Repository navigation

All pull requests

Labels (149)Milestones (4)New pull request

Open

(293)

Closed

(11,041)

Add MLFlowLogger.log_model to log and register models with MLflow
#22051Michael-Duck opened Oct 23, 2026
Fix ModelSummary.total_layer_params double-counting nested modules when max_depth != 1
#22050deep1blue opened Oct 22, 2026
Account for GradientAccumulationScheduler in Trainer.estimated_stepping_batches
#22049mottopanikeiku opened Oct 22, 20263 of 7 tasks
docs: resolve Fabric experimental policy links with intersphinx
#22047littlebullGit opened Oct 22, 2026Contributor
Fix ModelPruning._make_pruning_permanent_on_state_dict suffix stripping and O(M^2) sanitization
#22045deep1blue opened Oct 22, 20268 of 12 tasks
Fix _Linear8bitLt quantized guards and O(N^2) missing-keys hook in bitsandbytes plugin
#22044deep1blue opened Oct 22, 2026
Fix 'a optimizer' grammar in configure_optimizers docstring example
#22040aarushkandukoori opened Oct 21, 2026
fix(callbacks): check parameter identity instead of torch.equal in BaseFinetuning.filter_on_optimizer
#22039deep1blue opened Oct 21, 20265 tasks done
docs: highlight key progress-bar examples
#22038aryanbarde80 opened Oct 21, 2026
perf(callbacks): avoid O(G^2) duplicate scan and redundant string replacements in LearningRateMonitor
#22037deep1blue opened Oct 21, 20263 tasks done
Fix LearningRateMonitor crash for optimizers without weight decay
#22036RaphaelFakhri opened Oct 20, 20263 of 7 tasks
fix(checkpoint): avoid overwriting checkpoint after resume
#22034wpyaoxueros opened Oct 19, 2026
docs: say fast_dev_run=True runs one batch in the debugging guide
#22030ZainnQureshii opened Oct 19, 20262 of 7 tasks
docs: fix the default devices example in the Fabric arguments page
#22029ZainnQureshii opened Oct 19, 20262 of 7 tasks
Honor strict=False when loading sharded (distributed) checkpoints
#22028deep1blue opened Oct 19, 20264 tasks done
Fix 'a accelerator' grammar in AcceleratorRegistry docstrings
#22026aarushkandukoori opened Oct 15, 2026
fix: isolate AMP autocast cache clearing across threads
#22025Shoryamishra61 opened Oct 14, 20267 of 8 tasks
1
docs: highlight key lines in EarlyStopping examples
#22024Rahul5914 opened Oct 13, 20262 of 7 tasks
fix(data): support positional DataLoader arguments when re-instantiating
#22022dai123x opened Oct 12, 20263 tasks done
2
build(deps): bump actions/upload-artifact from 7.0.1 to 7.0.2
#22021dependabot[bot] opened Oct 12, 2026Bot
1
build(deps): bump actions/download-artifact from 8.0.1 to 8.0.2
#22020dependabot[bot] opened Oct 12, 2026Bot
1
Fix MLFlowLogger reading MLFLOW_TRACKING_URI at import time
#22019abhishekacharya200 opened Oct 11, 20263 tasks done
Fix WandbLogger ignoring dir when save_dir is not set
#22018abhishekacharya200 opened Oct 11, 20263 tasks done
Improve error message for mixed strategy namespaces
#22017Subodh-17 opened Oct 9, 2026
fix(strategies): report the world size as num_replicas for DDP and FSDP
#22016dai123x opened Oct 9, 20263 tasks done

Navigation Menu

Repository navigation

All pull requests

Add `MLFlowLogger.log_model` to log and register models with MLflow

Fix `ModelSummary.total_layer_params` double-counting nested modules when `max_depth != 1`

Account for `GradientAccumulationScheduler` in `Trainer.estimated_stepping_batches`

docs: resolve Fabric experimental policy links with intersphinx

Fix `ModelPruning._make_pruning_permanent_on_state_dict` suffix stripping and `O(M^2)` sanitization

Fix _Linear8bitLt quantized guards and O(N^2) missing-keys hook in bitsandbytes plugin

Fix 'a optimizer' grammar in configure_optimizers docstring example

fix(callbacks): check parameter identity instead of torch.equal in BaseFinetuning.filter_on_optimizer

docs: highlight key progress-bar examples

perf(callbacks): avoid O(G^2) duplicate scan and redundant string replacements in LearningRateMonitor

Fix `LearningRateMonitor` crash for optimizers without weight decay

fix(checkpoint): avoid overwriting checkpoint after resume

docs: say `fast_dev_run=True` runs one batch in the debugging guide

docs: fix the default `devices` example in the Fabric arguments page

Honor `strict=False` when loading sharded (distributed) checkpoints

Fix 'a accelerator' grammar in AcceleratorRegistry docstrings

fix: isolate AMP autocast cache clearing across threads

docs: highlight key lines in EarlyStopping examples

fix(data): support positional DataLoader arguments when re-instantiating

build(deps): bump actions/upload-artifact from 7.0.1 to 7.0.2

build(deps): bump actions/download-artifact from 8.0.1 to 8.0.2

Fix `MLFlowLogger` reading `MLFLOW_TRACKING_URI` at import time

Fix `WandbLogger` ignoring `dir` when `save_dir` is not set

Improve error message for mixed strategy namespaces

fix(strategies): report the world size as `num_replicas` for DDP and FSDP

Pull requests