| commit |
1.00 |
CUDAAccelerator.setup_device: fix unrelated device init by m |
|
Commit message contains explicit AI assi |
2026-09-14 |
| commit |
1.00 |
refactor: use public `LRScheduler` instead of private `_LRSc |
|
Commit message contains explicit AI assi |
2026-09-14 |
| commit |
1.00 |
Profile dataloader initialization in FitLoop and update outd |
|
Commit message contains explicit AI assi |
2026-07-02 |
| commit |
1.00 |
fix(FSDP): pass explicit CPU device to satisfy torch>=2.5 |
|
Commit message contains explicit AI assi |
2026-07-02 |
| commit |
1.00 |
Fix FLOPs inconsistency and add using_sparse_model flag (#21 |
|
Commit message contains explicit AI assi |
2026-06-08 |
| commit |
1.00 |
Remove unnecessary TODO comment in fit_loop.py (#21265) |
|
Commit message contains explicit AI assi |
2026-04-13 |
| commit |
1.00 |
Update issue templates and labeler for v2.6 release (#21598) |
|
Commit message contains explicit AI assi |
2026-03-30 |
| pr |
1.00 |
Recheck model FLOPs capability when reusing ThroughputMonito |
|
Explicitly states AI assistance was used |
2026-11-01 |
| pr |
1.00 |
Delegate async checkpoint teardown to the wrapped backend |
|
Explicitly states AI assistance was used |
2026-11-01 |
| pr |
1.00 |
Allow scheduled gradient accumulation across repeated fits |
|
Explicitly states AI assistance was used |
2026-11-01 |
| pr |
1.00 |
Restore LR-finder state when training raises |
|
Explicitly discloses that an AI assistan |
2026-11-01 |
| pr |
1.00 |
Preserve initial parameter snapshots across lottery-ticket r |
|
Explicitly discloses that an AI assistan |
2026-11-01 |
| pr |
1.00 |
Reset averaging counters when starting a fresh fit |
|
Explicitly discloses that an AI coding a |
2026-11-01 |
| pr |
1.00 |
Preserve original methods across repeated compile conversion |
|
Explicitly discloses AI development of t |
2026-11-01 |
| pr |
1.00 |
Restore WeightAveraging model roles after validation excepti |
|
Explicitly discloses that an AI assistan |
2026-11-01 |
| pr |
1.00 |
Account for `GradientAccumulationScheduler` in `Trainer.esti |
|
Explicitly states development was assist |
2026-10-22 |
| pr |
1.00 |
Stage AsyncCheckpointIO tensor snapshots on CPU |
|
Explicitly states that Codex helped impl |
2026-11-01 |
| pr |
1.00 |
Fix CLI config saving without a local logger directory |
|
PR body explicitly mentions AI collabora |
2026-10-02 |
| pr |
1.00 |
Support Python module execution in the Fabric CLI |
|
Explicitly states that Codex assisted wi |
2026-09-22 |
| pr |
1.00 |
Fix CSVLogger automatic step restarting at 0 after every flu |
|
Explicitly states that an AI coding assi |
2026-09-30 |
| pr |
1.00 |
Fix step scheduler frequency across epochs |
|
Explicitly states the contribution was p |
2026-08-30 |
| pr |
1.00 |
Fix CombinedLoader max_size_cycle with empty iterable |
|
Explicitly states that OpenAI Codex assi |
2026-08-11 |
| pr |
1.00 |
Restore stateful dataloader states when resuming from end-of |
|
Explicitly discloses that an AI coding a |
2026-10-29 |
| pr |
1.00 |
fix(checkpoint): avoid overwriting checkpoint after resume |
|
Explicitly discloses assistance from Cod |
2026-10-19 |
| pr |
1.00 |
Add sync_dist_op to LightningModule logging |
|
Explicitly discloses AI assistance with |
2026-08-09 |
| pr |
0.62 |
Fix `_EvaluationLoop._find_value` mixed nested/scalar lookup |
|
Highly polished, exhaustive bug narrativ |
2026-10-28 |
| pr |
0.58 |
fix(pytorch): handle empty YAML/CSV and non-dict overrides i |
|
Detailed enumeration and unusually unifo |
2026-10-28 |
| pr |
0.52 |
Fix negative base_seed in `pl_worker_init_function` and ensu |
|
Very polished edge-case explanations and |
2026-10-27 |
| pr |
0.45 |
Fix import error in computer_vision_fine_tuning.py example n |
|
Structured technical explanation is poli |
2026-10-27 |
| pr |
0.22 |
Fix 'responsible of' grammar in pruning callback docstrings |
|
The verification wording is slightly for |
2026-10-25 |
| pr |
0.15 |
ci: repair documentation checks and interrupted test cleanup |
|
Highly polished and comprehensive, but r |
2026-10-29 |
| pr |
0.15 |
Fix AdvancedProfiler.__reduce__ dump_stats state loss and Si |
|
Detailed, domain-specific explanation wi |
2026-10-24 |
| pr |
0.14 |
fix: isolate AMP autocast cache clearing across threads |
|
Polished technical prose, but details ar |
2026-10-14 |
| pr |
0.14 |
Fix Timer callback stale training elapsed time across multip |
|
Detailed lifecycle analysis is polished |
2026-10-24 |
| pr |
0.12 |
perf(utilities): use torch.stack in grad_norm to avoid per-p |
|
Technical motivation is polished but inc |
2026-10-25 |
| pr |
0.12 |
fix(callbacks): check parameter identity instead of torch.eq |
|
Polished technical prose, but specific c |
2026-10-21 |
| commit |
0.10 |
Add fsspec URL support for distributed checkpoints in ModelP |
|
Precise technical commit, domain-specifi |
2026-07-10 |
| pr |
0.10 |
Add `MLFlowLogger.log_model` to log and register models with |
|
Technical MLflow-specific rationale and |
2026-10-23 |
| pr |
0.10 |
Docs: Fix case-sensitive import in Fabric README |
|
Polished explanatory prose is generic bu |
2026-10-30 |
| pr |
0.10 |
perf(utilities): unwrap multi-decorator hooks and cache args |
|
Domain-specific profiling details and co |
2026-10-26 |
| pr |
0.09 |
fix(tuner): forward `max_val` to `_adjust_batch_size` and ev |
|
Concrete edge cases, issue references, a |
2026-10-25 |
| pr |
0.08 |
Add `from_fx()` and `to_unfx()` utilities for fx-converted m |
|
Detailed framework-specific explanation |
2026-10-09 |
| pr |
0.08 |
fix(data): support positional DataLoader arguments when re-i |
|
Detailed domain-specific debugging and t |
2026-10-12 |
| pr |
0.08 |
Expose weights_only for batch size tuning checkpoint restora |
|
Specific API, regression, and environmen |
2026-11-01 |
| pr |
0.08 |
docs: update SECURITY.md pointing reporters to support@light |
|
Brief and somewhat rough wording is more |
2026-10-30 |
| pr |
0.08 |
docs: resolve Fabric experimental policy links with intersph |
|
Specific technical references and minor |
2026-10-22 |
| pr |
0.08 |
fix(fabric): validate Throughput.update inputs before mutati |
|
Very polished technical prose, though th |
2026-10-29 |
| pr |
0.08 |
Fix SWA with untracked BatchNorm and avoid per-parameter hos |
|
Specific PyTorch behavior and implementa |
2026-10-26 |
| pr |
0.08 |
fix: clone async checkpoint tensors to CPU to prevent GPU OO |
|
Concise technical summary with concrete |
2026-03-31 |
| pr |
0.08 |
Fix `ModelSummary.total_layer_params` double-counting nested |
|
Specific implementation reasoning and re |
2026-10-22 |
| pr |
0.08 |
Fix silent data corruption when ranks log different metric k |
|
Structured design summary is technical a |
2025-12-20 |
| pr |
0.07 |
Fix LeafSpec FutureWarning emitted by CombinedLoader on torc |
|
Specific version behavior, reproduction |
2026-10-25 |
| pr |
0.06 |
feat(fabric): add return-type overloads for Fabric.setup, se |
|
Specific typing details and informal cav |
2026-10-07 |
| commit |
0.05 |
Docs: Pytorch: Optimization: correct return type of configur |
|
Terse, issue-specific wording with concr |
2026-09-17 |
| commit |
0.05 |
ci: drop PyTorch 2.1-2.5 and raise minimum to 2.6 (#21851) |
|
Concise, technical language with domain- |
2026-07-29 |
| commit |
0.05 |
Fixes mypy issues with `bitsandbytes > 0.48` (#21858) |
|
Brief, task-focused commit log entries w |
2026-07-29 |
| pr |
0.05 |
fix(profiler): gracefully handle stopping unstarted actions |
|
Informal caveats and concrete profiler b |
2026-10-04 |
| pr |
0.05 |
Fix `_log_hyperparams` key-collision check for shared `Tenso |
|
Detailed code-level references and issue |
2026-10-29 |
| pr |
0.05 |
fix: check the weight actually exists for LayerNorm in `_con |
|
Terse, issue-specific wording and an unf |
2026-06-03 |
| pr |
0.04 |
Avoid per-parameter device sync and redundant transfers in ` |
|
Technical performance analysis and proje |
2026-11-02 |
| pr |
0.04 |
perf(fabric): avoid per-element host syncs and use tuple sli |
|
Specific PyTorch synchronization details |
2026-11-02 |
| pr |
0.04 |
Improve error message for mixed strategy namespaces |
|
Brief, issue-focused description with co |
2026-10-09 |
| pr |
0.04 |
Defer logger flushing to logger policies |
|
Terse, repository-specific summary with |
2026-09-27 |
| pr |
0.04 |
Fix extract_batch_size crash on NumPy ndarray batches |
|
Concise technical explanation with speci |
2026-09-22 |
| pr |
0.04 |
Fix sync_dist epoch metrics inflated by floored integer batc |
|
Highly specific debugging details and do |
2026-07-16 |
| pr |
0.04 |
Fix `last.ckpt` not holding the latest state when the monito |
|
Detailed checkpoint behavior and regress |
2026-10-25 |
| pr |
0.04 |
Fix upgrade_checkpoint failure on PyTorch 2.6+ due to weight |
|
Highly specific PyTorch and Lightning de |
2026-10-31 |
| pr |
0.04 |
Fix `_load` treating local relative paths starting with `htt |
|
Contains concrete reproduction cases, AP |
2026-10-30 |
| pr |
0.04 |
Avoid mutating input `scheduling` dict and reject boolean ke |
|
Specific Python validation behavior and |
2026-10-30 |
| pr |
0.04 |
docs: fix linkcheck by replacing dead www.pytorchlightning.a |
|
Concrete linkcheck details and malformed |
2026-09-01 |
| pr |
0.04 |
Fix `_CheckpointConnector.__max_ckpt_version_in_folder` matc |
|
Specific root-cause analysis, filenames, |
2026-10-28 |
| pr |
0.04 |
Fix `ModelPruning._make_pruning_permanent_on_state_dict` suf |
|
Detailed implementation and complexity a |
2026-10-22 |
| pr |
0.04 |
Fix _Linear8bitLt quantized guards and O(N^2) missing-keys h |
|
Highly specific bitsandbytes details and |
2026-10-22 |
| commit |
0.03 |
fix(mypy): cast OmegaConf result in `load_hparams_from_yaml` |
|
Highly specific mypy and OmegaConf detai |
2026-08-25 |
| commit |
0.03 |
Remove dead PyTorch < 2.6 version guards (#21870) |
|
Detailed technical changelog with precis |
2026-08-07 |
| commit |
0.03 |
ci: remove the markdown link checker (#21888) |
|
Specific CI history and repository polic |
2026-08-07 |
| pr |
0.03 |
Avoid materializing full index lists in `UnrepeatedDistribut |
|
Detailed domain-specific optimization ra |
2026-11-03 |
| pr |
0.03 |
Add `val_check_interval_unit` to validate every N optimizer |
|
Issue-specific behavior, constraints, an |
2026-11-02 |
| pr |
0.03 |
fix: set sampler epoch before creating resumed iterator |
|
Terse technical description with concret |
2026-10-04 |
| pr |
0.03 |
Fix `LearningRateMonitor` crash for optimizers without weigh |
|
Contains precise optimizer behavior and |
2026-10-20 |
| pr |
0.03 |
Fix EMA updates after skipped AMP optimizer steps |
|
Specific AMP and EMA behavior with focus |
2026-10-01 |
| pr |
0.03 |
Fix tag forwarding and missing directory error in DeepSpeed |
|
Concrete DeepSpeed bugs, code references |
2026-11-01 |
| pr |
0.03 |
Reset `Timer` end timestamps when a stage restarts |
|
Specific timer behavior, test counts, an |
2026-10-08 |
| pr |
0.03 |
Fix WeightAveraging swapping in the un-updated average model |
|
Precise WeightAveraging lifecycle explan |
2026-05-22 |
| pr |
0.03 |
Fix SpikeDetection.load_state_dict referencing nonexistent s |
|
Specific code references, failure detail |
2026-06-02 |
| pr |
0.03 |
fix: resume from max_steps checkpoint saved on an epoch boun |
|
Specific loop-state diagnosis and compac |
2026-10-30 |
| pr |
0.03 |
Fix forward hook leak and state reset in `_FabricModule` cal |
|
Specific implementation details and info |
2026-10-31 |
| pr |
0.03 |
Fix `save_hyperparameters` crash when `__init__` has array/t |
|
Detailed tensor behavior, identity match |
2026-10-31 |
| pr |
0.03 |
ci: pin runner images to ubuntu-24.04, macos-26, windows-202 |
|
Detailed runner diagnostics, timings, an |
2026-10-07 |
| pr |
0.03 |
fix: allow `LearningRateMonitor` to be used without a logger |
|
Specific issue context and implementatio |
2026-10-30 |
| pr |
0.03 |
Honor `strict=False` when loading sharded (distributed) chec |
|
Detailed distributed-checkpoint behavior |
2026-10-19 |
| pr |
0.03 |
Fix TorchSyncBatchNorm.revert resetting eval-mode layers to |
|
Precise framework behavior, regression e |
2026-10-30 |
| pr |
0.03 |
Use `deque` for batch queue in `_PrefetchDataFetcher` |
|
Concise, domain-specific explanation wit |
2026-10-23 |
| pr |
0.03 |
Remove the call tracker's forward hooks and reset its flag o |
|
Concrete failure modes, cleanup behavior |
2026-09-09 |
| pr |
0.03 |
Fix the uneven first step in the lr_finder learning-rate lad |
|
Specific scheduler behavior and tests in |
2026-09-05 |
| pr |
0.03 |
Fix lr_finder suggestion discarding every sample when skip_e |
|
Precise bug analysis and regression deta |
2026-08-29 |
| commit |
0.02 |
Add log_key_prefix to Trainer to control the prefix for metr |
|
Technical and concise, with a normal hum |
2026-09-21 |
| commit |
0.02 |
lightning.pytorch.profilers: use os.PathLike (#21871) |
|
Specific technical changes and commit me |
2026-09-14 |
| commit |
0.02 |
fix(checkpoint): prevent arbitrary code execution via `_clas |
|
Detailed security analysis and repositor |
2026-09-07 |
| commit |
0.02 |
fix: `LightningCLI` emitting `jsonargparse` deprecation warn |
|
Specific compatibility constraints and t |
2026-09-01 |
| commit |
0.02 |
fix: warn when training batches are fewer than `accumulate_g |
|
Terse, domain-specific commit message wi |
2026-08-09 |
| pr |
0.02 |
Fix `MLFlowLogger` reading `MLFLOW_TRACKING_URI` at import t |
|
Concise, concrete bug explanation with s |
2026-10-11 |
| pr |
0.02 |
Fix DeepSpeedStrategy.save_checkpoint identity check for cli |
|
Specific technical explanation with a co |
2026-11-01 |
| pr |
0.02 |
Snapshot numerical NumPy arrays before asynchronous checkpoi |
|
Highly specific checkpoint, NumPy, and t |
2026-10-31 |
| pr |
0.02 |
fix(trainer): ensure model configure_callbacks ProgressBar o |
|
Brief, domain-specific changelog with co |
2026-09-14 |
| pr |
0.02 |
Fix `ckpt_path="last"` not finding last checkpoints with a t |
|
Detailed reproduction, regex behavior, a |
2026-09-30 |
| pr |
0.02 |
Fix BaseFinetuning re-freezing unfrozen modules on resume |
|
Detailed resume semantics, parameter-gro |
2026-08-14 |
| pr |
0.02 |
Fix stale batch indices in prediction writers for dataloader |
|
Concrete reproduction output, issue refe |
2026-10-27 |
| pr |
0.02 |
CI: update transformer engine & 1-bit adam DeepSpeed links i |
|
Brief, informal wording and grammatical |
2026-09-29 |
| pr |
0.02 |
Add Deepspeed Zero 3 MiCS support |
|
Grammar errors and terse implementation |
2024-12-03 |
| pr |
0.02 |
Fix dataloader not reloading when resuming from checkpoint |
|
Terse, concrete checkpoint behavior and |
2026-01-28 |
| commit |
0.01 |
chore: update changelogs and bump to 2.7.0.dev (#21949) |
|
Terse release-maintenance notes and vers |
2026-09-14 |
| commit |
0.01 |
Fix FSDPPrecision rejecting a scaler for 16-mixed precision |
|
Concise, domain-specific bug explanation |
2026-09-12 |
| commit |
0.01 |
docs: add 2.6 and 2.7 rows to the compatibility matrix (#218 |
|
Release and CI references are concrete a |
2026-09-01 |
| commit |
0.01 |
ci: bump torch ceiling to <2.14.0 and fix torch-tensorrt/Dee |
|
Highly specific CI failure analysis and |
2026-09-01 |
| commit |
0.01 |
Fix PermissionError swallowing in _atomic_save (#21799) |
|
Terse notes, typos, bot trailers, and co |
2026-09-01 |
| pr |
0.01 |
add `to_tensorrt` in documentations. |
|
Brief, informal documentation note with |
2026-07-19 |
| pr |
0.01 |
Enhance debugging_basic.rst with FLOPs and input sizes |
|
Concise, slightly informal documentation |
2026-11-01 |
| pr |
0.01 |
CI [doctest]: Don't match the progress bar width in the hook |
|
Contains only a title and inherited temp |
2026-10-29 |
| pr |
0.01 |
fix(mlflow): Enabling multiple callbacks for checkpoint repo |
|
Informal phrasing, typos, and direct per |
2025-02-11 |
| pr |
0.01 |
feature: add `sub_dir` to CSV logger constructor |
|
Mostly template content with minimal fre |
2025-08-11 |
| commit |
0.00 |
Avoid negative epoch totals for unlimited training (#21924) |
|
— |
2026-08-31 |
| commit |
0.00 |
Remove dead Python 3.9 guards and duplicate torch constant ( |
|
Concise, domain-specific commit message— |
2026-07-20 |
| commit |
0.00 |
Remove dead jsonargparse<4.39 skip guards (#21848) |
|
Brief message with typical engineering w |
2026-07-20 |
| commit |
0.00 |
Remove unreachable torchvision<0.14 fallback in get_torchvis |
|
Technical, succinct commit message; huma |
2026-07-20 |
| commit |
0.00 |
fix(tests): flaky model summary precision test on CPU (#2185 |
|
Detailed, domain-specific, with informal |
2026-07-20 |
| commit |
0.00 |
ci: add pytorch 2.12 and 2.13 to cpu tests (#21835) |
|
Commit messages are succinct and domain- |
2026-07-16 |
| commit |
0.00 |
Disable compile tracing at the LightningModule.log boundary |
|
Short, technical commit title showing hu |
2026-07-16 |
| commit |
0.00 |
fix(checkpoint): prevent arbitrary code execution via `_inst |
|
Concise technical language; no AI hallma |
2026-07-14 |
| commit |
0.00 |
ci: add pytorch 2.11 to cpu tests (#21679) |
|
Uses changelog commit style with technic |
2026-07-14 |
| commit |
0.00 |
Fix DeepSpeed checkpoint path validation for remote filesyst |
|
Technical details, terse phrasing, and r |
2026-07-14 |
| commit |
0.00 |
Stream atomic checkpoint saves to local disk instead of buff |
|
Contains domain jargon and a co-author t |
2026-07-14 |
| commit |
0.00 |
feat: support remote storage (fsspec) in checkpoint consolid |
|
Informal tone and minimal explanations; |
2026-07-14 |
| commit |
0.00 |
fix: remove catch protections for unscaled AMP types for gra |
|
Brief, domain-specific commit message wi |
2026-07-14 |
| commit |
0.00 |
Fix FSDP CPU offloading when shards are already on CPU (#218 |
|
Terse commit message, technical, no AI s |
2026-07-13 |
| commit |
0.00 |
ci: bump ubuntu in `readthedocs` (#21828) |
|
Single-word, terse commit message—typica |
2026-07-13 |
| commit |
0.00 |
Support remote storage (fsspec URLs) for FSDP checkpoints (# |
|
Standard detailed commit messages; no AI |
2026-07-08 |
| commit |
0.00 |
Fix: Add `weights_only` parameter to LearningRateFinder chec |
|
Commit messages use terse, technical lan |
2026-06-09 |
| commit |
0.00 |
jsonargparse 4.39 required for CLI, check at run-time (#2174 |
|
Commit message is concise and has a huma |
2026-06-09 |
| commit |
0.00 |
fix: AccumulateGrad stream mismatch warning when using DDP w |
|
Commit message is terse and uses informa |
2026-06-01 |
| commit |
0.00 |
feat: add `filter_keys` to log only specified device stats |
|
Changelog and commit text are terse and |
2026-06-01 |
| commit |
0.00 |
Fix torch.compile breaking toggle_optimizer / untoggle_optim |
|
Technical, detailed, with domain abbrevi |
2026-06-01 |
| commit |
0.00 |
build(deps): bump astral-sh/setup-uv from 7.6.0 to 8.1.0 (#2 |
|
— |
2026-06-01 |
| commit |
0.00 |
build(deps): bump pypa/gh-action-pypi-publish from 1.13.0 to |
|
— |
2026-06-01 |
| commit |
0.00 |
fix: use del instead of explicit shutdown in CombinedLoader. |
|
Commit message and changes are concise a |
2026-06-01 |
| commit |
0.00 |
docs: expand DDP metric synchronization guidance (#21685) |
|
— |
2026-05-27 |
| commit |
0.00 |
ci: temporarily disable litbot checkgroup (#21713) |
|
— |
2026-05-27 |
| commit |
0.00 |
ci: grant minimum permissions to probot (#21687) |
|
— |
2026-05-27 |
| commit |
0.00 |
[Docs] Fix Sphinx parameter permalink positioning (#21734) |
|
— |
2026-05-27 |
| commit |
0.00 |
Pin internal workflow actions to SHAs (5/n) (#21738) |
|
— |
2026-05-25 |
| commit |
0.00 |
Pin third-party CI workflow actions to SHAs (3/n) (#21736) |
|
— |
2026-05-25 |
| commit |
0.00 |
Pin disabled TPU workflow actions to SHAs (6/n) (#21739) |
|
— |
2026-05-25 |
| commit |
0.00 |
Pin cloud and release workflow actions to SHAs (4/n) (#21737 |
|
— |
2026-05-25 |
| commit |
0.00 |
Pin official GitHub workflow actions to SHAs (2/n) (#21735) |
|
— |
2026-05-25 |
| commit |
0.00 |
Pin PyPI publish action for trusted publishing (#21730) |
|
Terse, domain-specific commit message wi |
2026-05-20 |
| commit |
0.00 |
chore: prepare changelog for 2.6.4 release (#21729) |
|
Changelog details with informal, concise |
2026-05-20 |
| commit |
0.00 |
ci: add 2-day package install cooldown (supply-chain guard) |
|
Highly technical, succinct, and informal |
2026-05-19 |
| commit |
0.00 |
ci: delete all pull_request_target workflows (#21721) |
|
Brief, domain-focused, and informal comm |
2026-05-19 |
| commit |
0.00 |
Pin low-risk GitHub workflow actions to SHAs (1/n) (#21719) |
|
Concise technical commit message; human |
2026-05-18 |
| commit |
0.00 |
ci: bump auto-label-merge-conflicts commit hash (#21704) |
|
Commit message is terse and technical wi |
2026-05-11 |
| commit |
0.00 |
Setup package publishing via PyPi Trusted Publisher Manageme |
|
Concise title, domain-specific, non-AI p |
2026-05-05 |
| commit |
0.00 |
fix (ci): allow `probot` comment posting on fork PRs (#21670 |
|
Terse commit log, no AI hallmarks, co-au |
2026-04-28 |
| commit |
0.00 |
chore: add make command for running standalone tests (#21668 |
|
Commit messages are terse and domain-spe |
2026-04-21 |
| commit |
0.00 |
build(deps): update onnxscript requirement from <0.6.0,>=0.2 |
|
— |
2026-04-13 |
| commit |
0.00 |
build(deps): update fsspec requirement from <2026.3.0,>=2022 |
|
— |
2026-04-13 |
| commit |
0.00 |
Warn when FSDP auto-wrap policy splits tied weights (#21613) |
|
Commit message is concise, technical, an |
2026-04-13 |
| commit |
0.00 |
build(deps): bump Lightning-AI/probot from 1 to 5 (#21658) |
|
— |
2026-04-13 |
| commit |
0.00 |
[pre-commit.ci] pre-commit suggestions (#21653) |
|
— |
2026-04-13 |
| commit |
0.00 |
build(deps): bump mypy from 1.19.1 to 1.20.0 in /requirement |
|
— |
2026-04-13 |
| commit |
0.00 |
build(deps): update onnx requirement from <1.21.0,>1.12.0 to |
|
— |
2026-04-13 |
| commit |
0.00 |
build(deps): update requests requirement from <2.33.0 to <2. |
|
— |
2026-04-13 |
| commit |
0.00 |
build(deps): update sphinx-togglebutton requirement from <=0 |
|
— |
2026-04-13 |
| commit |
0.00 |
build(deps): update packaging requirement from <25.1 to <26. |
|
— |
2026-04-13 |
| commit |
0.00 |
Fix free-threaded cleanup race in _replace_dunder_methods (# |
|
Concise, technical commit message with n |
2026-04-01 |
| commit |
0.00 |
feat: add `suggest_integrations` flag to toggle `integration |
|
Brief and domain-specific; no AI signals |
2026-04-01 |
| commit |
0.00 |
fix: restore cached AMP step context after no_grad workaroun |
|
Terse, technical changelog; no signs of |
2026-04-01 |
| commit |
0.00 |
Add `log_key_prefix` parameter to `LearningRateMonitor` (#21 |
|
Technical language; clear domain experti |
2026-04-01 |
| commit |
0.00 |
build(deps): update ipython requirement from <9.12.0 to <9.1 |
|
— |
2026-03-30 |
| commit |
0.00 |
fix: set nonzero exit code for SIGTERMException (#21623) |
|
Technical explanation with Unix referenc |
2026-03-30 |
| commit |
0.00 |
build(deps): update importlib-metadata requirement from <9.0 |
|
— |
2026-03-30 |
| commit |
0.00 |
build(deps): bump codecov/codecov-action from 5 to 6 (#21625 |
|
— |
2026-03-30 |
| commit |
0.00 |
build(deps): update numpy requirement from <2.4.2,>2.1.0 to |
|
— |
2026-03-30 |
| commit |
0.00 |
drop deprecated stale.yml config (#21597) |
|
Very short, direct commit message, clear |
2026-03-30 |
| commit |
0.00 |
Docs: avoid unresolved torch.jit Sphinx refs to fix the CICD |
|
Concise, domain-specific documentation f |
2026-03-30 |
| commit |
0.00 |
build(deps): update fsspec[http] requirement from <2026.2.0, |
|
— |
2026-03-30 |
| commit |
0.00 |
build(deps): update onnxruntime requirement from <1.24.0,>=1 |
|
— |
2026-03-30 |
| commit |
0.00 |
build(deps): bump pytest-doctestplus from 1.7.0 to 1.7.1 in |
|
— |
2026-03-30 |
| commit |
0.00 |
build(deps): update ipython[notebook] requirement from <9.10 |
|
— |
2026-03-30 |
| commit |
0.00 |
build(deps): update jsonargparse[jsonnet,signatures] require |
|
— |
2026-03-30 |
| commit |
0.00 |
build(deps): bump coverage from 7.13.4 to 7.13.5 in /require |
|
— |
2026-03-30 |
| commit |
0.00 |
Use fs.pipe() for S3/GCS checkpoint uploads in _atomic_save |
|
Detailed technical changelog, domain-ric |
2026-03-21 |
| commit |
0.00 |
Fix LearningRateMonitor TypeError on MPS with float64 lr val |
|
Changelog is technical, references bugs, |
2026-03-20 |
| commit |
0.00 |
Prepare Release Note for 2.6.2 Release (#21592) |
|
Minimal text, release note update, bot c |
2026-03-19 |
| commit |
0.00 |
Remove NeptuneLogger integration from PyTorch Lightning (#21 |
|
Technical commit language and domain ter |
2026-03-19 |
| commit |
0.00 |
Update Litlogger Integration to latest version (#21591) |
|
Concise, informal commit messages; uses |
2026-03-19 |
| commit |
0.00 |
fix: FSDP mixed precision initializes params in fp32, not bf |
|
Terse and direct fix message common to h |
2026-03-17 |
| commit |
0.00 |
build(deps): update rich requirement from <14.3.0,>=12.3.0 t |
|
— |
2026-03-17 |
| pr |
0.00 |
build(deps): bump actions/setup-python from 6.2.0 to 7.1.0 |
|
— |
2026-10-26 |
| pr |
0.00 |
build(deps): bump actions/setup-python from 6.2.0 to 7.0.0 |
|
— |
2026-09-14 |
| pr |
0.00 |
build(deps): bump pypa/gh-action-pypi-publish from 1.14.0 to |
|
— |
2026-08-03 |