values_stack_nested_locked 84.3420μs 56.7229μs 17.6296 KOps/s 17.5435 KOps/s $\color{#35bf28}+0.49\%$
test_membership 6.2535μs 0.8414μs 1.1884 MOps/s 1.1731 MOps/s $\color{#35bf28}+1.30\%$
test_membership_nested 42.5910μs 2.9047μs 344.2705 KOps/s 344.1986 KOps/s $\color{#35bf28}+0.02\%$
test_membership_nested_leaf 23.6655μs 2.8403μs 352.0802 KOps/s 344.0739 KOps/s $\color{#35bf28}+2.33\%$
test_membership_stacked_nested 27.7710μs 2.9114μs 343.4817 KOps/s 340.8665 KOps/s $\color{#35bf28}+0.77\%$
test_membership_stacked_nested_leaf 29.0700μs 2.9101μs 343.6277 KOps/s 343.7837 KOps/s $\color{#d91a1a}-0.05\%$
test_membership_nested_last 37.2810μs 4.3794μs 228.3395 KOps/s 229.0503 KOps/s $\color{#d91a1a}-0.31\%$
test_membership_nested_leaf_last 27.9600μs 4.4100μs 226.7595 KOps/s 229.0495 KOps/s $\color{#d91a1a}-1.00\%$
test_membership_stacked_nested_last 28.0010μs 4.3981μs 227.3730 KOps/s 228.1836 KOps/s $\color{#d91a1a}-0.36\%$
test_membership_stacked_nested_leaf_last 23.6200μs 4.3801μs 228.3050 KOps/s 231.3253 KOps/s $\color{#d91a1a}-1.31\%$
test_nested_getleaf 64.3420μs 21.4121μs 46.7026 KOps/s 45.7537 KOps/s $\color{#35bf28}+2.07\%$
test_nested_get 46.4120μs 20.2269μs 49.4392 KOps/s 48.4438 KOps/s $\color{#35bf28}+2.05\%$
test_stacked_getleaf 47.8410μs 21.1888μs 47.1947 KOps/s 46.2824 KOps/s $\color{#35bf28}+1.97\%$
test_stacked_get 47.5710μs 19.9530μs 50.1177 KOps/s 48.9145 KOps/s $\color{#35bf28}+2.46\%$
test_nested_getitemleaf 44.9910μs 22.1257μs 45.1963 KOps/s 44.9963 KOps/s $\color{#35bf28}+0.44\%$
test_nested_getitem 47.4420μs 20.9971μs 47.6257 KOps/s 47.7165 KOps/s $\color{#d91a1a}-0.19\%$
test_stacked_getitemleaf 44.9910μs 21.7465μs 45.9844 KOps/s 45.3193 KOps/s $\color{#35bf28}+1.47\%$
test_stacked_getitem 49.9710μs 20.4765μs 48.8365 KOps/s 47.5727 KOps/s $\color{#35bf28}+2.66\%$
test_lock_nested 8.0096ms 0.4856ms 2.0595 KOps/s 2.1027 KOps/s $\color{#d91a1a}-2.06\%$
test_lock_stack_nested 0.5079ms 0.4794ms 2.0860 KOps/s 2.0671 KOps/s $\color{#35bf28}+0.91\%$
test_unlock_nested 0.4628ms 0.3877ms 2.5791 KOps/s 2.5818 KOps/s $\color{#d91a1a}-0.10\%$
test_unlock_stack_nested 0.4133ms 0.3881ms 2.5768 KOps/s 2.5524 KOps/s $\color{#35bf28}+0.96\%$
test_flatten_speed 0.2116ms 0.1229ms 8.1362 KOps/s 8.1266 KOps/s $\color{#35bf28}+0.12\%$
test_unflatten_speed 0.6403ms 0.5730ms 1.7451 KOps/s 1.7533 KOps/s $\color{#d91a1a}-0.47\%$
test_common_ops 0.8198ms 0.6887ms 1.4520 KOps/s 1.4120 KOps/s $\color{#35bf28}+2.83\%$
test_creation 0.1134ms 3.1628μs 316.1788 KOps/s 319.3746 KOps/s $\color{#d91a1a}-1.00\%$
test_creation_empty 37.6400μs 6.9891μs 143.0807 KOps/s 142.5106 KOps/s $\color{#35bf28}+0.40\%$
test_creation_nested_1 44.3210μs 11.4866μs 87.0579 KOps/s 86.2447 KOps/s $\color{#35bf28}+0.94\%$
test_creation_nested_2 40.5910μs 13.3227μs 75.0597 KOps/s 75.6237 KOps/s $\color{#d91a1a}-0.75\%$
test_creation_many_keys[10] 50.0910μs 21.0526μs 47.5002 KOps/s 47.9624 KOps/s $\color{#d91a1a}-0.96\%$
test_creation_many_keys[50] 0.1231ms 90.6602μs 11.0302 KOps/s 11.1334 KOps/s $\color{#d91a1a}-0.93\%$
test_creation_many_keys[100] 0.2433ms 0.1784ms 5.6058 KOps/s 5.7290 KOps/s $\color{#d91a1a}-2.15\%$
test_creation_nested_many_keys[10] 75.8520μs 45.0839μs 22.1809 KOps/s 22.1650 KOps/s $\color{#35bf28}+0.07\%$
test_creation_nested_many_keys[50] 0.2295ms 0.1845ms 5.4199 KOps/s 5.4752 KOps/s $\color{#d91a1a}-1.01\%$
test_clone 37.1410μs 12.9584μs 77.1702 KOps/s 75.8329 KOps/s $\color{#35bf28}+1.76\%$
test_getitem[int] 1.6551ms 15.0802μs 66.3120 KOps/s 59.2022 KOps/s $\textbf{\color{#35bf28}+12.01\%}$
test_getitem[slice_int] 0.1365ms 23.7083μs 42.1792 KOps/s 41.4642 KOps/s $\color{#35bf28}+1.72\%$
test_getitem[range] 0.1706ms 62.9922μs 15.8750 KOps/s 15.5264 KOps/s $\color{#35bf28}+2.24\%$
test_getitem[tuple] 0.1414ms 23.4986μs 42.5557 KOps/s 42.1878 KOps/s $\color{#35bf28}+0.87\%$
test_getitem[list] 0.1770ms 58.3999μs 17.1233 KOps/s 17.2310 KOps/s $\color{#d91a1a}-0.62\%$
test_setitem_dim[int] 45.2210μs 25.6901μs 38.9255 KOps/s 37.5268 KOps/s $\color{#35bf28}+3.73\%$
test_setitem_dim[slice_int] 62.0220μs 41.6525μs 24.0081 KOps/s 21.5734 KOps/s $\textbf{\color{#35bf28}+11.29\%}$
test_setitem_dim[range] 0.1188ms 93.0461μs 10.7474 KOps/s 10.0824 KOps/s $\textbf{\color{#35bf28}+6.60\%}$
test_setitem_dim[tuple] 61.8810μs 38.8619μs 25.7321 KOps/s 24.7074 KOps/s $\color{#35bf28}+4.15\%$
test_setitem 52.6910μs 17.0436μs 58.6731 KOps/s 56.7709 KOps/s $\color{#35bf28}+3.35\%$
test_set 50.0210μs 16.6114μs 60.1998 KOps/s 60.2161 KOps/s $\color{#d91a1a}-0.03\%$
test_set_shared 0.4900ms 0.2087ms 4.7907 KOps/s 4.7127 KOps/s $\color{#35bf28}+1.65\%$
test_update 0.2171ms 21.5427μs 46.4195 KOps/s 42.8757 KOps/s $\textbf{\color{#35bf28}+8.27\%}$
test_update_nested 70.5510μs 32.0994μs 31.1532 KOps/s 28.3676 KOps/s $\textbf{\color{#35bf28}+9.82\%}$
test_update__nested 0.4453ms 33.5694μs 29.7890 KOps/s 27.4757 KOps/s $\textbf{\color{#35bf28}+8.42\%}$
test_set_nested 48.0410μs 18.3060μs 54.6268 KOps/s 49.0427 KOps/s $\textbf{\color{#35bf28}+11.39\%}$
test_set_nested_new 54.0810μs 23.2400μs 43.0293 KOps/s 41.8610 KOps/s $\color{#35bf28}+2.79\%$
test_select 82.8620μs 39.5990μs 25.2532 KOps/s 23.3194 KOps/s $\textbf{\color{#35bf28}+8.29\%}$
test_select_nested 0.1075ms 73.4456μs 13.6155 KOps/s 13.3274 KOps/s $\color{#35bf28}+2.16\%$
test_exclude_nested 0.1234ms 92.0589μs 10.8626 KOps/s 10.7783 KOps/s $\color{#35bf28}+0.78\%$
test_empty[True] 0.4270ms 0.3952ms 2.5306 KOps/s 2.4924 KOps/s $\color{#35bf28}+1.53\%$
test_empty[False] 6.7928μs 1.3116μs 762.4027 KOps/s 758.4752 KOps/s $\color{#35bf28}+0.52\%$
test_to 0.1073ms 71.8591μs 13.9161 KOps/s 13.8457 KOps/s $\color{#35bf28}+0.51\%$
test_to_nonblocking 0.1114ms 65.7652μs 15.2056 KOps/s 14.7204 KOps/s $\color{#35bf28}+3.30\%$
test_unbind_speed 0.3611ms 0.3328ms 3.0052 KOps/s 2.9222 KOps/s $\color{#35bf28}+2.84\%$
test_unbind_speed_stack0 0.4325ms 0.3310ms 3.0213 KOps/s 2.9086 KOps/s $\color{#35bf28}+3.87\%$
test_unbind_speed_stack1 0.1046s 0.8302ms 1.2045 KOps/s 1.1790 KOps/s $\color{#35bf28}+2.16\%$
test_split 0.1045s 1.2685ms 788.3603 Ops/s 785.4857 Ops/s $\color{#35bf28}+0.37\%$
test_chunk 0.1042s 1.2113ms 825.5357 Ops/s 925.7243 Ops/s $\textbf{\color{#d91a1a}-10.82\%}$
test_to_cpu_blocking 19.5610ms 19.4699ms 51.3614 Ops/s 51.2460 Ops/s $\color{#35bf28}+0.23\%$
test_to_cpu_global_sync 11.3153ms 11.2089ms 89.2147 Ops/s 88.9615 Ops/s $\color{#35bf28}+0.28\%$
test_to_cpu_event_sync 12.4169ms 12.2060ms 81.9272 Ops/s 81.7532 Ops/s $\color{#35bf28}+0.21\%$
test_to_cpu_default 0.1162s 13.5028ms 74.0587 Ops/s 81.5824 Ops/s $\textbf{\color{#d91a1a}-9.22\%}$
test_consolidate[False-None] 4.1916ms 4.1436ms 241.3373 Ops/s 216.2412 Ops/s $\textbf{\color{#35bf28}+11.61\%}$
test_consolidate[default-None] 2.1165ms 2.0400ms 490.1936 Ops/s 481.6993 Ops/s $\color{#35bf28}+1.76\%$
test_consolidate[reduce-overhead-None] 2.0633ms 1.9585ms 510.5945 Ops/s 502.9528 Ops/s $\color{#35bf28}+1.52\%$
test_consolidate_njt[False-None] 9.0532ms 8.4354ms 118.5485 Ops/s 117.4811 Ops/s $\color{#35bf28}+0.91\%$
test_to[False-False-None] 2.2533ms 2.0982ms 476.6026 Ops/s 474.3908 Ops/s $\color{#35bf28}+0.47\%$
test_to[True-False-None] 2.2295ms 1.9431ms 514.6413 Ops/s 521.8367 Ops/s $\color{#d91a1a}-1.38\%$
test_to[within-False-None] 6.2958ms 6.1427ms 162.7945 Ops/s 161.5219 Ops/s $\color{#35bf28}+0.79\%$
test_to[True-default-None] 8.9159ms 8.6769ms 115.2479 Ops/s 111.5432 Ops/s $\color{#35bf28}+3.32\%$
test_to_njt[False-False-None] 8.5351ms 8.4208ms 118.7539 Ops/s 117.4619 Ops/s $\color{#35bf28}+1.10\%$
test_to_njt[True-False-None] 7.1137ms 6.9017ms 144.8910 Ops/s 142.7616 Ops/s $\color{#35bf28}+1.49\%$
test_to_njt[within-False-None] 16.5910ms 16.0934ms 62.1371 Ops/s 63.3895 Ops/s $\color{#d91a1a}-1.98\%$
test_creation[device0] 0.2648ms 0.1128ms 8.8640 KOps/s 8.3442 KOps/s $\textbf{\color{#35bf28}+6.23\%}$
test_creation_from_tensor 0.3524ms 0.1115ms 8.9712 KOps/s 8.5035 KOps/s $\textbf{\color{#35bf28}+5.50\%}$
test_add_one[memmap_tensor0] 0.2832ms 6.3914μs 156.4603 KOps/s 155.2456 KOps/s $\color{#35bf28}+0.78\%$
test_contiguous[memmap_tensor0] 22.7910μs 0.6650μs 1.5037 MOps/s 2.0821 MOps/s $\textbf{\color{#d91a1a}-27.78\%}$
test_stack[memmap_tensor0] 33.1010μs 4.6564μs 214.7590 KOps/s 213.9690 KOps/s $\color{#35bf28}+0.37\%$
test_memmaptd_index 1.0783ms 0.2672ms 3.7422 KOps/s 3.7229 KOps/s $\color{#35bf28}+0.52\%$
test_memmaptd_index_astensor 0.5397ms 0.3741ms 2.6733 KOps/s 2.6991 KOps/s $\color{#d91a1a}-0.96\%$
test_memmaptd_index_op 0.1717s 0.7346ms 1.3614 KOps/s 1.5993 KOps/s $\textbf{\color{#d91a1a}-14.88\%}$
test_serialize_model 0.1374s 0.1357s 7.3712 Ops/s 7.3638 Ops/s $\color{#35bf28}+0.10\%$
test_serialize_model_pickle 1.3498s 1.2104s 0.8262 Ops/s 0.8254 Ops/s $\color{#35bf28}+0.10\%$
test_serialize_weights 0.1359s 0.1343s 7.4455 Ops/s 7.4121 Ops/s $\color{#35bf28}+0.45\%$
test_serialize_weights_returnearly 0.4349s 87.0947ms 11.4818 Ops/s 6.1135 Ops/s $\textbf{\color{#35bf28}+87.81\%}$
test_serialize_weights_pickle 1.3670s 1.2137s 0.8239 Ops/s 0.8228 Ops/s $\color{#35bf28}+0.13\%$
test_reshape_pytree 0.2048ms 32.5229μs 30.7475 KOps/s 30.1843 KOps/s $\color{#35bf28}+1.87\%$
test_reshape_td 85.1820μs 45.6407μs 21.9103 KOps/s 21.5824 KOps/s $\color{#35bf28}+1.52\%$
test_view_pytree 0.2249ms 32.7026μs 30.5786 KOps/s 30.9028 KOps/s $\color{#d91a1a}-1.05\%$
test_view_td 95.2220μs 53.7514μs 18.6042 KOps/s 18.7825 KOps/s $\color{#d91a1a}-0.95\%$
test_unbind_pytree 0.2383ms 36.7252μs 27.2293 KOps/s 26.9398 KOps/s $\color{#35bf28}+1.07\%$
test_unbind_td 0.1877ms 49.0985μs 20.3672 KOps/s 20.1166 KOps/s $\color{#35bf28}+1.25\%$
test_split_pytree 0.2457ms 42.6031μs 23.4725 KOps/s 23.6021 KOps/s $\color{#d91a1a}-0.55\%$
test_split_td 0.1762ms 66.1993μs 15.1059 KOps/s 15.5585 KOps/s $\color{#d91a1a}-2.91\%$
test_add_pytree 0.2399ms 43.5221μs 22.9768 KOps/s 23.7221 KOps/s $\color{#d91a1a}-3.14\%$
test_add_td 98.3020μs 57.1603μs 17.4947 KOps/s 18.3790 KOps/s $\color{#d91a1a}-4.81\%$
test_compile_add_one_nested[tensordict-compile] 0.2188ms 0.1440ms 6.9422 KOps/s 6.6917 KOps/s $\color{#35bf28}+3.74\%$
test_compile_add_one_nested[tensordict-eager] 0.5628ms 0.2036ms 4.9120 KOps/s 4.9821 KOps/s $\color{#d91a1a}-1.41\%$
test_compile_add_one_nested[pytree-compile] 0.2409ms 0.1116ms 8.9601 KOps/s 9.1257 KOps/s $\color{#d91a1a}-1.81\%$
test_compile_add_one_nested[pytree-eager] 0.4344ms 0.1805ms 5.5394 KOps/s 5.5088 KOps/s $\color{#35bf28}+0.56\%$
test_compile_copy_nested[tensordict-compile] 0.3555ms 9.9295μs 100.7096 KOps/s 95.7275 KOps/s $\textbf{\color{#35bf28}+5.20\%}$
test_compile_copy_nested[tensordict-eager] 92.8020μs 53.4543μs 18.7076 KOps/s 18.1168 KOps/s $\color{#35bf28}+3.26\%$
test_compile_copy_nested[pytree-compile] 47.6210μs 9.8814μs 101.2007 KOps/s 100.5773 KOps/s $\color{#35bf28}+0.62\%$
test_compile_copy_nested[pytree-eager] 0.4379ms 69.5116μs 14.3861 KOps/s 14.3014 KOps/s $\color{#35bf28}+0.59\%$
test_compile_add_one_flat[tensordict-compile] 0.2461ms 0.1761ms 5.6799 KOps/s 5.4218 KOps/s $\color{#35bf28}+4.76\%$
test_compile_add_one_flat[tensordict-eager] 0.4032ms 0.2784ms 3.5919 KOps/s 3.5651 KOps/s $\color{#35bf28}+0.75\%$
test_compile_add_one_flat[tensorclass-compile] 0.1877ms 0.1168ms 8.5635 KOps/s 8.3119 KOps/s $\color{#35bf28}+3.03\%$
test_compile_add_one_flat[tensorclass-eager] 0.1172ms 72.8063μs 13.7351 KOps/s 13.7017 KOps/s $\color{#35bf28}+0.24\%$
test_compile_add_one_flat[pytree-compile] 0.2406ms 0.1570ms 6.3681 KOps/s 6.2404 KOps/s $\color{#35bf28}+2.05\%$
test_compile_add_one_flat[pytree-eager] 0.8022ms 0.5206ms 1.9209 KOps/s 1.8974 KOps/s $\color{#35bf28}+1.24\%$
test_compile_add_self_flat[tensordict-eager] 0.4566ms 0.3320ms 3.0124 KOps/s 3.0030 KOps/s $\color{#35bf28}+0.31\%$
test_compile_add_self_flat[tensordict-compile] 0.2358ms 0.1776ms 5.6315 KOps/s 5.2029 KOps/s $\textbf{\color{#35bf28}+8.24\%}$
test_compile_add_self_flat[tensorclass-eager] 0.1600ms 88.6042μs 11.2862 KOps/s 11.2657 KOps/s $\color{#35bf28}+0.18\%$
test_compile_add_self_flat[tensorclass-compile] 0.2041ms 0.1185ms 8.4407 KOps/s 7.9455 KOps/s $\textbf{\color{#35bf28}+6.23\%}$
test_compile_add_self_flat[pytree-eager] 0.6556ms 0.4307ms 2.3219 KOps/s 2.2916 KOps/s $\color{#35bf28}+1.32\%$
test_compile_add_self_flat[pytree-compile] 0.2238ms 0.1566ms 6.3857 KOps/s 6.1841 KOps/s $\color{#35bf28}+3.26\%$
test_compile_copy_flat[tensordict-compile] 63.8220μs 13.3414μs 74.9549 KOps/s 75.2878 KOps/s $\color{#d91a1a}-0.44\%$
test_compile_copy_flat[tensordict-eager] 83.1220μs 41.5372μs 24.0748 KOps/s 23.7402 KOps/s $\color{#35bf28}+1.41\%$
test_compile_copy_flat[pytree-compile] 45.4210μs 10.5691μs 94.6153 KOps/s 93.7651 KOps/s $\color{#35bf28}+0.91\%$
test_compile_copy_flat[pytree-eager] 0.4094ms 52.3430μs 19.1047 KOps/s 19.0328 KOps/s $\color{#35bf28}+0.38\%$
test_compile_assign_and_add[tensordict-compile] 2.0126ms 0.1748ms 5.7223 KOps/s 5.3835 KOps/s $\textbf{\color{#35bf28}+6.29\%}$
test_compile_assign_and_add[tensordict-eager] 3.3555ms 3.2265ms 309.9329 Ops/s 304.2294 Ops/s $\color{#35bf28}+1.87\%$
test_compile_assign_and_add[pytree-compile] 1.9657ms 0.1619ms 6.1754 KOps/s 6.0057 KOps/s $\color{#35bf28}+2.82\%$
test_compile_assign_and_add[pytree-eager] 2.9047ms 2.7804ms 359.6623 Ops/s 352.6493 Ops/s $\color{#35bf28}+1.99\%$
test_compile_indexing[tensor-tensordict-compile] 0.1743ms 0.1087ms 9.2009 KOps/s 8.8534 KOps/s $\color{#35bf28}+3.93\%$
test_compile_indexing[tensor-tensordict-eager] 0.3178ms 72.9189μs 13.7139 KOps/s 13.8823 KOps/s $\color{#d91a1a}-1.21\%$
test_compile_indexing[tensor-tensorclass-compile] 0.1519ms 95.5668μs 10.4639 KOps/s 10.3431 KOps/s $\color{#35bf28}+1.17\%$
test_compile_indexing[tensor-tensorclass-eager] 0.2445ms 44.7729μs 22.3349 KOps/s 22.7134 KOps/s $\color{#d91a1a}-1.67\%$
test_compile_indexing[tensor-pytree-compile] 0.1670ms 96.0470μs 10.4116 KOps/s 10.2706 KOps/s $\color{#35bf28}+1.37\%$
test_compile_indexing[tensor-pytree-eager] 0.2751ms 44.3264μs 22.5599 KOps/s 22.5660 KOps/s $\color{#d91a1a}-0.03\%$
test_compile_indexing[slice-tensordict-compile] 0.1035ms 56.9669μs 17.5541 KOps/s 17.1642 KOps/s $\color{#35bf28}+2.27\%$
test_compile_indexing[slice-tensordict-eager] 0.2369ms 27.2854μs 36.6497 KOps/s 36.9191 KOps/s $\color{#d91a1a}-0.73\%$
test_compile_indexing[slice-tensorclass-compile] 85.1420μs 44.7290μs 22.3568 KOps/s 22.4791 KOps/s $\color{#d91a1a}-0.54\%$
test_compile_indexing[slice-tensorclass-eager] 0.2613ms 22.2016μs 45.0418 KOps/s 44.4058 KOps/s $\color{#35bf28}+1.43\%$
test_compile_indexing[slice-pytree-compile] 79.8120μs 44.3111μs 22.5677 KOps/s 22.1405 KOps/s $\color{#35bf28}+1.93\%$
test_compile_indexing[slice-pytree-eager] 0.2696ms 22.4667μs 44.5104 KOps/s 44.1341 KOps/s $\color{#35bf28}+0.85\%$
test_compile_indexing[int-tensordict-compile] 0.1132ms 56.8187μs 17.5999 KOps/s 16.5707 KOps/s $\textbf{\color{#35bf28}+6.21\%}$
test_compile_indexing[int-tensordict-eager] 0.2847ms 27.2796μs 36.6575 KOps/s 35.8543 KOps/s $\color{#35bf28}+2.24\%$
test_compile_indexing[int-tensorclass-compile] 90.7930μs 44.6582μs 22.3923 KOps/s 22.1452 KOps/s $\color{#35bf28}+1.12\%$
test_compile_indexing[int-tensorclass-eager] 0.2710ms 22.4859μs 44.4724 KOps/s 43.8572 KOps/s $\color{#35bf28}+1.40\%$
test_compile_indexing[int-pytree-compile] 86.9030μs 45.0525μs 22.1963 KOps/s 21.8369 KOps/s $\color{#35bf28}+1.65\%$
test_compile_indexing[int-pytree-eager] 0.2805ms 22.4286μs 44.5860 KOps/s 44.2595 KOps/s $\color{#35bf28}+0.74\%$
test_compile_replace[single-eager] 0.1094ms 49.9107μs 20.0358 KOps/s 20.8875 KOps/s $\color{#d91a1a}-4.08\%$
test_compile_replace[single-compile] 0.1629ms 0.1049ms 9.5335 KOps/s 9.3743 KOps/s $\color{#35bf28}+1.70\%$
test_compile_replace[multi-eager] 0.6531ms 0.5730ms 1.7452 KOps/s 1.8169 KOps/s $\color{#d91a1a}-3.94\%$
test_compile_replace[multi-compile] 0.1667ms 0.1104ms 9.0590 KOps/s 8.8269 KOps/s $\color{#35bf28}+2.63\%$
test_compile_tc_getattr_20[eager] 0.2192ms 0.1671ms 5.9858 KOps/s 6.1325 KOps/s $\color{#d91a1a}-2.39\%$
test_compile_tc_getattr_20[compile] 0.1710ms 0.1183ms 8.4535 KOps/s 8.2693 KOps/s $\color{#35bf28}+2.23\%$
test_compile_clone_shallow[20-eager] 49.0210μs 19.3258μs 51.7443 KOps/s 50.6022 KOps/s $\color{#35bf28}+2.26\%$
test_compile_clone_shallow[20-compile] 45.9210μs 11.3389μs 88.1916 KOps/s 84.8608 KOps/s $\color{#35bf28}+3.92\%$
test_compile_clone_shallow[40-eager] 63.3320μs 33.7736μs 29.6089 KOps/s 28.8283 KOps/s $\color{#35bf28}+2.71\%$
test_compile_clone_shallow[40-compile] 45.3510μs 12.3984μs 80.6556 KOps/s 78.4398 KOps/s $\color{#35bf28}+2.82\%$
test_compile_clone_shallow[80-eager] 96.6720μs 62.6347μs 15.9656 KOps/s 15.4338 KOps/s $\color{#35bf28}+3.45\%$
test_compile_clone_shallow[80-compile] 58.0010μs 15.1017μs 66.2177 KOps/s 66.2592 KOps/s $\color{#d91a1a}-0.06\%$
test_compile_update_inplace[eager] 97.0120μs 58.0853μs 17.2160 KOps/s 16.7185 KOps/s $\color{#35bf28}+2.98\%$
test_compile_update_inplace[compile] 0.1958ms 0.1402ms 7.1307 KOps/s 6.9131 KOps/s $\color{#35bf28}+3.15\%$
test_mod_add[eager] 95.4930μs 47.6496μs 20.9865 KOps/s 19.5255 KOps/s $\textbf{\color{#35bf28}+7.48\%}$
test_mod_add[compile] 0.1846ms 0.1055ms 9.4764 KOps/s 9.4546 KOps/s $\color{#35bf28}+0.23\%$
test_mod_add[compile-overhead] 0.2354ms 0.1477ms 6.7720 KOps/s 6.5502 KOps/s $\color{#35bf28}+3.38\%$
test_mod_wrap[eager] 0.3996ms 0.3048ms 3.2808 KOps/s 3.3101 KOps/s $\color{#d91a1a}-0.88\%$
test_mod_wrap[compile] 0.4198ms 0.3432ms 2.9141 KOps/s 2.8024 KOps/s $\color{#35bf28}+3.99\%$
test_mod_wrap[compile-overhead] 7.1077ms 3.9279ms 254.5893 Ops/s 251.1453 Ops/s $\color{#35bf28}+1.37\%$
test_mod_wrap_and_backward[eager] 1.5867ms 1.4750ms 677.9521 Ops/s 608.1245 Ops/s $\textbf{\color{#35bf28}+11.48\%}$
test_mod_wrap_and_backward[compile] 1.5737ms 1.4295ms 699.5534 Ops/s 668.2758 Ops/s $\color{#35bf28}+4.68\%$
test_mod_wrap_and_backward[compile-overhead] 1.2528ms 0.8823ms 1.1334 KOps/s 1.1033 KOps/s $\color{#35bf28}+2.72\%$
test_seq_add[eager] 0.2078ms 0.1510ms 6.6238 KOps/s 6.4647 KOps/s $\color{#35bf28}+2.46\%$
test_seq_add[compile] 0.1496ms 0.1115ms 8.9670 KOps/s 8.1320 KOps/s $\textbf{\color{#35bf28}+10.27\%}$
test_seq_add[compile-overhead] 0.2532ms 0.1569ms 6.3734 KOps/s 6.2622 KOps/s $\color{#35bf28}+1.78\%$
test_seq_wrap[eager] 0.6053ms 0.5340ms 1.8727 KOps/s 1.8275 KOps/s $\color{#35bf28}+2.47\%$
test_seq_wrap[compile] 0.4360ms 0.3616ms 2.7656 KOps/s 2.6052 KOps/s $\textbf{\color{#35bf28}+6.16\%}$
test_seq_wrap[compile-overhead] 0.3270ms 0.2633ms 3.7975 KOps/s 3.7185 KOps/s $\color{#35bf28}+2.13\%$
test_func_call_runtime[False-eager] 0.9019ms 0.8290ms 1.2063 KOps/s 1.1935 KOps/s $\color{#35bf28}+1.07\%$
test_func_call_runtime[False-compile] 0.9587ms 0.9051ms 1.1048 KOps/s 1.0859 KOps/s $\color{#35bf28}+1.75\%$
test_func_call_runtime[False-compile-overhead] 0.5272ms 0.4568ms 2.1890 KOps/s 2.1342 KOps/s $\color{#35bf28}+2.57\%$
test_func_call_runtime[True-eager] 1.1527ms 1.0646ms 939.3448 Ops/s 930.8586 Ops/s $\color{#35bf28}+0.91\%$
test_func_call_runtime[True-compile] 1.0183ms 0.9164ms 1.0913 KOps/s 1.0740 KOps/s $\color{#35bf28}+1.61\%$
test_func_call_runtime[True-compile-overhead] 0.5472ms 0.4733ms 2.1130 KOps/s 2.0764 KOps/s $\color{#35bf28}+1.76\%$
test_func_call_cm_runtime[False-eager] 0.9427ms 0.8399ms 1.1906 KOps/s 1.2047 KOps/s $\color{#d91a1a}-1.17\%$
test_func_call_cm_runtime[False-compile] 0.9744ms 0.9064ms 1.1032 KOps/s 1.0857 KOps/s $\color{#35bf28}+1.61\%$
test_func_call_cm_runtime[False-compile-overhead] 0.5680ms 0.4613ms 2.1677 KOps/s 2.1402 KOps/s $\color{#35bf28}+1.28\%$
test_func_call_cm_runtime[True-eager] 1.4395ms 1.2229ms 817.7492 Ops/s 817.2534 Ops/s $\color{#35bf28}+0.06\%$
test_func_call_cm_runtime[True-compile] 1.0318ms 0.9503ms 1.0523 KOps/s 1.0220 KOps/s $\color{#35bf28}+2.96\%$
test_func_call_cm_runtime[True-compile-overhead] 0.5644ms 0.5067ms 1.9737 KOps/s 1.9272 KOps/s $\color{#35bf28}+2.41\%$
test_vmap_func_call_cm_runtime[eager] 2.8238ms 2.3536ms 424.8786 Ops/s 421.4405 Ops/s $\color{#35bf28}+0.82\%$
test_vmap_func_call_cm_runtime[compile] 1.0939ms 0.9685ms 1.0325 KOps/s 1.0070 KOps/s $\color{#35bf28}+2.53\%$
test_vmap_func_call_cm_runtime[compile-overhead] 0.6017ms 0.5109ms 1.9573 KOps/s 1.9183 KOps/s $\color{#35bf28}+2.03\%$
test_distributed 0.5766ms 0.1524ms 6.5617 KOps/s 6.4954 KOps/s $\color{#35bf28}+1.02\%$
test_tdmodule 0.1989ms 27.2471μs 36.7011 KOps/s 35.2637 KOps/s $\color{#35bf28}+4.08\%$
test_tdmodule_dispatch 76.3120μs 45.3772μs 22.0375 KOps/s 21.8602 KOps/s $\color{#35bf28}+0.81\%$
test_tdseq 47.3910μs 26.8989μs 37.1762 KOps/s 36.5113 KOps/s $\color{#35bf28}+1.82\%$
test_tdseq_dispatch 70.6710μs 47.3799μs 21.1060 KOps/s 20.6139 KOps/s $\color{#35bf28}+2.39\%$
test_instantiation_functorch 2.1968ms 2.0813ms 480.4625 Ops/s 481.0721 Ops/s $\color{#d91a1a}-0.13\%$
test_exec_functorch 0.2773ms 0.1801ms 5.5521 KOps/s 5.6186 KOps/s $\color{#d91a1a}-1.18\%$
test_exec_functional_call 0.2107ms 0.1608ms 6.2198 KOps/s 6.3007 KOps/s $\color{#d91a1a}-1.28\%$
test_exec_td_decorator 0.4662ms 0.2351ms 4.2539 KOps/s 4.2375 KOps/s $\color{#35bf28}+0.39\%$
test_vmap_mlp_speed_decorator[True-True] 0.9981ms 0.8142ms 1.2282 KOps/s 1.2105 KOps/s $\color{#35bf28}+1.46\%$
test_vmap_mlp_speed_decorator[True-False] 0.9969ms 0.8175ms 1.2232 KOps/s 1.2136 KOps/s $\color{#35bf28}+0.79\%$
test_vmap_mlp_speed_decorator[False-True] 0.8883ms 0.7046ms 1.4193 KOps/s 1.4087 KOps/s $\color{#35bf28}+0.75\%$
test_vmap_mlp_speed_decorator[False-False] 0.8798ms 0.7082ms 1.4120 KOps/s 1.4104 KOps/s $\color{#35bf28}+0.11\%$
test_vmap_transformer_speed_decorator[True-True] 21.1272ms 20.4359ms 48.9334 Ops/s 48.6663 Ops/s $\color{#35bf28}+0.55\%$
test_vmap_transformer_speed_decorator[True-False] 21.0323ms 20.4141ms 48.9858 Ops/s 48.6487 Ops/s $\color{#35bf28}+0.69\%$
test_vmap_transformer_speed_decorator[False-True] 20.4241ms 20.2334ms 49.4232 Ops/s 49.2322 Ops/s $\color{#35bf28}+0.39\%$
test_vmap_transformer_speed_decorator[False-False] 20.3955ms 20.2952ms 49.2727 Ops/s 49.2255 Ops/s $\color{#35bf28}+0.10\%$
test_to_module_speed[True] 1.6140ms 1.4713ms 679.6663 Ops/s 677.8393 Ops/s $\color{#35bf28}+0.27\%$
test_to_module_speed[False] 1.5721ms 1.4578ms 685.9877 Ops/s 682.7092 Ops/s $\color{#35bf28}+0.48\%$
test_tc_init 75.0610μs 43.5641μs 22.9547 KOps/s 21.9463 KOps/s $\color{#35bf28}+4.59\%$
test_tc_init_tensor_only 41.8710μs 9.8043μs 101.9957 KOps/s 101.0925 KOps/s $\color{#35bf28}+0.89\%$
test_tc_init_nested 0.1247ms 87.3811μs 11.4441 KOps/s 11.0941 KOps/s $\color{#35bf28}+3.16\%$
test_tc_init_many_fields 41.5110μs 16.3712μs 61.0827 KOps/s 60.8865 KOps/s $\color{#35bf28}+0.32\%$
test_tc_first_layer_tensor 20.3100μs 1.8305μs 546.2871 KOps/s 543.8532 KOps/s $\color{#35bf28}+0.45\%$
test_tc_first_layer_tensor_only 1.6576μs 0.3953μs 2.5299 MOps/s 2.5097 MOps/s $\color{#35bf28}+0.80\%$
test_tc_first_layer_tensor_set 29.2800μs 3.9485μs 253.2586 KOps/s 253.2819 KOps/s $-0.01\%$
test_tc_first_layer_tensor_only_set 27.9010μs 3.2645μs 306.3234 KOps/s 300.1246 KOps/s $\color{#35bf28}+2.07\%$
test_tc_first_layer_nontensor 27.9900μs 6.1878μs 161.6086 KOps/s 159.8698 KOps/s $\color{#35bf28}+1.09\%$
test_tc_second_layer_tensor 24.1510μs 4.4519μs 224.6228 KOps/s 221.6602 KOps/s $\color{#35bf28}+1.34\%$
test_tc_second_layer_nontensor 42.1810μs 8.6762μs 115.2573 KOps/s 112.6216 KOps/s $\color{#35bf28}+2.34\%$
test_unbind 0.2516s 16.2010ms 61.7246 Ops/s 66.4768 Ops/s $\textbf{\color{#d91a1a}-7.15\%}$
test_full_like 5.0500ms 4.3876ms 227.9174 Ops/s 228.5870 Ops/s $\color{#d91a1a}-0.29\%$
test_zeros_like 4.7507ms 4.3487ms 229.9561 Ops/s 60.1450 Ops/s $\textbf{\color{#35bf28}+282.34\%}$
test_ones_like 4.4839ms 4.3517ms 229.7942 Ops/s 60.0930 Ops/s $\textbf{\color{#35bf28}+282.40\%}$
test_clone 6.6881ms 6.3300ms 157.9780 Ops/s 56.3278 Ops/s $\textbf{\color{#35bf28}+180.46\%}$
test_squeeze 0.2041ms 14.2534μs 70.1585 KOps/s 70.4338 KOps/s $\color{#d91a1a}-0.39\%$
test_unsqueeze 0.1663ms 0.1102ms 9.0752 KOps/s 8.9566 KOps/s $\color{#35bf28}+1.32\%$
test_split 0.2485ms 0.1868ms 5.3527 KOps/s 5.4165 KOps/s $\color{#d91a1a}-1.18\%$
test_permute 0.2899ms 0.2131ms 4.6931 KOps/s 4.8608 KOps/s $\color{#d91a1a}-3.45\%$
test_stack 42.8185ms 42.4954ms 23.5319 Ops/s 19.5939 Ops/s $\textbf{\color{#35bf28}+20.10\%}$
test_cat 42.5590ms 42.4038ms 23.5828 Ops/s 19.7069 Ops/s $\textbf{\color{#35bf28}+19.67\%}$
test_sequential_tensordict 0.5455ms 0.2165ms 4.6189 KOps/s 4.5540 KOps/s $\color{#35bf28}+1.42\%$
test_sequential_graph_module 0.2512ms 0.1202ms 8.3198 KOps/s 8.3820 KOps/s $\color{#d91a1a}-0.74\%$
test_nested_tensordict 0.6791ms 0.2823ms 3.5422 KOps/s 3.5068 KOps/s $\color{#35bf28}+1.01\%$
test_nested_graph_module 0.1837ms 0.1267ms 7.8955 KOps/s 7.7692 KOps/s $\color{#35bf28}+1.63\%$

@vmoens vmoens mentioned this pull request Mar 11, 2026
Open
@github-actions

github-actions Bot commented Mar 11, 2026

Copy link
Copy Markdown
Contributor

PR Title Label Error

Unknown or invalid prefix [DTensor].

Current title: [DTensor] Add ModelTransferPlan for multi-parameter precomputed transfers

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.

@vmoens
Update
a05ba67 
[ghstack-poisoned]
@github-actions

github-actions Bot commented Mar 17, 2026

Copy link
Copy Markdown
Contributor

PR Title Label Error

Unknown or invalid prefix [DTensor].

Current title: [DTensor] Add ModelTransferPlan for multi-parameter precomputed transfers

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.

1 similar comment
@github-actions

github-actions Bot commented Mar 17, 2026

Copy link
Copy Markdown
Contributor

PR Title Label Error

Unknown or invalid prefix [DTensor].

Current title: [DTensor] Add ModelTransferPlan for multi-parameter precomputed transfers

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.

@vmoens
Update
61dfd11 
[ghstack-poisoned]
@github-actions

github-actions Bot commented Mar 17, 2026

Copy link
Copy Markdown
Contributor

PR Title Label Error

Unknown or invalid prefix [DTensor].

Current title: [DTensor] Add ModelTransferPlan for multi-parameter precomputed transfers

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.

@vmoens
Update
c956842 
[ghstack-poisoned]
@github-actions

github-actions Bot commented Mar 17, 2026

Copy link
Copy Markdown
Contributor

PR Title Label Error

Unknown or invalid prefix [DTensor].

Current title: [DTensor] Add ModelTransferPlan for multi-parameter precomputed transfers

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.

1 similar comment
@github-actions

github-actions Bot commented Mar 17, 2026

Copy link
Copy Markdown
Contributor

PR Title Label Error

Unknown or invalid prefix [DTensor].

Current title: [DTensor] Add ModelTransferPlan for multi-parameter precomputed transfers

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.

@vmoens
Update
c4a890e 
[ghstack-poisoned]
@github-actions

github-actions Bot commented Mar 17, 2026

Copy link
Copy Markdown
Contributor

PR Title Label Error

Unknown or invalid prefix [DTensor].

Current title: [DTensor] Add ModelTransferPlan for multi-parameter precomputed transfers

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.

Sign up for free to join this conversation on GitHub. Already have an account? Sign in to comment

Reviewers

No reviews

Assignees

No one assigned

Labels

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

Projects

None yet

Milestone

No milestone

Development

Successfully merging this pull request may close these issues.

1 participant

@vmoens