[Feature] Add VLA token policy and tracking primitives#3874
Merged
Conversation
🔗 Helpful Links🧪 See artifacts and rendered test results at hud.pytorch.org/pr/pytorch/rl/3874
Note: Links to docs will display an error until the docs builds have been completed. ✅ No FailuresAs of commit 7fadab0 with merge base 6364a19 ( This comment was automatically generated by Dr. CI and updates every 15 minutes. |
meta-cla
Bot
added
the
CLA Signed
github-actions
Bot
added
Documentation
This was referenced Jun 17, 2026
Contributor
Benchmark Results: PR
|
| Benchmark | main ops | PR ops | Change |
|---|---|---|---|
benchmarks/test_replaybuffer_benchmark.py::test_rb_iterate[TensorDictReplayBuffer-LazyTensorStorage-SamplerWithoutReplacement-10000] |
3,038 | 3,717 | +22.34% |
benchmarks/test_objectives_benchmarks.py::test_dqn_speed[True-backward] |
828.73 | 1,000 | +20.71% |
benchmarks/test_objectives_benchmarks.py::test_redq_deprec_speed[True-backward] |
121.42 | 140.30 | +15.55% |
benchmarks/test_replaybuffer_benchmark.py::test_rb_populate[TensorDictReplayBuffer-ListStorage-SamplerWithoutReplacement-400] |
47.81 | 55.12 | +15.29% |
benchmarks/test_replaybuffer_benchmark.py::test_rb_sample[TensorDictReplayBuffer-LazyTensorStorage-RandomSampler-10000] |
3,037 | 3,491 | +14.95% |
benchmarks/test_replaybuffer_benchmark.py::test_rb_populate[TensorDictReplayBuffer-LazyMemmapStorage-RandomSampler-400] |
480.46 | 544.31 | +13.29% |
benchmarks/test_envs_benchmark.py::test_cat_frames_functional[16-same] |
23.63 | 20.54 | -13.09% |
benchmarks/test_objectives_benchmarks.py::test_ddpg_speed[True-backward] |
372.76 | 420.70 | +12.86% |
benchmarks/test_replaybuffer_benchmark.py::test_rb_sample[TensorDictReplayBuffer-LazyMemmapStorage-RandomSampler-10000] |
3,370 | 2,952 | -12.40% |
benchmarks/test_replaybuffer_benchmark.py::test_rb_iterate[TensorDictPrioritizedReplayBuffer-LazyMemmapStorage-None-10000] |
1,967 | 2,204 | +12.08% |
benchmarks/test_replaybuffer_benchmark.py::test_rb_iterate[TensorDictPrioritizedReplayBuffer-LazyTensorStorage-None-10000] |
2,071 | 2,290 | +10.60% |
benchmarks/test_objectives_benchmarks.py::test_redq_deprec_speed[reduce-overhead-None] |
260.37 | 284.88 | +9.41% |
benchmarks/test_replaybuffer_benchmark.py::test_rb_iterate[TensorDictReplayBuffer-LazyTensorStorage-RandomSampler-10000] |
3,414 | 3,720 | +8.98% |
benchmarks/test_replaybuffer_benchmark.py::test_rb_sample[TensorDictReplayBuffer-LazyTensorStorage-sampler7-10000] |
841.88 | 768.83 | -8.68% |
benchmarks/test_objectives_benchmarks.py::test_values[td0_return_estimate-False-False] |
8,006 | 7,341 | -8.31% |
benchmarks/test_replaybuffer_benchmark.py::test_rb_sample[TensorDictReplayBuffer-LazyMemmapStorage-SamplerWithoutReplacement-10000] |
3,449 | 3,176 | -7.92% |
benchmarks/test_replaybuffer_benchmark.py::test_rb_iterate[TensorDictReplayBuffer-LazyMemmapStorage-RandomSampler-10000] |
3,094 | 3,310 | +6.98% |
benchmarks/test_storage_write_benchmark.py::TestStorageWriteBenchmark::test_storage_write_lazystack[100-img_shape2-large_img] |
392.87 | 419.30 | +6.73% |
benchmarks/test_objectives_benchmarks.py::test_redq_speed[reduce-overhead-None] |
213.02 | 226.71 | +6.43% |
benchmarks/test_storage_write_benchmark.py::TestStorageWriteBenchmark::test_collector_lazystack_then_write[100-img_shape2-large_img] |
375.03 | 398.96 | +6.38% |
benchmarks/test_replaybuffer_benchmark.py::test_rb_populate[TensorDictReplayBuffer-LazyMemmapStorage-SamplerWithoutReplacement-400] |
536.28 | 503.63 | -6.09% |
benchmarks/test_compressed_storage_benchmark.py::TestCompressedStorageBenchmark::test_tensor_to_bytestream_speed[pickle] |
12,539 | 11,848 | -5.51% |
benchmarks/test_storage_write_benchmark.py::TestStorageWriteBenchmark::test_collector_lazystack_then_write[100-img_shape1-atari] |
617.89 | 651.75 | +5.48% |
benchmarks/test_replaybuffer_benchmark.py::test_rb_sample[TensorDictReplayBuffer-LazyMemmapStorage-sampler6-10000] |
776.62 | 735.10 | -5.35% |
benchmarks/test_storage_write_benchmark.py::TestStorageWriteBenchmark::test_storage_write_contiguous[100-img_shape1-atari] |
5,083 | 4,812 | -5.33% |
benchmarks/test_rnn_reset_backends_benchmark.py::test_rnn_rollout_with_intermediate_resets[b256-t128-i32-h512-cudnn-False-0-gru] |
1.3485 | 1.2781 | -5.22% |
benchmarks/test_objectives_benchmarks.py::test_reinforce_speed[True-backward] |
119.70 | 125.89 | +5.17% |
benchmarks/test_replaybuffer_benchmark.py::test_rb_populate[TensorDictPrioritizedReplayBuffer-LazyMemmapStorage-None-400] |
480.47 | 455.87 | -5.12% |
benchmarks/test_rnn_reset_backends_benchmark.py::test_rnn_rollout_with_intermediate_resets[b256-t128-i32-h512-cudnn-True-0-gru] |
1.4491 | 1.3771 | -4.97% |
benchmarks/test_non_tensor_env_benchmark.py::test_non_tensor_env_rollout_speed[1000-single-True] |
1.4136 | 1.3441 | -4.92% |
benchmarks/test_objectives_benchmarks.py::test_sac_speed[True-backward] |
237.79 | 248.31 | +4.42% |
benchmarks/test_compressed_storage_benchmark.py::TestCompressedStorageBenchmark::test_tensor_to_bytestream_speed[torch.save] |
7,306 | 6,986 | -4.38% |
benchmarks/test_compressed_storage_benchmark.py::TestCompressedStorageBenchmark::test_tensor_to_bytestream_speed[safetensors] |
24,331 | 23,293 | -4.27% |
benchmarks/test_envs_benchmark.py::test_step_mdp_speed[True-True-True-True-True] |
24,744 | 23,748 | -4.02% |
benchmarks/test_envs_benchmark.py::test_step_mdp_speed[True-True-False-True-True] |
23,464 | 22,524 | -4.00% |
benchmarks/test_storage_write_benchmark.py::TestStorageWriteBenchmark::test_collector_lazystack_then_write[50-img_shape0-small] |
3,432 | 3,568 | +3.98% |
benchmarks/test_objectives_benchmarks.py::test_redq_speed[True-None] |
217.65 | 226.29 | +3.97% |
benchmarks/test_storage_write_benchmark.py::TestStorageWriteBenchmark::test_storage_write_lazystack[100-img_shape1-atari] |
679.85 | 705.80 | +3.82% |
benchmarks/test_envs_benchmark.py::test_transformed |
0.9116 | 0.9451 | +3.67% |
benchmarks/test_envs_benchmark.py::test_cat_frames_functional[4-constant] |
4,361 | 4,202 | -3.64% |
benchmarks/test_replaybuffer_benchmark.py::test_rb_extend_sample[ReplayBuffer-LazyTensorStorage-RandomSampler-10000-10000-100-False] |
55.76 | 53.78 | -3.56% |
benchmarks/test_rnn_reset_backends_benchmark.py::test_rnn_rollout_with_intermediate_resets[b256-t128-i32-h512-cudnn-False-0-lstm] |
0.8513 | 0.8219 | -3.45% |
benchmarks/test_storage_write_benchmark.py::TestCollectorIntegrationBenchmark::test_collector_without_rb[100-img_shape0-atari] |
29.57 | 30.59 | +3.43% |
benchmarks/test_storage_write_benchmark.py::TestStorageWriteBenchmark::test_collector_stack_then_write[200-img_shape3-large_batch] |
141.41 | 136.60 | -3.40% |
benchmarks/test_rnn_reset_backends_benchmark.py::test_rnn_rollout_with_intermediate_resets[b256-t128-i32-h512-scan-False-0-lstm] |
1.9307 | 1.9950 | +3.33% |
benchmarks/test_objectives_benchmarks.py::test_dqn_speed[reduce-overhead-None] |
1,793 | 1,853 | +3.32% |
benchmarks/test_envs_benchmark.py::test_step_mdp_speed[False-True-False-False-True] |
30,886 | 31,890 | +3.25% |
benchmarks/test_objectives_benchmarks.py::test_gae_speed[vec_generalized_advantage_estimate-True-1-512] |
671.69 | 649.99 | -3.23% |
benchmarks/test_objectives_benchmarks.py::test_td3_speed[True-None] |
535.45 | 552.71 | +3.22% |
benchmarks/test_objectives_benchmarks.py::test_ddpg_speed[False-None] |
341.13 | 352.08 | +3.21% |
benchmarks/test_objectives_benchmarks.py::test_cql_speed[True-backward] |
59.21 | 57.35 | -3.14% |
benchmarks/test_objectives_benchmarks.py::test_redq_speed[False-backward] |
55.67 | 53.94 | -3.11% |
benchmarks/test_envs_benchmark.py::test_step_mdp_speed[True-True-True-False-True] |
43,791 | 42,442 | -3.08% |
benchmarks/test_envs_benchmark.py::test_step_mdp_speed[True-True-False-False-True] |
39,445 | 38,246 | -3.04% |
benchmarks/test_objectives_benchmarks.py::test_values[generalized_advantage_estimate-True-True] |
100.62 | 97.61 | -3.00% |
benchmarks/test_replaybuffer_benchmark.py::test_rb_extend_sample[ReplayBuffer-LazyTensorStorage-RandomSampler-100000-10000-100-True] |
24.57 | 25.30 | +2.98% |
benchmarks/test_storage_write_benchmark.py::TestStorageWriteBenchmark::test_storage_write_lazystack[50-img_shape0-small] |
4,225 | 4,350 | +2.95% |
benchmarks/test_storage_write_benchmark.py::TestStorageWriteBenchmark::test_collector_stack_then_write[100-img_shape1-atari] |
275.12 | 283.17 | +2.93% |
benchmarks/test_replaybuffer_benchmark.py::test_rb_extend_sample[ReplayBuffer-LazyTensorStorage-RandomSampler-1000000-10000-100-True] |
23.46 | 24.14 | +2.92% |
benchmarks/test_storage_write_benchmark.py::TestStorageWriteBenchmark::test_collector_stack_then_write[100-img_shape2-large_img] |
176.97 | 171.89 | -2.87% |
benchmarks/test_envs_benchmark.py::test_step_mdp_speed[True-False-True-False-True] |
39,736 | 38,596 | -2.87% |
benchmarks/test_envs_benchmark.py::test_step_mdp_speed[True-False-True-True-True] |
21,806 | 21,184 | -2.85% |
benchmarks/test_objectives_benchmarks.py::test_cql_speed[reduce-overhead-None] |
82.26 | 84.57 | +2.81% |
benchmarks/test_objectives_benchmarks.py::test_gae_speed[vec_generalized_advantage_estimate-True-32-512] |
29.19 | 28.39 | -2.74% |
benchmarks/test_envs_benchmark.py::test_parallel |
0.9985 | 0.9730 | -2.55% |
benchmarks/test_storage_write_benchmark.py::TestStorageWriteBenchmark::test_storage_write_lazystack[200-img_shape3-large_batch] |
322.61 | 330.46 | +2.43% |
benchmarks/test_envs_benchmark.py::test_step_mdp_speed[False-False-False-False-True] |
28,735 | 29,429 | +2.42% |
benchmarks/test_objectives_benchmarks.py::test_iql_speed[True-None] |
113.04 | 115.77 | +2.42% |
benchmarks/test_envs_benchmark.py::test_step_mdp_speed[False-True-True-True-True] |
20,934 | 21,438 | +2.41% |
benchmarks/test_replaybuffer_benchmark.py::test_rb_iterate[TensorDictReplayBuffer-LazyMemmapStorage-SamplerWithoutReplacement-10000] |
3,326 | 3,406 | +2.40% |
benchmarks/test_collectors_benchmark.py::test_single |
9.1193 | 9.3326 | +2.34% |
benchmarks/test_storage_write_benchmark.py::TestStorageWriteBenchmark::test_collector_lazystack_then_write[200-img_shape3-large_batch] |
304.04 | 311.14 | +2.33% |
benchmarks/test_non_tensor_env_benchmark.py::test_non_tensor_env_rollout_speed[1000-serial-buffers-True] |
0.5181 | 0.5300 | +2.30% |
benchmarks/test_objectives_benchmarks.py::test_sac_speed[reduce-overhead-None] |
468.78 | 479.47 | +2.28% |
benchmarks/test_objectives_benchmarks.py::test_td3_speed[False-None] |
122.46 | 125.26 | +2.28% |
benchmarks/test_envs_benchmark.py::test_serial |
0.5865 | 0.5997 | +2.26% |
benchmarks/test_envs_benchmark.py::test_step_mdp_speed[True-False-False-True-True] |
20,495 | 20,035 | -2.24% |
benchmarks/test_storage_write_benchmark.py::TestCollectorIntegrationBenchmark::test_collector_with_rb[100-img_shape0-atari] |
26.36 | 26.94 | +2.23% |
benchmarks/test_storage_write_benchmark.py::TestStorageWriteBenchmark::test_storage_write_contiguous[200-img_shape3-large_batch] |
753.88 | 737.14 | -2.22% |
benchmarks/test_envs_benchmark.py::test_step_mdp_speed[False-True-False-True-True] |
20,444 | 19,993 | -2.21% |
benchmarks/test_replaybuffer_benchmark.py::test_rb_iterate[TensorDictPrioritizedReplayBuffer-ListStorage-None-4000] |
168.05 | 171.69 | +2.16% |
benchmarks/test_replaybuffer_benchmark.py::test_rb_iterate[TensorDictReplayBuffer-ListStorage-RandomSampler-4000] |
172.41 | 175.93 | +2.04% |
benchmarks/test_rnn_reset_backends_benchmark.py::test_rnn_rollout_with_intermediate_resets[b256-t128-i32-h512-scan-True-0-lstm] |
3.0091 | 3.0697 | +2.01% |
benchmarks/test_objectives_benchmarks.py::test_a2c_speed[False-backward] |
81.81 | 83.45 | +2.01% |
benchmarks/test_objectives_benchmarks.py::test_values[td1_return_estimate-False-False] |
38.56 | 37.79 | -2.00% |
benchmarks/test_collectors_benchmark.py::test_single_with_rb |
8.8145 | 8.9880 | +1.97% |
benchmarks/test_objectives_benchmarks.py::test_ppo_speed[False-backward] |
76.55 | 78.05 | +1.96% |
benchmarks/test_replaybuffer_benchmark.py::test_rb_iterate[TensorDictReplayBuffer-ListStorage-SamplerWithoutReplacement-4000] |
173.90 | 177.14 | +1.86% |
benchmarks/test_objectives_benchmarks.py::test_redq_deprec_speed[True-None] |
274.04 | 279.03 | +1.82% |
benchmarks/test_replaybuffer_benchmark.py::test_rb_sample[TensorDictReplayBuffer-ListStorage-SamplerWithoutReplacement-4000] |
170.54 | 173.60 | +1.79% |
benchmarks/test_envs_benchmark.py::test_step_mdp_speed[True-False-True-True-False] |
36,367 | 35,719 | -1.78% |
benchmarks/test_replaybuffer_benchmark.py::test_rb_populate[TensorDictReplayBuffer-ListStorage-RandomSampler-400] |
196.80 | 200.27 | +1.76% |
benchmarks/test_envs_benchmark.py::test_step_mdp_speed[True-True-True-True-False] |
43,867 | 43,097 | -1.76% |
benchmarks/test_objectives_benchmarks.py::test_iql_speed[False-None] |
49.12 | 49.98 | +1.75% |
benchmarks/test_objectives_benchmarks.py::test_cql_speed[False-None] |
37.49 | 38.13 | +1.72% |
benchmarks/test_objectives_benchmarks.py::test_ddpg_speed[True-None] |
697.70 | 709.69 | +1.72% |
benchmarks/test_storage_write_benchmark.py::TestStorageWriteBenchmark::test_storage_write_contiguous[100-img_shape2-large_img] |
562.52 | 552.88 | -1.71% |
benchmarks/test_objectives_benchmarks.py::test_ppo_speed[False-None] |
161.30 | 164.06 | +1.71% |
benchmarks/test_compressed_storage_benchmark.py::TestCompressedStorageBenchmark::test_tensor_to_bytestream_speed[numpy] |
375,070 | 381,394 | +1.69% |
benchmarks/test_non_tensor_env_benchmark.py::test_non_tensor_env_rollout_speed[1000-serial-no-buffers-False] |
0.6920 | 0.7032 | +1.63% |
benchmarks/test_objectives_benchmarks.py::test_reinforce_speed[False-None] |
209.07 | 212.46 | +1.62% |
benchmarks/test_replaybuffer_benchmark.py::test_rb_populate[TensorDictPrioritizedReplayBuffer-ListStorage-None-400] |
188.30 | 191.28 | +1.58% |
benchmarks/test_objectives_benchmarks.py::test_td3_speed[False-backward] |
90.84 | 92.28 | +1.58% |
benchmarks/test_envs_benchmark.py::test_step_mdp_speed[False-True-True-False-True] |
33,313 | 33,831 | +1.55% |
benchmarks/test_objectives_benchmarks.py::test_a2c_speed[False-None] |
176.03 | 178.73 | +1.54% |
benchmarks/test_objectives_benchmarks.py::test_values[td_lambda_return_estimate-True-False] |
25.92 | 25.53 | -1.53% |
benchmarks/test_envs_benchmark.py::test_step_mdp_speed[True-False-True-False-False] |
66,154 | 65,147 | -1.52% |
benchmarks/test_rnn_reset_backends_benchmark.py::test_rnn_rollout_with_intermediate_resets[b256-t128-i32-h512-scan-True-0-gru] |
4.1149 | 4.1771 | +1.51% |
benchmarks/test_objectives_benchmarks.py::test_ppo_speed[True-None] |
258.67 | 262.57 | +1.51% |
benchmarks/test_objectives_benchmarks.py::test_dqn_speed[True-None] |
1,758 | 1,785 | +1.49% |
benchmarks/test_objectives_benchmarks.py::test_cql_speed[False-backward] |
27.31 | 27.72 | +1.49% |
benchmarks/test_objectives_benchmarks.py::test_td3_speed[reduce-overhead-None] |
557.25 | 565.51 | +1.48% |
benchmarks/test_objectives_benchmarks.py::test_td3_speed[True-backward] |
272.22 | 276.13 | +1.44% |
benchmarks/test_objectives_benchmarks.py::test_ppo_speed[True-backward] |
112.27 | 113.86 | +1.41% |
benchmarks/test_non_tensor_env_benchmark.py::test_non_tensor_env_rollout_speed[1000-serial-buffers-False] |
0.6033 | 0.6118 | +1.40% |
benchmarks/test_non_tensor_env_benchmark.py::test_non_tensor_env_rollout_speed[1000-parallel-buffers-False] |
0.6070 | 0.6154 | +1.38% |
benchmarks/test_objectives_benchmarks.py::test_iql_speed[True-backward] |
59.51 | 58.69 | -1.37% |
benchmarks/test_envs_benchmark.py::test_step_mdp_speed[False-False-False-False-False] |
44,848 | 45,461 | +1.37% |
benchmarks/test_envs_benchmark.py::test_step_mdp_speed[True-False-False-False-True] |
35,542 | 35,061 | -1.35% |
benchmarks/test_envs_benchmark.py::test_step_mdp_speed[True-True-True-False-False] |
79,281 | 80,352 | +1.35% |
| ... | ... | ... | Showing 120 of 192 comparisons, sorted by absolute change. |
GPU
Compared 202 benchmarks. Regressions over 5%: 10. Improvements over 5%: 11.
| Benchmark | main ops | PR ops | Change |
|---|---|---|---|
benchmarks/test_replaybuffer_benchmark.py::test_rb_populate[TensorDictReplayBuffer-ListStorage-RandomSampler-400] |
44.53 | 187.36 | +320.71% |
benchmarks/test_replaybuffer_benchmark.py::test_rb_populate[TensorDictReplayBuffer-ListStorage-SamplerWithoutReplacement-400] |
196.50 | 46.75 | -76.21% |
benchmarks/test_objectives_benchmarks.py::test_iql_speed[reduce-overhead-None] |
78.11 | 102.50 | +31.22% |
benchmarks/test_replaybuffer_benchmark.py::test_rb_iterate[TensorDictReplayBuffer-LazyTensorStorage-SamplerWithoutReplacement-10000] |
2,948 | 3,784 | +28.35% |
benchmarks/test_replaybuffer_benchmark.py::test_rb_sample[TensorDictReplayBuffer-LazyTensorStorage-SamplerWithoutReplacement-10000] |
2,472 | 3,070 | +24.19% |
benchmarks/test_storage_write_benchmark.py::TestStorageWriteBenchmark::test_storage_write_contiguous[100-img_shape1-atari] |
3,993 | 4,585 | +14.84% |
benchmarks/test_replaybuffer_benchmark.py::test_rb_sample[TensorDictReplayBuffer-LazyTensorStorage-sampler7-10000] |
717.26 | 800.30 | +11.58% |
benchmarks/test_collectors_benchmark.py::test_single_with_rb_pixels |
5.3832 | 4.7612 | -11.56% |
benchmarks/test_objectives_benchmarks.py::test_reinforce_speed[True-backward] |
347.96 | 381.07 | +9.51% |
benchmarks/test_objectives_benchmarks.py::test_values[vec_generalized_advantage_estimate-True-True] |
303.04 | 276.42 | -8.78% |
benchmarks/test_replaybuffer_benchmark.py::test_rb_populate[TensorDictPrioritizedReplayBuffer-LazyMemmapStorage-None-400] |
472.54 | 512.44 | +8.44% |
benchmarks/test_replaybuffer_benchmark.py::test_rb_sample[TensorDictPrioritizedReplayBuffer-LazyTensorStorage-None-10000] |
2,263 | 2,082 | -8.02% |
benchmarks/test_replaybuffer_benchmark.py::test_rb_populate[TensorDictReplayBuffer-LazyMemmapStorage-RandomSampler-400] |
523.56 | 482.67 | -7.81% |
benchmarks/test_objectives_benchmarks.py::test_td3_speed[True-backward] |
377.24 | 406.39 | +7.73% |
benchmarks/test_replaybuffer_benchmark.py::test_rb_sample[TensorDictReplayBuffer-LazyMemmapStorage-sampler6-10000] |
806.47 | 749.58 | -7.05% |
benchmarks/test_envs_benchmark.py::test_step_mdp_speed[False-True-False-False-False] |
53,418 | 50,010 | -6.38% |
benchmarks/test_replaybuffer_benchmark.py::test_rb_iterate[TensorDictPrioritizedReplayBuffer-LazyTensorStorage-None-10000] |
2,323 | 2,185 | -5.92% |
benchmarks/test_envs_benchmark.py::test_step_mdp_speed[False-True-True-False-False] |
60,730 | 57,141 | -5.91% |
benchmarks/test_objectives_benchmarks.py::test_iql_speed[True-backward] |
241.58 | 255.34 | +5.70% |
benchmarks/test_storage_write_benchmark.py::TestStorageWriteBenchmark::test_collector_stack_then_write[200-img_shape3-large_batch] |
135.07 | 142.25 | +5.32% |
benchmarks/test_storage_write_benchmark.py::TestStorageWriteBenchmark::test_storage_write_contiguous[100-img_shape2-large_img] |
580.71 | 551.38 | -5.05% |
benchmarks/test_rnn_reset_backends_benchmark.py::test_rnn_rollout_with_intermediate_resets[b256-t128-i32-h512-scan-True-0-gru] |
47.47 | 49.79 | +4.89% |
benchmarks/test_replaybuffer_benchmark.py::test_rb_sample[TensorDictReplayBuffer-LazyMemmapStorage-RandomSampler-10000] |
3,104 | 3,255 | +4.86% |
benchmarks/test_envs_benchmark.py::test_step_mdp_speed[True-False-True-True-False] |
35,823 | 34,101 | -4.81% |
benchmarks/test_replaybuffer_benchmark.py::test_rb_iterate[TensorDictReplayBuffer-LazyMemmapStorage-RandomSampler-10000] |
2,888 | 3,025 | +4.77% |
benchmarks/test_replaybuffer_benchmark.py::test_rb_sample[TensorDictPrioritizedReplayBuffer-LazyMemmapStorage-None-10000] |
1,942 | 2,029 | +4.49% |
benchmarks/test_envs_benchmark.py::test_simple |
1.2603 | 1.2055 | -4.35% |
benchmarks/test_replaybuffer_benchmark.py::test_rb_iterate[TensorDictReplayBuffer-LazyTensorStorage-RandomSampler-10000] |
3,435 | 3,577 | +4.15% |
benchmarks/test_envs_benchmark.py::test_step_mdp_speed[True-False-False-False-False] |
57,078 | 54,741 | -4.10% |
benchmarks/test_compressed_storage_benchmark.py::TestCompressedStorageBenchmark::test_tensor_to_bytestream_speed[safetensors] |
24,106 | 23,129 | -4.05% |
benchmarks/test_objectives_benchmarks.py::test_reinforce_speed[False-backward] |
274.12 | 284.87 | +3.92% |
benchmarks/test_objectives_benchmarks.py::test_a2c_speed[True-None] |
755.55 | 726.84 | -3.80% |
benchmarks/test_envs_benchmark.py::test_step_mdp_speed[True-False-True-True-True] |
21,327 | 20,549 | -3.65% |
benchmarks/test_storage_write_benchmark.py::TestStorageWriteBenchmark::test_storage_write_lazystack[100-img_shape1-atari] |
696.55 | 720.90 | +3.50% |
benchmarks/test_envs_benchmark.py::test_step_mdp_speed[False-True-True-True-False] |
35,574 | 34,350 | -3.44% |
benchmarks/test_rnn_reset_backends_benchmark.py::test_rnn_rollout_with_intermediate_resets[b256-t128-i32-h512-scan-False-0-gru] |
22.86 | 23.65 | +3.42% |
benchmarks/test_envs_benchmark.py::test_step_mdp_speed[False-True-True-False-True] |
33,478 | 32,379 | -3.28% |
benchmarks/test_envs_benchmark.py::test_cat_frames_functional[4-constant] |
4,956 | 4,794 | -3.28% |
benchmarks/test_storage_write_benchmark.py::TestStorageWriteBenchmark::test_collector_lazystack_then_write[50-img_shape0-small] |
3,512 | 3,625 | +3.23% |
benchmarks/test_objectives_benchmarks.py::test_values[generalized_advantage_estimate-True-True] |
48.56 | 50.02 | +3.01% |
benchmarks/test_rnn_reset_backends_benchmark.py::test_rnn_rollout_with_intermediate_resets[b256-t128-i32-h512-scan-True-0-lstm] |
75.81 | 78.09 | +3.01% |
benchmarks/test_replaybuffer_benchmark.py::test_rb_extend_sample[ReplayBuffer-LazyTensorStorage-RandomSampler-1000000-10000-100-False] |
49.34 | 50.80 | +2.96% |
benchmarks/test_objectives_benchmarks.py::test_ppo_speed[True-backward] |
347.24 | 357.37 | +2.92% |
benchmarks/test_envs_benchmark.py::test_cat_frames_functional[16-constant] |
4,856 | 4,994 | +2.84% |
benchmarks/test_objectives_benchmarks.py::test_dqn_speed[True-backward] |
897.62 | 922.97 | +2.82% |
benchmarks/test_replaybuffer_benchmark.py::test_rb_extend_sample[ReplayBuffer-LazyTensorStorage-RandomSampler-1000000-10000-100-True] |
21.83 | 22.44 | +2.81% |
benchmarks/test_replaybuffer_benchmark.py::TestPrioritizedReplayBufferBenchmark::test_sampler_sample_scale[1000000-cuda] |
2,299 | 2,236 | -2.76% |
benchmarks/test_objectives_benchmarks.py::test_values[td1_return_estimate-False-False] |
20.36 | 20.92 | +2.76% |
benchmarks/test_objectives_benchmarks.py::test_gae_speed[generalized_advantage_estimate-False-1-512] |
48.58 | 49.86 | +2.63% |
benchmarks/test_replaybuffer_benchmark.py::test_rb_extend_sample[ReplayBuffer-LazyTensorStorage-RandomSampler-100000-10000-100-False] |
53.25 | 54.58 | +2.50% |
benchmarks/test_envs_benchmark.py::test_step_mdp_speed[True-True-False-False-True] |
38,643 | 37,678 | -2.50% |
benchmarks/test_replaybuffer_benchmark.py::test_rb_extend_sample[ReplayBuffer-LazyTensorStorage-RandomSampler-10000-10000-100-True] |
23.59 | 24.17 | +2.44% |
benchmarks/test_envs_benchmark.py::test_transformed |
0.7210 | 0.7035 | -2.43% |
benchmarks/test_objectives_benchmarks.py::test_dqn_speed[False-None] |
639.54 | 654.88 | +2.40% |
benchmarks/test_objectives_benchmarks.py::test_iql_speed[True-None] |
509.60 | 521.50 | +2.33% |
benchmarks/test_envs_benchmark.py::test_cat_frames_functional[4-same] |
6.5619 | 6.7119 | +2.29% |
benchmarks/test_replaybuffer_benchmark.py::test_rb_extend_sample[ReplayBuffer-LazyTensorStorage-RandomSampler-100000-10000-100-True] |
22.82 | 23.34 | +2.26% |
benchmarks/test_replaybuffer_benchmark.py::test_rb_sample[TensorDictReplayBuffer-LazyMemmapStorage-SamplerWithoutReplacement-10000] |
2,871 | 2,936 | +2.24% |
benchmarks/test_objectives_benchmarks.py::test_cql_speed[True-backward] |
222.95 | 227.88 | +2.21% |
benchmarks/test_storage_write_benchmark.py::TestStorageWriteBenchmark::test_storage_write_lazystack[50-img_shape0-small] |
4,377 | 4,473 | +2.20% |
benchmarks/test_compressed_storage_benchmark.py::TestCompressedStorageBenchmark::test_tensor_to_bytestream_speed[pickle] |
12,290 | 12,020 | -2.19% |
benchmarks/test_non_tensor_env_benchmark.py::test_non_tensor_env_rollout_speed[1000-single-True] |
1.3307 | 1.3595 | +2.16% |
benchmarks/test_objectives_benchmarks.py::test_ppo_speed[False-backward] |
134.00 | 136.89 | +2.16% |
benchmarks/test_storage_write_benchmark.py::TestStorageWriteBenchmark::test_collector_lazystack_then_write[100-img_shape1-atari] |
657.33 | 670.97 | +2.08% |
benchmarks/test_rnn_reset_backends_benchmark.py::test_rnn_rollout_with_intermediate_resets[b256-t128-i32-h512-scan-False-0-lstm] |
21.88 | 22.32 | +2.01% |
benchmarks/test_storage_write_benchmark.py::TestCollectorIntegrationBenchmark::test_collector_without_rb[100-img_shape0-atari] |
31.04 | 30.42 | -1.98% |
benchmarks/test_envs_benchmark.py::test_step_mdp_speed[True-True-False-False-False] |
64,716 | 63,448 | -1.96% |
benchmarks/test_storage_write_benchmark.py::TestStorageWriteBenchmark::test_storage_write_contiguous[200-img_shape3-large_batch] |
758.43 | 743.95 | -1.91% |
benchmarks/test_envs_benchmark.py::test_step_mdp_speed[False-False-True-False-True] |
31,044 | 30,458 | -1.89% |
benchmarks/test_envs_benchmark.py::test_cat_frames_functional[16-same] |
5.4972 | 5.5993 | +1.86% |
benchmarks/test_replaybuffer_benchmark.py::test_rb_sample[TensorDictReplayBuffer-LazyTensorStorage-RandomSampler-10000] |
3,637 | 3,705 | +1.86% |
benchmarks/test_envs_benchmark.py::test_step_mdp_speed[True-True-False-True-True] |
22,503 | 22,093 | -1.82% |
benchmarks/test_envs_benchmark.py::test_step_mdp_speed[True-True-False-True-False] |
39,040 | 38,332 | -1.81% |
benchmarks/test_replaybuffer_benchmark.py::test_rb_populate[TensorDictReplayBuffer-LazyMemmapStorage-SamplerWithoutReplacement-400] |
470.69 | 479.06 | +1.78% |
benchmarks/test_compressed_storage_benchmark.py::TestCompressedStorageBenchmark::test_tensor_to_bytestream_speed[untyped_storage] |
8.8843 | 9.0418 | +1.77% |
benchmarks/test_envs_benchmark.py::test_step_mdp_speed[False-False-True-True-False] |
29,705 | 29,198 | -1.70% |
benchmarks/test_non_tensor_env_benchmark.py::test_non_tensor_env_rollout_speed[1000-parallel-no-buffers-False] |
0.2280 | 0.2241 | -1.69% |
benchmarks/test_envs_benchmark.py::test_step_mdp_speed[True-False-False-True-True] |
20,138 | 19,801 | -1.67% |
benchmarks/test_collectors_benchmark.py::test_async_pixels |
10.80 | 10.97 | +1.66% |
benchmarks/test_objectives_benchmarks.py::test_reinforce_speed[False-None] |
401.90 | 395.28 | -1.65% |
benchmarks/test_envs_benchmark.py::test_step_mdp_speed[False-False-False-True-True] |
18,379 | 18,078 | -1.64% |
benchmarks/test_objectives_benchmarks.py::test_a2c_speed[True-backward] |
368.81 | 374.85 | +1.64% |
benchmarks/test_objectives_benchmarks.py::test_redq_deprec_speed[True-backward] |
263.62 | 267.87 | +1.61% |
benchmarks/test_envs_benchmark.py::test_step_mdp_speed[True-False-True-False-False] |
64,714 | 63,669 | -1.61% |
benchmarks/test_replaybuffer_benchmark.py::test_rb_populate[TensorDictReplayBuffer-LazyTensorStorage-RandomSampler-400] |
1,000 | 1,016 | +1.60% |
benchmarks/test_objectives_benchmarks.py::test_dqn_speed[reduce-overhead-None] |
1,927 | 1,897 | -1.58% |
benchmarks/test_storage_write_benchmark.py::TestStorageWriteBenchmark::test_collector_lazystack_then_write[100-img_shape2-large_img] |
389.80 | 395.95 | +1.58% |
benchmarks/test_objectives_benchmarks.py::test_sac_speed[True-backward] |
319.16 | 324.17 | +1.57% |
benchmarks/test_objectives_benchmarks.py::test_ppo_speed[reduce-overhead-None] |
808.39 | 820.99 | +1.56% |
benchmarks/test_non_tensor_env_benchmark.py::test_non_tensor_env_rollout_speed[1000-parallel-buffers-True] |
0.5360 | 0.5280 | -1.50% |
benchmarks/test_storage_write_benchmark.py::TestCollectorIntegrationBenchmark::test_collector_without_rb_cuda[200-img_shape1-large_batch] |
9.0032 | 8.8730 | -1.45% |
benchmarks/test_compressed_storage_benchmark.py::TestCompressedStorageBenchmark::test_tensor_to_bytestream_speed[numpy] |
370,860 | 376,158 | +1.43% |
benchmarks/test_replaybuffer_benchmark.py::test_rb_extend_sample[ReplayBuffer-LazyTensorStorage-RandomSampler-10000-10000-100-False] |
55.02 | 55.80 | +1.42% |
benchmarks/test_objectives_benchmarks.py::test_a2c_speed[reduce-overhead-None] |
871.13 | 883.36 | +1.40% |
benchmarks/test_objectives_benchmarks.py::test_ddpg_speed[True-None] |
830.25 | 841.89 | +1.40% |
benchmarks/test_objectives_benchmarks.py::test_ddpg_speed[True-backward] |
453.38 | 459.70 | +1.39% |
benchmarks/test_envs_benchmark.py::test_step_mdp_speed[True-True-True-True-True] |
23,178 | 23,499 | +1.39% |
benchmarks/test_storage_write_benchmark.py::TestCollectorIntegrationBenchmark::test_collector_with_rb[200-img_shape1-large_batch] |
13.72 | 13.53 | -1.37% |
benchmarks/test_objectives_benchmarks.py::test_values[td_lambda_return_estimate-True-False] |
12.29 | 12.46 | +1.36% |
benchmarks/test_objectives_benchmarks.py::test_iql_speed[False-backward] |
71.04 | 71.99 | +1.35% |
benchmarks/test_objectives_benchmarks.py::test_gae_speed[vec_generalized_advantage_estimate-True-1-512] |
1,287 | 1,304 | +1.35% |
benchmarks/test_storage_write_benchmark.py::TestCollectorIntegrationBenchmark::test_collector_without_rb[200-img_shape1-large_batch] |
15.56 | 15.35 | -1.31% |
benchmarks/test_objectives_benchmarks.py::test_td3_speed[False-backward] |
85.00 | 86.09 | +1.29% |
benchmarks/test_objectives_benchmarks.py::test_cql_speed[reduce-overhead-None] |
89.33 | 88.21 | -1.26% |
benchmarks/test_envs_benchmark.py::test_step_mdp_speed[True-False-False-False-True] |
35,302 | 34,868 | -1.23% |
benchmarks/test_envs_benchmark.py::test_serial |
0.4255 | 0.4307 | +1.22% |
benchmarks/test_storage_write_benchmark.py::TestStorageWriteBenchmark::test_collector_lazystack_then_write[200-img_shape3-large_batch] |
309.96 | 313.70 | +1.21% |
benchmarks/test_objectives_benchmarks.py::test_a2c_speed[False-backward] |
155.02 | 156.84 | +1.18% |
benchmarks/test_non_tensor_env_benchmark.py::test_non_tensor_env_rollout_speed[1000-serial-no-buffers-False] |
0.6909 | 0.6830 | -1.14% |
benchmarks/test_envs_benchmark.py::test_step_mdp_speed[False-True-True-True-True] |
21,014 | 20,777 | -1.13% |
benchmarks/test_envs_benchmark.py::test_step_mdp_speed[False-False-True-True-True] |
19,054 | 18,840 | -1.12% |
benchmarks/test_envs_benchmark.py::test_step_mdp_speed[False-True-False-True-False] |
32,670 | 32,315 | -1.09% |
benchmarks/test_objectives_benchmarks.py::test_sac_speed[reduce-overhead-None] |
100.76 | 99.71 | -1.05% |
benchmarks/test_objectives_benchmarks.py::test_redq_deprec_speed[False-backward] |
72.55 | 73.31 | +1.05% |
benchmarks/test_replaybuffer_benchmark.py::test_rb_iterate[TensorDictReplayBuffer-ListStorage-RandomSampler-4000] |
167.52 | 165.76 | -1.05% |
benchmarks/test_non_tensor_env_benchmark.py::test_non_tensor_env_rollout_speed[1000-parallel-buffers-False] |
0.6064 | 0.6126 | +1.03% |
benchmarks/test_non_tensor_env_benchmark.py::test_non_tensor_env_rollout_speed[1000-single-False] |
1.6204 | 1.6040 | -1.01% |
benchmarks/test_replaybuffer_benchmark.py::test_rb_populate[TensorDictPrioritizedReplayBuffer-ListStorage-None-400] |
183.95 | 182.09 | -1.01% |
benchmarks/test_storage_write_benchmark.py::TestCollectorIntegrationBenchmark::test_collector_with_rb[100-img_shape0-atari] |
26.77 | 26.51 | -0.97% |
benchmarks/test_collectors_benchmark.py::test_sync |
10.54 | 10.45 | -0.87% |
| ... | ... | ... | Showing 120 of 202 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):