Skip to content

[CI] Add checkbox options for wheel build variants (CPU/CUDA/ROCm)#3451

Merged
vmoens merged 1 commit intomainfrom
ci/cpu-only-wheel-builds
Feb 7, 2026
Merged

[CI] Add checkbox options for wheel build variants (CPU/CUDA/ROCm)#3451
vmoens merged 1 commit intomainfrom
ci/cpu-only-wheel-builds

Conversation

@vmoens
Copy link
Collaborator

@vmoens vmoens commented Feb 5, 2026

Summary

This PR adds workflow_dispatch inputs to select which wheel variants to build, presented as tickable checkboxes in the GitHub Actions UI:

  • build_cpu: Build CPU wheels (default: true)
  • build_cuda: Build CUDA wheels (default: false)
  • build_rocm: Build ROCm wheels, Linux only (default: false)

The default is CPU-only since TorchRL is a pure Python library and doesn't require CUDA-specific wheels. This significantly speeds up release dry-runs.

Changes

Updated workflows:

  • build-wheels-linux.yml: CPU, CUDA, ROCm options
  • build-wheels-windows.yml: CPU, CUDA options
  • build-wheels-m1.yml: CPU option only (macOS ARM64)
  • build-wheels-aarch64-linux.yml: CPU option only (no CUDA support)
  • release.yml: Updated to use boolean checkboxes instead of choice dropdown

How it works:

  • For workflow_dispatch (manual trigger): Boolean checkboxes appear in the UI
  • For workflow_call (from release.yml): String inputs (with-cpu, with-cuda, with-rocm) with values enable/disable
  • Parameters are passed to pytorch/test-infra's generate_binary_build_matrix.yml

Test plan

  • Trigger Build Linux Wheels workflow manually with CPU-only selected
  • Verify only CPU matrix is generated
  • Trigger with CUDA+CPU selected and verify both are generated

Made with Cursor

@vmoens @cursoragent
[CI] Add checkbox options for wheel build variants (CPU/CUDA/ROCm)
77facdf 
This adds workflow_dispatch inputs to select which wheel variants to build:
- build_cpu: Build CPU wheels (default: true)
- build_cuda: Build CUDA wheels (default: false)
- build_rocm: Build ROCm wheels, Linux only (default: false)

For workflow_call (used by release.yml), string inputs are used:
- with-cpu, with-cuda, with-rocm (enable/disable)

Default is CPU-only since TorchRL is a pure Python library and doesn't
require CUDA-specific wheels. This significantly speeds up dry-run builds.

Updated workflows:
- build-wheels-linux.yml: CPU, CUDA, ROCm options
- build-wheels-windows.yml: CPU, CUDA options
- build-wheels-m1.yml: CPU option only (macOS ARM64)
- build-wheels-aarch64-linux.yml: CPU option only (no CUDA support)
- release.yml: Updated to use boolean checkboxes instead of choice dropdown

Co-authored-by: Cursor <[email protected]>
@vmoens vmoens added the ciflow/binaries/all Build all binaries label Feb 5, 2026
@pytorch-bot
Copy link

pytorch-bot bot commented Feb 5, 2026 •
edited
Loading

🔗 Helpful Links

🧪 See artifacts and rendered test results at hud.pytorch.org/pr/pytorch/rl/3451

Note: Links to docs will display an error until the docs builds have been completed.

❌ 1 New Failure

As of commit 77facdf with merge base 838410c (image):

NEW FAILURE - The following job has failed:

This comment was automatically generated by Dr. CI and updates every 15 minutes.

@meta-cla meta-cla bot added the CLA Signed This label is managed by the Facebook bot. Authors need to sign the CLA before a PR can be reviewed. label Feb 5, 2026
@github-actions github-actions bot added the CI Has to do with CI setup (e.g. wheels & builds, tests...) label Feb 5, 2026
@github-actions
Copy link
Contributor

github-actions bot commented Feb 5, 2026

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

Total Benchmarks: 173. Improved: $\large\color{#35bf28}15$. Worsened: $\large\color{#d91a1a}10$.

Expand to view detailed results
Name Max Mean Ops Ops on Repo HEAD Change
test_tensor_to_bytestream_speed[pickle] 79.8447μs 78.8353μs 12.6847 KOps/s 12.6983 KOps/s $\color{#d91a1a}-0.11\%$
test_tensor_to_bytestream_speed[torch.save] 0.1366ms 0.1362ms 7.3399 KOps/s 7.3011 KOps/s $\color{#35bf28}+0.53\%$
test_tensor_to_bytestream_speed[untyped_storage] 0.1068s 0.1065s 9.3877 Ops/s 8.9122 Ops/s $\textbf{\color{#35bf28}+5.34\%}$
test_tensor_to_bytestream_speed[numpy] 2.5975μs 2.5938μs 385.5301 KOps/s 395.1288 KOps/s $\color{#d91a1a}-2.43\%$
test_tensor_to_bytestream_speed[safetensors] 36.8427μs 36.5947μs 27.3264 KOps/s 27.1537 KOps/s $\color{#35bf28}+0.64\%$
test_simple 0.5333s 0.5324s 1.8784 Ops/s 1.7931 Ops/s $\color{#35bf28}+4.76\%$
test_transformed 1.2242s 1.1279s 0.8866 Ops/s 0.8843 Ops/s $\color{#35bf28}+0.26\%$
test_serial 1.6342s 1.6302s 0.6134 Ops/s 0.6020 Ops/s $\color{#35bf28}+1.90\%$
test_parallel 1.2927s 1.1235s 0.8901 Ops/s 0.8527 Ops/s $\color{#35bf28}+4.38\%$
test_step_mdp_speed[True-True-True-True-True] 0.3449ms 42.9646μs 23.2750 KOps/s 23.0492 KOps/s $\color{#35bf28}+0.98\%$
test_step_mdp_speed[True-True-True-True-False] 49.1230μs 24.6958μs 40.4927 KOps/s 41.2690 KOps/s $\color{#d91a1a}-1.88\%$
test_step_mdp_speed[True-True-True-False-True] 56.6340μs 24.0387μs 41.5997 KOps/s 40.9726 KOps/s $\color{#35bf28}+1.53\%$
test_step_mdp_speed[True-True-True-False-False] 46.6630μs 13.3866μs 74.7018 KOps/s 73.9781 KOps/s $\color{#35bf28}+0.98\%$
test_step_mdp_speed[True-True-False-True-True] 73.5450μs 46.0279μs 21.7259 KOps/s 21.3749 KOps/s $\color{#35bf28}+1.64\%$
test_step_mdp_speed[True-True-False-True-False] 55.7040μs 27.0459μs 36.9742 KOps/s 37.0572 KOps/s $\color{#d91a1a}-0.22\%$
test_step_mdp_speed[True-True-False-False-True] 61.7140μs 26.9707μs 37.0773 KOps/s 36.7467 KOps/s $\color{#35bf28}+0.90\%$
test_step_mdp_speed[True-True-False-False-False] 41.2730μs 16.2076μs 61.6996 KOps/s 61.2215 KOps/s $\color{#35bf28}+0.78\%$
test_step_mdp_speed[True-False-True-True-True] 88.0460μs 49.5425μs 20.1847 KOps/s 19.7825 KOps/s $\color{#35bf28}+2.03\%$
test_step_mdp_speed[True-False-True-True-False] 63.7740μs 30.4406μs 32.8509 KOps/s 32.7237 KOps/s $\color{#35bf28}+0.39\%$
test_step_mdp_speed[True-False-True-False-True] 59.3530μs 27.1284μs 36.8617 KOps/s 36.5515 KOps/s $\color{#35bf28}+0.85\%$
test_step_mdp_speed[True-False-True-False-False] 48.5530μs 16.1577μs 61.8900 KOps/s 61.1010 KOps/s $\color{#35bf28}+1.29\%$
test_step_mdp_speed[True-False-False-True-True] 84.4650μs 51.9475μs 19.2502 KOps/s 19.0662 KOps/s $\color{#35bf28}+0.97\%$
test_step_mdp_speed[True-False-False-True-False] 59.6140μs 33.0042μs 30.2991 KOps/s 30.3528 KOps/s $\color{#d91a1a}-0.18\%$
test_step_mdp_speed[True-False-False-False-True] 60.7040μs 29.6193μs 33.7618 KOps/s 33.3623 KOps/s $\color{#35bf28}+1.20\%$
test_step_mdp_speed[True-False-False-False-False] 58.3140μs 18.9450μs 52.7844 KOps/s 52.8347 KOps/s $\color{#d91a1a}-0.10\%$
test_step_mdp_speed[False-True-True-True-True] 79.1250μs 49.3805μs 20.2509 KOps/s 20.0779 KOps/s $\color{#35bf28}+0.86\%$
test_step_mdp_speed[False-True-True-True-False] 97.9960μs 29.9223μs 33.4199 KOps/s 32.6247 KOps/s $\color{#35bf28}+2.44\%$
test_step_mdp_speed[False-True-True-False-True] 2.3715ms 31.3438μs 31.9042 KOps/s 32.2207 KOps/s $\color{#d91a1a}-0.98\%$
test_step_mdp_speed[False-True-True-False-False] 45.9430μs 18.0257μs 55.4765 KOps/s 55.0989 KOps/s $\color{#35bf28}+0.69\%$
test_step_mdp_speed[False-True-False-True-True] 88.0560μs 51.4606μs 19.4323 KOps/s 19.1347 KOps/s $\color{#35bf28}+1.56\%$
test_step_mdp_speed[False-True-False-True-False] 66.8740μs 32.9565μs 30.3430 KOps/s 30.6622 KOps/s $\color{#d91a1a}-1.04\%$
test_step_mdp_speed[False-True-False-False-True] 71.6750μs 33.0428μs 30.2638 KOps/s 29.9263 KOps/s $\color{#35bf28}+1.13\%$
test_step_mdp_speed[False-True-False-False-False] 57.2840μs 20.6939μs 48.3235 KOps/s 48.3186 KOps/s $\color{#35bf28}+0.01\%$
test_step_mdp_speed[False-False-True-True-True] 87.1250μs 53.8717μs 18.5626 KOps/s 18.3314 KOps/s $\color{#35bf28}+1.26\%$
test_step_mdp_speed[False-False-True-True-False] 69.5440μs 35.5965μs 28.0927 KOps/s 28.3818 KOps/s $\color{#d91a1a}-1.02\%$
test_step_mdp_speed[False-False-True-False-True] 0.1027ms 33.0570μs 30.2508 KOps/s 29.6710 KOps/s $\color{#35bf28}+1.95\%$
test_step_mdp_speed[False-False-True-False-False] 52.5230μs 20.6415μs 48.4460 KOps/s 48.7416 KOps/s $\color{#d91a1a}-0.61\%$
test_step_mdp_speed[False-False-False-True-True] 90.4860μs 56.3804μs 17.7366 KOps/s 17.5245 KOps/s $\color{#35bf28}+1.21\%$
test_step_mdp_speed[False-False-False-True-False] 73.3850μs 37.8169μs 26.4432 KOps/s 26.4474 KOps/s $\color{#d91a1a}-0.02\%$
test_step_mdp_speed[False-False-False-False-True] 69.5640μs 35.5739μs 28.1105 KOps/s 28.3097 KOps/s $\color{#d91a1a}-0.70\%$
test_step_mdp_speed[False-False-False-False-False] 87.4760μs 22.9740μs 43.5275 KOps/s 43.6606 KOps/s $\color{#d91a1a}-0.30\%$
test_non_tensor_env_rollout_speed[1000-single-True] 0.8541s 0.7547s 1.3251 Ops/s 1.3143 Ops/s $\color{#35bf28}+0.82\%$
test_non_tensor_env_rollout_speed[1000-single-False] 0.7187s 0.6215s 1.6090 Ops/s 1.5926 Ops/s $\color{#35bf28}+1.04\%$
test_non_tensor_env_rollout_speed[1000-serial-no-buffers-True] 1.7224s 1.6474s 0.6070 Ops/s 0.6042 Ops/s $\color{#35bf28}+0.46\%$
test_non_tensor_env_rollout_speed[1000-serial-no-buffers-False] 1.5068s 1.4296s 0.6995 Ops/s 0.6978 Ops/s $\color{#35bf28}+0.25\%$
test_non_tensor_env_rollout_speed[1000-serial-buffers-True] 1.9613s 1.8822s 0.5313 Ops/s 0.5257 Ops/s $\color{#35bf28}+1.06\%$
test_non_tensor_env_rollout_speed[1000-serial-buffers-False] 1.7471s 1.6680s 0.5995 Ops/s 0.5922 Ops/s $\color{#35bf28}+1.24\%$
test_non_tensor_env_rollout_speed[1000-parallel-no-buffers-True] 4.6892s 4.6127s 0.2168 Ops/s 0.2181 Ops/s $\color{#d91a1a}-0.59\%$
test_non_tensor_env_rollout_speed[1000-parallel-no-buffers-False] 4.5067s 4.4489s 0.2248 Ops/s 0.2240 Ops/s $\color{#35bf28}+0.33\%$
test_non_tensor_env_rollout_speed[1000-parallel-buffers-True] 2.1315s 1.9677s 0.5082 Ops/s 0.5146 Ops/s $\color{#d91a1a}-1.23\%$
test_non_tensor_env_rollout_speed[1000-parallel-buffers-False] 1.7179s 1.6419s 0.6090 Ops/s 0.6054 Ops/s $\color{#35bf28}+0.61\%$
test_values[generalized_advantage_estimate-True-True] 10.5168ms 9.8487ms 101.5363 Ops/s 103.2040 Ops/s $\color{#d91a1a}-1.62\%$
test_values[vec_generalized_advantage_estimate-True-True] 21.7597ms 17.7560ms 56.3190 Ops/s 57.0249 Ops/s $\color{#d91a1a}-1.24\%$
test_values[td0_return_estimate-False-False] 0.2259ms 0.1342ms 7.4499 KOps/s 8.1524 KOps/s $\textbf{\color{#d91a1a}-8.62\%}$
test_values[td1_return_estimate-False-False] 26.7724ms 26.0506ms 38.3868 Ops/s 38.3025 Ops/s $\color{#35bf28}+0.22\%$
test_values[vec_td1_return_estimate-False-False] 22.4593ms 17.8354ms 56.0684 Ops/s 56.4710 Ops/s $\color{#d91a1a}-0.71\%$
test_values[td_lambda_return_estimate-True-False] 39.8524ms 38.3694ms 26.0624 Ops/s 25.9879 Ops/s $\color{#35bf28}+0.29\%$
test_values[vec_td_lambda_return_estimate-True-False] 21.0015ms 17.7674ms 56.2828 Ops/s 56.9443 Ops/s $\color{#d91a1a}-1.16\%$
test_gae_speed[generalized_advantage_estimate-False-1-512] 8.6783ms 8.6135ms 116.0970 Ops/s 116.2943 Ops/s $\color{#d91a1a}-0.17\%$
test_gae_speed[vec_generalized_advantage_estimate-True-1-512] 1.6726ms 1.4528ms 688.3134 Ops/s 685.0953 Ops/s $\color{#35bf28}+0.47\%$
test_gae_speed[vec_generalized_advantage_estimate-False-1-512] 0.5167ms 0.4029ms 2.4822 KOps/s 2.4843 KOps/s $\color{#d91a1a}-0.08\%$
test_gae_speed[vec_generalized_advantage_estimate-True-32-512] 35.4723ms 34.9485ms 28.6135 Ops/s 32.6057 Ops/s $\textbf{\color{#d91a1a}-12.24\%}$
test_gae_speed[vec_generalized_advantage_estimate-False-32-512] 1.8281ms 1.6828ms 594.2454 Ops/s 588.5981 Ops/s $\color{#35bf28}+0.96\%$
test_dqn_speed[False-None] 1.4790ms 1.3671ms 731.4748 Ops/s 739.7090 Ops/s $\color{#d91a1a}-1.11\%$
test_dqn_speed[False-backward] 1.9378ms 1.8520ms 539.9617 Ops/s 540.4967 Ops/s $\color{#d91a1a}-0.10\%$
test_dqn_speed[True-None] 0.6465ms 0.5327ms 1.8772 KOps/s 1.7813 KOps/s $\textbf{\color{#35bf28}+5.39\%}$
test_dqn_speed[True-backward] 1.0403ms 0.9827ms 1.0176 KOps/s 821.7749 Ops/s $\textbf{\color{#35bf28}+23.84\%}$
test_dqn_speed[reduce-overhead-None] 0.6334ms 0.5437ms 1.8393 KOps/s 1.8370 KOps/s $\color{#35bf28}+0.13\%$
test_ddpg_speed[False-None] 3.2330ms 2.8714ms 348.2605 Ops/s 359.6526 Ops/s $\color{#d91a1a}-3.17\%$
test_ddpg_speed[False-backward] 4.0337ms 3.9526ms 252.9976 Ops/s 253.6366 Ops/s $\color{#d91a1a}-0.25\%$
test_ddpg_speed[True-None] 1.6349ms 1.4175ms 705.4829 Ops/s 711.0128 Ops/s $\color{#d91a1a}-0.78\%$
test_ddpg_speed[True-backward] 2.4790ms 2.3623ms 423.3212 Ops/s 354.8300 Ops/s $\textbf{\color{#35bf28}+19.30\%}$
test_ddpg_speed[reduce-overhead-None] 1.5265ms 1.3804ms 724.4054 Ops/s 716.7538 Ops/s $\color{#35bf28}+1.07\%$
test_sac_speed[False-None] 8.5984ms 7.9107ms 126.4118 Ops/s 128.7192 Ops/s $\color{#d91a1a}-1.79\%$
test_sac_speed[False-backward] 11.4794ms 10.9340ms 91.4578 Ops/s 91.8529 Ops/s $\color{#d91a1a}-0.43\%$
test_sac_speed[True-None] 2.2854ms 2.1202ms 471.6528 Ops/s 464.4811 Ops/s $\color{#35bf28}+1.54\%$
test_sac_speed[True-backward] 4.1330ms 3.9941ms 250.3716 Ops/s 220.4053 Ops/s $\textbf{\color{#35bf28}+13.60\%}$
test_sac_speed[reduce-overhead-None] 2.3045ms 2.1137ms 473.1081 Ops/s 452.8944 Ops/s $\color{#35bf28}+4.46\%$
test_redq_speed[False-None] 16.5442ms 10.8192ms 92.4279 Ops/s 85.0440 Ops/s $\textbf{\color{#35bf28}+8.68\%}$
test_redq_speed[False-backward] 18.3943ms 17.7742ms 56.2613 Ops/s 57.4033 Ops/s $\color{#d91a1a}-1.99\%$
test_redq_speed[True-None] 4.6303ms 4.4449ms 224.9761 Ops/s 220.3366 Ops/s $\color{#35bf28}+2.11\%$
test_redq_speed[True-backward] 10.1069ms 9.8611ms 101.4089 Ops/s 103.0878 Ops/s $\color{#d91a1a}-1.63\%$
test_redq_speed[reduce-overhead-None] 4.7784ms 4.4334ms 225.5630 Ops/s 219.9848 Ops/s $\color{#35bf28}+2.54\%$
test_redq_deprec_speed[False-None] 11.1714ms 10.7547ms 92.9824 Ops/s 92.4264 Ops/s $\color{#35bf28}+0.60\%$
test_redq_deprec_speed[False-backward] 15.7467ms 15.4282ms 64.8165 Ops/s 63.8367 Ops/s $\color{#35bf28}+1.53\%$
test_redq_deprec_speed[True-None] 4.0806ms 3.6775ms 271.9229 Ops/s 270.0849 Ops/s $\color{#35bf28}+0.68\%$
test_redq_deprec_speed[True-backward] 7.8540ms 7.5678ms 132.1385 Ops/s 125.7537 Ops/s $\textbf{\color{#35bf28}+5.08\%}$
test_redq_deprec_speed[reduce-overhead-None] 4.2396ms 3.6538ms 273.6848 Ops/s 273.0911 Ops/s $\color{#35bf28}+0.22\%$
test_td3_speed[False-None] 8.0195ms 7.8456ms 127.4604 Ops/s 128.0432 Ops/s $\color{#d91a1a}-0.46\%$
test_td3_speed[False-backward] 11.1170ms 10.6248ms 94.1197 Ops/s 94.8116 Ops/s $\color{#d91a1a}-0.73\%$
test_td3_speed[True-None] 1.8612ms 1.8282ms 546.9752 Ops/s 536.3342 Ops/s $\color{#35bf28}+1.98\%$
test_td3_speed[True-backward] 3.7433ms 3.6207ms 276.1923 Ops/s 242.5286 Ops/s $\textbf{\color{#35bf28}+13.88\%}$
test_td3_speed[reduce-overhead-None] 1.8925ms 1.7728ms 564.0822 Ops/s 554.5962 Ops/s $\color{#35bf28}+1.71\%$
test_cql_speed[False-None] 29.6653ms 25.9123ms 38.5917 Ops/s 39.0474 Ops/s $\color{#d91a1a}-1.17\%$
test_cql_speed[False-backward] 38.3689ms 34.8713ms 28.6769 Ops/s 29.1523 Ops/s $\color{#d91a1a}-1.63\%$
test_cql_speed[True-None] 12.6496ms 12.3235ms 81.1461 Ops/s 80.1235 Ops/s $\color{#35bf28}+1.28\%$
test_cql_speed[True-backward] 18.7846ms 18.2694ms 54.7364 Ops/s 59.4833 Ops/s $\textbf{\color{#d91a1a}-7.98\%}$
test_cql_speed[reduce-overhead-None] 12.6821ms 12.4482ms 80.3329 Ops/s 79.9320 Ops/s $\color{#35bf28}+0.50\%$
test_a2c_speed[False-None] 5.4748ms 5.2547ms 190.3071 Ops/s 187.8660 Ops/s $\color{#35bf28}+1.30\%$
test_a2c_speed[False-backward] 11.8248ms 11.6533ms 85.8129 Ops/s 85.2589 Ops/s $\color{#35bf28}+0.65\%$
test_a2c_speed[True-None] 3.8325ms 3.7007ms 270.2217 Ops/s 256.0255 Ops/s $\textbf{\color{#35bf28}+5.54\%}$
test_a2c_speed[True-backward] 8.9549ms 8.5363ms 117.1475 Ops/s 111.4515 Ops/s $\textbf{\color{#35bf28}+5.11\%}$
test_a2c_speed[reduce-overhead-None] 3.8699ms 3.6837ms 271.4636 Ops/s 270.1411 Ops/s $\color{#35bf28}+0.49\%$
test_ppo_speed[False-None] 6.0145ms 5.8418ms 171.1793 Ops/s 168.8546 Ops/s $\color{#35bf28}+1.38\%$
test_ppo_speed[False-backward] 12.5565ms 12.3191ms 81.1746 Ops/s 81.1097 Ops/s $\color{#35bf28}+0.08\%$
test_ppo_speed[True-None] 3.7857ms 3.6284ms 275.6022 Ops/s 273.5647 Ops/s $\color{#35bf28}+0.74\%$
test_ppo_speed[True-backward] 8.5603ms 8.3022ms 120.4500 Ops/s 119.0861 Ops/s $\color{#35bf28}+1.15\%$
test_ppo_speed[reduce-overhead-None] 3.7087ms 3.6063ms 277.2894 Ops/s 276.7931 Ops/s $\color{#35bf28}+0.18\%$
test_reinforce_speed[False-None] 5.4980ms 4.5120ms 221.6335 Ops/s 221.4213 Ops/s $\color{#35bf28}+0.10\%$
test_reinforce_speed[False-backward] 7.4275ms 7.2432ms 138.0613 Ops/s 137.1018 Ops/s $\color{#35bf28}+0.70\%$
test_reinforce_speed[True-None] 3.0474ms 2.8821ms 346.9640 Ops/s 338.0506 Ops/s $\color{#35bf28}+2.64\%$
test_reinforce_speed[True-backward] 8.0289ms 7.7474ms 129.0749 Ops/s 117.1770 Ops/s $\textbf{\color{#35bf28}+10.15\%}$
test_reinforce_speed[reduce-overhead-None] 3.0522ms 2.8603ms 349.6162 Ops/s 335.7122 Ops/s $\color{#35bf28}+4.14\%$
test_iql_speed[False-None] 19.8915ms 19.2128ms 52.0488 Ops/s 51.1515 Ops/s $\color{#35bf28}+1.75\%$
test_iql_speed[False-backward] 35.3161ms 30.0253ms 33.3052 Ops/s 33.0502 Ops/s $\color{#35bf28}+0.77\%$
test_iql_speed[True-None] 8.7686ms 8.4577ms 118.2357 Ops/s 114.2159 Ops/s $\color{#35bf28}+3.52\%$
test_iql_speed[True-backward] 17.1746ms 16.6889ms 59.9201 Ops/s 59.9585 Ops/s $\color{#d91a1a}-0.06\%$
test_iql_speed[reduce-overhead-None] 8.7683ms 8.5209ms 117.3579 Ops/s 113.4285 Ops/s $\color{#35bf28}+3.46\%$
test_rb_sample[TensorDictReplayBuffer-ListStorage-RandomSampler-4000] 5.9885ms 5.8936ms 169.6766 Ops/s 171.8013 Ops/s $\color{#d91a1a}-1.24\%$
test_rb_sample[TensorDictReplayBuffer-LazyMemmapStorage-RandomSampler-10000] 2.0001ms 0.2955ms 3.3842 KOps/s 3.0198 KOps/s $\textbf{\color{#35bf28}+12.07\%}$
test_rb_sample[TensorDictReplayBuffer-LazyTensorStorage-RandomSampler-10000] 0.4711ms 0.2558ms 3.9090 KOps/s 3.3248 KOps/s $\textbf{\color{#35bf28}+17.57\%}$
test_rb_sample[TensorDictReplayBuffer-ListStorage-SamplerWithoutReplacement-4000] 5.9077ms 5.7021ms 175.3740 Ops/s 177.0160 Ops/s $\color{#d91a1a}-0.93\%$
test_rb_sample[TensorDictReplayBuffer-LazyMemmapStorage-SamplerWithoutReplacement-10000] 1.9263ms 0.3220ms 3.1052 KOps/s 3.6660 KOps/s $\textbf{\color{#d91a1a}-15.30\%}$
test_rb_sample[TensorDictReplayBuffer-LazyTensorStorage-SamplerWithoutReplacement-10000] 0.5136ms 0.2849ms 3.5096 KOps/s 3.9217 KOps/s $\textbf{\color{#d91a1a}-10.51\%}$
test_rb_sample[TensorDictReplayBuffer-LazyMemmapStorage-sampler6-10000] 1.4623ms 1.2140ms 823.6975 Ops/s 826.0165 Ops/s $\color{#d91a1a}-0.28\%$
test_rb_sample[TensorDictReplayBuffer-LazyTensorStorage-sampler7-10000] 1.3584ms 1.1262ms 887.9493 Ops/s 883.6538 Ops/s $\color{#35bf28}+0.49\%$
test_rb_sample[TensorDictPrioritizedReplayBuffer-ListStorage-None-4000] 9.8085ms 5.9856ms 167.0678 Ops/s 170.2359 Ops/s $\color{#d91a1a}-1.86\%$
test_rb_sample[TensorDictPrioritizedReplayBuffer-LazyMemmapStorage-None-10000] 1.7655ms 0.4246ms 2.3552 KOps/s 2.3008 KOps/s $\color{#35bf28}+2.36\%$
test_rb_sample[TensorDictPrioritizedReplayBuffer-LazyTensorStorage-None-10000] 0.6679ms 0.4126ms 2.4235 KOps/s 2.4487 KOps/s $\color{#d91a1a}-1.03\%$
test_rb_iterate[TensorDictReplayBuffer-ListStorage-RandomSampler-4000] 5.7807ms 5.6929ms 175.6580 Ops/s 175.3809 Ops/s $\color{#35bf28}+0.16\%$
test_rb_iterate[TensorDictReplayBuffer-LazyMemmapStorage-RandomSampler-10000] 1.0366ms 0.2772ms 3.6070 KOps/s 3.5874 KOps/s $\color{#35bf28}+0.55\%$
test_rb_iterate[TensorDictReplayBuffer-LazyTensorStorage-RandomSampler-10000] 0.5107ms 0.3314ms 3.0178 KOps/s 3.6025 KOps/s $\textbf{\color{#d91a1a}-16.23\%}$
test_rb_iterate[TensorDictReplayBuffer-ListStorage-SamplerWithoutReplacement-4000] 5.8921ms 5.6660ms 176.4905 Ops/s 176.9980 Ops/s $\color{#d91a1a}-0.29\%$
test_rb_iterate[TensorDictReplayBuffer-LazyMemmapStorage-SamplerWithoutReplacement-10000] 1.1727ms 0.3294ms 3.0356 KOps/s 2.7041 KOps/s $\textbf{\color{#35bf28}+12.26\%}$
test_rb_iterate[TensorDictReplayBuffer-LazyTensorStorage-SamplerWithoutReplacement-10000] 0.5547ms 0.3297ms 3.0334 KOps/s 3.5598 KOps/s $\textbf{\color{#d91a1a}-14.79\%}$
test_rb_iterate[TensorDictPrioritizedReplayBuffer-ListStorage-None-4000] 6.2846ms 5.8580ms 170.7057 Ops/s 171.8811 Ops/s $\color{#d91a1a}-0.68\%$
test_rb_iterate[TensorDictPrioritizedReplayBuffer-LazyMemmapStorage-None-10000] 0.9333ms 0.4521ms 2.2120 KOps/s 650.2763 Ops/s $\textbf{\color{#35bf28}+240.16\%}$
test_rb_iterate[TensorDictPrioritizedReplayBuffer-LazyTensorStorage-None-10000] 0.6527ms 0.4493ms 2.2256 KOps/s 2.1330 KOps/s $\color{#35bf28}+4.34\%$
test_rb_populate[TensorDictReplayBuffer-ListStorage-RandomSampler-400] 6.4405ms 4.9698ms 201.2151 Ops/s 201.5604 Ops/s $\color{#d91a1a}-0.17\%$
test_rb_populate[TensorDictReplayBuffer-LazyMemmapStorage-RandomSampler-400] 12.8160ms 1.9333ms 517.2388 Ops/s 586.3411 Ops/s $\textbf{\color{#d91a1a}-11.79\%}$
test_rb_populate[TensorDictReplayBuffer-LazyTensorStorage-RandomSampler-400] 1.2399ms 0.8874ms 1.1269 KOps/s 1.1319 KOps/s $\color{#d91a1a}-0.44\%$
test_rb_populate[TensorDictReplayBuffer-ListStorage-SamplerWithoutReplacement-400] 6.5327ms 4.9299ms 202.8422 Ops/s 199.3997 Ops/s $\color{#35bf28}+1.73\%$
test_rb_populate[TensorDictReplayBuffer-LazyMemmapStorage-SamplerWithoutReplacement-400] 4.9924ms 1.8304ms 546.3230 Ops/s 547.0981 Ops/s $\color{#d91a1a}-0.14\%$
test_rb_populate[TensorDictReplayBuffer-LazyTensorStorage-SamplerWithoutReplacement-400] 1.3457ms 0.8542ms 1.1707 KOps/s 1.1572 KOps/s $\color{#35bf28}+1.17\%$
test_rb_populate[TensorDictPrioritizedReplayBuffer-ListStorage-None-400] 0.5440s 16.0211ms 62.4178 Ops/s 59.9364 Ops/s $\color{#35bf28}+4.14\%$
test_rb_populate[TensorDictPrioritizedReplayBuffer-LazyMemmapStorage-None-400] 8.0561ms 2.0042ms 498.9416 Ops/s 531.6949 Ops/s $\textbf{\color{#d91a1a}-6.16\%}$
test_rb_populate[TensorDictPrioritizedReplayBuffer-LazyTensorStorage-None-400] 2.5121ms 1.0211ms 979.3366 Ops/s 964.4696 Ops/s $\color{#35bf28}+1.54\%$
test_rb_extend_sample[ReplayBuffer-LazyTensorStorage-RandomSampler-10000-10000-100-True] 40.3526ms 35.4794ms 28.1854 Ops/s 28.1622 Ops/s $\color{#35bf28}+0.08\%$
test_rb_extend_sample[ReplayBuffer-LazyTensorStorage-RandomSampler-10000-10000-100-False] 19.1435ms 17.6715ms 56.5884 Ops/s 56.6993 Ops/s $\color{#d91a1a}-0.20\%$
test_rb_extend_sample[ReplayBuffer-LazyTensorStorage-RandomSampler-100000-10000-100-True] 40.0582ms 36.2599ms 27.5787 Ops/s 27.3043 Ops/s $\color{#35bf28}+1.00\%$
test_rb_extend_sample[ReplayBuffer-LazyTensorStorage-RandomSampler-100000-10000-100-False] 19.8750ms 18.0076ms 55.5320 Ops/s 55.7609 Ops/s $\color{#d91a1a}-0.41\%$
test_rb_extend_sample[ReplayBuffer-LazyTensorStorage-RandomSampler-1000000-10000-100-True] 39.7831ms 37.9531ms 26.3483 Ops/s 25.9758 Ops/s $\color{#35bf28}+1.43\%$
test_rb_extend_sample[ReplayBuffer-LazyTensorStorage-RandomSampler-1000000-10000-100-False] 20.5407ms 19.3663ms 51.6362 Ops/s 52.3028 Ops/s $\color{#d91a1a}-1.27\%$
test_storage_write_lazystack[50-img_shape0-small] 0.8353ms 0.2147ms 4.6574 KOps/s 4.6547 KOps/s $\color{#35bf28}+0.06\%$
test_storage_write_lazystack[100-img_shape1-atari] 1.6935ms 1.3756ms 726.9313 Ops/s 731.9877 Ops/s $\color{#d91a1a}-0.69\%$
test_storage_write_lazystack[100-img_shape2-large_img] 2.7304ms 2.3078ms 433.3058 Ops/s 421.8570 Ops/s $\color{#35bf28}+2.71\%$
test_storage_write_lazystack[200-img_shape3-large_batch] 2.9999ms 2.8619ms 349.4169 Ops/s 347.7104 Ops/s $\color{#35bf28}+0.49\%$
test_storage_write_contiguous[50-img_shape0-small] 0.2161ms 0.1289ms 7.7574 KOps/s 7.7231 KOps/s $\color{#35bf28}+0.44\%$
test_storage_write_contiguous[100-img_shape1-atari] 0.6018ms 0.1833ms 5.4544 KOps/s 5.6516 KOps/s $\color{#d91a1a}-3.49\%$
test_storage_write_contiguous[100-img_shape2-large_img] 2.1516ms 1.7767ms 562.8380 Ops/s 572.6249 Ops/s $\color{#d91a1a}-1.71\%$
test_storage_write_contiguous[200-img_shape3-large_batch] 1.6768ms 1.3780ms 725.6914 Ops/s 763.4776 Ops/s $\color{#d91a1a}-4.95\%$
test_collector_stack_then_write[50-img_shape0-small] 1.6168ms 1.0899ms 917.5210 Ops/s 908.3056 Ops/s $\color{#35bf28}+1.01\%$
test_collector_stack_then_write[100-img_shape1-atari] 3.6968ms 3.6163ms 276.5224 Ops/s 288.5424 Ops/s $\color{#d91a1a}-4.17\%$
test_collector_stack_then_write[100-img_shape2-large_img] 5.6556ms 5.5166ms 181.2722 Ops/s 173.3291 Ops/s $\color{#35bf28}+4.58\%$
test_collector_stack_then_write[200-img_shape3-large_batch] 7.2926ms 6.9566ms 143.7482 Ops/s 139.3269 Ops/s $\color{#35bf28}+3.17\%$
test_collector_lazystack_then_write[50-img_shape0-small] 0.4095ms 0.2666ms 3.7510 KOps/s 3.7338 KOps/s $\color{#35bf28}+0.46\%$
test_collector_lazystack_then_write[100-img_shape1-atari] 1.6048ms 1.4743ms 678.2792 Ops/s 672.4954 Ops/s $\color{#35bf28}+0.86\%$
test_collector_lazystack_then_write[100-img_shape2-large_img] 2.8333ms 2.4205ms 413.1339 Ops/s 400.7692 Ops/s $\color{#35bf28}+3.09\%$
test_collector_lazystack_then_write[200-img_shape3-large_batch] 3.3341ms 3.0501ms 327.8629 Ops/s 324.1382 Ops/s $\color{#35bf28}+1.15\%$
test_collector_without_rb[100-img_shape0-atari] 33.0094ms 32.7091ms 30.5725 Ops/s 30.2886 Ops/s $\color{#35bf28}+0.94\%$
test_collector_without_rb[200-img_shape1-large_batch] 64.8979ms 64.6154ms 15.4762 Ops/s 15.3832 Ops/s $\color{#35bf28}+0.60\%$
test_collector_with_rb[100-img_shape0-atari] 37.8473ms 37.1278ms 26.9340 Ops/s 26.7012 Ops/s $\color{#35bf28}+0.87\%$
test_collector_with_rb[200-img_shape1-large_batch] 0.6624s 0.1156s 8.6533 Ops/s 13.6177 Ops/s $\textbf{\color{#d91a1a}-36.46\%}$

@github-actions
Copy link
Contributor

github-actions bot commented Feb 5, 2026

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

Total Benchmarks: 172. Improved: $\large\color{#35bf28}12$. Worsened: $\large\color{#d91a1a}14$.

Expand to view detailed results
Name Max Mean Ops Ops on Repo HEAD Change
test_tensor_to_bytestream_speed[pickle] 81.4205μs 80.4073μs 12.4367 KOps/s 12.4880 KOps/s $\color{#d91a1a}-0.41\%$
test_tensor_to_bytestream_speed[torch.save] 0.1468ms 0.1456ms 6.8676 KOps/s 7.2572 KOps/s $\textbf{\color{#d91a1a}-5.37\%}$
test_tensor_to_bytestream_speed[untyped_storage] 0.1055s 0.1054s 9.4912 Ops/s 9.5493 Ops/s $\color{#d91a1a}-0.61\%$
test_tensor_to_bytestream_speed[numpy] 2.6048μs 2.5855μs 386.7772 KOps/s 400.6032 KOps/s $\color{#d91a1a}-3.45\%$
test_tensor_to_bytestream_speed[safetensors] 37.5730μs 37.4059μs 26.7337 KOps/s 26.2385 KOps/s $\color{#35bf28}+1.89\%$
test_simple 0.7971s 0.7893s 1.2669 Ops/s 1.2383 Ops/s $\color{#35bf28}+2.31\%$
test_transformed 1.5369s 1.4481s 0.6906 Ops/s 0.6920 Ops/s $\color{#d91a1a}-0.20\%$
test_serial 2.4465s 2.3229s 0.4305 Ops/s 0.4258 Ops/s $\color{#35bf28}+1.09\%$
test_parallel 2.0543s 1.9433s 0.5146 Ops/s 0.5015 Ops/s $\color{#35bf28}+2.62\%$
test_step_mdp_speed[True-True-True-True-True] 0.3738ms 44.7474μs 22.3476 KOps/s 22.6317 KOps/s $\color{#d91a1a}-1.26\%$
test_step_mdp_speed[True-True-True-True-False] 57.5010μs 25.0772μs 39.8769 KOps/s 40.3732 KOps/s $\color{#d91a1a}-1.23\%$
test_step_mdp_speed[True-True-True-False-True] 53.2710μs 24.8601μs 40.2250 KOps/s 40.9017 KOps/s $\color{#d91a1a}-1.65\%$
test_step_mdp_speed[True-True-True-False-False] 50.3610μs 13.6949μs 73.0199 KOps/s 73.4920 KOps/s $\color{#d91a1a}-0.64\%$
test_step_mdp_speed[True-True-False-True-True] 77.2810μs 47.9695μs 20.8466 KOps/s 20.9460 KOps/s $\color{#d91a1a}-0.47\%$
test_step_mdp_speed[True-True-False-True-False] 72.6920μs 27.8170μs 35.9493 KOps/s 36.3510 KOps/s $\color{#d91a1a}-1.11\%$
test_step_mdp_speed[True-True-False-False-True] 57.7210μs 27.8967μs 35.8466 KOps/s 36.5205 KOps/s $\color{#d91a1a}-1.85\%$
test_step_mdp_speed[True-True-False-False-False] 47.5110μs 16.7029μs 59.8697 KOps/s 60.2538 KOps/s $\color{#d91a1a}-0.64\%$
test_step_mdp_speed[True-False-True-True-True] 82.9420μs 50.3172μs 19.8739 KOps/s 19.7472 KOps/s $\color{#35bf28}+0.64\%$
test_step_mdp_speed[True-False-True-True-False] 57.7810μs 30.4147μs 32.8788 KOps/s 32.9579 KOps/s $\color{#d91a1a}-0.24\%$
test_step_mdp_speed[True-False-True-False-True] 65.3810μs 27.6910μs 36.1129 KOps/s 36.7649 KOps/s $\color{#d91a1a}-1.77\%$
test_step_mdp_speed[True-False-True-False-False] 42.9810μs 16.6649μs 60.0063 KOps/s 60.6802 KOps/s $\color{#d91a1a}-1.11\%$
test_step_mdp_speed[True-False-False-True-True] 85.0910μs 52.5693μs 19.0225 KOps/s 18.9973 KOps/s $\color{#35bf28}+0.13\%$
test_step_mdp_speed[True-False-False-True-False] 73.4710μs 33.3141μs 30.0173 KOps/s 30.3026 KOps/s $\color{#d91a1a}-0.94\%$
test_step_mdp_speed[True-False-False-False-True] 60.7010μs 30.4420μs 32.8493 KOps/s 33.3728 KOps/s $\color{#d91a1a}-1.57\%$
test_step_mdp_speed[True-False-False-False-False] 50.4010μs 19.1792μs 52.1397 KOps/s 52.2854 KOps/s $\color{#d91a1a}-0.28\%$
test_step_mdp_speed[False-True-True-True-True] 80.7620μs 50.2189μs 19.9128 KOps/s 19.7893 KOps/s $\color{#35bf28}+0.62\%$
test_step_mdp_speed[False-True-True-True-False] 76.7620μs 30.5112μs 32.7749 KOps/s 32.9109 KOps/s $\color{#d91a1a}-0.41\%$
test_step_mdp_speed[False-True-True-False-True] 2.2935ms 31.7114μs 31.5344 KOps/s 31.9719 KOps/s $\color{#d91a1a}-1.37\%$
test_step_mdp_speed[False-True-True-False-False] 78.2620μs 18.2005μs 54.9435 KOps/s 54.8790 KOps/s $\color{#35bf28}+0.12\%$
test_step_mdp_speed[False-True-False-True-True] 91.8510μs 52.8610μs 18.9176 KOps/s 18.9789 KOps/s $\color{#d91a1a}-0.32\%$
test_step_mdp_speed[False-True-False-True-False] 60.3310μs 33.1861μs 30.1331 KOps/s 30.4104 KOps/s $\color{#d91a1a}-0.91\%$
test_step_mdp_speed[False-True-False-False-True] 61.7310μs 34.2854μs 29.1669 KOps/s 29.2313 KOps/s $\color{#d91a1a}-0.22\%$
test_step_mdp_speed[False-True-False-False-False] 48.9810μs 20.9176μs 47.8066 KOps/s 47.6947 KOps/s $\color{#35bf28}+0.23\%$
test_step_mdp_speed[False-False-True-True-True] 94.4210μs 55.6245μs 17.9777 KOps/s 17.9518 KOps/s $\color{#35bf28}+0.14\%$
test_step_mdp_speed[False-False-True-True-False] 63.6510μs 36.7334μs 27.2231 KOps/s 27.5773 KOps/s $\color{#d91a1a}-1.28\%$
test_step_mdp_speed[False-False-True-False-True] 61.5210μs 33.9655μs 29.4416 KOps/s 29.0612 KOps/s $\color{#35bf28}+1.31\%$
test_step_mdp_speed[False-False-True-False-False] 47.2610μs 20.9381μs 47.7598 KOps/s 47.8897 KOps/s $\color{#d91a1a}-0.27\%$
test_step_mdp_speed[False-False-False-True-True] 0.1049ms 57.4914μs 17.3939 KOps/s 17.4145 KOps/s $\color{#d91a1a}-0.12\%$
test_step_mdp_speed[False-False-False-True-False] 67.3710μs 38.6126μs 25.8983 KOps/s 26.0426 KOps/s $\color{#d91a1a}-0.55\%$
test_step_mdp_speed[False-False-False-False-True] 73.9010μs 36.4115μs 27.4638 KOps/s 27.4703 KOps/s $\color{#d91a1a}-0.02\%$
test_step_mdp_speed[False-False-False-False-False] 59.6910μs 23.4154μs 42.7069 KOps/s 42.8284 KOps/s $\color{#d91a1a}-0.28\%$
test_non_tensor_env_rollout_speed[1000-single-True] 0.8672s 0.7735s 1.2928 Ops/s 1.3002 Ops/s $\color{#d91a1a}-0.57\%$
test_non_tensor_env_rollout_speed[1000-single-False] 0.7309s 0.6346s 1.5758 Ops/s 1.5752 Ops/s $\color{#35bf28}+0.04\%$
test_non_tensor_env_rollout_speed[1000-serial-no-buffers-True] 1.7542s 1.6809s 0.5949 Ops/s 0.6002 Ops/s $\color{#d91a1a}-0.88\%$
test_non_tensor_env_rollout_speed[1000-serial-no-buffers-False] 1.5271s 1.4505s 0.6894 Ops/s 0.6919 Ops/s $\color{#d91a1a}-0.36\%$
test_non_tensor_env_rollout_speed[1000-serial-buffers-True] 2.0063s 1.9309s 0.5179 Ops/s 0.5226 Ops/s $\color{#d91a1a}-0.91\%$
test_non_tensor_env_rollout_speed[1000-serial-buffers-False] 1.7939s 1.7077s 0.5856 Ops/s 0.5903 Ops/s $\color{#d91a1a}-0.79\%$
test_non_tensor_env_rollout_speed[1000-parallel-no-buffers-True] 4.8815s 4.7051s 0.2125 Ops/s 0.2147 Ops/s $\color{#d91a1a}-1.02\%$
test_non_tensor_env_rollout_speed[1000-parallel-no-buffers-False] 4.5182s 4.4889s 0.2228 Ops/s 0.2247 Ops/s $\color{#d91a1a}-0.85\%$
test_non_tensor_env_rollout_speed[1000-parallel-buffers-True] 2.1901s 2.0263s 0.4935 Ops/s 0.5093 Ops/s $\color{#d91a1a}-3.11\%$
test_non_tensor_env_rollout_speed[1000-parallel-buffers-False] 1.8044s 1.7171s 0.5824 Ops/s 0.5994 Ops/s $\color{#d91a1a}-2.84\%$
test_values[generalized_advantage_estimate-True-True] 20.3560ms 19.8157ms 50.4649 Ops/s 50.2415 Ops/s $\color{#35bf28}+0.44\%$
test_values[vec_generalized_advantage_estimate-True-True] 0.1312s 3.5365ms 282.7692 Ops/s 260.0067 Ops/s $\textbf{\color{#35bf28}+8.75\%}$
test_values[td0_return_estimate-False-False] 0.1056ms 81.0726μs 12.3346 KOps/s 12.2469 KOps/s $\color{#35bf28}+0.72\%$
test_values[td1_return_estimate-False-False] 47.8497ms 47.3225ms 21.1316 Ops/s 21.1371 Ops/s $\color{#d91a1a}-0.03\%$
test_values[vec_td1_return_estimate-False-False] 1.2888ms 1.0768ms 928.6867 Ops/s 930.9480 Ops/s $\color{#d91a1a}-0.24\%$
test_values[td_lambda_return_estimate-True-False] 82.7761ms 77.9436ms 12.8298 Ops/s 12.9116 Ops/s $\color{#d91a1a}-0.63\%$
test_values[vec_td_lambda_return_estimate-True-False] 1.2762ms 1.0739ms 931.1970 Ops/s 933.7830 Ops/s $\color{#d91a1a}-0.28\%$
test_gae_speed[generalized_advantage_estimate-False-1-512] 20.6964ms 20.0213ms 49.9469 Ops/s 49.8236 Ops/s $\color{#35bf28}+0.25\%$
test_gae_speed[vec_generalized_advantage_estimate-True-1-512] 1.0181ms 0.7426ms 1.3467 KOps/s 1.3352 KOps/s $\color{#35bf28}+0.86\%$
test_gae_speed[vec_generalized_advantage_estimate-False-1-512] 0.8111ms 0.6914ms 1.4464 KOps/s 1.4970 KOps/s $\color{#d91a1a}-3.37\%$
test_gae_speed[vec_generalized_advantage_estimate-True-32-512] 1.5397ms 1.4811ms 675.1576 Ops/s 675.3605 Ops/s $\color{#d91a1a}-0.03\%$
test_gae_speed[vec_generalized_advantage_estimate-False-32-512] 0.7563ms 0.7068ms 1.4149 KOps/s 1.4665 KOps/s $\color{#d91a1a}-3.52\%$
test_dqn_speed[False-None] 1.7042ms 1.5313ms 653.0193 Ops/s 654.1960 Ops/s $\color{#d91a1a}-0.18\%$
test_dqn_speed[False-backward] 2.2275ms 2.1776ms 459.2257 Ops/s 456.7836 Ops/s $\color{#35bf28}+0.53\%$
test_dqn_speed[True-None] 1.2691ms 0.5779ms 1.7304 KOps/s 1.7379 KOps/s $\color{#d91a1a}-0.43\%$
test_dqn_speed[True-backward] 1.1756ms 1.1136ms 897.9808 Ops/s 920.2857 Ops/s $\color{#d91a1a}-2.42\%$
test_dqn_speed[reduce-overhead-None] 0.6696ms 0.5993ms 1.6687 KOps/s 1.6668 KOps/s $\color{#35bf28}+0.11\%$
test_ddpg_speed[False-None] 3.4565ms 2.8966ms 345.2309 Ops/s 344.6884 Ops/s $\color{#35bf28}+0.16\%$
test_ddpg_speed[False-backward] 4.6512ms 4.2377ms 235.9763 Ops/s 236.9833 Ops/s $\color{#d91a1a}-0.42\%$
test_ddpg_speed[True-None] 1.5001ms 1.3191ms 758.1044 Ops/s 759.1369 Ops/s $\color{#d91a1a}-0.14\%$
test_ddpg_speed[True-backward] 2.4093ms 2.3506ms 425.4306 Ops/s 426.6445 Ops/s $\color{#d91a1a}-0.28\%$
test_ddpg_speed[reduce-overhead-None] 1.7581ms 1.3298ms 751.9844 Ops/s 753.1660 Ops/s $\color{#d91a1a}-0.16\%$
test_sac_speed[False-None] 8.9690ms 8.3981ms 119.0747 Ops/s 119.6360 Ops/s $\color{#d91a1a}-0.47\%$
test_sac_speed[False-backward] 12.0086ms 11.5877ms 86.2987 Ops/s 87.0131 Ops/s $\color{#d91a1a}-0.82\%$
test_sac_speed[True-None] 1.9902ms 1.7997ms 555.6627 Ops/s 554.8104 Ops/s $\color{#35bf28}+0.15\%$
test_sac_speed[True-backward] 3.5775ms 3.4267ms 291.8290 Ops/s 290.5914 Ops/s $\color{#35bf28}+0.43\%$
test_sac_speed[reduce-overhead-None] 19.6462ms 10.9994ms 90.9140 Ops/s 89.9366 Ops/s $\color{#35bf28}+1.09\%$
test_redq_deprec_speed[False-None] 10.0744ms 9.3597ms 106.8415 Ops/s 106.4448 Ops/s $\color{#35bf28}+0.37\%$
test_redq_deprec_speed[False-backward] 12.9058ms 12.5428ms 79.7267 Ops/s 80.0975 Ops/s $\color{#d91a1a}-0.46\%$
test_redq_deprec_speed[True-None] 2.9100ms 2.4981ms 400.3002 Ops/s 399.2739 Ops/s $\color{#35bf28}+0.26\%$
test_redq_deprec_speed[True-backward] 4.1513ms 4.0836ms 244.8813 Ops/s 237.3992 Ops/s $\color{#35bf28}+3.15\%$
test_redq_deprec_speed[reduce-overhead-None] 16.3911ms 10.0087ms 99.9132 Ops/s 100.1635 Ops/s $\color{#d91a1a}-0.25\%$
test_td3_speed[False-None] 8.5558ms 8.3365ms 119.9551 Ops/s 114.6523 Ops/s $\color{#35bf28}+4.63\%$
test_td3_speed[False-backward] 11.1594ms 10.6596ms 93.8120 Ops/s 91.9886 Ops/s $\color{#35bf28}+1.98\%$
test_td3_speed[True-None] 1.6736ms 1.6135ms 619.7695 Ops/s 618.7020 Ops/s $\color{#35bf28}+0.17\%$
test_td3_speed[True-backward] 3.1959ms 3.0676ms 325.9850 Ops/s 307.4743 Ops/s $\textbf{\color{#35bf28}+6.02\%}$
test_td3_speed[reduce-overhead-None] 72.8567ms 24.7258ms 40.4436 Ops/s 39.8571 Ops/s $\color{#35bf28}+1.47\%$
test_cql_speed[False-None] 18.0885ms 17.4007ms 57.4689 Ops/s 57.7281 Ops/s $\color{#d91a1a}-0.45\%$
test_cql_speed[False-backward] 23.9344ms 22.7337ms 43.9876 Ops/s 43.4164 Ops/s $\color{#35bf28}+1.32\%$
test_cql_speed[True-None] 3.5056ms 3.3514ms 298.3843 Ops/s 309.2655 Ops/s $\color{#d91a1a}-3.52\%$
test_cql_speed[True-backward] 5.8083ms 5.4539ms 183.3567 Ops/s 180.9588 Ops/s $\color{#35bf28}+1.33\%$
test_cql_speed[reduce-overhead-None] 19.0171ms 11.9491ms 83.6886 Ops/s 82.7305 Ops/s $\color{#35bf28}+1.16\%$
test_a2c_speed[False-None] 4.0511ms 3.2512ms 307.5744 Ops/s 308.3398 Ops/s $\color{#d91a1a}-0.25\%$
test_a2c_speed[False-backward] 6.8568ms 6.4394ms 155.2938 Ops/s 153.2819 Ops/s $\color{#35bf28}+1.31\%$
test_a2c_speed[True-None] 1.4823ms 1.3217ms 756.5998 Ops/s 749.0466 Ops/s $\color{#35bf28}+1.01\%$
test_a2c_speed[True-backward] 3.1956ms 3.1028ms 322.2913 Ops/s 337.2081 Ops/s $\color{#d91a1a}-4.42\%$
test_a2c_speed[reduce-overhead-None] 1.0587ms 0.9867ms 1.0135 KOps/s 1.0176 KOps/s $\color{#d91a1a}-0.41\%$
test_ppo_speed[False-None] 4.0960ms 3.8899ms 257.0774 Ops/s 262.3012 Ops/s $\color{#d91a1a}-1.99\%$
test_ppo_speed[False-backward] 7.4877ms 7.2995ms 136.9957 Ops/s 142.9171 Ops/s $\color{#d91a1a}-4.14\%$
test_ppo_speed[True-None] 1.6684ms 1.4263ms 701.0961 Ops/s 704.3291 Ops/s $\color{#d91a1a}-0.46\%$
test_ppo_speed[True-backward] 3.1278ms 3.0577ms 327.0387 Ops/s 305.6006 Ops/s $\textbf{\color{#35bf28}+7.02\%}$
test_ppo_speed[reduce-overhead-None] 1.1264ms 1.0387ms 962.7528 Ops/s 927.5657 Ops/s $\color{#35bf28}+3.79\%$
test_reinforce_speed[False-None] 2.4024ms 2.2827ms 438.0731 Ops/s 435.3758 Ops/s $\color{#35bf28}+0.62\%$
test_reinforce_speed[False-backward] 3.7417ms 3.3171ms 301.4643 Ops/s 288.1385 Ops/s $\color{#35bf28}+4.62\%$
test_reinforce_speed[True-None] 1.4037ms 1.2834ms 779.1863 Ops/s 791.2317 Ops/s $\color{#d91a1a}-1.52\%$
test_reinforce_speed[True-backward] 2.9071ms 2.8618ms 349.4315 Ops/s 335.4034 Ops/s $\color{#35bf28}+4.18\%$
test_reinforce_speed[reduce-overhead-None] 0.4515s 10.3703ms 96.4295 Ops/s 103.4872 Ops/s $\textbf{\color{#d91a1a}-6.82\%}$
test_iql_speed[False-None] 10.0428ms 9.5396ms 104.8258 Ops/s 105.3766 Ops/s $\color{#d91a1a}-0.52\%$
test_iql_speed[False-backward] 13.7422ms 13.2646ms 75.3888 Ops/s 75.5955 Ops/s $\color{#d91a1a}-0.27\%$
test_iql_speed[True-None] 2.4906ms 2.1634ms 462.2259 Ops/s 460.1564 Ops/s $\color{#35bf28}+0.45\%$
test_iql_speed[True-backward] 4.8566ms 4.6802ms 213.6641 Ops/s 208.7999 Ops/s $\color{#35bf28}+2.33\%$
test_iql_speed[reduce-overhead-None] 18.1877ms 10.6586ms 93.8212 Ops/s 95.3685 Ops/s $\color{#d91a1a}-1.62\%$
test_rb_sample[TensorDictReplayBuffer-ListStorage-RandomSampler-4000] 6.4236ms 6.0245ms 165.9895 Ops/s 167.2136 Ops/s $\color{#d91a1a}-0.73\%$
test_rb_sample[TensorDictReplayBuffer-LazyMemmapStorage-RandomSampler-10000] 0.9586ms 0.3109ms 3.2167 KOps/s 3.0035 KOps/s $\textbf{\color{#35bf28}+7.10\%}$
test_rb_sample[TensorDictReplayBuffer-LazyTensorStorage-RandomSampler-10000] 0.7119ms 0.3084ms 3.2422 KOps/s 3.5237 KOps/s $\textbf{\color{#d91a1a}-7.99\%}$
test_rb_sample[TensorDictReplayBuffer-ListStorage-SamplerWithoutReplacement-4000] 6.1662ms 5.9065ms 169.3049 Ops/s 172.4523 Ops/s $\color{#d91a1a}-1.83\%$
test_rb_sample[TensorDictReplayBuffer-LazyMemmapStorage-SamplerWithoutReplacement-10000] 0.6911ms 0.2806ms 3.5640 KOps/s 2.9128 KOps/s $\textbf{\color{#35bf28}+22.36\%}$
test_rb_sample[TensorDictReplayBuffer-LazyTensorStorage-SamplerWithoutReplacement-10000] 0.5610ms 0.2618ms 3.8193 KOps/s 3.0514 KOps/s $\textbf{\color{#35bf28}+25.17\%}$
test_rb_sample[TensorDictReplayBuffer-LazyMemmapStorage-sampler6-10000] 1.6099ms 1.2561ms 796.1411 Ops/s 736.3407 Ops/s $\textbf{\color{#35bf28}+8.12\%}$
test_rb_sample[TensorDictReplayBuffer-LazyTensorStorage-sampler7-10000] 1.6897ms 1.3100ms 763.3626 Ops/s 856.4126 Ops/s $\textbf{\color{#d91a1a}-10.87\%}$
test_rb_sample[TensorDictPrioritizedReplayBuffer-ListStorage-None-4000] 6.2498ms 6.0869ms 164.2883 Ops/s 169.1256 Ops/s $\color{#d91a1a}-2.86\%$
test_rb_sample[TensorDictPrioritizedReplayBuffer-LazyMemmapStorage-None-10000] 2.3670ms 0.4910ms 2.0366 KOps/s 2.3183 KOps/s $\textbf{\color{#d91a1a}-12.15\%}$
test_rb_sample[TensorDictPrioritizedReplayBuffer-LazyTensorStorage-None-10000] 0.7360ms 0.5049ms 1.9804 KOps/s 2.4075 KOps/s $\textbf{\color{#d91a1a}-17.74\%}$
test_rb_iterate[TensorDictReplayBuffer-ListStorage-RandomSampler-4000] 6.0684ms 5.9709ms 167.4795 Ops/s 172.2128 Ops/s $\color{#d91a1a}-2.75\%$
test_rb_iterate[TensorDictReplayBuffer-LazyMemmapStorage-RandomSampler-10000] 1.8464ms 0.3517ms 2.8433 KOps/s 3.5011 KOps/s $\textbf{\color{#d91a1a}-18.79\%}$
test_rb_iterate[TensorDictReplayBuffer-LazyTensorStorage-RandomSampler-10000] 0.6547ms 0.3429ms 2.9166 KOps/s 3.6854 KOps/s $\textbf{\color{#d91a1a}-20.86\%}$
test_rb_iterate[TensorDictReplayBuffer-ListStorage-SamplerWithoutReplacement-4000] 6.1843ms 5.9108ms 169.1824 Ops/s 173.0765 Ops/s $\color{#d91a1a}-2.25\%$
test_rb_iterate[TensorDictReplayBuffer-LazyMemmapStorage-SamplerWithoutReplacement-10000] 0.5742ms 0.3002ms 3.3311 KOps/s 3.1577 KOps/s $\textbf{\color{#35bf28}+5.49\%}$
test_rb_iterate[TensorDictReplayBuffer-LazyTensorStorage-SamplerWithoutReplacement-10000] 0.7344ms 0.2981ms 3.3551 KOps/s 3.3390 KOps/s $\color{#35bf28}+0.48\%$
test_rb_iterate[TensorDictPrioritizedReplayBuffer-ListStorage-None-4000] 6.2421ms 6.1065ms 163.7596 Ops/s 166.8099 Ops/s $\color{#d91a1a}-1.83\%$
test_rb_iterate[TensorDictPrioritizedReplayBuffer-LazyMemmapStorage-None-10000] 0.8270ms 0.5151ms 1.9412 KOps/s 628.8264 Ops/s $\textbf{\color{#35bf28}+208.71\%}$
test_rb_iterate[TensorDictPrioritizedReplayBuffer-LazyTensorStorage-None-10000] 0.7613ms 0.5113ms 1.9558 KOps/s 1.9874 KOps/s $\color{#d91a1a}-1.59\%$
test_rb_populate[TensorDictReplayBuffer-ListStorage-RandomSampler-400] 0.6493s 17.9953ms 55.5700 Ops/s 195.2799 Ops/s $\textbf{\color{#d91a1a}-71.54\%}$
test_rb_populate[TensorDictReplayBuffer-LazyMemmapStorage-RandomSampler-400] 12.8473ms 2.0207ms 494.8884 Ops/s 561.8926 Ops/s $\textbf{\color{#d91a1a}-11.92\%}$
test_rb_populate[TensorDictReplayBuffer-LazyTensorStorage-RandomSampler-400] 7.3088ms 1.2276ms 814.5903 Ops/s 1.0633 KOps/s $\textbf{\color{#d91a1a}-23.39\%}$
test_rb_populate[TensorDictReplayBuffer-ListStorage-SamplerWithoutReplacement-400] 6.5479ms 5.0432ms 198.2868 Ops/s 195.1380 Ops/s $\color{#35bf28}+1.61\%$
test_rb_populate[TensorDictReplayBuffer-LazyMemmapStorage-SamplerWithoutReplacement-400] 13.0155ms 2.1312ms 469.2155 Ops/s 523.9608 Ops/s $\textbf{\color{#d91a1a}-10.45\%}$
test_rb_populate[TensorDictReplayBuffer-LazyTensorStorage-SamplerWithoutReplacement-400] 3.1807ms 1.1877ms 841.9945 Ops/s 1.0700 KOps/s $\textbf{\color{#d91a1a}-21.31\%}$
test_rb_populate[TensorDictPrioritizedReplayBuffer-ListStorage-None-400] 0.5800s 16.7512ms 59.6974 Ops/s 50.9932 Ops/s $\textbf{\color{#35bf28}+17.07\%}$
test_rb_populate[TensorDictPrioritizedReplayBuffer-LazyMemmapStorage-None-400] 8.3425ms 2.1123ms 473.4215 Ops/s 504.9190 Ops/s $\textbf{\color{#d91a1a}-6.24\%}$
test_rb_populate[TensorDictPrioritizedReplayBuffer-LazyTensorStorage-None-400] 1.8821ms 1.1050ms 905.0111 Ops/s 911.6802 Ops/s $\color{#d91a1a}-0.73\%$
test_rb_extend_sample[ReplayBuffer-LazyTensorStorage-RandomSampler-10000-10000-100-True] 38.8016ms 35.7201ms 27.9954 Ops/s 27.4065 Ops/s $\color{#35bf28}+2.15\%$
test_rb_extend_sample[ReplayBuffer-LazyTensorStorage-RandomSampler-10000-10000-100-False] 19.9808ms 18.1314ms 55.1530 Ops/s 55.2283 Ops/s $\color{#d91a1a}-0.14\%$
test_rb_extend_sample[ReplayBuffer-LazyTensorStorage-RandomSampler-100000-10000-100-True] 39.8093ms 36.9912ms 27.0334 Ops/s 26.9669 Ops/s $\color{#35bf28}+0.25\%$
test_rb_extend_sample[ReplayBuffer-LazyTensorStorage-RandomSampler-100000-10000-100-False] 20.5393ms 18.4647ms 54.1575 Ops/s 54.8590 Ops/s $\color{#d91a1a}-1.28\%$
test_rb_extend_sample[ReplayBuffer-LazyTensorStorage-RandomSampler-1000000-10000-100-True] 40.5833ms 38.7985ms 25.7742 Ops/s 25.6723 Ops/s $\color{#35bf28}+0.40\%$
test_rb_extend_sample[ReplayBuffer-LazyTensorStorage-RandomSampler-1000000-10000-100-False] 21.7671ms 20.2500ms 49.3827 Ops/s 50.6132 Ops/s $\color{#d91a1a}-2.43\%$
test_storage_write_lazystack[50-img_shape0-small] 1.0006ms 0.2221ms 4.5016 KOps/s 4.4977 KOps/s $\color{#35bf28}+0.09\%$
test_storage_write_lazystack[100-img_shape1-atari] 1.5845ms 1.4013ms 713.6302 Ops/s 725.8364 Ops/s $\color{#d91a1a}-1.68\%$
test_storage_write_lazystack[100-img_shape2-large_img] 2.7472ms 2.3519ms 425.1898 Ops/s 432.4285 Ops/s $\color{#d91a1a}-1.67\%$
test_storage_write_lazystack[200-img_shape3-large_batch] 3.1111ms 2.9436ms 339.7219 Ops/s 336.1664 Ops/s $\color{#35bf28}+1.06\%$
test_storage_write_contiguous[50-img_shape0-small] 0.5140ms 0.1510ms 6.6210 KOps/s 6.5392 KOps/s $\color{#35bf28}+1.25\%$
test_storage_write_contiguous[100-img_shape1-atari] 0.3709ms 0.2003ms 4.9930 KOps/s 4.3056 KOps/s $\textbf{\color{#35bf28}+15.97\%}$
test_storage_write_contiguous[100-img_shape2-large_img] 2.1094ms 1.7096ms 584.9481 Ops/s 575.5107 Ops/s $\color{#35bf28}+1.64\%$
test_storage_write_contiguous[200-img_shape3-large_batch] 1.4558ms 1.2548ms 796.9086 Ops/s 713.0464 Ops/s $\textbf{\color{#35bf28}+11.76\%}$
test_collector_stack_then_write[50-img_shape0-small] 1.3163ms 1.1446ms 873.6625 Ops/s 872.4753 Ops/s $\color{#35bf28}+0.14\%$
test_collector_stack_then_write[100-img_shape1-atari] 3.7373ms 3.6305ms 275.4460 Ops/s 274.3290 Ops/s $\color{#35bf28}+0.41\%$
test_collector_stack_then_write[100-img_shape2-large_img] 5.8865ms 5.7090ms 175.1618 Ops/s 175.3049 Ops/s $\color{#d91a1a}-0.08\%$
test_collector_stack_then_write[200-img_shape3-large_batch] 8.8719ms 7.1431ms 139.9955 Ops/s 140.4597 Ops/s $\color{#d91a1a}-0.33\%$
test_collector_lazystack_then_write[50-img_shape0-small] 0.4511ms 0.2763ms 3.6187 KOps/s 3.6681 KOps/s $\color{#d91a1a}-1.35\%$
test_collector_lazystack_then_write[100-img_shape1-atari] 1.7248ms 1.5557ms 642.8094 Ops/s 670.5309 Ops/s $\color{#d91a1a}-4.13\%$
test_collector_lazystack_then_write[100-img_shape2-large_img] 2.8950ms 2.4709ms 404.7124 Ops/s 407.9006 Ops/s $\color{#d91a1a}-0.78\%$
test_collector_lazystack_then_write[200-img_shape3-large_batch] 3.5055ms 3.1685ms 315.6065 Ops/s 314.4630 Ops/s $\color{#35bf28}+0.36\%$
test_collector_without_rb[100-img_shape0-atari] 34.6886ms 34.1017ms 29.3240 Ops/s 29.1473 Ops/s $\color{#35bf28}+0.61\%$
test_collector_without_rb[200-img_shape1-large_batch] 68.0964ms 67.1940ms 14.8823 Ops/s 14.9753 Ops/s $\color{#d91a1a}-0.62\%$
test_collector_with_rb[100-img_shape0-atari] 39.6858ms 39.0313ms 25.6205 Ops/s 25.9016 Ops/s $\color{#d91a1a}-1.09\%$
test_collector_with_rb[200-img_shape1-large_batch] 77.6058ms 76.6970ms 13.0383 Ops/s 13.2611 Ops/s $\color{#d91a1a}-1.68\%$
test_collector_without_rb_cuda[100-img_shape0-atari] 57.7638ms 57.2541ms 17.4660 Ops/s 17.9752 Ops/s $\color{#d91a1a}-2.83\%$
test_collector_without_rb_cuda[200-img_shape1-large_batch] 0.1151s 0.1144s 8.7393 Ops/s 9.0191 Ops/s $\color{#d91a1a}-3.10\%$
test_collector_with_rb_cuda[100-img_shape0-atari] 59.3615ms 57.9909ms 17.2441 Ops/s 17.3108 Ops/s $\color{#d91a1a}-0.39\%$
test_collector_with_rb_cuda[200-img_shape1-large_batch] 0.1183s 0.1160s 8.6194 Ops/s 8.6667 Ops/s $\color{#d91a1a}-0.55\%$

@vmoens vmoens merged commit bec4498 into main Feb 7, 2026
295 of 296 checks passed
@vmoens vmoens deleted the ci/cpu-only-wheel-builds branch February 7, 2026 21:51
Sign up for free to join this conversation on GitHub. Already have an account? Sign in to comment

Reviewers

No reviews

Assignees

No one assigned

Labels

CI Has to do with CI setup (e.g. wheels & builds, tests...) ciflow/binaries/all Build all binaries ciflow/rocm-mi300 CLA Signed This label is managed by the Facebook bot. Authors need to sign the CLA before a PR can be reviewed. module: rocm

Projects

None yet

Milestone

No milestone

Development

Successfully merging this pull request may close these issues.

1 participant

@vmoens