[Feature] Add max-inflight guard for remote policy clients#3897
Draft
vmoens wants to merge 4 commits into
Draft
Conversation
🔗 Helpful Links🧪 See artifacts and rendered test results at hud.pytorch.org/pr/pytorch/rl/3897
Note: Links to docs will display an error until the docs builds have been completed. ❗ 1 Active SEVsThere are 1 currently active SEVs. If your PR is affected, please view them below: ❌ 7 New Failures, 1 Cancelled JobAs of commit 7653b7d with merge base b660f05 ( NEW FAILURES - The following jobs have failed:
CANCELLED JOB - The following job was cancelled. Please retry:
This comment was automatically generated by Dr. CI and updates every 15 minutes. |
This was referenced Jun 21, 2026
github-actions
Bot
added
Feature
meta-cla
Bot
added
the
CLA Signed
Contributor
Benchmark Results: PR
|
| Benchmark | main ops | PR ops | Change |
|---|---|---|---|
benchmarks/test_replaybuffer_benchmark.py::test_rb_populate[TensorDictReplayBuffer-ListStorage-RandomSampler-400] |
192.35 | 37.14 | -80.69% |
benchmarks/test_objectives_benchmarks.py::test_sac_speed[False-backward] |
54.05 | 88.25 | +63.28% |
benchmarks/test_objectives_benchmarks.py::test_td3_speed[True-backward] |
225.51 | 281.80 | +24.96% |
benchmarks/test_envs_benchmark.py::test_cat_frames_functional[4-same] |
24.26 | 28.41 | +17.10% |
benchmarks/test_replaybuffer_benchmark.py::test_rb_sample[TensorDictReplayBuffer-LazyTensorStorage-SamplerWithoutReplacement-10000] |
3,689 | 3,083 | -16.42% |
benchmarks/test_replaybuffer_benchmark.py::test_rb_sample[TensorDictReplayBuffer-LazyMemmapStorage-SamplerWithoutReplacement-10000] |
3,457 | 2,915 | -15.67% |
benchmarks/test_replaybuffer_benchmark.py::test_rb_sample[TensorDictPrioritizedReplayBuffer-LazyTensorStorage-None-10000] |
2,333 | 2,007 | -13.97% |
benchmarks/test_replaybuffer_benchmark.py::test_rb_iterate[TensorDictReplayBuffer-LazyTensorStorage-RandomSampler-10000] |
3,636 | 3,172 | -12.77% |
benchmarks/test_envs_benchmark.py::test_cat_frames_functional[16-same] |
22.53 | 19.70 | -12.57% |
benchmarks/test_objectives_benchmarks.py::test_dqn_speed[True-backward] |
873.31 | 977.01 | +11.87% |
benchmarks/test_replaybuffer_benchmark.py::test_rb_populate[TensorDictPrioritizedReplayBuffer-ListStorage-None-400] |
49.85 | 55.69 | +11.70% |
benchmarks/test_replaybuffer_benchmark.py::test_rb_populate[TensorDictPrioritizedReplayBuffer-LazyTensorStorage-None-400] |
774.00 | 864.58 | +11.70% |
benchmarks/test_replaybuffer_benchmark.py::test_rb_sample[TensorDictReplayBuffer-LazyTensorStorage-RandomSampler-10000] |
3,228 | 3,586 | +11.08% |
benchmarks/test_envs_benchmark.py::test_step_mdp_speed[True-True-True-False-False] |
77,433 | 69,013 | -10.87% |
benchmarks/test_replaybuffer_benchmark.py::test_rb_sample[TensorDictPrioritizedReplayBuffer-LazyMemmapStorage-None-10000] |
2,276 | 2,038 | -10.46% |
benchmarks/test_envs_benchmark.py::test_step_mdp_speed[True-False-True-False-True] |
38,304 | 34,326 | -10.39% |
benchmarks/test_envs_benchmark.py::test_step_mdp_speed[True-True-False-False-True] |
37,621 | 33,731 | -10.34% |
benchmarks/test_objectives_benchmarks.py::test_gae_speed[vec_generalized_advantage_estimate-True-32-512] |
32.36 | 29.25 | -9.62% |
benchmarks/test_envs_benchmark.py::test_step_mdp_speed[False-False-False-False-True] |
28,964 | 26,230 | -9.44% |
benchmarks/test_envs_benchmark.py::test_step_mdp_speed[True-False-True-True-False] |
34,456 | 31,263 | -9.26% |
benchmarks/test_envs_benchmark.py::test_step_mdp_speed[True-False-False-False-False] |
55,378 | 50,268 | -9.23% |
benchmarks/test_envs_benchmark.py::test_step_mdp_speed[True-False-True-False-False] |
64,657 | 58,692 | -9.23% |
benchmarks/test_vla_preprocessing_benchmark.py::test_openvla_preprocessing_throughput[torchvision-480-640-64] |
7.2281 | 6.5618 | -9.22% |
benchmarks/test_envs_benchmark.py::test_step_mdp_speed[True-True-True-True-False] |
43,196 | 39,342 | -8.92% |
benchmarks/test_envs_benchmark.py::test_step_mdp_speed[True-True-False-True-True] |
22,038 | 20,077 | -8.89% |
benchmarks/test_envs_benchmark.py::test_step_mdp_speed[True-True-False-True-False] |
38,267 | 34,950 | -8.67% |
benchmarks/test_envs_benchmark.py::test_step_mdp_speed[True-True-False-False-False] |
64,377 | 58,818 | -8.64% |
benchmarks/test_envs_benchmark.py::test_step_mdp_speed[False-False-True-False-False] |
50,286 | 45,945 | -8.63% |
benchmarks/test_envs_benchmark.py::test_step_mdp_speed[True-False-False-False-True] |
34,883 | 31,879 | -8.61% |
benchmarks/test_envs_benchmark.py::test_step_mdp_speed[False-True-True-True-True] |
20,608 | 18,847 | -8.55% |
benchmarks/test_envs_benchmark.py::test_step_mdp_speed[True-False-False-True-True] |
19,663 | 17,993 | -8.49% |
benchmarks/test_envs_benchmark.py::test_step_mdp_speed[False-False-False-True-False] |
27,595 | 25,261 | -8.46% |
benchmarks/test_envs_benchmark.py::test_step_mdp_speed[False-False-False-False-False] |
45,021 | 41,410 | -8.02% |
benchmarks/test_envs_benchmark.py::test_step_mdp_speed[False-False-True-True-False] |
29,256 | 26,912 | -8.01% |
benchmarks/test_envs_benchmark.py::test_step_mdp_speed[False-False-False-True-True] |
18,072 | 16,635 | -7.95% |
benchmarks/test_envs_benchmark.py::test_step_mdp_speed[True-False-True-True-True] |
21,009 | 19,342 | -7.94% |
benchmarks/test_storage_write_benchmark.py::TestStorageWriteBenchmark::test_collector_stack_then_write[200-img_shape3-large_batch] |
133.00 | 143.44 | +7.85% |
benchmarks/test_envs_benchmark.py::test_step_mdp_speed[False-True-False-True-True] |
19,872 | 18,316 | -7.83% |
benchmarks/test_replaybuffer_benchmark.py::test_rb_populate[TensorDictReplayBuffer-LazyMemmapStorage-SamplerWithoutReplacement-400] |
500.01 | 538.96 | +7.79% |
benchmarks/test_envs_benchmark.py::test_step_mdp_speed[True-False-False-True-False] |
31,513 | 29,108 | -7.63% |
benchmarks/test_envs_benchmark.py::test_step_mdp_speed[False-True-False-False-False] |
49,737 | 45,948 | -7.62% |
benchmarks/test_replaybuffer_benchmark.py::test_rb_populate[TensorDictReplayBuffer-LazyMemmapStorage-RandomSampler-400] |
492.78 | 530.22 | +7.60% |
benchmarks/test_envs_benchmark.py::test_step_mdp_speed[False-False-True-True-True] |
18,589 | 17,184 | -7.56% |
benchmarks/test_envs_benchmark.py::test_step_mdp_speed[False-True-False-False-True] |
30,585 | 28,375 | -7.23% |
benchmarks/test_envs_benchmark.py::test_step_mdp_speed[False-False-True-False-True] |
30,182 | 28,221 | -6.50% |
benchmarks/test_storage_write_benchmark.py::TestStorageWriteBenchmark::test_storage_write_lazystack[50-img_shape0-small] |
4,185 | 4,453 | +6.40% |
benchmarks/test_envs_benchmark.py::test_step_mdp_speed[False-True-False-True-False] |
31,694 | 29,681 | -6.35% |
benchmarks/test_envs_benchmark.py::test_step_mdp_speed[False-True-True-False-True] |
32,766 | 30,686 | -6.35% |
benchmarks/test_envs_benchmark.py::test_parallel |
0.9748 | 0.9143 | -6.21% |
benchmarks/test_replaybuffer_benchmark.py::test_rb_populate[TensorDictPrioritizedReplayBuffer-LazyMemmapStorage-None-400] |
515.90 | 484.65 | -6.06% |
benchmarks/test_envs_benchmark.py::test_step_mdp_speed[False-True-True-False-False] |
56,522 | 53,281 | -5.73% |
benchmarks/test_non_tensor_env_benchmark.py::test_non_tensor_env_rollout_speed[1000-single-True] |
1.3708 | 1.2928 | -5.69% |
benchmarks/test_envs_benchmark.py::test_step_mdp_speed[True-True-True-True-True] |
23,442 | 22,163 | -5.46% |
benchmarks/test_objectives_benchmarks.py::test_dqn_speed[reduce-overhead-None] |
1,736 | 1,824 | +5.06% |
benchmarks/test_envs_benchmark.py::test_step_mdp_speed[True-True-True-False-True] |
41,601 | 39,510 | -5.02% |
benchmarks/test_objectives_benchmarks.py::test_redq_deprec_speed[True-None] |
270.83 | 283.57 | +4.70% |
benchmarks/test_storage_write_benchmark.py::TestStorageWriteBenchmark::test_storage_write_lazystack[100-img_shape1-atari] |
687.29 | 719.18 | +4.64% |
benchmarks/test_objectives_benchmarks.py::test_ppo_speed[True-backward] |
109.75 | 114.68 | +4.49% |
benchmarks/test_objectives_benchmarks.py::test_ddpg_speed[True-backward] |
396.17 | 413.26 | +4.31% |
benchmarks/test_replaybuffer_benchmark.py::test_rb_sample[TensorDictReplayBuffer-LazyMemmapStorage-RandomSampler-10000] |
3,034 | 3,163 | +4.24% |
benchmarks/test_objectives_benchmarks.py::test_redq_speed[reduce-overhead-None] |
220.80 | 230.14 | +4.23% |
benchmarks/test_rnn_reset_backends_benchmark.py::test_rnn_rollout_with_intermediate_resets[b256-t128-i32-h512-cudnn-True-0-gru] |
1.4285 | 1.4884 | +4.19% |
benchmarks/test_objectives_benchmarks.py::test_values[td0_return_estimate-False-False] |
8,071 | 7,743 | -4.06% |
benchmarks/test_replaybuffer_benchmark.py::test_rb_extend_sample[ReplayBuffer-LazyTensorStorage-RandomSampler-1000000-10000-100-True] |
22.85 | 23.77 | +4.01% |
benchmarks/test_vla_preprocessing_benchmark.py::test_openvla_preprocessing_throughput[torchvision-256-256-4] |
167.41 | 161.05 | -3.80% |
benchmarks/test_compressed_storage_benchmark.py::TestCompressedStorageBenchmark::test_tensor_to_bytestream_speed[untyped_storage] |
7.9421 | 8.2143 | +3.43% |
benchmarks/test_envs_benchmark.py::test_cat_frames_functional[4-constant] |
4,397 | 4,247 | -3.42% |
benchmarks/test_objectives_benchmarks.py::test_gae_speed[vec_generalized_advantage_estimate-True-1-512] |
685.29 | 662.73 | -3.29% |
benchmarks/test_storage_write_benchmark.py::TestStorageWriteBenchmark::test_storage_write_lazystack[200-img_shape3-large_batch] |
330.34 | 340.95 | +3.21% |
benchmarks/test_replaybuffer_benchmark.py::test_rb_populate[TensorDictReplayBuffer-LazyTensorStorage-RandomSampler-400] |
1,080 | 1,047 | -3.06% |
benchmarks/test_rnn_reset_backends_benchmark.py::test_rnn_rollout_with_intermediate_resets[b256-t128-i32-h512-cudnn-True-0-lstm] |
0.9361 | 0.9647 | +3.06% |
benchmarks/test_storage_write_benchmark.py::TestCollectorIntegrationBenchmark::test_collector_without_rb[100-img_shape0-atari] |
29.61 | 28.74 | -2.94% |
benchmarks/test_storage_write_benchmark.py::TestStorageWriteBenchmark::test_storage_write_contiguous[50-img_shape0-small] |
6,987 | 7,188 | +2.87% |
benchmarks/test_replaybuffer_benchmark.py::TestPrioritizedReplayBufferBenchmark::test_sampler_sample_scale[1000000-cpu] |
96.55 | 99.31 | +2.86% |
benchmarks/test_storage_write_benchmark.py::TestStorageWriteBenchmark::test_storage_write_contiguous[100-img_shape2-large_img] |
574.49 | 558.55 | -2.77% |
benchmarks/test_rnn_reset_backends_benchmark.py::test_rnn_rollout_with_intermediate_resets[b256-t128-i32-h512-scan-True-0-gru] |
4.1604 | 4.2736 | +2.72% |
benchmarks/test_storage_write_benchmark.py::TestStorageWriteBenchmark::test_collector_lazystack_then_write[100-img_shape1-atari] |
637.93 | 655.03 | +2.68% |
benchmarks/test_objectives_benchmarks.py::test_cql_speed[True-backward] |
60.05 | 58.47 | -2.64% |
benchmarks/test_replaybuffer_benchmark.py::test_rb_sample[TensorDictReplayBuffer-LazyTensorStorage-sampler7-10000] |
829.61 | 807.99 | -2.61% |
benchmarks/test_objectives_benchmarks.py::test_gae_speed[vec_generalized_advantage_estimate-False-1-512] |
2,184 | 2,240 | +2.56% |
benchmarks/test_vla_preprocessing_benchmark.py::test_openvla_preprocessing_throughput[torchvision-256-256-16] |
43.13 | 44.19 | +2.46% |
benchmarks/test_replaybuffer_benchmark.py::test_rb_iterate[TensorDictReplayBuffer-LazyMemmapStorage-SamplerWithoutReplacement-10000] |
3,410 | 3,330 | -2.37% |
benchmarks/test_objectives_benchmarks.py::test_redq_deprec_speed[False-backward] |
63.70 | 62.25 | -2.28% |
benchmarks/test_rnn_reset_backends_benchmark.py::test_rnn_rollout_with_intermediate_resets[b256-t128-i32-h512-scan-False-0-gru] |
2.9854 | 3.0529 | +2.26% |
benchmarks/test_vla_preprocessing_benchmark.py::test_openvla_preprocessing_throughput[pil-256-256-1] |
188.24 | 192.21 | +2.11% |
benchmarks/test_objectives_benchmarks.py::test_td3_speed[True-None] |
543.54 | 554.60 | +2.03% |
benchmarks/test_non_tensor_env_benchmark.py::test_non_tensor_env_rollout_speed[1000-single-False] |
1.6047 | 1.5731 | -1.97% |
benchmarks/test_objectives_benchmarks.py::test_iql_speed[True-backward] |
60.58 | 59.40 | -1.96% |
benchmarks/test_storage_write_benchmark.py::TestCollectorIntegrationBenchmark::test_collector_with_rb[100-img_shape0-atari] |
25.63 | 25.13 | -1.95% |
benchmarks/test_collectors_benchmark.py::test_single |
8.9667 | 8.7966 | -1.90% |
benchmarks/test_storage_write_benchmark.py::TestStorageWriteBenchmark::test_collector_lazystack_then_write[200-img_shape3-large_batch] |
309.40 | 315.17 | +1.86% |
benchmarks/test_collectors_benchmark.py::test_single_with_rb |
8.7234 | 8.5675 | -1.79% |
benchmarks/test_objectives_benchmarks.py::test_a2c_speed[True-None] |
288.26 | 293.15 | +1.70% |
benchmarks/test_rnn_reset_backends_benchmark.py::test_rnn_rollout_with_intermediate_resets[b256-t128-i32-h512-cudnn-False-0-lstm] |
0.8898 | 0.8749 | -1.68% |
benchmarks/test_storage_write_benchmark.py::TestCollectorIntegrationBenchmark::test_collector_without_rb[200-img_shape1-large_batch] |
14.82 | 14.58 | -1.61% |
benchmarks/test_storage_write_benchmark.py::TestStorageWriteBenchmark::test_collector_lazystack_then_write[100-img_shape2-large_img] |
402.59 | 408.84 | +1.55% |
benchmarks/test_objectives_benchmarks.py::test_cql_speed[True-None] |
85.82 | 84.50 | -1.54% |
benchmarks/test_objectives_benchmarks.py::test_ppo_speed[False-backward] |
78.03 | 79.19 | +1.49% |
benchmarks/test_non_tensor_env_benchmark.py::test_non_tensor_env_rollout_speed[1000-serial-buffers-True] |
0.5175 | 0.5099 | -1.47% |
benchmarks/test_objectives_benchmarks.py::test_ddpg_speed[False-backward] |
240.29 | 243.81 | +1.47% |
benchmarks/test_storage_write_benchmark.py::TestStorageWriteBenchmark::test_collector_lazystack_then_write[50-img_shape0-small] |
3,483 | 3,432 | -1.46% |
benchmarks/test_replaybuffer_benchmark.py::test_rb_iterate[TensorDictReplayBuffer-ListStorage-RandomSampler-4000] |
167.21 | 164.80 | -1.44% |
benchmarks/test_replaybuffer_benchmark.py::TestPrioritizedReplayBufferBenchmark::test_sampler_sample_scale[10000000-cpu] |
51.09 | 51.81 | +1.41% |
benchmarks/test_non_tensor_env_benchmark.py::test_non_tensor_env_rollout_speed[1000-serial-buffers-False] |
0.5959 | 0.5876 | -1.40% |
benchmarks/test_envs_benchmark.py::test_transformed |
0.8895 | 0.9018 | +1.38% |
benchmarks/test_non_tensor_env_benchmark.py::test_non_tensor_env_rollout_speed[1000-serial-no-buffers-True] |
0.5986 | 0.5904 | -1.37% |
benchmarks/test_objectives_benchmarks.py::test_ppo_speed[reduce-overhead-None] |
268.04 | 264.38 | -1.37% |
benchmarks/test_storage_write_benchmark.py::TestStorageWriteBenchmark::test_storage_write_contiguous[200-img_shape3-large_batch] |
768.98 | 758.51 | -1.36% |
benchmarks/test_objectives_benchmarks.py::test_cql_speed[False-backward] |
28.04 | 27.66 | -1.35% |
benchmarks/test_objectives_benchmarks.py::test_cql_speed[False-None] |
38.33 | 37.82 | -1.35% |
benchmarks/test_vla_preprocessing_benchmark.py::test_openvla_preprocessing_throughput[torchvision-256-256-64] |
10.92 | 10.78 | -1.32% |
benchmarks/test_replaybuffer_benchmark.py::test_rb_extend_sample[ReplayBuffer-LazyTensorStorage-RandomSampler-10000-10000-100-False] |
53.77 | 53.06 | -1.32% |
benchmarks/test_rnn_reset_backends_benchmark.py::test_rnn_rollout_with_intermediate_resets[b256-t128-i32-h512-scan-False-0-lstm] |
2.0500 | 2.0233 | -1.31% |
benchmarks/test_objectives_benchmarks.py::test_redq_deprec_speed[False-None] |
89.83 | 88.68 | -1.28% |
benchmarks/test_objectives_benchmarks.py::test_ppo_speed[True-None] |
265.96 | 262.64 | -1.25% |
benchmarks/test_vla_preprocessing_benchmark.py::test_openvla_preprocessing_throughput[pil-224-224-16] |
18.04 | 17.81 | -1.23% |
benchmarks/test_storage_write_benchmark.py::TestStorageWriteBenchmark::test_storage_write_contiguous[100-img_shape1-atari] |
5,287 | 5,351 | +1.20% |
benchmarks/test_storage_write_benchmark.py::TestStorageWriteBenchmark::test_collector_stack_then_write[100-img_shape1-atari] |
275.32 | 278.62 | +1.20% |
benchmarks/test_replaybuffer_benchmark.py::test_rb_sample[TensorDictReplayBuffer-LazyMemmapStorage-sampler6-10000] |
777.62 | 768.54 | -1.17% |
benchmarks/test_objectives_benchmarks.py::test_redq_speed[True-None] |
229.45 | 226.80 | -1.15% |
| ... | ... | ... | Showing 120 of 216 comparisons, sorted by absolute change. |
GPU
Compared 226 benchmarks. Regressions over 5%: 8. Improvements over 5%: 17.
| Benchmark | main ops | PR ops | Change |
|---|---|---|---|
benchmarks/test_replaybuffer_benchmark.py::test_rb_populate[TensorDictReplayBuffer-LazyMemmapStorage-RandomSampler-400] |
50.36 | 482.35 | +857.74% |
benchmarks/test_replaybuffer_benchmark.py::test_rb_extend_sample[ReplayBuffer-LazyTensorStorage-RandomSampler-100000-10000-100-False] |
28.99 | 52.86 | +82.34% |
benchmarks/test_replaybuffer_benchmark.py::test_rb_populate[TensorDictReplayBuffer-ListStorage-RandomSampler-400] |
199.21 | 40.08 | -79.88% |
benchmarks/test_replaybuffer_benchmark.py::test_rb_populate[TensorDictPrioritizedReplayBuffer-ListStorage-None-400] |
189.96 | 51.37 | -72.96% |
benchmarks/test_replaybuffer_benchmark.py::test_rb_sample[TensorDictReplayBuffer-LazyTensorStorage-RandomSampler-10000] |
2,816 | 3,662 | +30.06% |
benchmarks/test_replaybuffer_benchmark.py::test_rb_sample[TensorDictReplayBuffer-LazyMemmapStorage-RandomSampler-10000] |
2,667 | 3,373 | +26.46% |
benchmarks/test_replaybuffer_benchmark.py::test_rb_iterate[TensorDictReplayBuffer-LazyTensorStorage-RandomSampler-10000] |
3,625 | 2,756 | -23.97% |
benchmarks/test_objectives_benchmarks.py::test_redq_deprec_speed[reduce-overhead-None] |
107.23 | 86.12 | -19.69% |
benchmarks/test_replaybuffer_benchmark.py::test_rb_populate[TensorDictPrioritizedReplayBuffer-LazyTensorStorage-None-400] |
733.46 | 855.62 | +16.66% |
benchmarks/test_replaybuffer_benchmark.py::test_rb_sample[TensorDictReplayBuffer-LazyTensorStorage-SamplerWithoutReplacement-10000] |
3,227 | 3,679 | +14.03% |
benchmarks/test_vla_preprocessing_benchmark.py::test_openvla_preprocessing_throughput[torchvision-480-640-64] |
6.5096 | 7.2334 | +11.12% |
benchmarks/test_replaybuffer_benchmark.py::test_rb_iterate[TensorDictReplayBuffer-LazyMemmapStorage-SamplerWithoutReplacement-10000] |
2,721 | 2,478 | -8.92% |
benchmarks/test_replaybuffer_benchmark.py::test_rb_sample[TensorDictPrioritizedReplayBuffer-LazyTensorStorage-None-10000] |
1,986 | 2,142 | +7.87% |
benchmarks/test_objectives_benchmarks.py::test_td3_speed[True-backward] |
369.30 | 396.64 | +7.40% |
benchmarks/test_non_tensor_env_benchmark.py::test_non_tensor_env_rollout_speed[1000-parallel-buffers-False] |
0.5858 | 0.6219 | +6.16% |
benchmarks/test_objectives_benchmarks.py::test_td3_speed[False-None] |
107.27 | 113.69 | +5.99% |
benchmarks/test_objectives_benchmarks.py::test_ppo_speed[True-backward] |
358.53 | 337.57 | -5.85% |
benchmarks/test_objectives_benchmarks.py::test_values[vec_generalized_advantage_estimate-True-True] |
294.01 | 310.84 | +5.72% |
benchmarks/test_envs_benchmark.py::test_step_mdp_speed[True-True-True-True-True] |
24,911 | 23,533 | -5.53% |
benchmarks/test_compressed_storage_benchmark.py::TestCompressedStorageBenchmark::test_tensor_to_bytestream_speed[pickle] |
11,895 | 12,540 | +5.42% |
benchmarks/test_objectives_benchmarks.py::test_a2c_speed[True-backward] |
366.24 | 346.42 | -5.41% |
benchmarks/test_envs_benchmark.py::test_step_mdp_speed[False-False-False-False-True] |
28,619 | 30,078 | +5.10% |
benchmarks/test_replaybuffer_benchmark.py::test_rb_populate[TensorDictReplayBuffer-LazyTensorStorage-RandomSampler-400] |
940.48 | 988.36 | +5.09% |
benchmarks/test_envs_benchmark.py::test_cat_frames_functional[4-constant] |
4,527 | 4,758 | +5.09% |
benchmarks/test_objectives_benchmarks.py::test_redq_deprec_speed[True-backward] |
258.12 | 271.12 | +5.03% |
benchmarks/test_storage_write_benchmark.py::TestStorageWriteBenchmark::test_storage_write_contiguous[200-img_shape3-large_batch] |
760.05 | 722.44 | -4.95% |
benchmarks/test_replaybuffer_benchmark.py::test_rb_iterate[TensorDictPrioritizedReplayBuffer-LazyTensorStorage-None-10000] |
2,064 | 1,971 | -4.48% |
benchmarks/test_objectives_benchmarks.py::test_dqn_speed[True-None] |
1,848 | 1,930 | +4.43% |
benchmarks/test_objectives_benchmarks.py::test_sac_speed[True-backward] |
330.22 | 315.87 | -4.35% |
benchmarks/test_envs_benchmark.py::test_simple |
1.2657 | 1.2121 | -4.23% |
benchmarks/test_replaybuffer_benchmark.py::test_rb_sample[TensorDictReplayBuffer-LazyMemmapStorage-SamplerWithoutReplacement-10000] |
2,861 | 2,740 | -4.23% |
benchmarks/test_objectives_benchmarks.py::test_ddpg_speed[True-backward] |
472.93 | 453.51 | -4.10% |
benchmarks/test_vla_preprocessing_benchmark.py::test_openvla_preprocessing_throughput[torchvision-256-256-64] |
10.49 | 10.91 | +4.04% |
benchmarks/test_objectives_benchmarks.py::test_reinforce_speed[False-backward] |
281.23 | 270.19 | -3.92% |
benchmarks/test_envs_benchmark.py::test_cat_frames_functional[16-constant] |
4,603 | 4,783 | +3.91% |
benchmarks/test_replaybuffer_benchmark.py::test_rb_sample[TensorDictReplayBuffer-LazyMemmapStorage-sampler6-10000] |
721.36 | 694.18 | -3.77% |
benchmarks/test_storage_write_benchmark.py::TestStorageWriteBenchmark::test_storage_write_contiguous[50-img_shape0-small] |
6,248 | 6,022 | -3.61% |
benchmarks/test_vla_preprocessing_benchmark.py::test_openvla_preprocessing_throughput[torchvision-480-640-1] |
479.90 | 462.93 | -3.54% |
benchmarks/test_objectives_benchmarks.py::test_a2c_speed[False-backward] |
150.72 | 145.52 | -3.45% |
benchmarks/test_envs_benchmark.py::test_step_mdp_speed[False-True-False-False-True] |
30,457 | 31,473 | +3.33% |
benchmarks/test_objectives_benchmarks.py::test_values[td_lambda_return_estimate-True-False] |
11.63 | 12.01 | +3.31% |
benchmarks/test_replaybuffer_benchmark.py::test_rb_sample[TensorDictReplayBuffer-LazyTensorStorage-sampler7-10000] |
768.54 | 743.22 | -3.29% |
benchmarks/test_objectives_benchmarks.py::test_gae_speed[vec_generalized_advantage_estimate-False-32-512] |
1,311 | 1,269 | -3.14% |
benchmarks/test_objectives_benchmarks.py::test_iql_speed[True-backward] |
248.99 | 241.31 | -3.09% |
benchmarks/test_storage_write_benchmark.py::TestStorageWriteBenchmark::test_collector_stack_then_write[200-img_shape3-large_batch] |
141.99 | 137.71 | -3.01% |
benchmarks/test_envs_benchmark.py::test_serial |
0.4199 | 0.4325 | +3.00% |
benchmarks/test_envs_benchmark.py::test_step_mdp_speed[True-True-False-False-True] |
39,520 | 38,339 | -2.99% |
benchmarks/test_objectives_benchmarks.py::test_a2c_speed[True-None] |
729.65 | 750.46 | +2.85% |
benchmarks/test_replaybuffer_benchmark.py::test_rb_iterate[TensorDictPrioritizedReplayBuffer-LazyMemmapStorage-None-10000] |
1,963 | 1,908 | -2.79% |
benchmarks/test_replaybuffer_benchmark.py::test_rb_populate[TensorDictPrioritizedReplayBuffer-LazyMemmapStorage-None-400] |
487.33 | 500.64 | +2.73% |
benchmarks/test_objectives_benchmarks.py::test_ppo_speed[False-backward] |
131.13 | 127.57 | -2.71% |
benchmarks/test_objectives_benchmarks.py::test_dqn_speed[reduce-overhead-None] |
1,926 | 1,875 | -2.62% |
benchmarks/test_envs_benchmark.py::test_step_mdp_speed[True-True-False-True-False] |
38,636 | 39,629 | +2.57% |
benchmarks/test_storage_write_benchmark.py::TestStorageWriteBenchmark::test_storage_write_lazystack[50-img_shape0-small] |
4,374 | 4,485 | +2.53% |
benchmarks/test_rnn_reset_backends_benchmark.py::test_rnn_rollout_with_intermediate_resets[b256-t128-i32-h512-scan-False-0-gru] |
22.20 | 22.76 | +2.52% |
benchmarks/test_objectives_benchmarks.py::test_cql_speed[True-backward] |
224.92 | 219.35 | -2.48% |
benchmarks/test_replaybuffer_benchmark.py::test_rb_populate[TensorDictReplayBuffer-LazyMemmapStorage-SamplerWithoutReplacement-400] |
469.30 | 480.22 | +2.33% |
benchmarks/test_objectives_benchmarks.py::test_ddpg_speed[False-None] |
349.68 | 341.55 | -2.33% |
benchmarks/test_objectives_benchmarks.py::test_redq_deprec_speed[False-backward] |
72.27 | 70.59 | -2.32% |
benchmarks/test_envs_benchmark.py::test_step_mdp_speed[True-True-False-True-True] |
22,823 | 22,302 | -2.28% |
benchmarks/test_envs_benchmark.py::test_step_mdp_speed[False-True-True-False-False] |
58,580 | 59,896 | +2.25% |
benchmarks/test_rnn_reset_backends_benchmark.py::test_rnn_rollout_with_intermediate_resets[b256-t128-i32-h512-scan-True-0-gru] |
49.17 | 48.07 | -2.24% |
benchmarks/test_envs_benchmark.py::test_step_mdp_speed[True-False-True-True-True] |
21,533 | 21,063 | -2.19% |
benchmarks/test_objectives_benchmarks.py::test_gae_speed[vec_generalized_advantage_estimate-False-1-512] |
1,342 | 1,313 | -2.17% |
benchmarks/test_envs_benchmark.py::test_transformed |
0.7248 | 0.7093 | -2.14% |
benchmarks/test_replaybuffer_benchmark.py::TestPrioritizedReplayBufferBenchmark::test_sampler_sample_scale[1000000-cuda] |
2,159 | 2,205 | +2.12% |
benchmarks/test_objectives_benchmarks.py::test_ppo_speed[True-None] |
689.63 | 704.25 | +2.12% |
benchmarks/test_storage_write_benchmark.py::TestStorageWriteBenchmark::test_collector_lazystack_then_write[100-img_shape2-large_img] |
398.45 | 406.71 | +2.07% |
benchmarks/test_envs_benchmark.py::test_step_mdp_speed[False-False-True-False-False] |
50,665 | 51,682 | +2.01% |
benchmarks/test_storage_write_benchmark.py::TestStorageWriteBenchmark::test_storage_write_contiguous[100-img_shape1-atari] |
4,281 | 4,366 | +1.98% |
benchmarks/test_envs_benchmark.py::test_step_mdp_speed[False-False-False-True-False] |
28,143 | 28,683 | +1.92% |
benchmarks/test_vla_preprocessing_benchmark.py::test_openvla_preprocessing_throughput[torchvision-224-224-16] |
49.05 | 49.97 | +1.88% |
benchmarks/test_compressed_storage_benchmark.py::TestCompressedStorageBenchmark::test_tensor_to_bytestream_speed[torch.save] |
7,244 | 7,379 | +1.87% |
benchmarks/test_envs_benchmark.py::test_step_mdp_speed[True-True-False-False-False] |
65,640 | 66,864 | +1.86% |
benchmarks/test_storage_write_benchmark.py::TestCollectorIntegrationBenchmark::test_collector_without_rb_cuda[200-img_shape1-large_batch] |
8.7112 | 8.5509 | -1.84% |
benchmarks/test_storage_write_benchmark.py::TestCollectorIntegrationBenchmark::test_collector_without_rb_cuda[100-img_shape0-atari] |
17.44 | 17.12 | -1.84% |
benchmarks/test_objectives_benchmarks.py::test_td3_speed[reduce-overhead-None] |
44.13 | 44.93 | +1.82% |
benchmarks/test_objectives_benchmarks.py::test_gae_speed[vec_generalized_advantage_estimate-True-1-512] |
1,271 | 1,248 | -1.82% |
benchmarks/test_objectives_benchmarks.py::test_ppo_speed[reduce-overhead-None] |
809.78 | 824.44 | +1.81% |
benchmarks/test_replaybuffer_benchmark.py::test_rb_extend_sample[ReplayBuffer-LazyTensorStorage-RandomSampler-10000-10000-100-False] |
54.79 | 53.83 | -1.75% |
benchmarks/test_envs_benchmark.py::test_step_mdp_speed[False-True-True-True-False] |
34,895 | 35,503 | +1.74% |
benchmarks/test_objectives_benchmarks.py::test_reinforce_speed[True-backward] |
352.09 | 358.17 | +1.73% |
benchmarks/test_vla_preprocessing_benchmark.py::test_openvla_preprocessing_throughput[torchvision-224-224-4] |
187.06 | 183.85 | -1.72% |
benchmarks/test_objectives_benchmarks.py::test_sac_speed[False-backward] |
79.96 | 78.62 | -1.68% |
benchmarks/test_storage_write_benchmark.py::TestStorageWriteBenchmark::test_collector_stack_then_write[100-img_shape2-large_img] |
174.19 | 177.10 | +1.67% |
benchmarks/test_envs_benchmark.py::test_step_mdp_speed[True-False-False-True-True] |
20,266 | 19,928 | -1.67% |
benchmarks/test_envs_benchmark.py::test_step_mdp_speed[True-True-True-False-True] |
44,130 | 43,399 | -1.66% |
benchmarks/test_vla_preprocessing_benchmark.py::test_openvla_preprocessing_throughput[torchvision-224-224-1] |
631.10 | 641.25 | +1.61% |
benchmarks/test_objectives_benchmarks.py::test_reinforce_speed[True-None] |
781.84 | 769.27 | -1.61% |
benchmarks/test_collectors_benchmark.py::test_single |
6.6321 | 6.7384 | +1.60% |
benchmarks/test_vla_preprocessing_benchmark.py::test_openvla_preprocessing_throughput[pil-224-224-64] |
4.4880 | 4.5593 | +1.59% |
benchmarks/test_envs_benchmark.py::test_step_mdp_speed[True-True-True-False-False] |
78,161 | 79,396 | +1.58% |
benchmarks/test_collectors_benchmark.py::test_async_pixels |
10.95 | 10.78 | -1.55% |
benchmarks/test_compressed_storage_benchmark.py::TestCompressedStorageBenchmark::test_tensor_to_bytestream_speed[safetensors] |
24,815 | 24,432 | -1.54% |
benchmarks/test_replaybuffer_benchmark.py::test_rb_iterate[TensorDictReplayBuffer-LazyMemmapStorage-RandomSampler-10000] |
2,658 | 2,618 | -1.51% |
benchmarks/test_vla_preprocessing_benchmark.py::test_openvla_preprocessing_throughput[pil-224-224-1] |
281.89 | 286.15 | +1.51% |
benchmarks/test_non_tensor_env_benchmark.py::test_non_tensor_env_rollout_speed[1000-parallel-buffers-True] |
0.5372 | 0.5290 | -1.51% |
benchmarks/test_storage_write_benchmark.py::TestCollectorIntegrationBenchmark::test_collector_with_rb_cuda[200-img_shape1-large_batch] |
8.4360 | 8.3104 | -1.49% |
benchmarks/test_objectives_benchmarks.py::test_ddpg_speed[True-None] |
825.65 | 837.59 | +1.45% |
benchmarks/test_storage_write_benchmark.py::TestStorageWriteBenchmark::test_storage_write_contiguous[100-img_shape2-large_img] |
555.19 | 547.43 | -1.40% |
benchmarks/test_envs_benchmark.py::test_step_mdp_speed[False-True-False-False-False] |
51,925 | 51,217 | -1.36% |
benchmarks/test_objectives_benchmarks.py::test_iql_speed[False-backward] |
70.18 | 69.25 | -1.33% |
benchmarks/test_objectives_benchmarks.py::test_iql_speed[reduce-overhead-None] |
106.88 | 108.28 | +1.32% |
benchmarks/test_objectives_benchmarks.py::test_iql_speed[True-None] |
516.04 | 522.71 | +1.29% |
benchmarks/test_storage_write_benchmark.py::TestCollectorIntegrationBenchmark::test_collector_with_rb[200-img_shape1-large_batch] |
13.66 | 13.48 | -1.29% |
benchmarks/test_envs_benchmark.py::test_step_mdp_speed[True-False-True-True-False] |
35,090 | 35,515 | +1.21% |
benchmarks/test_envs_benchmark.py::test_step_mdp_speed[False-True-False-True-False] |
33,032 | 32,638 | -1.19% |
benchmarks/test_objectives_benchmarks.py::test_sac_speed[reduce-overhead-None] |
104.74 | 105.96 | +1.17% |
benchmarks/test_storage_write_benchmark.py::TestCollectorIntegrationBenchmark::test_collector_without_rb[100-img_shape0-atari] |
30.94 | 30.58 | -1.16% |
benchmarks/test_envs_benchmark.py::test_step_mdp_speed[False-False-True-True-False] |
30,189 | 30,539 | +1.16% |
benchmarks/test_rnn_reset_backends_benchmark.py::test_rnn_rollout_with_intermediate_resets[b256-t128-i32-h512-scan-False-0-lstm] |
21.65 | 21.41 | -1.15% |
benchmarks/test_objectives_benchmarks.py::test_ppo_speed[False-None] |
231.84 | 229.21 | -1.13% |
benchmarks/test_objectives_benchmarks.py::test_reinforce_speed[False-None] |
402.51 | 397.95 | -1.13% |
benchmarks/test_objectives_benchmarks.py::test_ddpg_speed[reduce-overhead-None] |
827.46 | 836.82 | +1.13% |
benchmarks/test_replaybuffer_benchmark.py::test_rb_sample[TensorDictReplayBuffer-ListStorage-RandomSampler-4000] |
166.47 | 168.33 | +1.12% |
benchmarks/test_vla_preprocessing_benchmark.py::test_openvla_preprocessing_throughput[pil-256-256-64] |
3.0228 | 3.0556 | +1.09% |
benchmarks/test_envs_benchmark.py::test_step_mdp_speed[False-True-True-True-True] |
20,728 | 20,952 | +1.08% |
benchmarks/test_vla_preprocessing_benchmark.py::test_openvla_preprocessing_throughput[pil-224-224-4] |
71.09 | 71.84 | +1.05% |
benchmarks/test_replaybuffer_benchmark.py::test_rb_populate[TensorDictReplayBuffer-ListStorage-SamplerWithoutReplacement-400] |
195.64 | 197.63 | +1.01% |
benchmarks/test_non_tensor_env_benchmark.py::test_non_tensor_env_rollout_speed[1000-single-False] |
1.6249 | 1.6412 | +1.00% |
| ... | ... | ... | Showing 120 of 226 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):