Search results

MLflow model format required for the Register Model

featureIs an improvement or enhancementIs an improvement or enhancement

needs triageWaiting to be triaged by maintainersWaiting to be triaged by maintainers

Status: Open.

#21437 In Lightning-AI/pytorch-lightning;

· GustavoDCC opened on Dec 20, 2025

bf16-mixed causing issues with fused AdamW

bugSomething isn't workingSomething isn't working

needs triageWaiting to be triaged by maintainersWaiting to be triaged by maintainers

ver: 2.5.x

Status: Open.

#21435 In Lightning-AI/pytorch-lightning;

· vsandwar-sumer opened on Dec 20, 2025

MLFlow Logger Tests Not Passing

bugSomething isn't workingSomething isn't working

needs triageWaiting to be triaged by maintainersWaiting to be triaged by maintainers

ver: 2.5.x

Status: Open.

#21433 In Lightning-AI/pytorch-lightning;

· alexeatscake opened on Dec 19, 2025

Trainer.save_checkpoint() occasionally saves corrupt checkpoint in Docker/WSL

bugSomething isn't workingSomething isn't working

checkpointingRelated to checkpointingRelated to checkpointing

dockers

ver: 2.5.x

Status: Open.

#21431 In Lightning-AI/pytorch-lightning;

· stf976 opened on Dec 18, 2025

Optimizer.step called after on_epoch_end

bugSomething isn't workingSomething isn't working

optimizer

ver: 2.5.x

Status: Open.

#21428 In Lightning-AI/pytorch-lightning;

· RorryB opened on Dec 18, 2025

how to implement a customer precision with torchao

3rd partyRelated to a 3rd-partyRelated to a 3rd-party

bugSomething isn't workingSomething isn't working

ver: 2.5.x

Status: Open.

#21425 In Lightning-AI/pytorch-lightning;

· qmpzzpmq opened on Dec 17, 2025

Trainer.predict and Trainer.test reset model state to evaluation

bugSomething isn't workingSomething isn't working

ver: 2.5.x

working as intendedWorking as intendedWorking as intended

Status: Open.

#21410 In Lightning-AI/pytorch-lightning;

· sfo opened on Dec 10, 2025

DDP Validation Metric Logging got Misplaced Silently When Processes Different Metric Keys on Different Devices

bugSomething isn't workingSomething isn't working

strategy: ddpDistributedDataParallelDistributedDataParallel

ver: 2.4.x

Status: Open.

#21409 In Lightning-AI/pytorch-lightning;

· worldlife123 opened on Dec 7, 2025

Checkpointing in signal handlers for SLURM auto-requeueing leads to intermittent failures

bugSomething isn't workingSomething isn't working

needs triageWaiting to be triaged by maintainersWaiting to be triaged by maintainers

ver: 2.5.x

Status: Open.

#21406 In Lightning-AI/pytorch-lightning;

· martenlienen opened on Dec 5, 2025

Make the default auto-requeue signal in SLURM environments configurable

featureIs an improvement or enhancementIs an improvement or enhancement

needs triageWaiting to be triaged by maintainersWaiting to be triaged by maintainers

Status: Open.

#21404 In Lightning-AI/pytorch-lightning;

· martenlienen opened on Dec 5, 2025

tensor size mismatch/crash trying FSDP with Llama 3.x model(s)

bugSomething isn't workingSomething isn't working

strategy: fsdpFully Sharded Data ParallelFully Sharded Data Parallel

ver: 2.5.x

Status: Open.

#21403 In Lightning-AI/pytorch-lightning;

· hpcpony opened on Dec 4, 2025

[Help needed] How to pass <code>weights_only</code> param to <code>tuner</code>

featureIs an improvement or enhancementIs an improvement or enhancement

tuner

Status: Open.

#21401 In Lightning-AI/pytorch-lightning;

· phoeenniixx opened on Dec 2, 2025