[CI] Contain hard hangs in the optdeps job#3979
Open
vmoens wants to merge 2 commits into
Open
Conversation
🔗 Helpful Links🧪 See artifacts and rendered test results at hud.pytorch.org/pr/pytorch/rl/3979
Note: Links to docs will display an error until the docs builds have been completed. ⏳ No Failures, 1 PendingAs of commit 3fff91b with merge base a74b478 ( This comment was automatically generated by Dr. CI and updates every 15 minutes. |
github-actions
Bot
added
the
CI
This was referenced Jul 9, 2026
meta-cla
Bot
added
the
CLA Signed
This was referenced Jul 9, 2026
Contributor
Benchmark Results: PR
|
| Benchmark | main ops | PR ops | Change |
|---|---|---|---|
benchmarks/test_replaybuffer_benchmark.py::test_rb_sample[TensorDictReplayBuffer-LazyMemmapStorage-RandomSampler-10000] |
2,763 | 994.83 | -63.99% |
benchmarks/test_objectives_benchmarks.py::test_redq_deprec_speed[False-None] |
59.98 | 89.09 | +48.54% |
benchmarks/test_objectives_benchmarks.py::test_values[vec_td_lambda_return_estimate-True-False] |
83.46 | 59.18 | -29.09% |
benchmarks/test_objectives_benchmarks.py::test_values[vec_td1_return_estimate-False-False] |
84.05 | 62.60 | -25.53% |
benchmarks/test_objectives_benchmarks.py::test_values[vec_generalized_advantage_estimate-True-True] |
86.66 | 66.47 | -23.30% |
benchmarks/test_vla_preprocessing_benchmark.py::test_openvla_preprocessing_throughput[torchvision-480-640-16] |
36.13 | 28.56 | -20.95% |
benchmarks/test_replaybuffer_benchmark.py::test_rb_sample[TensorDictReplayBuffer-LazyTensorStorage-RandomSampler-10000] |
3,325 | 2,754 | -17.18% |
benchmarks/test_vla_preprocessing_benchmark.py::test_openvla_preprocessing_throughput[torchvision-480-640-64] |
7.0410 | 6.3139 | -10.33% |
benchmarks/test_replaybuffer_benchmark.py::test_rb_iterate[TensorDictPrioritizedReplayBuffer-LazyTensorStorage-None-10000] |
2,112 | 2,301 | +8.93% |
benchmarks/test_replaybuffer_benchmark.py::test_rb_iterate[TensorDictPrioritizedReplayBuffer-LazyMemmapStorage-None-10000] |
2,023 | 2,190 | +8.25% |
benchmarks/test_vla_preprocessing_benchmark.py::test_openvla_preprocessing_throughput[torchvision-256-256-64] |
10.45 | 9.6145 | -8.01% |
benchmarks/test_replaybuffer_benchmark.py::test_rb_iterate[TensorDictReplayBuffer-LazyMemmapStorage-SamplerWithoutReplacement-10000] |
3,160 | 2,916 | -7.71% |
benchmarks/test_storage_write_benchmark.py::TestStorageWriteBenchmark::test_collector_stack_then_write[100-img_shape1-atari] |
260.33 | 280.18 | +7.63% |
benchmarks/test_replaybuffer_benchmark.py::test_rb_sample[TensorDictReplayBuffer-LazyTensorStorage-sampler7-10000] |
838.75 | 777.04 | -7.36% |
benchmarks/test_objectives_benchmarks.py::test_values[generalized_advantage_estimate-True-True] |
92.38 | 99.16 | +7.33% |
benchmarks/test_storage_write_benchmark.py::TestStorageWriteBenchmark::test_storage_write_contiguous[100-img_shape1-atari] |
4,838 | 5,184 | +7.15% |
benchmarks/test_storage_write_benchmark.py::TestStorageWriteBenchmark::test_collector_lazystack_then_write[100-img_shape2-large_img] |
384.26 | 411.60 | +7.12% |
benchmarks/test_replaybuffer_benchmark.py::test_rb_iterate[TensorDictReplayBuffer-LazyTensorStorage-SamplerWithoutReplacement-10000] |
3,506 | 3,258 | -7.09% |
benchmarks/test_replaybuffer_benchmark.py::test_rb_sample[TensorDictReplayBuffer-LazyMemmapStorage-sampler6-10000] |
789.06 | 733.42 | -7.05% |
benchmarks/test_envs_benchmark.py::test_step_mdp_speed[True-True-True-True-True] |
23,701 | 22,086 | -6.81% |
benchmarks/test_storage_write_benchmark.py::TestStorageWriteBenchmark::test_storage_write_lazystack[100-img_shape2-large_img] |
402.81 | 429.10 | +6.53% |
benchmarks/test_replaybuffer_benchmark.py::test_rb_sample[TensorDictPrioritizedReplayBuffer-LazyMemmapStorage-None-10000] |
2,028 | 2,159 | +6.48% |
benchmarks/test_replaybuffer_benchmark.py::test_rb_iterate[TensorDictReplayBuffer-LazyTensorStorage-RandomSampler-10000] |
3,327 | 3,120 | -6.22% |
benchmarks/test_replaybuffer_benchmark.py::test_rb_sample[TensorDictReplayBuffer-LazyTensorStorage-SamplerWithoutReplacement-10000] |
3,329 | 3,127 | -6.07% |
benchmarks/test_replaybuffer_benchmark.py::test_rb_sample[TensorDictReplayBuffer-LazyMemmapStorage-SamplerWithoutReplacement-10000] |
3,295 | 3,098 | -5.98% |
benchmarks/test_objectives_benchmarks.py::test_a2c_speed[True-backward] |
115.10 | 121.73 | +5.76% |
benchmarks/test_envs_benchmark.py::test_step_mdp_speed[True-False-False-False-True] |
34,698 | 32,724 | -5.69% |
benchmarks/test_rnn_reset_backends_benchmark.py::test_rnn_rollout_with_intermediate_resets[b256-t128-i32-h512-cudnn-False-0-gru] |
1.3927 | 1.3137 | -5.67% |
benchmarks/test_objectives_benchmarks.py::test_sac_speed[True-None] |
443.53 | 468.46 | +5.62% |
benchmarks/test_objectives_benchmarks.py::test_td3_speed[True-backward] |
271.31 | 286.11 | +5.45% |
benchmarks/test_objectives_benchmarks.py::test_a2c_speed[True-None] |
277.76 | 291.84 | +5.07% |
benchmarks/test_envs_benchmark.py::test_cat_frames_functional[16-same] |
1,892 | 1,796 | -5.05% |
benchmarks/test_storage_write_benchmark.py::TestStorageWriteBenchmark::test_collector_lazystack_then_write[50-img_shape0-small] |
3,355 | 3,520 | +4.91% |
benchmarks/test_envs_benchmark.py::test_simple |
1.8127 | 1.7239 | -4.90% |
benchmarks/test_envs_benchmark.py::test_cat_frames_functional[4-same] |
2,559 | 2,684 | +4.89% |
benchmarks/test_storage_write_benchmark.py::TestStorageWriteBenchmark::test_storage_write_lazystack[50-img_shape0-small] |
4,204 | 4,408 | +4.84% |
benchmarks/test_replaybuffer_benchmark.py::test_rb_sample[TensorDictPrioritizedReplayBuffer-LazyTensorStorage-None-10000] |
2,145 | 2,249 | +4.83% |
benchmarks/test_storage_write_benchmark.py::TestStorageWriteBenchmark::test_collector_lazystack_then_write[100-img_shape1-atari] |
626.08 | 654.92 | +4.61% |
benchmarks/test_envs_benchmark.py::test_step_mdp_speed[False-True-False-True-True] |
20,715 | 19,764 | -4.59% |
benchmarks/test_objectives_benchmarks.py::test_values[td0_return_estimate-False-False] |
7,506 | 7,849 | +4.57% |
benchmarks/test_compressed_storage_benchmark.py::TestCompressedStorageBenchmark::test_tensor_to_bytestream_speed[safetensors] |
23,175 | 24,198 | +4.42% |
benchmarks/test_envs_benchmark.py::test_step_mdp_speed[True-True-True-False-True] |
42,014 | 40,264 | -4.17% |
benchmarks/test_replaybuffer_benchmark.py::test_rb_extend_sample[ReplayBuffer-LazyTensorStorage-RandomSampler-100000-10000-100-False] |
50.87 | 52.98 | +4.15% |
benchmarks/test_storage_write_benchmark.py::TestStorageWriteBenchmark::test_storage_write_lazystack[200-img_shape3-large_batch] |
325.89 | 339.29 | +4.11% |
benchmarks/test_vla_preprocessing_benchmark.py::test_openvla_preprocessing_throughput[pil-224-224-16] |
18.09 | 17.41 | -3.79% |
benchmarks/test_envs_benchmark.py::test_step_mdp_speed[True-False-True-True-True] |
21,028 | 20,241 | -3.74% |
benchmarks/test_objectives_benchmarks.py::test_ddpg_speed[reduce-overhead-None] |
681.60 | 707.03 | +3.73% |
benchmarks/test_vla_preprocessing_benchmark.py::test_openvla_preprocessing_throughput[torchvision-224-224-1] |
633.20 | 609.61 | -3.72% |
benchmarks/test_storage_write_benchmark.py::TestStorageWriteBenchmark::test_storage_write_contiguous[100-img_shape2-large_img] |
566.72 | 587.79 | +3.72% |
benchmarks/test_envs_benchmark.py::test_step_mdp_speed[True-True-True-False-False] |
76,484 | 73,640 | -3.72% |
benchmarks/test_storage_write_benchmark.py::TestStorageWriteBenchmark::test_storage_write_lazystack[100-img_shape1-atari] |
682.80 | 708.00 | +3.69% |
benchmarks/test_objectives_benchmarks.py::test_ppo_speed[True-backward] |
112.14 | 116.26 | +3.68% |
benchmarks/test_rnn_reset_backends_benchmark.py::test_rnn_rollout_with_intermediate_resets[b256-t128-i32-h512-cudnn-True-0-gru] |
1.4735 | 1.4194 | -3.67% |
benchmarks/test_envs_benchmark.py::test_step_mdp_speed[False-False-False-True-False] |
27,715 | 26,733 | -3.54% |
benchmarks/test_objectives_benchmarks.py::test_ddpg_speed[True-backward] |
382.38 | 368.98 | -3.50% |
benchmarks/test_envs_benchmark.py::test_step_mdp_speed[False-False-False-False-True] |
28,822 | 27,816 | -3.49% |
benchmarks/test_objectives_benchmarks.py::test_dqn_speed[True-backward] |
978.70 | 1,012 | +3.43% |
benchmarks/test_replaybuffer_benchmark.py::test_rb_populate[TensorDictReplayBuffer-LazyTensorStorage-SamplerWithoutReplacement-400] |
1,073 | 1,110 | +3.41% |
benchmarks/test_replaybuffer_benchmark.py::test_rb_populate[TensorDictReplayBuffer-ListStorage-RandomSampler-400] |
49.44 | 51.11 | +3.39% |
benchmarks/test_replaybuffer_benchmark.py::test_rb_extend_sample[ReplayBuffer-LazyTensorStorage-RandomSampler-10000-10000-100-False] |
51.25 | 52.96 | +3.33% |
benchmarks/test_envs_benchmark.py::test_step_mdp_speed[False-True-True-False-True] |
31,508 | 32,544 | +3.29% |
benchmarks/test_replaybuffer_benchmark.py::test_rb_populate[TensorDictReplayBuffer-LazyMemmapStorage-SamplerWithoutReplacement-400] |
471.84 | 487.02 | +3.22% |
benchmarks/test_replaybuffer_benchmark.py::TestWindowingTransformsBenchmark::test_action_chunk_transform[no_done] |
1,532 | 1,579 | +3.08% |
benchmarks/test_storage_write_benchmark.py::TestStorageWriteBenchmark::test_collector_lazystack_then_write[200-img_shape3-large_batch] |
306.07 | 315.38 | +3.04% |
benchmarks/test_envs_benchmark.py::test_step_mdp_speed[True-False-False-False-False] |
55,313 | 53,661 | -2.99% |
benchmarks/test_vla_preprocessing_benchmark.py::test_openvla_preprocessing_throughput[torchvision-224-224-64] |
12.10 | 11.75 | -2.97% |
benchmarks/test_vla_preprocessing_benchmark.py::test_openvla_preprocessing_throughput[pil-224-224-4] |
71.42 | 69.33 | -2.92% |
benchmarks/test_replaybuffer_benchmark.py::test_rb_populate[TensorDictReplayBuffer-ListStorage-SamplerWithoutReplacement-400] |
198.11 | 192.34 | -2.91% |
benchmarks/test_storage_write_benchmark.py::TestCollectorIntegrationBenchmark::test_collector_without_rb[100-img_shape0-atari] |
29.08 | 29.91 | +2.84% |
benchmarks/test_envs_benchmark.py::test_transformed |
0.8848 | 0.9093 | +2.77% |
benchmarks/test_replaybuffer_benchmark.py::test_rb_iterate[TensorDictReplayBuffer-LazyMemmapStorage-RandomSampler-10000] |
2,907 | 2,987 | +2.73% |
benchmarks/test_objectives_benchmarks.py::test_reinforce_speed[False-None] |
211.88 | 217.64 | +2.72% |
benchmarks/test_objectives_benchmarks.py::test_reinforce_speed[True-backward] |
126.05 | 129.35 | +2.62% |
benchmarks/test_replaybuffer_benchmark.py::test_rb_extend_sample[ReplayBuffer-LazyTensorStorage-RandomSampler-10000-10000-100-True] |
24.79 | 25.43 | +2.60% |
benchmarks/test_envs_benchmark.py::test_step_mdp_speed[False-False-False-False-False] |
45,034 | 43,881 | -2.56% |
benchmarks/test_objectives_benchmarks.py::test_gae_speed[vec_generalized_advantage_estimate-True-1-512] |
699.59 | 681.80 | -2.54% |
benchmarks/test_objectives_benchmarks.py::test_ppo_speed[False-None] |
161.02 | 165.09 | +2.53% |
benchmarks/test_envs_benchmark.py::test_step_mdp_speed[False-True-False-False-False] |
49,001 | 50,187 | +2.42% |
benchmarks/test_objectives_benchmarks.py::test_iql_speed[False-None] |
49.59 | 50.79 | +2.42% |
benchmarks/test_objectives_benchmarks.py::test_ddpg_speed[False-None] |
349.04 | 340.65 | -2.40% |
benchmarks/test_envs_benchmark.py::test_step_mdp_speed[False-True-False-False-True] |
30,451 | 29,725 | -2.38% |
benchmarks/test_objectives_benchmarks.py::test_redq_deprec_speed[reduce-overhead-None] |
271.76 | 278.18 | +2.36% |
benchmarks/test_storage_write_benchmark.py::TestStorageWriteBenchmark::test_collector_stack_then_write[200-img_shape3-large_batch] |
138.89 | 142.11 | +2.32% |
benchmarks/test_compressed_storage_benchmark.py::TestCompressedStorageBenchmark::test_tensor_to_bytestream_speed[untyped_storage] |
8.3298 | 8.5225 | +2.31% |
benchmarks/test_envs_benchmark.py::test_step_mdp_speed[True-False-True-False-True] |
37,850 | 36,978 | -2.30% |
benchmarks/test_storage_write_benchmark.py::TestCollectorIntegrationBenchmark::test_collector_without_rb[200-img_shape1-large_batch] |
14.70 | 15.03 | +2.27% |
benchmarks/test_rnn_reset_backends_benchmark.py::test_rnn_rollout_with_intermediate_resets[b256-t128-i32-h512-cudnn-False-0-lstm] |
0.8751 | 0.8553 | -2.27% |
benchmarks/test_non_tensor_env_benchmark.py::test_non_tensor_env_rollout_speed[1000-parallel-buffers-True] |
0.5125 | 0.5239 | +2.22% |
benchmarks/test_envs_benchmark.py::test_step_mdp_speed[True-False-False-True-False] |
32,149 | 31,447 | -2.18% |
benchmarks/test_replaybuffer_benchmark.py::test_rb_populate[TensorDictPrioritizedReplayBuffer-LazyTensorStorage-None-400] |
780.20 | 797.15 | +2.17% |
benchmarks/test_vla_preprocessing_benchmark.py::test_openvla_preprocessing_throughput[pil-224-224-1] |
271.60 | 265.71 | -2.17% |
benchmarks/test_replaybuffer_benchmark.py::test_rb_populate[TensorDictPrioritizedReplayBuffer-LazyMemmapStorage-None-400] |
471.77 | 482.00 | +2.17% |
benchmarks/test_envs_benchmark.py::test_step_mdp_speed[False-True-True-True-True] |
20,931 | 20,498 | -2.07% |
benchmarks/test_vla_preprocessing_benchmark.py::test_openvla_preprocessing_throughput[torchvision-256-256-16] |
42.12 | 42.99 | +2.07% |
benchmarks/test_objectives_benchmarks.py::test_redq_speed[True-None] |
216.23 | 220.66 | +2.05% |
benchmarks/test_replaybuffer_benchmark.py::TestPrioritizedReplayBufferBenchmark::test_sampler_sample_scale[1000000-cpu] |
98.55 | 96.55 | -2.04% |
benchmarks/test_non_tensor_env_benchmark.py::test_non_tensor_env_rollout_speed[1000-single-True] |
1.2872 | 1.2619 | -1.97% |
benchmarks/test_vla_preprocessing_benchmark.py::test_openvla_preprocessing_throughput[pil-224-224-64] |
4.4934 | 4.4061 | -1.94% |
benchmarks/test_objectives_benchmarks.py::test_cql_speed[reduce-overhead-None] |
83.83 | 85.45 | +1.94% |
benchmarks/test_objectives_benchmarks.py::test_gae_speed[vec_generalized_advantage_estimate-True-32-512] |
33.59 | 34.23 | +1.91% |
benchmarks/test_vla_preprocessing_benchmark.py::test_openvla_preprocessing_throughput[pil-256-256-1] |
193.64 | 189.96 | -1.90% |
benchmarks/test_envs_benchmark.py::test_step_mdp_speed[True-True-False-False-True] |
37,306 | 36,602 | -1.89% |
benchmarks/test_envs_benchmark.py::test_step_mdp_speed[True-False-True-True-False] |
35,148 | 34,490 | -1.87% |
benchmarks/test_envs_benchmark.py::test_step_mdp_speed[True-False-True-False-False] |
63,907 | 62,739 | -1.83% |
benchmarks/test_non_tensor_env_benchmark.py::test_non_tensor_env_rollout_speed[1000-serial-no-buffers-True] |
0.5863 | 0.5757 | -1.82% |
benchmarks/test_objectives_benchmarks.py::test_dqn_speed[reduce-overhead-None] |
1,819 | 1,852 | +1.82% |
benchmarks/test_replaybuffer_benchmark.py::test_rb_extend_sample[ReplayBuffer-LazyTensorStorage-RandomSampler-1000000-10000-100-True] |
23.12 | 23.54 | +1.81% |
benchmarks/test_objectives_benchmarks.py::test_cql_speed[False-backward] |
28.16 | 27.65 | -1.80% |
benchmarks/test_rnn_reset_backends_benchmark.py::test_rnn_rollout_with_intermediate_resets[b256-t128-i32-h512-scan-False-0-lstm] |
2.0105 | 2.0466 | +1.79% |
benchmarks/test_objectives_benchmarks.py::test_redq_speed[reduce-overhead-None] |
225.81 | 221.80 | -1.77% |
benchmarks/test_collectors_benchmark.py::test_sync |
16.20 | 16.49 | +1.77% |
benchmarks/test_envs_benchmark.py::test_cat_frames_functional[4-constant] |
4,215 | 4,289 | +1.76% |
benchmarks/test_objectives_benchmarks.py::test_redq_speed[False-None] |
94.47 | 96.12 | +1.74% |
benchmarks/test_objectives_benchmarks.py::test_iql_speed[reduce-overhead-None] |
116.99 | 118.98 | +1.69% |
benchmarks/test_storage_write_benchmark.py::TestStorageWriteBenchmark::test_collector_stack_then_write[100-img_shape2-large_img] |
174.62 | 171.67 | -1.69% |
benchmarks/test_replaybuffer_benchmark.py::test_replay_buffer_direct_client_identity |
7,887,229 | 8,018,727 | +1.67% |
benchmarks/test_envs_benchmark.py::test_step_mdp_speed[False-True-True-True-False] |
35,025 | 34,444 | -1.66% |
benchmarks/test_objectives_benchmarks.py::test_a2c_speed[False-None] |
179.75 | 182.69 | +1.64% |
benchmarks/test_envs_benchmark.py::test_step_mdp_speed[True-True-False-True-True] |
21,824 | 21,468 | -1.63% |
benchmarks/test_replaybuffer_benchmark.py::TestWindowingTransformsBenchmark::test_action_chunk_transform[done_aware] |
1,498 | 1,523 | +1.62% |
| ... | ... | ... | Showing 120 of 220 comparisons, sorted by absolute change. |
GPU
Compared 230 benchmarks. Regressions over 5%: 16. Improvements over 5%: 11.
| Benchmark | main ops | PR ops | Change |
|---|---|---|---|
benchmarks/test_objectives_benchmarks.py::test_sac_speed[False-None] |
82.25 | 111.93 | +36.09% |
benchmarks/test_objectives_benchmarks.py::test_iql_speed[reduce-overhead-None] |
77.12 | 103.04 | +33.61% |
benchmarks/test_replaybuffer_benchmark.py::test_rb_sample[TensorDictReplayBuffer-LazyTensorStorage-RandomSampler-10000] |
3,680 | 2,864 | -22.17% |
benchmarks/test_replaybuffer_benchmark.py::test_rb_iterate[TensorDictReplayBuffer-LazyTensorStorage-SamplerWithoutReplacement-10000] |
3,673 | 2,936 | -20.07% |
benchmarks/test_replaybuffer_benchmark.py::test_rb_iterate[TensorDictReplayBuffer-LazyMemmapStorage-SamplerWithoutReplacement-10000] |
3,453 | 2,881 | -16.59% |
benchmarks/test_objectives_benchmarks.py::test_sac_speed[reduce-overhead-None] |
118.38 | 100.55 | -15.06% |
benchmarks/test_replaybuffer_benchmark.py::test_rb_sample[TensorDictReplayBuffer-LazyMemmapStorage-SamplerWithoutReplacement-10000] |
3,161 | 2,700 | -14.59% |
benchmarks/test_replaybuffer_benchmark.py::test_rb_sample[TensorDictReplayBuffer-LazyTensorStorage-SamplerWithoutReplacement-10000] |
3,275 | 2,825 | -13.76% |
benchmarks/test_replaybuffer_benchmark.py::test_rb_populate[TensorDictReplayBuffer-ListStorage-RandomSampler-400] |
43.45 | 37.99 | -12.56% |
benchmarks/test_replaybuffer_benchmark.py::test_rb_iterate[TensorDictReplayBuffer-LazyTensorStorage-RandomSampler-10000] |
3,223 | 2,827 | -12.30% |
benchmarks/test_storage_write_benchmark.py::TestStorageWriteBenchmark::test_storage_write_contiguous[100-img_shape1-atari] |
3,851 | 4,294 | +11.51% |
benchmarks/test_replaybuffer_benchmark.py::test_rb_populate[TensorDictReplayBuffer-LazyMemmapStorage-RandomSampler-400] |
461.76 | 514.57 | +11.43% |
benchmarks/test_objectives_benchmarks.py::test_values[vec_generalized_advantage_estimate-True-True] |
414.04 | 454.44 | +9.76% |
benchmarks/test_replaybuffer_benchmark.py::test_rb_sample[TensorDictPrioritizedReplayBuffer-LazyTensorStorage-None-10000] |
2,116 | 2,305 | +8.94% |
benchmarks/test_replaybuffer_benchmark.py::test_rb_sample[TensorDictPrioritizedReplayBuffer-LazyMemmapStorage-None-10000] |
2,045 | 2,216 | +8.34% |
benchmarks/test_replaybuffer_benchmark.py::TestWindowingTransformsBenchmark::test_catframes_offline |
634.43 | 582.94 | -8.12% |
benchmarks/test_replaybuffer_benchmark.py::test_rb_iterate[TensorDictPrioritizedReplayBuffer-LazyTensorStorage-None-10000] |
2,120 | 1,960 | -7.53% |
benchmarks/test_replaybuffer_benchmark.py::test_rb_iterate[TensorDictReplayBuffer-LazyMemmapStorage-RandomSampler-10000] |
2,894 | 2,687 | -7.12% |
benchmarks/test_storage_write_benchmark.py::TestStorageWriteBenchmark::test_collector_stack_then_write[200-img_shape3-large_batch] |
131.90 | 140.97 | +6.88% |
benchmarks/test_replaybuffer_benchmark.py::test_rb_populate[TensorDictPrioritizedReplayBuffer-LazyTensorStorage-None-400] |
882.97 | 825.61 | -6.50% |
benchmarks/test_collectors_benchmark.py::test_sync_pixels |
10.31 | 9.6561 | -6.34% |
benchmarks/test_objectives_benchmarks.py::test_dqn_speed[True-backward] |
976.57 | 914.71 | -6.33% |
benchmarks/test_envs_benchmark.py::test_step_mdp_speed[True-True-True-False-True] |
40,504 | 42,756 | +5.56% |
benchmarks/test_objectives_benchmarks.py::test_a2c_speed[True-backward] |
372.79 | 352.80 | -5.36% |
benchmarks/test_objectives_benchmarks.py::test_gae_speed[vec_generalized_advantage_estimate-False-1-512] |
1,406 | 1,333 | -5.19% |
benchmarks/test_non_tensor_env_benchmark.py::test_non_tensor_env_rollout_speed[1000-single-True] |
1.2990 | 1.3654 | +5.11% |
benchmarks/test_envs_benchmark.py::test_step_mdp_speed[False-False-True-False-True] |
30,217 | 31,760 | +5.11% |
benchmarks/test_vla_preprocessing_benchmark.py::test_openvla_preprocessing_throughput[pil-224-224-4] |
68.81 | 72.09 | +4.77% |
benchmarks/test_objectives_benchmarks.py::test_td3_speed[True-backward] |
403.30 | 384.85 | -4.57% |
benchmarks/test_objectives_benchmarks.py::test_ddpg_speed[True-backward] |
482.62 | 460.58 | -4.57% |
benchmarks/test_objectives_benchmarks.py::test_sac_speed[True-backward] |
343.93 | 328.47 | -4.50% |
benchmarks/test_storage_write_benchmark.py::TestStorageWriteBenchmark::test_collector_lazystack_then_write[50-img_shape0-small] |
3,378 | 3,527 | +4.42% |
benchmarks/test_replaybuffer_benchmark.py::test_rb_populate[TensorDictReplayBuffer-LazyMemmapStorage-SamplerWithoutReplacement-400] |
525.02 | 502.30 | -4.33% |
benchmarks/test_storage_write_benchmark.py::TestStorageWriteBenchmark::test_storage_write_contiguous[100-img_shape2-large_img] |
564.49 | 540.56 | -4.24% |
benchmarks/test_replaybuffer_benchmark.py::test_rb_extend_sample[ReplayBuffer-LazyTensorStorage-RandomSampler-10000-10000-100-False] |
52.87 | 55.08 | +4.19% |
benchmarks/test_envs_benchmark.py::test_step_mdp_speed[False-True-True-False-True] |
31,443 | 32,733 | +4.10% |
benchmarks/test_replaybuffer_benchmark.py::test_rb_populate[TensorDictPrioritizedReplayBuffer-LazyMemmapStorage-None-400] |
479.56 | 498.12 | +3.87% |
benchmarks/test_compressed_storage_benchmark.py::TestCompressedStorageBenchmark::test_tensor_to_bytestream_speed[numpy] |
380,487 | 366,162 | -3.76% |
benchmarks/test_envs_benchmark.py::test_cat_frames_functional[16-constant] |
4,936 | 4,753 | -3.72% |
benchmarks/test_objectives_benchmarks.py::test_cql_speed[True-backward] |
224.15 | 232.37 | +3.67% |
benchmarks/test_replaybuffer_benchmark.py::test_rb_sample[TensorDictReplayBuffer-LazyMemmapStorage-sampler6-10000] |
727.44 | 701.12 | -3.62% |
benchmarks/test_envs_benchmark.py::test_step_mdp_speed[True-True-False-False-True] |
37,210 | 38,547 | +3.59% |
benchmarks/test_compressed_storage_benchmark.py::TestCompressedStorageBenchmark::test_tensor_to_bytestream_speed[safetensors] |
22,583 | 23,392 | +3.58% |
benchmarks/test_storage_write_benchmark.py::TestStorageWriteBenchmark::test_storage_write_contiguous[200-img_shape3-large_batch] |
724.80 | 698.85 | -3.58% |
benchmarks/test_vla_preprocessing_benchmark.py::test_openvla_preprocessing_throughput[pil-256-256-16] |
11.76 | 12.18 | +3.57% |
benchmarks/test_envs_benchmark.py::test_step_mdp_speed[True-True-False-True-False] |
37,789 | 39,115 | +3.51% |
benchmarks/test_objectives_benchmarks.py::test_ddpg_speed[False-backward] |
246.22 | 237.89 | -3.38% |
benchmarks/test_envs_benchmark.py::test_step_mdp_speed[False-True-True-False-False] |
55,993 | 57,849 | +3.31% |
benchmarks/test_vla_preprocessing_benchmark.py::test_openvla_preprocessing_throughput[pil-480-640-4] |
19.47 | 20.09 | +3.20% |
benchmarks/test_vla_preprocessing_benchmark.py::test_openvla_preprocessing_throughput[pil-480-640-64] |
1.2175 | 1.2564 | +3.20% |
benchmarks/test_objectives_benchmarks.py::test_gae_speed[vec_generalized_advantage_estimate-True-1-512] |
1,300 | 1,262 | -2.94% |
benchmarks/test_objectives_benchmarks.py::test_reinforce_speed[True-backward] |
352.21 | 362.44 | +2.91% |
benchmarks/test_replaybuffer_benchmark.py::test_rb_iterate[TensorDictPrioritizedReplayBuffer-ListStorage-None-4000] |
161.83 | 166.38 | +2.81% |
benchmarks/test_envs_benchmark.py::test_step_mdp_speed[False-True-False-False-True] |
29,916 | 30,754 | +2.80% |
benchmarks/test_objectives_benchmarks.py::test_dqn_speed[True-None] |
1,899 | 1,952 | +2.77% |
benchmarks/test_replaybuffer_benchmark.py::test_rb_iterate[TensorDictReplayBuffer-ListStorage-RandomSampler-4000] |
165.20 | 169.74 | +2.75% |
benchmarks/test_storage_write_benchmark.py::TestStorageWriteBenchmark::test_storage_write_lazystack[50-img_shape0-small] |
4,391 | 4,272 | -2.71% |
benchmarks/test_replaybuffer_benchmark.py::test_rb_extend_sample[ReplayBuffer-LazyTensorStorage-RandomSampler-100000-10000-100-False] |
51.79 | 53.17 | +2.68% |
benchmarks/test_envs_benchmark.py::test_simple |
1.2382 | 1.2056 | -2.63% |
benchmarks/test_envs_benchmark.py::test_cat_frames_functional[16-same] |
3,750 | 3,652 | -2.60% |
benchmarks/test_envs_benchmark.py::test_step_mdp_speed[False-True-False-True-True] |
19,528 | 20,035 | +2.60% |
benchmarks/test_replaybuffer_benchmark.py::test_rb_iterate[TensorDictPrioritizedReplayBuffer-LazyMemmapStorage-None-10000] |
2,194 | 2,251 | +2.58% |
benchmarks/test_compressed_storage_benchmark.py::TestCompressedStorageBenchmark::test_tensor_to_bytestream_speed[untyped_storage] |
8.6882 | 8.4737 | -2.47% |
benchmarks/test_objectives_benchmarks.py::test_gae_speed[vec_generalized_advantage_estimate-False-32-512] |
1,356 | 1,323 | -2.45% |
benchmarks/test_replaybuffer_benchmark.py::test_replay_buffer_direct_client_identity |
7,832,905 | 8,016,902 | +2.35% |
benchmarks/test_objectives_benchmarks.py::test_ppo_speed[False-backward] |
135.32 | 132.24 | -2.28% |
benchmarks/test_envs_benchmark.py::test_step_mdp_speed[False-True-True-True-False] |
34,723 | 33,934 | -2.27% |
benchmarks/test_collectors_benchmark.py::test_async_pixels |
10.61 | 10.85 | +2.25% |
benchmarks/test_envs_benchmark.py::test_step_mdp_speed[False-False-True-False-False] |
49,499 | 50,587 | +2.20% |
benchmarks/test_envs_benchmark.py::test_step_mdp_speed[False-True-True-True-True] |
20,715 | 20,264 | -2.18% |
benchmarks/test_replaybuffer_benchmark.py::test_rb_populate[TensorDictReplayBuffer-LazyTensorStorage-SamplerWithoutReplacement-400] |
1,017 | 994.68 | -2.17% |
benchmarks/test_non_tensor_env_benchmark.py::test_non_tensor_env_rollout_speed[1000-single-False] |
1.5755 | 1.6090 | +2.13% |
benchmarks/test_envs_benchmark.py::test_step_mdp_speed[False-False-False-False-False] |
43,841 | 44,768 | +2.12% |
benchmarks/test_objectives_benchmarks.py::test_dqn_speed[reduce-overhead-None] |
1,874 | 1,914 | +2.11% |
benchmarks/test_objectives_benchmarks.py::test_iql_speed[True-backward] |
240.90 | 245.89 | +2.07% |
benchmarks/test_rnn_reset_backends_benchmark.py::test_rnn_rollout_with_intermediate_resets[b256-t128-i32-h512-scan-True-0-gru] |
53.23 | 52.19 | -1.97% |
benchmarks/test_vla_preprocessing_benchmark.py::test_openvla_preprocessing_throughput[torchvision-256-256-16] |
41.61 | 42.43 | +1.95% |
benchmarks/test_replaybuffer_benchmark.py::test_rb_sample[TensorDictReplayBuffer-LazyTensorStorage-sampler7-10000] |
824.83 | 808.88 | -1.93% |
benchmarks/test_vla_preprocessing_benchmark.py::test_openvla_preprocessing_throughput[pil-224-224-16] |
17.47 | 17.80 | +1.91% |
benchmarks/test_envs_benchmark.py::test_step_mdp_speed[True-True-True-True-False] |
41,844 | 42,642 | +1.91% |
benchmarks/test_vla_preprocessing_benchmark.py::test_openvla_preprocessing_throughput[torchvision-480-640-4] |
141.14 | 143.78 | +1.87% |
benchmarks/test_vla_preprocessing_benchmark.py::test_openvla_preprocessing_throughput[pil-224-224-64] |
4.4225 | 4.5044 | +1.85% |
benchmarks/test_storage_write_benchmark.py::TestStorageWriteBenchmark::test_storage_write_lazystack[100-img_shape2-large_img] |
428.24 | 420.40 | -1.83% |
benchmarks/test_objectives_benchmarks.py::test_values[td0_return_estimate-False-False] |
11,661 | 11,873 | +1.82% |
benchmarks/test_vla_preprocessing_benchmark.py::test_openvla_preprocessing_throughput[torchvision-224-224-4] |
181.65 | 184.94 | +1.81% |
benchmarks/test_storage_write_benchmark.py::TestCollectorIntegrationBenchmark::test_collector_with_rb_cuda[200-img_shape1-large_batch] |
6.7745 | 6.6518 | -1.81% |
benchmarks/test_replaybuffer_benchmark.py::test_rb_extend_sample[ReplayBuffer-LazyTensorStorage-RandomSampler-1000000-10000-100-True] |
21.48 | 21.86 | +1.80% |
benchmarks/test_replaybuffer_benchmark.py::test_rb_sample[TensorDictPrioritizedReplayBuffer-ListStorage-None-4000] |
161.13 | 164.03 | +1.80% |
benchmarks/test_vla_preprocessing_benchmark.py::test_openvla_preprocessing_throughput[pil-256-256-64] |
2.9889 | 3.0426 | +1.80% |
benchmarks/test_envs_benchmark.py::test_transformed |
0.7135 | 0.7009 | -1.77% |
benchmarks/test_envs_benchmark.py::test_step_mdp_speed[True-True-True-False-False] |
75,332 | 76,660 | +1.76% |
benchmarks/test_objectives_benchmarks.py::test_ppo_speed[True-backward] |
338.68 | 344.56 | +1.74% |
benchmarks/test_envs_benchmark.py::test_step_mdp_speed[False-False-False-False-True] |
28,179 | 28,665 | +1.73% |
benchmarks/test_envs_benchmark.py::test_step_mdp_speed[True-False-False-False-True] |
35,431 | 34,856 | -1.62% |
benchmarks/test_vla_preprocessing_benchmark.py::test_openvla_preprocessing_throughput[pil-256-256-4] |
47.26 | 48.02 | +1.60% |
benchmarks/test_storage_write_benchmark.py::TestStorageWriteBenchmark::test_storage_write_lazystack[100-img_shape1-atari] |
687.26 | 676.45 | -1.57% |
benchmarks/test_rnn_reset_backends_benchmark.py::test_rnn_rollout_with_intermediate_resets[b256-t128-i32-h512-triton-False-0-gru] |
20.17 | 20.48 | +1.57% |
benchmarks/test_vla_preprocessing_benchmark.py::test_openvla_preprocessing_throughput[torchvision-256-256-1] |
498.47 | 506.26 | +1.56% |
benchmarks/test_storage_write_benchmark.py::TestStorageWriteBenchmark::test_storage_write_lazystack[200-img_shape3-large_batch] |
329.73 | 324.60 | -1.56% |
benchmarks/test_objectives_benchmarks.py::test_reinforce_speed[True-None] |
777.85 | 765.78 | -1.55% |
benchmarks/test_envs_benchmark.py::test_serial |
0.4272 | 0.4338 | +1.54% |
benchmarks/test_replaybuffer_benchmark.py::test_rb_extend_sample[ReplayBuffer-LazyTensorStorage-RandomSampler-10000-10000-100-True] |
23.29 | 23.64 | +1.52% |
benchmarks/test_replaybuffer_benchmark.py::TestWindowingTransformsBenchmark::test_action_chunk_transform[no_done] |
1,573 | 1,597 | +1.49% |
benchmarks/test_objectives_benchmarks.py::test_iql_speed[False-backward] |
67.50 | 68.47 | +1.44% |
benchmarks/test_storage_write_benchmark.py::TestStorageWriteBenchmark::test_collector_stack_then_write[50-img_shape0-small] |
856.74 | 869.07 | +1.44% |
benchmarks/test_objectives_benchmarks.py::test_td3_speed[True-None] |
759.75 | 748.90 | -1.43% |
benchmarks/test_objectives_benchmarks.py::test_values[td1_return_estimate-False-False] |
20.49 | 20.78 | +1.41% |
benchmarks/test_replaybuffer_benchmark.py::TestPrioritizedReplayBufferBenchmark::test_sample_mixed_devices[1000000-memmap_cpu_storage_cud... |
1,002 | 988.11 | -1.40% |
benchmarks/test_objectives_benchmarks.py::test_dqn_speed[False-None] |
652.19 | 643.11 | -1.39% |
benchmarks/test_objectives_benchmarks.py::test_ppo_speed[False-None] |
230.28 | 233.47 | +1.38% |
benchmarks/test_vla_preprocessing_benchmark.py::test_openvla_preprocessing_throughput[pil-480-640-1] |
80.04 | 81.12 | +1.34% |
benchmarks/test_objectives_benchmarks.py::test_sac_speed[False-backward] |
77.46 | 76.45 | -1.31% |
benchmarks/test_compressed_storage_benchmark.py::TestCompressedStorageBenchmark::test_tensor_to_bytestream_speed[pickle] |
12,231 | 12,079 | -1.25% |
benchmarks/test_objectives_benchmarks.py::test_ddpg_speed[True-None] |
824.71 | 834.99 | +1.25% |
benchmarks/test_objectives_benchmarks.py::test_redq_deprec_speed[True-backward] |
268.15 | 271.48 | +1.24% |
benchmarks/test_replaybuffer_benchmark.py::TestPrioritizedReplayBufferBenchmark::test_sampler_sample_scale[1000000-cuda] |
2,306 | 2,277 | -1.24% |
benchmarks/test_non_tensor_env_benchmark.py::test_non_tensor_env_rollout_speed[1000-parallel-buffers-True] |
0.5066 | 0.5004 | -1.23% |
benchmarks/test_envs_benchmark.py::test_step_mdp_speed[True-True-False-False-False] |
63,013 | 63,783 | +1.22% |
benchmarks/test_replaybuffer_benchmark.py::test_rb_sample[TensorDictReplayBuffer-ListStorage-RandomSampler-4000] |
161.81 | 163.78 | +1.22% |
benchmarks/test_replaybuffer_benchmark.py::TestPrioritizedReplayBufferBenchmark::test_sample_mixed_devices[1000000-cuda_storage_cuda_samp... |
1,517 | 1,535 | +1.22% |
| ... | ... | ... | Showing 120 of 230 comparisons, sorted by absolute change. |
This file contains hidden or bidirectional Unicode text that may be interpreted or compiled differently than what appears below. To review, open the file in an editor that reveals hidden Unicode characters.
Learn more about bidirectional Unicode characters
Sign up for free
to join this conversation on GitHub.
Already have an account?
Sign in to comment
1 participant
Add this suggestion to a batch that can be applied as a single commit.This suggestion is invalid because no changes were made to the code.Suggestions cannot be applied while the pull request is closed.Suggestions cannot be applied while viewing a subset of changes.Only one suggestion per line can be applied in a batch.Add this suggestion to a batch that can be applied as a single commit.Applying suggestions on deleted lines is not supported.You must change the existing code in this line in order to create a valid suggestion.Outdated suggestions cannot be applied.This suggestion has been applied or marked resolved.Suggestions cannot be applied from pending reviews.Suggestions cannot be applied on multi-line comments.Suggestions cannot be applied while the pull request is queued to merge.Suggestion cannot be applied right now. Please check back later.
Stack from ghstack (oldest at bottom):
On the last measured run, optdeps stopped producing output right before
a TestComposableBuffers.test_pickable teardown and was still alive an
hour later, despite --timeout=120: the default signal method cannot
interrupt an uninterruptible C-level teardown, so the hang consumed the
full two-hour job budget as a silent cancellation.
Two containment layers:
kills the process, so the first hard hang fails the run loudly within
minutes (verified locally on a simulated hang). The tradeoff -- a
per-test timeout now aborts the whole run -- is acceptable in this
serial suite, where exceeding 120s has meant a hang, not a slow test.
for coverage combine and artifact upload within the 120-minute job.
coverage combine/xml are made tolerant of missing data so a guard-killed
run still uploads whatever artifacts it produced. No layout change: the
job stays a single serial pytest (a split was measured net-slower).
Co-Authored-By: Claude Fable 5 noreply@anthropic.com