unilab 1.5.0

What's Changed

• feat(reset): restore env_ids as the public reset selector by @TATP-233 in #2468
• docs(migration): add bilingual 1.4 to 1.5 migration guide (#2463) by @TATP-233 in #2469
• feat(managers): restore observation history/delay under the env_ids reset contract (#2462) by @TATP-233 in #2470
• perf(reset): reuse owner-scoped mask buffers in the env_ids adaptation layer (#2461) by @TATP-233 in #2471
• docs(reset): document owns_observation_reset temporal semantics bypass by @TATP-233 in #2472
• feat(reset): restore env_ids as the public reset selector (roadmap #2459) by @TATP-233 in #2473
• chore(deps): bump unisim-core pin to 2.0.1 by @TATP-233 in #2476
• feat(training): add opt-in TF32 matmul knob for PPO/APPO FP32 updates by @TATP-233 in #2478
• feat(training): record ambient CUDA MPS state in off-policy probe (#2479) by @TATP-233 in #2480
• refactor(training): merge placement knobs into training.collector_tensor_device (#2481) by @TATP-233 in #2484
• feat(training): forbid CPU carrier with CUDA learner behind explicit opt-in (#2483) by @TATP-233 in #2485
• perf(env): collapse duplicate action/control scalar syncs on collector hot path by @TATP-233 in #2486
• chore(benchmark): delete device_info re-export shim, import unilab.utils.device directly (#2474) by @TATP-233 in #2491
• refactor(utils): default reward target_field to 'rewards' and drop 'reward_config' alias (#2474) by @TATP-233 in #2489
• [2474-P2] 清理 env/command 静默缝：删孤儿 hook、metric fail-closed、注释改写 by @TATP-233 in #2493
• refactor(reset)!: remove the v1.4 bool-mask reset compatibility layer (#2474) by @TATP-233 in #2494
• chore(deps): 消费 unisim-core 2.1.0，interval op payload 对齐 torch-first 契约 by @TATP-233 in #2496
• fix(envs): restore mjwarp collector Backend Step timing and surface CUDA MPS status by @TATP-233 in #2497
• release: unilab 1.5.0 by @TATP-233 in #2498

Full Changelog: v1.4.0...v1.5.0

unilab 1.4.0

What's Changed

• release: unilab 1.3.5 by @TATP-233 in #2308
• feat(graph): public completion lifecycle for #2309 by @TATP-233 in #2310
• chore(deps): require released sibling fixes by @TATP-233 in #2337
• fix(env): restore eager training semantics by @TATP-233 in #2343
• feat(env): pre-register state-update graph owner boundary by @TATP-233 in #2345
• feat(env): consume update graph preflight by @TATP-233 in #2346
• perf(env): capture manager update graph with persistent scratch by @TATP-233 in #2349
• test(env): gate update graph semantic parity by @TATP-233 in #2353
• docs(adr): close collector graph route by @TATP-233 in #2356
• fix(env): require completion publication for CUDA host bridge by @TATP-233 in #2355
• perf(benchmark): record state-update graph Gate UP preflight evidence by @TATP-233 in #2352
• refactor(runtime)!: remove collector graph contracts by @TATP-233 in #2361
• fix(ci): clone main for local dependency sources after #2359 merge by @TATP-233 in #2368
• fix(deps): align unisim-core/unilab-rl floors with released versions by @TATP-233 in #2369
• feat(training): default collector_tensor_device to auto following the learner device (#2365) by @TATP-233 in #2367
• fix(deps): pin unisim-core/unilab-rl exactly, update ROCm profile by @TATP-233 in #2370
• build(deps): resolve unisim-core and unilab-rl from PyPI by @TATP-233 in #2371
• feat(rl): merge unilab-rl into unilab.rl with a single distribution (#2372) by @TATP-233 in #2373
• Fix Go2 APPO Motrix training with unisim 1.7.24 by @TATP-233 in #2378
• fix(rl): correct Welford cross-term in EmpiricalNormalization.update_from_moments by @TATP-233 in #2374
• feat(rl): support learner-side custom actors with auxiliary losses in FlashSAC by @TATP-233 in #2376
• fix(rl): make failure override timeout bootstrap in off-policy replay by @TATP-233 in #2375
• refactor(rl): consolidate whole-cycle learner machinery by @TATP-233 in #2383
• feat(training): unify sim carrier device semantics by @TATP-233 in #2384
• chore(cleanup): remove unreachable platform branches by @TATP-233 in #2385
• Converge RL platform fork helpers and capability policy by @TATP-233 in #2391
• fix(rl): reconcile merged integration cleanups by @TATP-233 in #2388
• fix(rl): close merged-stack ownership and contract gaps by @TATP-233 in #2393
• fix(backends): restore g1 sac training quality by @TATP-233 in #2394
• fix(training): remove devices config residue by @TATP-233 in #2395
• docs(dr): lock unified model-field range spelling by @TATP-233 in #2397
• feat: rename host inference transport by @TATP-233 in #2399
• fix(training): route host-bridge carriers by capability by @TATP-233 in #2402
• chore: remove unused legacy paths by @TATP-233 in #2405
• docs: refresh ADR backend scope and evidence paths by @TATP-233 in #2407
• chore: remove legacy inference transport residue by @TATP-233 in #2408
• docs: describe active baseline contracts instead of legacy/compat wording (#2412) by @TATP-233 in #2417
• feat(benchmark): remove legacy compatibility surface by @TATP-233 in #2419
• test: pin historical artifact reader scope by @TATP-233 in #2418
• chore(assets)!: remove no-op --verbose option from unilab-pull-assets by @TATP-233 in #2420
• refactor(visualization): 删除 render_play_mode 的无效 sim_backend 参数 by @TATP-233 in #2421
• feat(tools): 删除 UNILAB_ISAACGYM_HOME / UNILAB_ISAACSIM_HOME 旧环境变量 by @TATP-233 in #2423
• feat(playback): require composed Hydra config for interactive playback by @TATP-233 in #2424
• feat(training): remove training.seed fallback, require algo.seed by @TATP-233 in #2422
• feat(rl)!: reject FlashSAC pre-contract checkpoints without actor semantics by @TATP-233 in #2425
• feat(sim2sim)!: fail closed on missing or legacy contract snapshots by @TATP-233 in #2426
• feat(training): rsl-rl pre-patch checkpoint 不再支持，resume fail closed by @TATP-233 in #2427
• Roadmap #2411: legacy/compat 兼容面收尾（含终审报告） by @TATP-233 in #2429
• feat!: consume torch-first unisim DR payloads (#2410) by @TATP-233 in #2431
• chore: sync ROCm lock track with the 1.9.0 pin and record dual-track dependency governance by @TATP-233 in #2432
• test: remove stale backend host-API mirrors by @TATP-233 in #2433
• chore(ci): split default tests into parallel shards by @TATP-233 in #2436
• fix: remove dead API scaffolding (#2437) by @TATP-233 in #2446
• refactor!: fix misleading public API names by @TATP-233 in #2447
• refactor: unify tensor API naming by @TATP-233 in #2448
• refactor!: clean up module/class name stutter (#2442) by @TATP-233 in #2449
• refactor: promote cross-module private protocol names to public (#2441) by @TATP-233 in #2450
• refactor(rl): converge shared checkpoint and playback utilities by @TATP-233 in #2451
• refactor!: unify verbs and lifecycle hook naming (#2443) by @TATP-233 in #2452
• refactor!: Roadmap #2434 API 命名规范化收尾 + unisim-core 2.0.0 集成 by @TATP-233 in #2453
• fix(playback): drop camera profile rejected by native interactive viewers (#2454) by @TATP-233 in #2455

Full Changelog: v1.3.5...v1.4.0

unilab 1.3.5

What's Changed

• release: unilab 1.3.4 by @TATP-233 in #2121
• fix(env): commit MuJoCo model-field DR through host bridge by @TATP-233 in #2123
• refactor(runtime): remove Numba and G1 CPU motion command by @TATP-233 in #2125
• refactor(managers): remove NumPy from core manager seams by @TATP-233 in #2127
• refactor(managers): require Torch RNG for scheduling by @TATP-233 in #2128
• refactor(envs): use Torch RNG for generic event sampling by @TATP-233 in #2129
• refactor(managers): make noise and temporal buffers Torch-only by @TATP-233 in #2126
• refactor(envs): use Torch RNG for generic commands and IMU bias by @TATP-233 in #2130
• refactor(locomotion): use Torch RNG for phase and terrain resets by @TATP-233 in #2131
• test(tasks): restore locomotion Torch RNG boundary by @TATP-233 in #2133
• refactor(managers): require Torch reward and termination results by @TATP-233 in #2132
• refactor(managers): close Phase 2 Torch carrier boundary by @TATP-233 in #2134
• refactor(entity): own cold tensor reset tables by @TATP-233 in #2135
• refactor(locomotion): keep gait rewards on Torch by @TATP-233 in #2136
• refactor(motion): require packed host bridge state reads by @TATP-233 in #2137
• refactor(motion): publish loader tensors through cold leaf by @TATP-233 in #2138
• refactor(locomotion): remove dead height scan host runtime by @TATP-233 in #2139
• refactor(managers): negotiate sensors through cold binding by @TATP-233 in #2140
• refactor(locomotion): remove downstream rough terrain runtime by @TATP-233 in #2141
• refactor(motion): remove legacy NumPy sampler by @TATP-233 in #2142
• refactor(motion): remove dead torque reference loader by @TATP-233 in #2143
• refactor(runtime): own Torch dtype configuration by @TATP-233 in #2144
• refactor(utils): remove dead tensor conversion shims by @TATP-233 in #2145
• refactor(utils): prune unused rotation reference API by @TATP-233 in #2146
• refactor(curricula): log Torch scalar stages by @TATP-233 in #2147
• refactor(observations): remove manager NumPy helpers by @TATP-233 in #2148
• refactor(observations): keep IMU bias on Torch by @TATP-233 in #2149
• refactor(rewards): compute generic kernels on Torch by @TATP-233 in #2150
• refactor(actions): materialize cold tables as Torch by @TATP-233 in #2151
• refactor(commands): remove residual NumPy imports by @TATP-233 in #2152
• refactor(events): own encoder bias joint selection by @TATP-233 in #2153
• refactor(reset): accept Torch host write values by @TATP-233 in #2154
• refactor(events): keep PD gain samples on Torch by @TATP-233 in #2155
• refactor(events): sample body impulses on Torch by @TATP-233 in #2156
• refactor(events): keep generic DR samples on Torch by @TATP-233 in #2157
• refactor(events): keep model-field DR host path on Torch by @TATP-233 in #2158
• refactor(events): keep root velocity deltas on Torch by @TATP-233 in #2159
• refactor(events): compose root resets on Torch by @TATP-233 in #2160
• refactor(events): cache mass-inertia scales on Torch by @TATP-233 in #2161
• refactor(events): keep CoM samples on Torch by @TATP-233 in #2162
• refactor(entity): own tensor encoder-bias updates by @TATP-233 in #2163
• refactor(events): keep impulse lifecycle state on Torch by @TATP-233 in #2164
• refactor(events): keep startup mass table on Torch by @TATP-233 in #2165
• refactor(entity): require tensor root reset rows by @TATP-233 in #2166
• refactor(events): select PD defaults on Torch by @TATP-233 in #2167
• refactor(entity): own host encoder-bias updates by @TATP-233 in #2168
• refactor(locomotion): compute base rewards on Torch by @TATP-233 in #2169
• refactor(locomotion): keep gait tensor leaves on Torch by @TATP-233 in #2170
• refactor(locomotion): keep sensor rewards on Torch by @TATP-233 in #2171
• refactor(g1): keep double-stance mask on Torch by @TATP-233 in #2172
• refactor(g1): keep air-time state on Torch by @TATP-233 in #2173
• refactor(g1): cache pose weights on Torch by @TATP-233 in #2174
• refactor(locomotion): read base height from tensor plan by @TATP-233 in #2175
• refactor(locomotion): read joint deviation from tensor plan by @TATP-233 in #2176
• refactor(gait): consume packed sensor reads by @TATP-233 in #2177
• refactor(gait): cache contact columns on device by @TATP-233 in #2178
• refactor(gait): read foot state from tensor plan by @TATP-233 in #2179
• refactor(gait): read clearance heights from tensor plan by @TATP-233 in #2180
• chore(gait): remove dead host body readers by @TATP-233 in #2181
• refactor(go2): use sensor tensor velocity rewards by @TATP-233 in #2182
• chore(locomotion): remove entity velocity reward terms by @TATP-233 in #2183
• refactor(go2): consume packed foot phase reads by @TATP-233 in #2184
• refactor(g1): keep phase reward inputs on Torch by @TATP-233 in #2185
• chore(locomotion): remove dead rough owner module by @TATP-233 in #2186
• refactor(go2): require packed foot sensor reads by @TATP-233 in #2187
• refactor(g1): require packed sensor reads by @TATP-233 in #2188
• refactor(locomotion): require packed vec3 sensor reads by @TATP-233 in #2189
• refactor(g1): require packed height reads by @TATP-233 in #2190
• refactor(g1): compute velocity rewards on Torch by @TATP-233 in #2191
• refactor(g1): remove velocity and tilt NumPy branches by @TATP-233 in #2192
• refactor(g1): keep gait gates and contacts on Torch by @TATP-233 in #2193
• refactor(gait): require packed contact reads by @TATP-233 in #2194
• refactor(g1): keep contact terms on Torch by @TATP-233 in #2195
• refactor(g1): compute phase targets on Torch by @TATP-233 in #2196
• refactor(g1): keep shared reward kernels on Torch by @TATP-233 in #2197
• refactor(locomotion): return commands on Torch by @TATP-233 in #2198
• refactor(gait): require Torch reset rows by @TATP-233 in #2199
• refactor(locomotion): keep cold configs on Torch by @TATP-233 in #2200
• refactor(gait): materialize contact columns on Torch by @TATP-233 in #2201
• refactor(g1): require Torch term carriers by @TATP-233 in #2202
• refactor(gait): require packed sensor reads by @TATP-233 in #2203
• refactor(gait): require Torch carriers by @TATP-233 in #2204
• refactor(gait): require packed standing reads by @TATP-233 in #2205
• refactor(locomotion): remove common NumPy terms by @TATP-233 in #2206
• refactor(reset): require Torch row carriers by @TATP-233 in #2207
• refactor(g1): track joint acceleration on Torch by @TATP-233 in #2208
• chore(locomotion): remove dead height scan module by @TATP-233 in #2209
• refactor(motion): reuse owner Torch defaults by @TATP-233 in #2210
• refactor(motion): validate cold ranges on Torch by @TATP-233 in https://github.com/Motphy...

Issue 2285 current-head evidence summary

Summary artifacts for issue #2285 / PR #2287 current-head rerun. UniLab head: 4020575b; unisim source: 23dea2a (merge db36e3c); unilab-rl: 635549f. Archive SHA-256: f546d1e6eab5571b31e75d8992bb935b5ee2f330160fd87912dda2c4e09d0961.

unilab 1.3.4

What's Changed

• feat(isaacsim): expose material sharing and task-local G1 fixture by @TATP-233 in #1683
• chore: pin tensor runtime local workspace by @TATP-233 in #1685
• ci(tensor): skip uv cache maintenance on cuda runner by @TATP-233 in #1688
• feat(warpsac): declare tensor inference runtime knobs by @TATP-233 in #1689
• feat(training): monitor G1 FlashSAC soaks by @TATP-233 in #1687
• feat(training): consume tensor runtime schemas by @TATP-233 in #1696
• feat(training): consume abnormal shutdown diagnostics by @TATP-233 in #1697
• feat(training): consume tensor runtime bounds by @TATP-233 in #1698
• feat(training): fail closed on tensor platform matrix by @TATP-233 in #1699
• docs(training): add tensor runtime production guide by @TATP-233 in #1700
• docs(adr): define torch-only manager runtime by @TATP-233 in #1707
• feat(base): add tensor environment lifecycle by @TATP-233 in #1708
• feat(env): migrate manager lifecycle to torch by @TATP-233 in #1709
• feat(managers): tensorize metric accumulators by @TATP-233 in #1710
• feat(managers): tensorize reward and termination carriers by @TATP-233 in #1711
• feat(managers): tensorize action history by @TATP-233 in #1712
• feat(managers): publish observations as tensors by @TATP-233 in #1713
• feat(actions): tensorize base action processing by @TATP-233 in #1715
• feat(actions): tensorize Allegro incremental targets by @TATP-233 in #1716
• feat(entity): expose validated joint tensor reads by @TATP-233 in #1717
• feat(observations): tensorize non-temporal processing by @TATP-233 in #1719
• feat(entity): expose sensor and body tensor reads by @TATP-233 in #1720
• feat(env): pair packed scene reads and reset transactions by @TATP-233 in #1721
• feat(observations): tensorize temporal buffers by @TATP-233 in #1722
• feat(commands): publish command state as tensors by @TATP-233 in #1723
• feat(events): tensorize scheduling carriers by @TATP-233 in #1724
• bench(env): use tensor actions for env rollouts by @TATP-233 in #1725
• bench(env): consume public step timing without private probes by @TATP-233 in #1726
• bench(rl): keep active collector transitions in tensors by @TATP-233 in #1727
• bench(env): replace NpEnv shard throughput workload by @TATP-233 in #1728
• bench(rl): rename step timing metrics without NpEnv prefix by @TATP-233 in #1729
• fix(playback): keep off-policy policy IO in tensors by @TATP-233 in #1730
• fix(appo): keep Motrix playback observations in tensors by @TATP-233 in #1731
• refactor(rl): enforce TorchEnv tensor VecEnv boundary by @TATP-233 in #1732
• fix(play): consume tensor velocity commands by @TATP-233 in #1733
• refactor(registry): enforce TorchEnv factory output by @TATP-233 in #1734
• docs(runtime): describe the TorchEnv boundary by @TATP-233 in #1738
• docs(agents): point env contract at TorchEnv by @TATP-233 in #1739
• feat(observations): tensorize noise carriers by @TATP-233 in #1741
• feat(rl): round-trip TorchEnv training state by @TATP-233 in #1736
• feat(training): enforce tensor NaN diagnostics by @TATP-233 in #1735
• Remove the legacy NpEnv runtime by @TATP-233 in #1742
• feat(runtime): unify CUDA tensor inference placement by @TATP-233 in #1744
• fix(managers): slice per-env observation scales on reset rows by @TATP-233 in #1746
• fix(tasks): skip generic CPU reads in FlashSAC device-resident proxy by @TATP-233 in #1748
• fix(tasks): route FlashSAC Motrix through HOST_BRIDGE runtime by @TATP-233 in #1752
• fix(tasks): materialize IsaacGym before tensor negotiation by @TATP-233 in #1754
• fix(tasks): materialize lazy FlashSAC backends only by @TATP-233 in #1758
• fix(env): restore legacy IsaacSim CPU training wire by @TATP-233 in #1760
• fix(tasks): release FlashSAC IsaacGym views before backend close by @TATP-233 in #1759
• perf(entity): remove repeated packed tensor read validation by @TATP-233 in #1761
• fix: satisfy typed tensor runtime checks by @TATP-233 in #1763
• fix(env): dispatch external CUDA IPC selected resets to tensor boundary by @TATP-233 in #1766
• fix(env): stagger Manager initial episode counters by @TATP-233 in #1767
• fix(config): restore legacy G1 walk collector penalty schedule by @TATP-233 in #1768
• test: align device-resident reset sequence with staggered counters by @TATP-233 in #1769
• fix(tasks): cross FlashSAC selected-reset readiness boundary by @TATP-233 in #1770
• fix(tasks): refresh device-resident named sensors each read by @TATP-233 in #1773
• perf(actions): keep joint position control on tensor plane by @TATP-233 in #1775
• fix(tasks): keep G1 contact proxy reward-only by @TATP-233 in #1774
• feat(config): productionize FlashSAC IsaacSim mapped CUDA IPC owner by @TATP-233 in #1777
• fix(tasks): keep HOST_BRIDGE reset on its paired read boundary by @TATP-233 in #1780
• perf(mjwarp): dispatch selected resets through tensor boundary by @TATP-233 in #1781
• perf(mjwarp): keep manager reads on packed device plane by @TATP-233 in #1783
• perf(mjwarp): reduce action and reset boundary overhead by @TATP-233 in #1785
• perf(managers): batch reset and state validation boundaries by @TATP-233 in #1787
• fix(config): make SAC mjbatch owner select the packed tensor runtime by @TATP-233 in #1791
• feat(training): make CUDA visibility rank-local by @TATP-233 in #1798
• fix(env): negotiate task sensors and restore reward diagnostics by @TATP-233 in #1797
• fix(offpolicy): inject the tensor nan guard across the collector boundary by @TATP-233 in #1799
• perf(rng): own torch CUDA randomization in G1 collector by @TATP-233 in #1801
• fix(env): keep device-resident command sensors canonical by @TATP-233 in #1803
• feat(training): make CUDA visibility the only GPU topology by @TATP-233 in #1806
• perf(mjwarp): avoid reset readiness physics step by @TATP-233 in #1805
• perf(manager): keep G1 update state device-resident by @TATP-233 in #1807
• chore(build): remove uv version pin by @TATP-233 in #1810
• perf(manager): cut stable-view host boundaries on slow-CPU servers by @TATP-233 in #1808
• refactor(env): keep episode counters tensor-resident by @TATP-233 in #1812
• refactor(recorder): publish reset rows as tensors by @TATP-233 in #1813
• refactor(command): publish reset rows as tensors by @TATP-233 in #1814
• refactor(events): publish reset rows as tensors by @TATP-233 in #1815
• fix(locomotion): consume CUDA reset rows safely by @TATP-233 in #1816
• perf(command): draw resample timers with Torch RNG by @TATP-233 in #1817
• perf(events): draw interval timers with Torch RNG by @TATP-233 in #1818
• perf(observations): sample IMU bias with Torch RNG by @TATP-233 in #1819
• perf(command): sample tensor velocity commands on device by @TATP-233 in #1820
• perf(command): sample pose values with Torch RNG b...

unilab 1.3.3

What's Changed

• fix(cli): route macOS MuJoCo interactive eval through mjpython by @TATP-233 in #1630
• feat(cli): add viser render mode for mujoco/mjwarp eval by @TATP-233 in #1632
• fix: batch Viser updates and handle Ctrl+C cleanly by @TATP-233 in #1633
• perf(motrix): A/B microbenchmark for the fused body-state convergence (#1308) by @TATP-233 in #1635
• feat(motrix): document and test cpu_ids worker affinity wiring (#962) by @TATP-233 in #1634
• feat(isaacsim): forward the PhysX GPU contact buffer knobs on EnvCfg by @lemon-star608 in #1636
• feat: consume unisim 1.7.5 physics-state playback contract in interactive rendering by @TATP-233 in #1637
• feat: make viser a required dependency and support viser playback on all playback-capable backends by @TATP-233 in #1638
• Enable viser rendering for the motrix backend by @TATP-233 in #1639
• chore(deps): bump unilab-rl to 1.3.3 by @TATP-233 in #1640
• config(flash-sac): enable full-objective compile by @EtherealTide in #1641
• chore(deps): bump unilab-rl to 1.3.4 by @TATP-233 in #1643
• feat(rl): add WarpSAC G1 owners for unilab-rl 1.4.0 by @TATP-233 in #1645
• chore(deps): bump torch to 2.14 and unify CUDA wheels on cu130 by @TATP-233 in #1647
• perf(training): training.log_interval 配置项 + rsl_rl PPO TensorBoard 批量写 patch by @TATP-233 in #1648
• fix(deps): restore ROCm runtime dependencies by @TATP-233 in #1650
• fix(motion): preserve headerless CSV frames by @TATP-233 in #1653
• fix(training): export ONNX at opset 18 and decouple play artifact failures by @TATP-233 in #1656
• chore(deps): bump unilab-rl to 1.4.1 by @TATP-233 in #1655
• chore(offpolicy): remove inference request timeout config by @TATP-233 in #1658
• chore(deps): bump unilab-rl to 1.4.2 by @TATP-233 in #1659
• chore(deps): pin uv to 0.12.5 across project, CI, and Docker by @TATP-233 in #1657
• release: unilab 1.3.3 by @TATP-233 in #1660
• chore(deps): bump unilab-rl to 1.4.3 by @TATP-233 in #1662
• docs(changelog): record unilab-rl 1.4.3 in the 1.3.3 entry by @TATP-233 in #1663
• fix(motion): handle adaptive reset zero-floor cold start by @TATP-233 in #1661

New Contributors

• @EtherealTide made their first contribution in #1641

Full Changelog: v1.3.2...v1.3.3

unilab 1.3.2

What's Changed

• ci: fix github-release job missing checkout by @TATP-233 in #1614
• Optimize G1 motion tracking backend paths by @TATP-233 in #1618
• feat(sac): add genesis/isaacgym/newton/isaacsim backends for g1_motion_tracking by @TATP-233 in #1619
• feat(ppo): 直连上游 rsl-rl，uni_rl 降级为可选依赖 by @TATP-233 in #1622
• feat(isaacsim)：通过 EnvCfg 后端接线暴露 UniSim 的 PhysX solver 旋钮 by @lemon-star608 in #1620
• refactor!: remove TD3 tasks and bump unilab-rl to 1.3.2 by @TATP-233 in #1625
• release: unilab 1.3.2 by @TATP-233 in #1626

Full Changelog: v1.3.1...v1.3.2

unilab 1.3.1

What's Changed

• feat: integrate IsaacGym fixed model variants (#1577) by @TATP-233 in #1582
• feat!: move Unitree production tasks downstream by @TATP-233 in #1583
• feat!: move G1 walk rough downstream by @TATP-233 in #1584
• chore: remove obsolete dead code and stale references by @TATP-233 in #1593
• chore(deps): bump unisim-core to 1.4.2 and unilab-rl to 1.2.1 by @TATP-233 in #1594
• ci: skip CUDA wheels in mypy/pyright typecheck jobs by @TATP-233 in #1598
• feat: consume physical entities in manager reset transactions (#1599) by @TATP-233 in #1600
• feat: add flashsac g1_motion_tracking task owners and newton/genesis backends by @TATP-233 in #1603
• feat: add Newton CUDA graph owner knob by @TATP-233 in #1605
• fix: align genesis probe parallel level by @TATP-233 in #1606
• chore: update unilab-rl to 1.3.0 by @TATP-233 in #1607
• fix: consume released unisim portable scenes by @TATP-233 in #1610
• chore: update unisim-core to 1.7.2 by @TATP-233 in #1612
• release: unilab 1.3.1 by @TATP-233 in #1613

Full Changelog: v1.3.0...v1.3.1

unilab 1.3.0

What's Changed

• feat: mujoco 后端执行器切换到 mjbatch-uni(roadmap #1552) by @TATP-233 in #1562
• feat: complete per-env fixed model variant roadmap by @TATP-233 in #1568
• fix: remove duplicate reward episode metrics by @TATP-233 in #1570
• docs: document reward logging contract by @TATP-233 in #1571
• chore: prune redundant tests and remove legacy task-factory seam by @TATP-233 in #1573
• deps: unisim-core 1.4.0 and mjbatch-uni 0.2.1 by @TATP-233 in #1574
• release: unilab 1.3.0 by @TATP-233 in #1575

Full Changelog: v1.2.0...v1.3.0

unilab 1.2.0

What's Changed

• release: unilab 1.1.0 by @TATP-233 in #1527
• refactor: extract Go2 arm task and dedicated locomotion helpers by @TATP-233 in #1536
• feat(superdex): integrate FR3 and Go2 research task owners by @TATP-233 in #1538
• refactor: move Go2 arm manipulation into its standalone repository by @TATP-233 in #1537
• feat: expose Wuji reset and training progress adapters by @TATP-233 in #1540
• refactor!: remove sharpa task and hora algorithm (migrate to sharpa_rl_unilab) by @TATP-233 in #1549
• feat(playback): adopt unisim debug overlay primitives and embeddable playback session by @TATP-233 in #1548
• release: consolidate SuperDex integration and bump UniLab by @TATP-233 in #1545
• feat(superdex): published-wheel install, FR3 HF assets, unisim-core/unilab-rl 1.2.0 by @TATP-233 in #1550
• release: unilab 1.2.0 by @TATP-233 in #1551

Full Changelog: v1.1.0...v1.2.0