Skip to content

[CI] Selective PR test matrix gated by changed-files + ciflow/* labels#3674

Open
vmoens wants to merge 1 commit intomainfrom
ci/run-selected-tests
Open

[CI] Selective PR test matrix gated by changed-files + ciflow/* labels#3674
vmoens wants to merge 1 commit intomainfrom
ci/run-selected-tests

Conversation

@vmoens
Copy link
Copy Markdown
Collaborator

@vmoens vmoens commented Apr 27, 2026

Summary

PRs currently run a heavy matrix on every push: 5 Python versions × CPU, 3 GPU shards, tests-olddeps, tests-optdeps, tests-stable-gpu (3 more shards) + distributed, plus SOTA / tutorials / Windows. This PR prunes that matrix on PRs to only what the diff demands, with ciflow/* labels as opt-in escape hatches. The full matrix is unchanged on push to main / nightly / release/*.

The label-gating pattern already used by test-linux-libs.yml, test-linux-llm.yml, and test-linux-habitat.yml is the model — this PR extends it to the heavy workflows and adds a small prepare job that maps the diff to which GPU shards / peer workflows should run.

Behaviour

Track PR default Full trigger
tests-cpu Python 3.12 only push / ciflow/cpu-matrix / ciflow/full
tests-gpu only the shard(s) whose paths changed (shard 3 fallback) push / ciflow/gpu / ciflow/full
tests-gpu-distributed only if distributed code changed push / ciflow/distributed / ciflow/full
tests-olddeps skipped push / ciflow/olddeps / ciflow/full
tests-optdeps skipped push / ciflow/optdeps / ciflow/full
tests-stable-gpu* skipped push / ciflow/stable / ciflow/full
test-linux-sota only if sota-implementations/** or torchrl/trainers/** changed push / ciflow/sota / ciflow/full
test-linux-tutorials only if tutorials/** changed push / ciflow/tutorials / ciflow/full
test-windows-optdepts only if torchrl/** / test/** / setup.py / pyproject.toml changed push / ciflow/windows / ciflow/full

Docs-only / tutorials/-only / sota-implementations/-only PRs skip the relevant heavy workflows entirely via paths-ignore:.

Implementation

  • .github/scripts/ci_decide.sh — centralised precedence logic (push > ciflow/full > per-track labels + file flags). Unit-tested by test_ci_decide.sh (40 assertions, all passing locally).
  • .github/workflows/test-linux.yml — new prepare job using tj-actions/changed-files@v45 to compute shard1/shard2/shard3/distributed change flags. All test jobs gain needs: prepare + if: gates; tests-cpu and tests-gpu matrices are now driven by fromJSON(needs.prepare.outputs.…).
  • Peer workflows (test-linux-sota.yml, test-linux-tutorials.yml, test-windows-optdepts.yml) — small changes job + label/file if: gate.
  • run_all.sh — new optional TORCHRL_TEST_PATHS env var that short-circuits the shard dispatch and passes its value verbatim to pytest. Useful for workflow_dispatch runs and local reproduction.
  • .github/labels.yml + create_labels.sh — 10 new ciflow/* labels registered (already created on the repo).
  • .github/CI.md — end-to-end docs for contributors.

Test plan

  • `bash .github/scripts/test_ci_decide.sh` → 40 passed, 0 failed
  • `actionlint` clean on all 4 modified workflows
  • `bash -n` clean on `run_all.sh`
  • All 10 `ciflow/*` labels created on `pytorch/rl`
  • After merge, open a draft PR touching only `torchrl/envs/transforms/vec_norm.py`. Expect: `tests-cpu` matrix is `["3.12"]`; `tests-gpu` matrix is `["1"]`; `tests-olddeps` / `tests-optdeps` / `tests-stable-gpu*` / sota / tutorials / windows all skipped.
  • After merge, open a draft PR editing only a `.md`. Expect: `test-linux.yml`, `test-linux-sota.yml`, `test-linux-tutorials.yml`, `test-windows-optdepts.yml` do not enqueue.
  • After merge, open a draft PR with `ciflow/full`. Expect: pre-change baseline matrix runs.
  • Verify the next push to `main` runs the full matrix (no regression vs. today's behaviour).

Rollback

All changes are additive `needs:` / `if:` conditions plus a new `prepare` job; reverting this PR restores current behaviour.

@vmoens
[CI] Selective PR test matrix gated by changed-files + ciflow/* labels
1ed531b 
PRs now run only the tracks relevant to the diff (Python 3.12 CPU + a
filtered subset of GPU shards), with the full matrix preserved on push to
main / nightly / release/*. Maintainers can pull individual tracks back in
on a PR by applying ciflow/full, ciflow/cpu-matrix, ciflow/gpu, ciflow/stable,
ciflow/olddeps, ciflow/optdeps, ciflow/distributed, ciflow/sota,
ciflow/tutorials, or ciflow/windows.

- New prepare job in test-linux.yml uses tj-actions/changed-files to compute
  shard flags; .github/scripts/ci_decide.sh centralises the precedence logic
  (push > ciflow/full > per-track labels + file flags) and is unit-tested by
  test_ci_decide.sh (40 assertions).
- Heavy peer workflows (test-linux-sota, test-linux-tutorials,
  test-windows-optdepts) gain a small changes job + if: gate using the same
  pattern.
- run_all.sh gains an optional TORCHRL_TEST_PATHS env var that short-circuits
  the shard dispatch and runs pytest against arbitrary paths/expressions, for
  workflow_dispatch and local reproduction.
- New ciflow/* labels registered in labels.yml and added to create_labels.sh.
- .github/CI.md documents the gating model end-to-end.
@pytorch-bot
Copy link
Copy Markdown

pytorch-bot Bot commented Apr 27, 2026 •
edited
Loading

🔗 Helpful Links

🧪 See artifacts and rendered test results at hud.pytorch.org/pr/pytorch/rl/3674

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

❗ 1 Active SEVs

There are 1 currently active SEVs. If your PR is affected, please view them below:

❌ 2 New Failures

As of commit 1ed531b with merge base 306cb92 (image):

NEW FAILURES - The following jobs have 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 Apr 27, 2026
@github-actions github-actions Bot added the CI Has to do with CI setup (e.g. wheels & builds, tests...) label Apr 27, 2026
@github-actions
Copy link
Copy Markdown
Contributor

github-actions Bot commented Apr 27, 2026

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

Total Benchmarks: 172. Improved: $\large\color{#35bf28}16$. Worsened: $\large\color{#d91a1a}7$.

Expand to view detailed results
Name Max Mean Ops Ops on Repo HEAD Change
test_tensor_to_bytestream_speed[pickle] 79.2588μs 78.6170μs 12.7199 KOps/s 12.4515 KOps/s $\color{#35bf28}+2.16\%$
test_tensor_to_bytestream_speed[torch.save] 0.1380ms 0.1372ms 7.2903 KOps/s 7.2745 KOps/s $\color{#35bf28}+0.22\%$
test_tensor_to_bytestream_speed[untyped_storage] 0.1041s 0.1038s 9.6306 Ops/s 9.7535 Ops/s $\color{#d91a1a}-1.26\%$
test_tensor_to_bytestream_speed[numpy] 2.4741μs 2.4687μs 405.0666 KOps/s 401.0089 KOps/s $\color{#35bf28}+1.01\%$
test_tensor_to_bytestream_speed[safetensors] 38.1069μs 37.9405μs 26.3571 KOps/s 28.1660 KOps/s $\textbf{\color{#d91a1a}-6.42\%}$
test_simple 0.5314s 0.5309s 1.8836 Ops/s 1.7817 Ops/s $\textbf{\color{#35bf28}+5.72\%}$
test_transformed 1.0678s 1.0613s 0.9422 Ops/s 0.9145 Ops/s $\color{#35bf28}+3.03\%$
test_serial 1.6902s 1.6813s 0.5948 Ops/s 0.5963 Ops/s $\color{#d91a1a}-0.26\%$
test_parallel 1.1185s 1.0298s 0.9711 Ops/s 0.9951 Ops/s $\color{#d91a1a}-2.42\%$
test_step_mdp_speed[True-True-True-True-True] 0.4568ms 41.3272μs 24.1971 KOps/s 24.6501 KOps/s $\color{#d91a1a}-1.84\%$
test_step_mdp_speed[True-True-True-True-False] 0.4516ms 22.8452μs 43.7729 KOps/s 44.6476 KOps/s $\color{#d91a1a}-1.96\%$
test_step_mdp_speed[True-True-True-False-True] 58.5810μs 23.5595μs 42.4457 KOps/s 43.4322 KOps/s $\color{#d91a1a}-2.27\%$
test_step_mdp_speed[True-True-True-False-False] 0.4489ms 12.4289μs 80.4574 KOps/s 79.6056 KOps/s $\color{#35bf28}+1.07\%$
test_step_mdp_speed[True-True-False-True-True] 0.4739ms 43.0758μs 23.2149 KOps/s 22.5174 KOps/s $\color{#35bf28}+3.10\%$
test_step_mdp_speed[True-True-False-True-False] 54.6210μs 24.8474μs 40.2456 KOps/s 40.0571 KOps/s $\color{#35bf28}+0.47\%$
test_step_mdp_speed[True-True-False-False-True] 0.4513ms 25.6519μs 38.9834 KOps/s 38.5796 KOps/s $\color{#35bf28}+1.05\%$
test_step_mdp_speed[True-True-False-False-False] 0.4352ms 15.2032μs 65.7755 KOps/s 65.6278 KOps/s $\color{#35bf28}+0.23\%$
test_step_mdp_speed[True-False-True-True-True] 70.4220μs 46.1096μs 21.6875 KOps/s 21.5726 KOps/s $\color{#35bf28}+0.53\%$
test_step_mdp_speed[True-False-True-True-False] 0.4579ms 27.6516μs 36.1643 KOps/s 36.6850 KOps/s $\color{#d91a1a}-1.42\%$
test_step_mdp_speed[True-False-True-False-True] 0.4520ms 25.5616μs 39.1212 KOps/s 37.9302 KOps/s $\color{#35bf28}+3.14\%$
test_step_mdp_speed[True-False-True-False-False] 48.4110μs 15.1873μs 65.8447 KOps/s 65.2775 KOps/s $\color{#35bf28}+0.87\%$
test_step_mdp_speed[True-False-False-True-True] 0.4808ms 48.5107μs 20.6140 KOps/s 20.3249 KOps/s $\color{#35bf28}+1.42\%$
test_step_mdp_speed[True-False-False-True-False] 0.4606ms 30.4417μs 32.8497 KOps/s 32.9522 KOps/s $\color{#d91a1a}-0.31\%$
test_step_mdp_speed[True-False-False-False-True] 0.4574ms 28.6272μs 34.9318 KOps/s 34.8378 KOps/s $\color{#35bf28}+0.27\%$
test_step_mdp_speed[True-False-False-False-False] 54.0010μs 17.7051μs 56.4808 KOps/s 56.3036 KOps/s $\color{#35bf28}+0.31\%$
test_step_mdp_speed[False-True-True-True-True] 0.4689ms 45.9624μs 21.7569 KOps/s 21.3154 KOps/s $\color{#35bf28}+2.07\%$
test_step_mdp_speed[False-True-True-True-False] 2.3549ms 27.7872μs 35.9878 KOps/s 36.0292 KOps/s $\color{#d91a1a}-0.11\%$
test_step_mdp_speed[False-True-True-False-True] 0.4534ms 29.5408μs 33.8514 KOps/s 33.2144 KOps/s $\color{#35bf28}+1.92\%$
test_step_mdp_speed[False-True-True-False-False] 54.8310μs 16.4862μs 60.6570 KOps/s 59.9761 KOps/s $\color{#35bf28}+1.14\%$
test_step_mdp_speed[False-True-False-True-True] 0.4706ms 48.0782μs 20.7994 KOps/s 20.2121 KOps/s $\color{#35bf28}+2.91\%$
test_step_mdp_speed[False-True-False-True-False] 0.4514ms 29.7689μs 33.5921 KOps/s 32.6416 KOps/s $\color{#35bf28}+2.91\%$
test_step_mdp_speed[False-True-False-False-True] 64.4310μs 31.3534μs 31.8944 KOps/s 31.9804 KOps/s $\color{#d91a1a}-0.27\%$
test_step_mdp_speed[False-True-False-False-False] 0.4407ms 18.8495μs 53.0519 KOps/s 52.4271 KOps/s $\color{#35bf28}+1.19\%$
test_step_mdp_speed[False-False-True-True-True] 0.1250ms 49.8532μs 20.0589 KOps/s 19.2780 KOps/s $\color{#35bf28}+4.05\%$
test_step_mdp_speed[False-False-True-True-False] 65.5310μs 32.3879μs 30.8758 KOps/s 30.7874 KOps/s $\color{#35bf28}+0.29\%$
test_step_mdp_speed[False-False-True-False-True] 59.6710μs 31.2557μs 31.9942 KOps/s 31.3337 KOps/s $\color{#35bf28}+2.11\%$
test_step_mdp_speed[False-False-True-False-False] 95.3820μs 18.5985μs 53.7678 KOps/s 52.3202 KOps/s $\color{#35bf28}+2.77\%$
test_step_mdp_speed[False-False-False-True-True] 0.4646ms 52.3980μs 19.0847 KOps/s 18.8150 KOps/s $\color{#35bf28}+1.43\%$
test_step_mdp_speed[False-False-False-True-False] 0.4576ms 35.2812μs 28.3437 KOps/s 28.1009 KOps/s $\color{#35bf28}+0.86\%$
test_step_mdp_speed[False-False-False-False-True] 62.0510μs 33.1475μs 30.1682 KOps/s 30.2262 KOps/s $\color{#d91a1a}-0.19\%$
test_step_mdp_speed[False-False-False-False-False] 48.7010μs 21.5029μs 46.5054 KOps/s 46.4907 KOps/s $\color{#35bf28}+0.03\%$
test_non_tensor_env_rollout_speed[1000-single-True] 0.7078s 0.7007s 1.4271 Ops/s 1.3622 Ops/s $\color{#35bf28}+4.77\%$
test_non_tensor_env_rollout_speed[1000-single-False] 0.6997s 0.5891s 1.6976 Ops/s 1.6627 Ops/s $\color{#35bf28}+2.10\%$
test_non_tensor_env_rollout_speed[1000-serial-no-buffers-True] 1.6918s 1.6094s 0.6213 Ops/s 0.6167 Ops/s $\color{#35bf28}+0.75\%$
test_non_tensor_env_rollout_speed[1000-serial-no-buffers-False] 1.4751s 1.3867s 0.7211 Ops/s 0.7140 Ops/s $\color{#35bf28}+1.00\%$
test_non_tensor_env_rollout_speed[1000-serial-buffers-True] 1.9441s 1.8535s 0.5395 Ops/s 0.5367 Ops/s $\color{#35bf28}+0.53\%$
test_non_tensor_env_rollout_speed[1000-serial-buffers-False] 1.7157s 1.6316s 0.6129 Ops/s 0.6062 Ops/s $\color{#35bf28}+1.11\%$
test_non_tensor_env_rollout_speed[1000-parallel-no-buffers-True] 4.6482s 4.5495s 0.2198 Ops/s 0.2184 Ops/s $\color{#35bf28}+0.65\%$
test_non_tensor_env_rollout_speed[1000-parallel-no-buffers-False] 4.4761s 4.3451s 0.2301 Ops/s 0.2306 Ops/s $\color{#d91a1a}-0.20\%$
test_non_tensor_env_rollout_speed[1000-parallel-buffers-True] 1.9173s 1.8422s 0.5428 Ops/s 0.5399 Ops/s $\color{#35bf28}+0.55\%$
test_non_tensor_env_rollout_speed[1000-parallel-buffers-False] 1.6975s 1.5874s 0.6300 Ops/s 0.6438 Ops/s $\color{#d91a1a}-2.14\%$
test_values[generalized_advantage_estimate-True-True] 10.1695ms 9.8800ms 101.2146 Ops/s 99.7525 Ops/s $\color{#35bf28}+1.47\%$
test_values[vec_generalized_advantage_estimate-True-True] 19.8827ms 17.6602ms 56.6246 Ops/s 56.9592 Ops/s $\color{#d91a1a}-0.59\%$
test_values[td0_return_estimate-False-False] 0.2174ms 0.1269ms 7.8831 KOps/s 7.9852 KOps/s $\color{#d91a1a}-1.28\%$
test_values[td1_return_estimate-False-False] 28.3369ms 26.6391ms 37.5388 Ops/s 37.5453 Ops/s $\color{#d91a1a}-0.02\%$
test_values[vec_td1_return_estimate-False-False] 18.6128ms 17.7487ms 56.3421 Ops/s 56.6305 Ops/s $\color{#d91a1a}-0.51\%$
test_values[td_lambda_return_estimate-True-False] 41.9088ms 39.8318ms 25.1056 Ops/s 25.3695 Ops/s $\color{#d91a1a}-1.04\%$
test_values[vec_td_lambda_return_estimate-True-False] 18.0571ms 17.6434ms 56.6783 Ops/s 56.5201 Ops/s $\color{#35bf28}+0.28\%$
test_gae_speed[generalized_advantage_estimate-False-1-512] 8.9240ms 8.7017ms 114.9203 Ops/s 114.6010 Ops/s $\color{#35bf28}+0.28\%$
test_gae_speed[vec_generalized_advantage_estimate-True-1-512] 1.7533ms 1.4717ms 679.4653 Ops/s 676.5096 Ops/s $\color{#35bf28}+0.44\%$
test_gae_speed[vec_generalized_advantage_estimate-False-1-512] 0.6435ms 0.4025ms 2.4848 KOps/s 2.4945 KOps/s $\color{#d91a1a}-0.39\%$
test_gae_speed[vec_generalized_advantage_estimate-True-32-512] 35.5986ms 34.7957ms 28.7392 Ops/s 29.1356 Ops/s $\color{#d91a1a}-1.36\%$
test_gae_speed[vec_generalized_advantage_estimate-False-32-512] 1.9649ms 1.6999ms 588.2539 Ops/s 589.4209 Ops/s $\color{#d91a1a}-0.20\%$
test_dqn_speed[False-None] 1.4589ms 1.3617ms 734.3720 Ops/s 732.0193 Ops/s $\color{#35bf28}+0.32\%$
test_dqn_speed[False-backward] 1.9333ms 1.8575ms 538.3509 Ops/s 534.0848 Ops/s $\color{#35bf28}+0.80\%$
test_dqn_speed[True-None] 0.9628ms 0.5590ms 1.7889 KOps/s 1.7826 KOps/s $\color{#35bf28}+0.35\%$
test_dqn_speed[True-backward] 1.0534ms 1.0112ms 988.8764 Ops/s 967.1607 Ops/s $\color{#35bf28}+2.25\%$
test_dqn_speed[reduce-overhead-None] 0.9606ms 0.5395ms 1.8534 KOps/s 1.8136 KOps/s $\color{#35bf28}+2.19\%$
test_ddpg_speed[False-None] 3.1573ms 2.7983ms 357.3567 Ops/s 363.5839 Ops/s $\color{#d91a1a}-1.71\%$
test_ddpg_speed[False-backward] 4.3850ms 3.9991ms 250.0577 Ops/s 254.7583 Ops/s $\color{#d91a1a}-1.85\%$
test_ddpg_speed[True-None] 1.6126ms 1.4393ms 694.7610 Ops/s 689.6279 Ops/s $\color{#35bf28}+0.74\%$
test_ddpg_speed[True-backward] 2.4937ms 2.4221ms 412.8698 Ops/s 413.1064 Ops/s $\color{#d91a1a}-0.06\%$
test_ddpg_speed[reduce-overhead-None] 1.7942ms 1.4123ms 708.0615 Ops/s 700.8302 Ops/s $\color{#35bf28}+1.03\%$
test_sac_speed[False-None] 8.5084ms 7.8825ms 126.8633 Ops/s 125.1330 Ops/s $\color{#35bf28}+1.38\%$
test_sac_speed[False-backward] 11.8353ms 11.0752ms 90.2918 Ops/s 90.2084 Ops/s $\color{#35bf28}+0.09\%$
test_sac_speed[True-None] 2.5563ms 2.1622ms 462.4984 Ops/s 448.5891 Ops/s $\color{#35bf28}+3.10\%$
test_sac_speed[True-backward] 4.2517ms 4.0736ms 245.4839 Ops/s 225.7241 Ops/s $\textbf{\color{#35bf28}+8.75\%}$
test_sac_speed[reduce-overhead-None] 2.5515ms 2.1432ms 466.5921 Ops/s 453.0109 Ops/s $\color{#35bf28}+3.00\%$
test_redq_speed[False-None] 14.2016ms 10.4605ms 95.5976 Ops/s 95.3473 Ops/s $\color{#35bf28}+0.26\%$
test_redq_speed[False-backward] 19.5858ms 17.6427ms 56.6807 Ops/s 56.4498 Ops/s $\color{#35bf28}+0.41\%$
test_redq_speed[True-None] 5.0087ms 4.5810ms 218.2924 Ops/s 211.2615 Ops/s $\color{#35bf28}+3.33\%$
test_redq_speed[reduce-overhead-None] 5.6995ms 4.6120ms 216.8243 Ops/s 219.0296 Ops/s $\color{#d91a1a}-1.01\%$
test_redq_deprec_speed[False-None] 11.6729ms 10.9799ms 91.0752 Ops/s 90.5841 Ops/s $\color{#35bf28}+0.54\%$
test_redq_deprec_speed[False-backward] 17.1397ms 15.6393ms 63.9416 Ops/s 63.2256 Ops/s $\color{#35bf28}+1.13\%$
test_redq_deprec_speed[True-None] 4.1510ms 3.6766ms 271.9869 Ops/s 270.5650 Ops/s $\color{#35bf28}+0.53\%$
test_redq_deprec_speed[True-backward] 7.9315ms 7.4721ms 133.8307 Ops/s 131.7414 Ops/s $\color{#35bf28}+1.59\%$
test_redq_deprec_speed[reduce-overhead-None] 3.7562ms 3.6148ms 276.6431 Ops/s 277.9835 Ops/s $\color{#d91a1a}-0.48\%$
test_td3_speed[False-None] 8.1628ms 7.9058ms 126.4894 Ops/s 126.0876 Ops/s $\color{#35bf28}+0.32\%$
test_td3_speed[False-backward] 10.8828ms 10.6687ms 93.7323 Ops/s 93.6880 Ops/s $\color{#35bf28}+0.05\%$
test_td3_speed[True-None] 1.8720ms 1.8267ms 547.4404 Ops/s 548.7830 Ops/s $\color{#d91a1a}-0.24\%$
test_td3_speed[True-backward] 3.7031ms 3.5906ms 278.5030 Ops/s 239.0137 Ops/s $\textbf{\color{#35bf28}+16.52\%}$
test_td3_speed[reduce-overhead-None] 1.8212ms 1.7833ms 560.7608 Ops/s 560.8812 Ops/s $\color{#d91a1a}-0.02\%$
test_cql_speed[False-None] 29.4684ms 25.8870ms 38.6295 Ops/s 38.4896 Ops/s $\color{#35bf28}+0.36\%$
test_cql_speed[False-backward] 37.9435ms 35.0793ms 28.5068 Ops/s 28.6536 Ops/s $\color{#d91a1a}-0.51\%$
test_cql_speed[True-None] 12.7688ms 12.4812ms 80.1202 Ops/s 79.8636 Ops/s $\color{#35bf28}+0.32\%$
test_cql_speed[True-backward] 18.4967ms 18.0967ms 55.2587 Ops/s 55.3521 Ops/s $\color{#d91a1a}-0.17\%$
test_cql_speed[reduce-overhead-None] 13.1021ms 12.5353ms 79.7747 Ops/s 78.8754 Ops/s $\color{#35bf28}+1.14\%$
test_a2c_speed[False-None] 5.6306ms 5.3919ms 185.4636 Ops/s 187.4510 Ops/s $\color{#d91a1a}-1.06\%$
test_a2c_speed[False-backward] 12.1634ms 11.7925ms 84.7994 Ops/s 85.2986 Ops/s $\color{#d91a1a}-0.59\%$
test_a2c_speed[True-None] 4.2438ms 3.8902ms 257.0546 Ops/s 256.2717 Ops/s $\color{#35bf28}+0.31\%$
test_a2c_speed[True-backward] 8.8783ms 8.6707ms 115.3311 Ops/s 112.2510 Ops/s $\color{#35bf28}+2.74\%$
test_a2c_speed[reduce-overhead-None] 4.1940ms 3.8760ms 258.0012 Ops/s 258.4619 Ops/s $\color{#d91a1a}-0.18\%$
test_ppo_speed[False-None] 6.3673ms 5.9932ms 166.8558 Ops/s 167.0476 Ops/s $\color{#d91a1a}-0.11\%$
test_ppo_speed[False-backward] 12.8590ms 12.5257ms 79.8360 Ops/s 79.9074 Ops/s $\color{#d91a1a}-0.09\%$
test_ppo_speed[True-None] 4.0081ms 3.8305ms 261.0636 Ops/s 255.1276 Ops/s $\color{#35bf28}+2.33\%$
test_ppo_speed[True-backward] 9.1290ms 8.7966ms 113.6802 Ops/s 114.3283 Ops/s $\color{#d91a1a}-0.57\%$
test_ppo_speed[reduce-overhead-None] 4.2569ms 3.8090ms 262.5376 Ops/s 262.0976 Ops/s $\color{#35bf28}+0.17\%$
test_reinforce_speed[False-None] 4.9448ms 4.5143ms 221.5166 Ops/s 218.1581 Ops/s $\color{#35bf28}+1.54\%$
test_reinforce_speed[False-backward] 7.6204ms 7.3594ms 135.8800 Ops/s 135.0911 Ops/s $\color{#35bf28}+0.58\%$
test_reinforce_speed[True-None] 3.2258ms 3.0042ms 332.8703 Ops/s 331.3155 Ops/s $\color{#35bf28}+0.47\%$
test_reinforce_speed[True-backward] 8.2007ms 7.9260ms 126.1670 Ops/s 118.4799 Ops/s $\textbf{\color{#35bf28}+6.49\%}$
test_reinforce_speed[reduce-overhead-None] 3.3687ms 2.9863ms 334.8663 Ops/s 334.0941 Ops/s $\color{#35bf28}+0.23\%$
test_iql_speed[False-None] 21.2454ms 19.9893ms 50.0267 Ops/s 50.4064 Ops/s $\color{#d91a1a}-0.75\%$
test_iql_speed[False-backward] 30.9458ms 30.2098ms 33.1019 Ops/s 33.3635 Ops/s $\color{#d91a1a}-0.78\%$
test_iql_speed[True-None] 9.0016ms 8.5673ms 116.7228 Ops/s 116.9477 Ops/s $\color{#d91a1a}-0.19\%$
test_iql_speed[True-backward] 17.0685ms 16.8164ms 59.4658 Ops/s 58.0324 Ops/s $\color{#35bf28}+2.47\%$
test_iql_speed[reduce-overhead-None] 9.0671ms 8.5718ms 116.6611 Ops/s 110.2414 Ops/s $\textbf{\color{#35bf28}+5.82\%}$
test_rb_sample[TensorDictReplayBuffer-ListStorage-RandomSampler-4000] 6.0777ms 5.8627ms 170.5697 Ops/s 166.8266 Ops/s $\color{#35bf28}+2.24\%$
test_rb_sample[TensorDictReplayBuffer-LazyMemmapStorage-RandomSampler-10000] 3.0482ms 0.2911ms 3.4350 KOps/s 3.0634 KOps/s $\textbf{\color{#35bf28}+12.13\%}$
test_rb_sample[TensorDictReplayBuffer-LazyTensorStorage-RandomSampler-10000] 0.5256ms 0.2703ms 3.7001 KOps/s 3.0756 KOps/s $\textbf{\color{#35bf28}+20.31\%}$
test_rb_sample[TensorDictReplayBuffer-ListStorage-SamplerWithoutReplacement-4000] 5.9609ms 5.6508ms 176.9660 Ops/s 174.6431 Ops/s $\color{#35bf28}+1.33\%$
test_rb_sample[TensorDictReplayBuffer-LazyMemmapStorage-SamplerWithoutReplacement-10000] 1.8018ms 0.2893ms 3.4570 KOps/s 3.0468 KOps/s $\textbf{\color{#35bf28}+13.46\%}$
test_rb_sample[TensorDictReplayBuffer-LazyTensorStorage-SamplerWithoutReplacement-10000] 0.4987ms 0.2684ms 3.7256 KOps/s 2.8844 KOps/s $\textbf{\color{#35bf28}+29.16\%}$
test_rb_sample[TensorDictReplayBuffer-LazyMemmapStorage-sampler6-10000] 1.7037ms 1.2653ms 790.3235 Ops/s 745.7970 Ops/s $\textbf{\color{#35bf28}+5.97\%}$
test_rb_sample[TensorDictReplayBuffer-LazyTensorStorage-sampler7-10000] 1.6164ms 1.1628ms 860.0074 Ops/s 780.1531 Ops/s $\textbf{\color{#35bf28}+10.24\%}$
test_rb_sample[TensorDictPrioritizedReplayBuffer-ListStorage-None-4000] 10.1620ms 5.9355ms 168.4779 Ops/s 169.9821 Ops/s $\color{#d91a1a}-0.88\%$
test_rb_sample[TensorDictPrioritizedReplayBuffer-LazyMemmapStorage-None-10000] 0.8468ms 0.4270ms 2.3418 KOps/s 1.8763 KOps/s $\textbf{\color{#35bf28}+24.81\%}$
test_rb_sample[TensorDictPrioritizedReplayBuffer-LazyTensorStorage-None-10000] 0.7317ms 0.4088ms 2.4464 KOps/s 2.0777 KOps/s $\textbf{\color{#35bf28}+17.75\%}$
test_rb_iterate[TensorDictReplayBuffer-ListStorage-RandomSampler-4000] 5.6775ms 5.5873ms 178.9773 Ops/s 174.3806 Ops/s $\color{#35bf28}+2.64\%$
test_rb_iterate[TensorDictReplayBuffer-LazyMemmapStorage-RandomSampler-10000] 0.9570s 1.9446ms 514.2356 Ops/s 2.8024 KOps/s $\textbf{\color{#d91a1a}-81.65\%}$
test_rb_iterate[TensorDictReplayBuffer-LazyTensorStorage-RandomSampler-10000] 0.5552ms 0.3517ms 2.8437 KOps/s 2.9982 KOps/s $\textbf{\color{#d91a1a}-5.15\%}$
test_rb_iterate[TensorDictReplayBuffer-ListStorage-SamplerWithoutReplacement-4000] 5.8369ms 5.6082ms 178.3101 Ops/s 175.5063 Ops/s $\color{#35bf28}+1.60\%$
test_rb_iterate[TensorDictReplayBuffer-LazyMemmapStorage-SamplerWithoutReplacement-10000] 1.6849ms 0.3448ms 2.8999 KOps/s 2.8895 KOps/s $\color{#35bf28}+0.36\%$
test_rb_iterate[TensorDictReplayBuffer-LazyTensorStorage-SamplerWithoutReplacement-10000] 0.6402ms 0.3115ms 3.2107 KOps/s 2.8556 KOps/s $\textbf{\color{#35bf28}+12.43\%}$
test_rb_iterate[TensorDictPrioritizedReplayBuffer-ListStorage-None-4000] 6.0099ms 5.8366ms 171.3333 Ops/s 170.9969 Ops/s $\color{#35bf28}+0.20\%$
test_rb_iterate[TensorDictPrioritizedReplayBuffer-LazyMemmapStorage-None-10000] 1.1078ms 0.4902ms 2.0401 KOps/s 1.9724 KOps/s $\color{#35bf28}+3.43\%$
test_rb_iterate[TensorDictPrioritizedReplayBuffer-LazyTensorStorage-None-10000] 0.6224ms 0.4386ms 2.2801 KOps/s 2.1942 KOps/s $\color{#35bf28}+3.91\%$
test_rb_populate[TensorDictReplayBuffer-ListStorage-RandomSampler-400] 6.4483ms 4.9905ms 200.3809 Ops/s 46.5901 Ops/s $\textbf{\color{#35bf28}+330.09\%}$
test_rb_populate[TensorDictReplayBuffer-LazyMemmapStorage-RandomSampler-400] 11.0998ms 2.1832ms 458.0409 Ops/s 516.8839 Ops/s $\textbf{\color{#d91a1a}-11.38\%}$
test_rb_populate[TensorDictReplayBuffer-LazyTensorStorage-RandomSampler-400] 1.1481ms 0.9132ms 1.0951 KOps/s 805.8727 Ops/s $\textbf{\color{#35bf28}+35.89\%}$
test_rb_populate[TensorDictReplayBuffer-ListStorage-SamplerWithoutReplacement-400] 0.6523s 17.9946ms 55.5721 Ops/s 197.8238 Ops/s $\textbf{\color{#d91a1a}-71.91\%}$
test_rb_populate[TensorDictReplayBuffer-LazyMemmapStorage-SamplerWithoutReplacement-400] 11.5047ms 1.9948ms 501.3119 Ops/s 539.7421 Ops/s $\textbf{\color{#d91a1a}-7.12\%}$
test_rb_populate[TensorDictReplayBuffer-LazyTensorStorage-SamplerWithoutReplacement-400] 0.9893ms 0.8658ms 1.1550 KOps/s 1.1180 KOps/s $\color{#35bf28}+3.31\%$
test_rb_populate[TensorDictPrioritizedReplayBuffer-ListStorage-None-400] 10.6443ms 5.2846ms 189.2278 Ops/s 190.3927 Ops/s $\color{#d91a1a}-0.61\%$
test_rb_populate[TensorDictPrioritizedReplayBuffer-LazyMemmapStorage-None-400] 3.9350ms 1.9035ms 525.3411 Ops/s 524.0821 Ops/s $\color{#35bf28}+0.24\%$
test_rb_populate[TensorDictPrioritizedReplayBuffer-LazyTensorStorage-None-400] 2.9639ms 1.0689ms 935.5753 Ops/s 933.2646 Ops/s $\color{#35bf28}+0.25\%$
test_rb_extend_sample[ReplayBuffer-LazyTensorStorage-RandomSampler-10000-10000-100-True] 42.7497ms 38.5126ms 25.9655 Ops/s 25.6081 Ops/s $\color{#35bf28}+1.40\%$
test_rb_extend_sample[ReplayBuffer-LazyTensorStorage-RandomSampler-10000-10000-100-False] 19.3769ms 17.9086ms 55.8390 Ops/s 54.7816 Ops/s $\color{#35bf28}+1.93\%$
test_rb_extend_sample[ReplayBuffer-LazyTensorStorage-RandomSampler-100000-10000-100-True] 41.6192ms 39.5878ms 25.2603 Ops/s 24.7764 Ops/s $\color{#35bf28}+1.95\%$
test_rb_extend_sample[ReplayBuffer-LazyTensorStorage-RandomSampler-100000-10000-100-False] 20.1495ms 18.4732ms 54.1326 Ops/s 54.2453 Ops/s $\color{#d91a1a}-0.21\%$
test_rb_extend_sample[ReplayBuffer-LazyTensorStorage-RandomSampler-1000000-10000-100-True] 42.5024ms 40.9969ms 24.3921 Ops/s 23.3577 Ops/s $\color{#35bf28}+4.43\%$
test_rb_extend_sample[ReplayBuffer-LazyTensorStorage-RandomSampler-1000000-10000-100-False] 21.4578ms 19.7553ms 50.6192 Ops/s 51.5081 Ops/s $\color{#d91a1a}-1.73\%$
test_storage_write_lazystack[50-img_shape0-small] 0.8634ms 0.2182ms 4.5831 KOps/s 4.3652 KOps/s $\color{#35bf28}+4.99\%$
test_storage_write_lazystack[100-img_shape1-atari] 1.5520ms 1.3766ms 726.4142 Ops/s 711.9213 Ops/s $\color{#35bf28}+2.04\%$
test_storage_write_lazystack[100-img_shape2-large_img] 2.7521ms 2.3145ms 432.0672 Ops/s 432.9618 Ops/s $\color{#d91a1a}-0.21\%$
test_storage_write_lazystack[200-img_shape3-large_batch] 3.0844ms 2.8861ms 346.4870 Ops/s 338.0426 Ops/s $\color{#35bf28}+2.50\%$
test_storage_write_contiguous[50-img_shape0-small] 0.2155ms 0.1332ms 7.5099 KOps/s 7.4800 KOps/s $\color{#35bf28}+0.40\%$
test_storage_write_contiguous[100-img_shape1-atari] 0.3353ms 0.1822ms 5.4886 KOps/s 5.2315 KOps/s $\color{#35bf28}+4.92\%$
test_storage_write_contiguous[100-img_shape2-large_img] 1.9050ms 1.7767ms 562.8393 Ops/s 577.4944 Ops/s $\color{#d91a1a}-2.54\%$
test_storage_write_contiguous[200-img_shape3-large_batch] 1.3975ms 1.2768ms 783.2076 Ops/s 780.7633 Ops/s $\color{#35bf28}+0.31\%$
test_collector_stack_then_write[50-img_shape0-small] 1.2073ms 1.0964ms 912.1090 Ops/s 909.1688 Ops/s $\color{#35bf28}+0.32\%$
test_collector_stack_then_write[100-img_shape1-atari] 7.5149ms 3.5359ms 282.8097 Ops/s 278.1414 Ops/s $\color{#35bf28}+1.68\%$
test_collector_stack_then_write[100-img_shape2-large_img] 11.1685ms 5.6906ms 175.7278 Ops/s 178.2876 Ops/s $\color{#d91a1a}-1.44\%$
test_collector_stack_then_write[200-img_shape3-large_batch] 14.9593ms 7.0552ms 141.7404 Ops/s 143.2483 Ops/s $\color{#d91a1a}-1.05\%$
test_collector_lazystack_then_write[50-img_shape0-small] 0.4420ms 0.2758ms 3.6258 KOps/s 3.6472 KOps/s $\color{#d91a1a}-0.59\%$
test_collector_lazystack_then_write[100-img_shape1-atari] 1.7026ms 1.5136ms 660.6624 Ops/s 661.2937 Ops/s $\color{#d91a1a}-0.10\%$
test_collector_lazystack_then_write[100-img_shape2-large_img] 2.5748ms 2.4444ms 409.0909 Ops/s 411.2209 Ops/s $\color{#d91a1a}-0.52\%$
test_collector_lazystack_then_write[200-img_shape3-large_batch] 3.4247ms 3.0886ms 323.7662 Ops/s 316.1881 Ops/s $\color{#35bf28}+2.40\%$
test_collector_without_rb[100-img_shape0-atari] 32.2848ms 31.8345ms 31.4125 Ops/s 31.0017 Ops/s $\color{#35bf28}+1.33\%$
test_collector_without_rb[200-img_shape1-large_batch] 64.2693ms 62.6401ms 15.9642 Ops/s 15.4474 Ops/s $\color{#35bf28}+3.35\%$
test_collector_with_rb[100-img_shape0-atari] 0.7122s 60.5420ms 16.5175 Ops/s 26.7551 Ops/s $\textbf{\color{#d91a1a}-38.26\%}$
test_collector_with_rb[200-img_shape1-large_batch] 72.2606ms 71.6638ms 13.9541 Ops/s 13.7638 Ops/s $\color{#35bf28}+1.38\%$

@github-actions
Copy link
Copy Markdown
Contributor

github-actions Bot commented Apr 27, 2026

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

Total Benchmarks: 172. Improved: $\large\color{#35bf28}16$. Worsened: $\large\color{#d91a1a}4$.

Expand to view detailed results
Name Max Mean Ops Ops on Repo HEAD Change
test_tensor_to_bytestream_speed[pickle] 82.0306μs 80.8881μs 12.3628 KOps/s 12.4316 KOps/s $\color{#d91a1a}-0.55\%$
test_tensor_to_bytestream_speed[torch.save] 0.1419ms 0.1414ms 7.0710 KOps/s 7.0053 KOps/s $\color{#35bf28}+0.94\%$
test_tensor_to_bytestream_speed[untyped_storage] 0.1118s 0.1113s 8.9847 Ops/s 8.9324 Ops/s $\color{#35bf28}+0.59\%$
test_tensor_to_bytestream_speed[numpy] 2.7088μs 2.7041μs 369.8030 KOps/s 377.8188 KOps/s $\color{#d91a1a}-2.12\%$
test_tensor_to_bytestream_speed[safetensors] 37.3690μs 37.0870μs 26.9636 KOps/s 26.2202 KOps/s $\color{#35bf28}+2.84\%$
test_simple 0.7913s 0.7895s 1.2666 Ops/s 1.2100 Ops/s $\color{#35bf28}+4.68\%$
test_transformed 1.3925s 1.3838s 0.7226 Ops/s 0.7024 Ops/s $\color{#35bf28}+2.88\%$
test_serial 2.3188s 2.3166s 0.4317 Ops/s 0.4242 Ops/s $\color{#35bf28}+1.75\%$
test_parallel 1.9330s 1.8384s 0.5439 Ops/s 0.5570 Ops/s $\color{#d91a1a}-2.35\%$
test_step_mdp_speed[True-True-True-True-True] 0.1506ms 40.5242μs 24.6766 KOps/s 23.6588 KOps/s $\color{#35bf28}+4.30\%$
test_step_mdp_speed[True-True-True-True-False] 51.5100μs 22.8872μs 43.6925 KOps/s 43.8075 KOps/s $\color{#d91a1a}-0.26\%$
test_step_mdp_speed[True-True-True-False-True] 52.0410μs 23.3435μs 42.8385 KOps/s 43.2028 KOps/s $\color{#d91a1a}-0.84\%$
test_step_mdp_speed[True-True-True-False-False] 37.6600μs 12.6896μs 78.8044 KOps/s 79.8916 KOps/s $\color{#d91a1a}-1.36\%$
test_step_mdp_speed[True-True-False-True-True] 67.6210μs 43.8373μs 22.8116 KOps/s 22.7503 KOps/s $\color{#35bf28}+0.27\%$
test_step_mdp_speed[True-True-False-True-False] 51.8510μs 25.4447μs 39.3010 KOps/s 40.0821 KOps/s $\color{#d91a1a}-1.95\%$
test_step_mdp_speed[True-True-False-False-True] 54.2110μs 25.8040μs 38.7537 KOps/s 38.7063 KOps/s $\color{#35bf28}+0.12\%$
test_step_mdp_speed[True-True-False-False-False] 45.4800μs 15.3071μs 65.3290 KOps/s 65.8244 KOps/s $\color{#d91a1a}-0.75\%$
test_step_mdp_speed[True-False-True-True-True] 79.8410μs 47.1072μs 21.2282 KOps/s 21.1963 KOps/s $\color{#35bf28}+0.15\%$
test_step_mdp_speed[True-False-True-True-False] 63.1110μs 28.2127μs 35.4450 KOps/s 35.8260 KOps/s $\color{#d91a1a}-1.06\%$
test_step_mdp_speed[True-False-True-False-True] 57.0110μs 26.4272μs 37.8397 KOps/s 38.6676 KOps/s $\color{#d91a1a}-2.14\%$
test_step_mdp_speed[True-False-True-False-False] 44.6900μs 15.3484μs 65.1532 KOps/s 65.9954 KOps/s $\color{#d91a1a}-1.28\%$
test_step_mdp_speed[True-False-False-True-True] 88.9310μs 49.4873μs 20.2072 KOps/s 20.3055 KOps/s $\color{#d91a1a}-0.48\%$
test_step_mdp_speed[True-False-False-True-False] 57.9610μs 30.6947μs 32.5789 KOps/s 33.3069 KOps/s $\color{#d91a1a}-2.19\%$
test_step_mdp_speed[True-False-False-False-True] 55.6010μs 29.3089μs 34.1193 KOps/s 34.5927 KOps/s $\color{#d91a1a}-1.37\%$
test_step_mdp_speed[True-False-False-False-False] 51.9510μs 17.7883μs 56.2168 KOps/s 56.7239 KOps/s $\color{#d91a1a}-0.89\%$
test_step_mdp_speed[False-True-True-True-True] 83.7320μs 46.9616μs 21.2940 KOps/s 21.4754 KOps/s $\color{#d91a1a}-0.84\%$
test_step_mdp_speed[False-True-True-True-False] 2.2320ms 28.5570μs 35.0177 KOps/s 35.9777 KOps/s $\color{#d91a1a}-2.67\%$
test_step_mdp_speed[False-True-True-False-True] 60.8710μs 29.9585μs 33.3795 KOps/s 32.9669 KOps/s $\color{#35bf28}+1.25\%$
test_step_mdp_speed[False-True-True-False-False] 42.1410μs 16.9440μs 59.0181 KOps/s 58.6474 KOps/s $\color{#35bf28}+0.63\%$
test_step_mdp_speed[False-True-False-True-True] 76.7010μs 49.2680μs 20.2971 KOps/s 20.0147 KOps/s $\color{#35bf28}+1.41\%$
test_step_mdp_speed[False-True-False-True-False] 70.5710μs 30.8570μs 32.4075 KOps/s 32.7704 KOps/s $\color{#d91a1a}-1.11\%$
test_step_mdp_speed[False-True-False-False-True] 67.9510μs 32.2027μs 31.0533 KOps/s 30.9145 KOps/s $\color{#35bf28}+0.45\%$
test_step_mdp_speed[False-True-False-False-False] 45.7010μs 19.5403μs 51.1764 KOps/s 51.1722 KOps/s $+0.01\%$
test_step_mdp_speed[False-False-True-True-True] 84.6410μs 52.7539μs 18.9560 KOps/s 19.2394 KOps/s $\color{#d91a1a}-1.47\%$
test_step_mdp_speed[False-False-True-True-False] 59.8010μs 33.5940μs 29.7672 KOps/s 30.4637 KOps/s $\color{#d91a1a}-2.29\%$
test_step_mdp_speed[False-False-True-False-True] 58.9510μs 32.4712μs 30.7965 KOps/s 31.5037 KOps/s $\color{#d91a1a}-2.24\%$
test_step_mdp_speed[False-False-True-False-False] 43.5210μs 19.4166μs 51.5023 KOps/s 51.6393 KOps/s $\color{#d91a1a}-0.27\%$
test_step_mdp_speed[False-False-False-True-True] 78.4810μs 55.2671μs 18.0939 KOps/s 18.3917 KOps/s $\color{#d91a1a}-1.62\%$
test_step_mdp_speed[False-False-False-True-False] 61.7710μs 35.9432μs 27.8217 KOps/s 28.4126 KOps/s $\color{#d91a1a}-2.08\%$
test_step_mdp_speed[False-False-False-False-True] 71.7920μs 34.4338μs 29.0412 KOps/s 29.0483 KOps/s $\color{#d91a1a}-0.02\%$
test_step_mdp_speed[False-False-False-False-False] 46.8610μs 22.0463μs 45.3590 KOps/s 45.7797 KOps/s $\color{#d91a1a}-0.92\%$
test_non_tensor_env_rollout_speed[1000-single-True] 0.7274s 0.7186s 1.3916 Ops/s 1.3411 Ops/s $\color{#35bf28}+3.76\%$
test_non_tensor_env_rollout_speed[1000-single-False] 0.7110s 0.6079s 1.6450 Ops/s 1.6493 Ops/s $\color{#d91a1a}-0.26\%$
test_non_tensor_env_rollout_speed[1000-serial-no-buffers-True] 1.7199s 1.6435s 0.6084 Ops/s 0.6084 Ops/s $-0.00\%$
test_non_tensor_env_rollout_speed[1000-serial-no-buffers-False] 1.5153s 1.4283s 0.7001 Ops/s 0.6985 Ops/s $\color{#35bf28}+0.23\%$
test_non_tensor_env_rollout_speed[1000-serial-buffers-True] 1.9872s 1.9023s 0.5257 Ops/s 0.5270 Ops/s $\color{#d91a1a}-0.25\%$
test_non_tensor_env_rollout_speed[1000-serial-buffers-False] 1.7668s 1.6774s 0.5962 Ops/s 0.5951 Ops/s $\color{#35bf28}+0.18\%$
test_non_tensor_env_rollout_speed[1000-parallel-no-buffers-True] 4.7515s 4.6615s 0.2145 Ops/s 0.2148 Ops/s $\color{#d91a1a}-0.13\%$
test_non_tensor_env_rollout_speed[1000-parallel-no-buffers-False] 4.5370s 4.4263s 0.2259 Ops/s 0.2239 Ops/s $\color{#35bf28}+0.92\%$
test_non_tensor_env_rollout_speed[1000-parallel-buffers-True] 1.9499s 1.8772s 0.5327 Ops/s 0.5246 Ops/s $\color{#35bf28}+1.54\%$
test_non_tensor_env_rollout_speed[1000-parallel-buffers-False] 1.6611s 1.5772s 0.6340 Ops/s 0.6374 Ops/s $\color{#d91a1a}-0.53\%$
test_values[generalized_advantage_estimate-True-True] 20.6902ms 20.3310ms 49.1859 Ops/s 49.5241 Ops/s $\color{#d91a1a}-0.68\%$
test_values[vec_generalized_advantage_estimate-True-True] 0.1366s 3.6561ms 273.5154 Ops/s 265.1883 Ops/s $\color{#35bf28}+3.14\%$
test_values[td0_return_estimate-False-False] 0.1233ms 85.2106μs 11.7356 KOps/s 12.0123 KOps/s $\color{#d91a1a}-2.30\%$
test_values[td1_return_estimate-False-False] 51.2510ms 48.7393ms 20.5173 Ops/s 20.1553 Ops/s $\color{#35bf28}+1.80\%$
test_values[vec_td1_return_estimate-False-False] 1.3488ms 1.0918ms 915.9276 Ops/s 913.2059 Ops/s $\color{#35bf28}+0.30\%$
test_values[td_lambda_return_estimate-True-False] 83.6943ms 81.6643ms 12.2453 Ops/s 12.2233 Ops/s $\color{#35bf28}+0.18\%$
test_values[vec_td_lambda_return_estimate-True-False] 1.3418ms 1.0941ms 913.9860 Ops/s 914.6745 Ops/s $\color{#d91a1a}-0.08\%$
test_gae_speed[generalized_advantage_estimate-False-1-512] 22.4782ms 22.1070ms 45.2346 Ops/s 48.5539 Ops/s $\textbf{\color{#d91a1a}-6.84\%}$
test_gae_speed[vec_generalized_advantage_estimate-True-1-512] 1.0144ms 0.7567ms 1.3215 KOps/s 1.3189 KOps/s $\color{#35bf28}+0.20\%$
test_gae_speed[vec_generalized_advantage_estimate-False-1-512] 0.7863ms 0.7029ms 1.4226 KOps/s 1.4677 KOps/s $\color{#d91a1a}-3.07\%$
test_gae_speed[vec_generalized_advantage_estimate-True-32-512] 1.6629ms 1.5080ms 663.1283 Ops/s 671.1680 Ops/s $\color{#d91a1a}-1.20\%$
test_gae_speed[vec_generalized_advantage_estimate-False-32-512] 0.7759ms 0.7150ms 1.3987 KOps/s 1.4352 KOps/s $\color{#d91a1a}-2.55\%$
test_dqn_speed[False-None] 1.7244ms 1.6024ms 624.0552 Ops/s 622.5630 Ops/s $\color{#35bf28}+0.24\%$
test_dqn_speed[False-backward] 2.4427ms 2.2708ms 440.3781 Ops/s 439.5773 Ops/s $\color{#35bf28}+0.18\%$
test_dqn_speed[True-None] 0.9273ms 0.6238ms 1.6030 KOps/s 1.5697 KOps/s $\color{#35bf28}+2.13\%$
test_dqn_speed[True-backward] 1.1971ms 1.1692ms 855.2579 Ops/s 835.1321 Ops/s $\color{#35bf28}+2.41\%$
test_dqn_speed[reduce-overhead-None] 0.6652ms 0.6123ms 1.6333 KOps/s 1.5350 KOps/s $\textbf{\color{#35bf28}+6.40\%}$
test_ddpg_speed[False-None] 3.4626ms 3.0229ms 330.8087 Ops/s 324.6221 Ops/s $\color{#35bf28}+1.91\%$
test_ddpg_speed[False-backward] 4.7699ms 4.3243ms 231.2507 Ops/s 230.2357 Ops/s $\color{#35bf28}+0.44\%$
test_ddpg_speed[True-None] 1.5291ms 1.3961ms 716.2822 Ops/s 712.5990 Ops/s $\color{#35bf28}+0.52\%$
test_ddpg_speed[True-backward] 2.5289ms 2.4195ms 413.3053 Ops/s 389.4797 Ops/s $\textbf{\color{#35bf28}+6.12\%}$
test_ddpg_speed[reduce-overhead-None] 1.4779ms 1.3866ms 721.1832 Ops/s 704.4889 Ops/s $\color{#35bf28}+2.37\%$
test_sac_speed[False-None] 8.9080ms 8.5036ms 117.5972 Ops/s 116.3390 Ops/s $\color{#35bf28}+1.08\%$
test_sac_speed[False-backward] 12.0144ms 11.5490ms 86.5877 Ops/s 85.6936 Ops/s $\color{#35bf28}+1.04\%$
test_sac_speed[True-None] 2.3042ms 1.9552ms 511.4681 Ops/s 508.6984 Ops/s $\color{#35bf28}+0.54\%$
test_sac_speed[True-backward] 4.1220ms 3.6314ms 275.3732 Ops/s 275.4775 Ops/s $\color{#d91a1a}-0.04\%$
test_sac_speed[reduce-overhead-None] 17.2688ms 10.1369ms 98.6494 Ops/s 96.3816 Ops/s $\color{#35bf28}+2.35\%$
test_redq_deprec_speed[False-None] 10.4135ms 9.5682ms 104.5134 Ops/s 104.0618 Ops/s $\color{#35bf28}+0.43\%$
test_redq_deprec_speed[False-backward] 13.1400ms 12.6798ms 78.8655 Ops/s 78.6265 Ops/s $\color{#35bf28}+0.30\%$
test_redq_deprec_speed[True-None] 3.1976ms 2.7687ms 361.1758 Ops/s 354.7120 Ops/s $\color{#35bf28}+1.82\%$
test_redq_deprec_speed[True-backward] 4.9688ms 4.4895ms 222.7401 Ops/s 222.2511 Ops/s $\color{#35bf28}+0.22\%$
test_redq_deprec_speed[reduce-overhead-None] 14.6501ms 9.7482ms 102.5828 Ops/s 101.7234 Ops/s $\color{#35bf28}+0.84\%$
test_td3_speed[False-None] 8.6662ms 8.3671ms 119.5155 Ops/s 118.4528 Ops/s $\color{#35bf28}+0.90\%$
test_td3_speed[False-backward] 11.4050ms 10.9988ms 90.9193 Ops/s 90.6641 Ops/s $\color{#35bf28}+0.28\%$
test_td3_speed[True-None] 1.7635ms 1.7187ms 581.8467 Ops/s 572.9943 Ops/s $\color{#35bf28}+1.54\%$
test_td3_speed[True-backward] 3.3861ms 3.3020ms 302.8476 Ops/s 296.8796 Ops/s $\color{#35bf28}+2.01\%$
test_td3_speed[reduce-overhead-None] 99.8028ms 26.2316ms 38.1220 Ops/s 37.4315 Ops/s $\color{#35bf28}+1.84\%$
test_cql_speed[False-None] 18.1539ms 17.7501ms 56.3378 Ops/s 55.7626 Ops/s $\color{#35bf28}+1.03\%$
test_cql_speed[False-backward] 23.7908ms 23.3569ms 42.8139 Ops/s 42.3647 Ops/s $\color{#35bf28}+1.06\%$
test_cql_speed[True-None] 3.6497ms 3.4883ms 286.6717 Ops/s 278.9512 Ops/s $\color{#35bf28}+2.77\%$
test_cql_speed[True-backward] 9.1024ms 6.1288ms 163.1633 Ops/s 168.0468 Ops/s $\color{#d91a1a}-2.91\%$
test_cql_speed[reduce-overhead-None] 17.5311ms 11.9765ms 83.4971 Ops/s 81.8423 Ops/s $\color{#35bf28}+2.02\%$
test_a2c_speed[False-None] 3.4612ms 3.3580ms 297.7983 Ops/s 298.7816 Ops/s $\color{#d91a1a}-0.33\%$
test_a2c_speed[False-backward] 6.9836ms 6.5040ms 153.7517 Ops/s 152.6939 Ops/s $\color{#35bf28}+0.69\%$
test_a2c_speed[True-None] 1.6771ms 1.4757ms 677.6575 Ops/s 685.4017 Ops/s $\color{#d91a1a}-1.13\%$
test_a2c_speed[True-backward] 3.4872ms 3.3215ms 301.0726 Ops/s 311.4539 Ops/s $\color{#d91a1a}-3.33\%$
test_a2c_speed[reduce-overhead-None] 1.1891ms 1.1190ms 893.6775 Ops/s 896.3309 Ops/s $\color{#d91a1a}-0.30\%$
test_ppo_speed[False-None] 4.2688ms 4.1001ms 243.8951 Ops/s 242.1110 Ops/s $\color{#35bf28}+0.74\%$
test_ppo_speed[False-backward] 7.8442ms 7.4548ms 134.1424 Ops/s 135.0358 Ops/s $\color{#d91a1a}-0.66\%$
test_ppo_speed[True-None] 1.7537ms 1.6332ms 612.3101 Ops/s 609.7996 Ops/s $\color{#35bf28}+0.41\%$
test_ppo_speed[True-backward] 3.6382ms 3.5377ms 282.6674 Ops/s 278.6589 Ops/s $\color{#35bf28}+1.44\%$
test_ppo_speed[reduce-overhead-None] 1.2979ms 1.1801ms 847.3887 Ops/s 830.2176 Ops/s $\color{#35bf28}+2.07\%$
test_reinforce_speed[False-None] 2.5151ms 2.4071ms 415.4458 Ops/s 415.3195 Ops/s $\color{#35bf28}+0.03\%$
test_reinforce_speed[False-backward] 3.8428ms 3.4379ms 290.8731 Ops/s 281.6058 Ops/s $\color{#35bf28}+3.29\%$
test_reinforce_speed[True-None] 1.5579ms 1.4766ms 677.2101 Ops/s 682.9928 Ops/s $\color{#d91a1a}-0.85\%$
test_reinforce_speed[True-backward] 3.3267ms 3.1974ms 312.7546 Ops/s 296.4748 Ops/s $\textbf{\color{#35bf28}+5.49\%}$
test_reinforce_speed[reduce-overhead-None] 0.6790s 10.4286ms 95.8897 Ops/s 111.2260 Ops/s $\textbf{\color{#d91a1a}-13.79\%}$
test_iql_speed[False-None] 9.9833ms 9.7583ms 102.4771 Ops/s 102.0052 Ops/s $\color{#35bf28}+0.46\%$
test_iql_speed[False-backward] 13.9747ms 13.5521ms 73.7890 Ops/s 72.4229 Ops/s $\color{#35bf28}+1.89\%$
test_iql_speed[True-None] 2.5723ms 2.3482ms 425.8583 Ops/s 420.0694 Ops/s $\color{#35bf28}+1.38\%$
test_iql_speed[True-backward] 5.4834ms 4.9782ms 200.8768 Ops/s 192.0583 Ops/s $\color{#35bf28}+4.59\%$
test_iql_speed[reduce-overhead-None] 16.7456ms 10.0445ms 99.5570 Ops/s 97.6947 Ops/s $\color{#35bf28}+1.91\%$
test_rb_sample[TensorDictReplayBuffer-ListStorage-RandomSampler-4000] 6.2009ms 6.0023ms 166.6039 Ops/s 163.6810 Ops/s $\color{#35bf28}+1.79\%$
test_rb_sample[TensorDictReplayBuffer-LazyMemmapStorage-RandomSampler-10000] 2.7019ms 0.2918ms 3.4273 KOps/s 2.9625 KOps/s $\textbf{\color{#35bf28}+15.69\%}$
test_rb_sample[TensorDictReplayBuffer-LazyTensorStorage-RandomSampler-10000] 0.5063ms 0.2698ms 3.7066 KOps/s 2.8376 KOps/s $\textbf{\color{#35bf28}+30.63\%}$
test_rb_sample[TensorDictReplayBuffer-ListStorage-SamplerWithoutReplacement-4000] 6.1141ms 5.8800ms 170.0676 Ops/s 170.4773 Ops/s $\color{#d91a1a}-0.24\%$
test_rb_sample[TensorDictReplayBuffer-LazyMemmapStorage-SamplerWithoutReplacement-10000] 1.9867ms 0.3334ms 2.9994 KOps/s 2.9793 KOps/s $\color{#35bf28}+0.68\%$
test_rb_sample[TensorDictReplayBuffer-LazyTensorStorage-SamplerWithoutReplacement-10000] 0.5630ms 0.3275ms 3.0533 KOps/s 3.1439 KOps/s $\color{#d91a1a}-2.88\%$
test_rb_sample[TensorDictReplayBuffer-LazyMemmapStorage-sampler6-10000] 1.7234ms 1.4391ms 694.8963 Ops/s 690.9860 Ops/s $\color{#35bf28}+0.57\%$
test_rb_sample[TensorDictReplayBuffer-LazyTensorStorage-sampler7-10000] 1.5969ms 1.3416ms 745.3512 Ops/s 770.1278 Ops/s $\color{#d91a1a}-3.22\%$
test_rb_sample[TensorDictPrioritizedReplayBuffer-ListStorage-None-4000] 9.5609ms 6.1180ms 163.4529 Ops/s 164.6113 Ops/s $\color{#d91a1a}-0.70\%$
test_rb_sample[TensorDictPrioritizedReplayBuffer-LazyMemmapStorage-None-10000] 0.9859ms 0.4382ms 2.2823 KOps/s 2.0234 KOps/s $\textbf{\color{#35bf28}+12.80\%}$
test_rb_sample[TensorDictPrioritizedReplayBuffer-LazyTensorStorage-None-10000] 0.7478ms 0.4165ms 2.4010 KOps/s 2.0054 KOps/s $\textbf{\color{#35bf28}+19.73\%}$
test_rb_iterate[TensorDictReplayBuffer-ListStorage-RandomSampler-4000] 5.9781ms 5.8379ms 171.2939 Ops/s 169.0529 Ops/s $\color{#35bf28}+1.33\%$
test_rb_iterate[TensorDictReplayBuffer-LazyMemmapStorage-RandomSampler-10000] 0.6867ms 0.2882ms 3.4695 KOps/s 2.5699 KOps/s $\textbf{\color{#35bf28}+35.01\%}$
test_rb_iterate[TensorDictReplayBuffer-LazyTensorStorage-RandomSampler-10000] 0.5286ms 0.2717ms 3.6808 KOps/s 2.7124 KOps/s $\textbf{\color{#35bf28}+35.70\%}$
test_rb_iterate[TensorDictReplayBuffer-ListStorage-SamplerWithoutReplacement-4000] 6.0715ms 5.8050ms 172.2639 Ops/s 168.9180 Ops/s $\color{#35bf28}+1.98\%$
test_rb_iterate[TensorDictReplayBuffer-LazyMemmapStorage-SamplerWithoutReplacement-10000] 1.5803ms 0.2866ms 3.4897 KOps/s 2.6057 KOps/s $\textbf{\color{#35bf28}+33.93\%}$
test_rb_iterate[TensorDictReplayBuffer-LazyTensorStorage-SamplerWithoutReplacement-10000] 0.6122ms 0.2885ms 3.4665 KOps/s 2.7053 KOps/s $\textbf{\color{#35bf28}+28.14\%}$
test_rb_iterate[TensorDictPrioritizedReplayBuffer-ListStorage-None-4000] 8.5142ms 6.0177ms 166.1755 Ops/s 165.5075 Ops/s $\color{#35bf28}+0.40\%$
test_rb_iterate[TensorDictPrioritizedReplayBuffer-LazyMemmapStorage-None-10000] 1.3170ms 0.5207ms 1.9206 KOps/s 1.8524 KOps/s $\color{#35bf28}+3.68\%$
test_rb_iterate[TensorDictPrioritizedReplayBuffer-LazyTensorStorage-None-10000] 0.8038ms 0.5091ms 1.9643 KOps/s 2.0093 KOps/s $\color{#d91a1a}-2.24\%$
test_rb_populate[TensorDictReplayBuffer-ListStorage-RandomSampler-400] 0.9783s 24.5379ms 40.7533 Ops/s 34.3034 Ops/s $\textbf{\color{#35bf28}+18.80\%}$
test_rb_populate[TensorDictReplayBuffer-LazyMemmapStorage-RandomSampler-400] 8.8227ms 2.0591ms 485.6603 Ops/s 514.8506 Ops/s $\textbf{\color{#d91a1a}-5.67\%}$
test_rb_populate[TensorDictReplayBuffer-LazyTensorStorage-RandomSampler-400] 3.5101ms 1.0355ms 965.7174 Ops/s 855.8421 Ops/s $\textbf{\color{#35bf28}+12.84\%}$
test_rb_populate[TensorDictReplayBuffer-ListStorage-SamplerWithoutReplacement-400] 7.6304ms 5.1959ms 192.4584 Ops/s 192.0031 Ops/s $\color{#35bf28}+0.24\%$
test_rb_populate[TensorDictReplayBuffer-LazyMemmapStorage-SamplerWithoutReplacement-400] 3.9477ms 1.8625ms 536.9015 Ops/s 560.4372 Ops/s $\color{#d91a1a}-4.20\%$
test_rb_populate[TensorDictReplayBuffer-LazyTensorStorage-SamplerWithoutReplacement-400] 1.1777ms 0.9769ms 1.0237 KOps/s 720.2249 Ops/s $\textbf{\color{#35bf28}+42.13\%}$
test_rb_populate[TensorDictPrioritizedReplayBuffer-ListStorage-None-400] 6.8210ms 5.2789ms 189.4323 Ops/s 187.0897 Ops/s $\color{#35bf28}+1.25\%$
test_rb_populate[TensorDictPrioritizedReplayBuffer-LazyMemmapStorage-None-400] 4.3144ms 2.0508ms 487.6109 Ops/s 465.6887 Ops/s $\color{#35bf28}+4.71\%$
test_rb_populate[TensorDictPrioritizedReplayBuffer-LazyTensorStorage-None-400] 3.8766ms 1.1754ms 850.7844 Ops/s 777.1068 Ops/s $\textbf{\color{#35bf28}+9.48\%}$
test_rb_extend_sample[ReplayBuffer-LazyTensorStorage-RandomSampler-10000-10000-100-True] 43.1204ms 39.8291ms 25.1073 Ops/s 24.9846 Ops/s $\color{#35bf28}+0.49\%$
test_rb_extend_sample[ReplayBuffer-LazyTensorStorage-RandomSampler-10000-10000-100-False] 19.8097ms 18.3935ms 54.3669 Ops/s 53.3061 Ops/s $\color{#35bf28}+1.99\%$
test_rb_extend_sample[ReplayBuffer-LazyTensorStorage-RandomSampler-100000-10000-100-True] 43.2940ms 41.1524ms 24.2999 Ops/s 17.3652 Ops/s $\textbf{\color{#35bf28}+39.93\%}$
test_rb_extend_sample[ReplayBuffer-LazyTensorStorage-RandomSampler-100000-10000-100-False] 20.0976ms 18.7048ms 53.4622 Ops/s 53.5695 Ops/s $\color{#d91a1a}-0.20\%$
test_rb_extend_sample[ReplayBuffer-LazyTensorStorage-RandomSampler-1000000-10000-100-True] 44.5805ms 42.9597ms 23.2776 Ops/s 23.2058 Ops/s $\color{#35bf28}+0.31\%$
test_rb_extend_sample[ReplayBuffer-LazyTensorStorage-RandomSampler-1000000-10000-100-False] 21.5574ms 20.4623ms 48.8704 Ops/s 49.2075 Ops/s $\color{#d91a1a}-0.69\%$
test_storage_write_lazystack[50-img_shape0-small] 0.8605ms 0.2209ms 4.5261 KOps/s 4.3805 KOps/s $\color{#35bf28}+3.32\%$
test_storage_write_lazystack[100-img_shape1-atari] 1.5684ms 1.4015ms 713.5079 Ops/s 714.8308 Ops/s $\color{#d91a1a}-0.19\%$
test_storage_write_lazystack[100-img_shape2-large_img] 2.6390ms 2.2901ms 436.6536 Ops/s 437.7063 Ops/s $\color{#d91a1a}-0.24\%$
test_storage_write_lazystack[200-img_shape3-large_batch] 3.0860ms 2.9017ms 344.6300 Ops/s 338.8066 Ops/s $\color{#35bf28}+1.72\%$
test_storage_write_contiguous[50-img_shape0-small] 0.2496ms 0.1687ms 5.9272 KOps/s 5.9764 KOps/s $\color{#d91a1a}-0.82\%$
test_storage_write_contiguous[100-img_shape1-atari] 0.3958ms 0.2740ms 3.6494 KOps/s 4.3753 KOps/s $\textbf{\color{#d91a1a}-16.59\%}$
test_storage_write_contiguous[100-img_shape2-large_img] 1.9403ms 1.7735ms 563.8670 Ops/s 560.3013 Ops/s $\color{#35bf28}+0.64\%$
test_storage_write_contiguous[200-img_shape3-large_batch] 1.5466ms 1.4072ms 710.6168 Ops/s 744.1700 Ops/s $\color{#d91a1a}-4.51\%$
test_collector_stack_then_write[50-img_shape0-small] 1.2750ms 1.1633ms 859.6579 Ops/s 858.7995 Ops/s $\color{#35bf28}+0.10\%$
test_collector_stack_then_write[100-img_shape1-atari] 4.1013ms 3.6067ms 277.2600 Ops/s 273.0822 Ops/s $\color{#35bf28}+1.53\%$
test_collector_stack_then_write[100-img_shape2-large_img] 9.7425ms 5.7736ms 173.2032 Ops/s 174.9033 Ops/s $\color{#d91a1a}-0.97\%$
test_collector_stack_then_write[200-img_shape3-large_batch] 15.2165ms 7.1507ms 139.8455 Ops/s 135.1323 Ops/s $\color{#35bf28}+3.49\%$
test_collector_lazystack_then_write[50-img_shape0-small] 0.5332ms 0.2771ms 3.6087 KOps/s 3.5925 KOps/s $\color{#35bf28}+0.45\%$
test_collector_lazystack_then_write[100-img_shape1-atari] 1.7332ms 1.5197ms 658.0431 Ops/s 654.3500 Ops/s $\color{#35bf28}+0.56\%$
test_collector_lazystack_then_write[100-img_shape2-large_img] 2.5476ms 2.3934ms 417.8192 Ops/s 417.1555 Ops/s $\color{#35bf28}+0.16\%$
test_collector_lazystack_then_write[200-img_shape3-large_batch] 3.2938ms 3.1023ms 322.3458 Ops/s 319.4007 Ops/s $\color{#35bf28}+0.92\%$
test_collector_without_rb[100-img_shape0-atari] 33.5582ms 33.1597ms 30.1571 Ops/s 29.8698 Ops/s $\color{#35bf28}+0.96\%$
test_collector_without_rb[200-img_shape1-large_batch] 65.3875ms 65.1449ms 15.3504 Ops/s 15.3130 Ops/s $\color{#35bf28}+0.24\%$
test_collector_with_rb[100-img_shape0-atari] 38.4318ms 37.8334ms 26.4317 Ops/s 26.2825 Ops/s $\color{#35bf28}+0.57\%$
test_collector_with_rb[200-img_shape1-large_batch] 74.3119ms 73.8096ms 13.5484 Ops/s 13.0676 Ops/s $\color{#35bf28}+3.68\%$
test_collector_without_rb_cuda[100-img_shape0-atari] 56.4042ms 56.1888ms 17.7971 Ops/s 17.2663 Ops/s $\color{#35bf28}+3.07\%$
test_collector_without_rb_cuda[200-img_shape1-large_batch] 0.1156s 0.1129s 8.8606 Ops/s 8.6845 Ops/s $\color{#35bf28}+2.03\%$
test_collector_with_rb_cuda[100-img_shape0-atari] 60.5848ms 58.8633ms 16.9885 Ops/s 16.9433 Ops/s $\color{#35bf28}+0.27\%$
test_collector_with_rb_cuda[200-img_shape1-large_batch] 0.1198s 0.1179s 8.4847 Ops/s 8.5387 Ops/s $\color{#d91a1a}-0.63\%$

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

Projects

None yet

Milestone

No milestone

Development

Successfully merging this pull request may close these issues.

1 participant

@vmoens