Mosaicml

Composer now supports the recently-released PyTorch [version 2.3.0](https://dev-discuss.pytorch.org/t/pytorch-2-3-0-general-availability/2033)! Please raise any issues with us so we can address them.

Bug Fixes
* Fixing checks for device microbatch size for sequence parallelism in [3200](https://github.com/mosaicml/composer/pull/3200)
* Fixing token logging in [3206](https://github.com/mosaicml/composer/pull/3206)
* Search for run name in MLFlowLogger in [3215](https://github.com/mosaicml/composer/pull/3215)
* Fix FQN names with activation checkpointing in [3210](https://github.com/mosaicml/composer/pull/3210)
* Strict weight matching for checkpoint loading in [3219](https://github.com/mosaicml/composer/pull/3219)

What's Changed
* Bump transformers by dakinggg in https://github.com/mosaicml/composer/pull/3197
* Add deprecation warnings for ICL datasets/helper functions/metrics by bmosaicml in https://github.com/mosaicml/composer/pull/3125
* Bump traitlets from 5.14.2 to 5.14.3 by dependabot in https://github.com/mosaicml/composer/pull/3204
* Raise LR schedule warnings only when necessary by snarayan21 in https://github.com/mosaicml/composer/pull/3207
* Add torch 2.3 support by mvpatel2000 in https://github.com/mosaicml/composer/pull/3209
* Add torch 2.3 CI/CD by mvpatel2000 in https://github.com/mosaicml/composer/pull/3211
* Fix daily test images by mvpatel2000 in https://github.com/mosaicml/composer/pull/3212
* Try FAv2 2.5.7 from source by mvpatel2000 in https://github.com/mosaicml/composer/pull/3213
* Update tests by mvpatel2000 in https://github.com/mosaicml/composer/pull/3217
* Fix torch 2.3 GPU tests by mvpatel2000 in https://github.com/mosaicml/composer/pull/3218
* Use flash-attn 2.5.8 with no build isolation in docker images by snarayan21 in https://github.com/mosaicml/composer/pull/3224
* Add a torch.cuda.empty_cache() in utils.save_checkpoint by bfontain in https://github.com/mosaicml/composer/pull/3216
* Require 2 steps for GS object store by mvpatel2000 in https://github.com/mosaicml/composer/pull/3228
* Add `rename_metrics` to Mlflow logger by hanlint in https://github.com/mosaicml/composer/pull/3225
* Fix daily tests by mvpatel2000 in https://github.com/mosaicml/composer/pull/3229
* Change precision for daily tests by mvpatel2000 in https://github.com/mosaicml/composer/pull/3231
* Create new Mlflow run by default and introduce `run_group` by chenmoneygithub in https://github.com/mosaicml/composer/pull/3208
* Fix daily test pt 4 by mvpatel2000 in https://github.com/mosaicml/composer/pull/3233
* Deprecate and bump version to 0.22 by mvpatel2000 in https://github.com/mosaicml/composer/pull/3230
* Fix daily tests v5 by mvpatel2000 in https://github.com/mosaicml/composer/pull/3234
* Fix daily v6 by mvpatel2000 in https://github.com/mosaicml/composer/pull/3235
* fix daily tests v7 by mvpatel2000 in https://github.com/mosaicml/composer/pull/3236
* Raise the daily test timeout by dakinggg in https://github.com/mosaicml/composer/pull/3241
* Accelerate GPU tests by mvpatel2000 in https://github.com/mosaicml/composer/pull/3237
* Make sharded checkpoint loading backwards-compatible by snarayan21 in https://github.com/mosaicml/composer/pull/3240


**Full Changelog**: https://github.com/mosaicml/composer/compare/v0.21.3...v0.22.0

What's New

Bug Fixes

**1. Increased Robustness to Checkpoint Loading**

We've patched several edge cases in loading sharded checkpoints, especially with DTensors, which should decrease memory usage when loading checkpoints. We've also hardened retry logic against object cloud failure, ensuring higher robustness to transient network issues.

What's Changed
* Raise daily test timeout by mvpatel2000 in https://github.com/mosaicml/composer/pull/3172
* fix remote file naming by cli99 in https://github.com/mosaicml/composer/pull/3173
* [fix] DTensor + SHARD_GRAD_OP + use_orig_params by bigning in https://github.com/mosaicml/composer/pull/3175
* Bump db sdk by dakinggg in https://github.com/mosaicml/composer/pull/3176
* Build latest pytorch nightly images by dakinggg in https://github.com/mosaicml/composer/pull/3179
* Add FP8 TransformerEngine activation checkpointing by cli99 in https://github.com/mosaicml/composer/pull/3156
* Enabling the computation of validation loss and other metrics when using sequence parallelism by ShashankMosaicML in https://github.com/mosaicml/composer/pull/3183
* Update mosaic_fsdp_utils.py by vchiley in https://github.com/mosaicml/composer/pull/3185
* Fix the FSDP.optim_state_dict_to_load OOM by bigning in https://github.com/mosaicml/composer/pull/3184
* Revert "Update mosaic_fsdp_utils.py" by vchiley in https://github.com/mosaicml/composer/pull/3187
* Bump databricks-sdk from 0.24.0 to 0.25.1 by dependabot in https://github.com/mosaicml/composer/pull/3190
* Add version tag to local builds by mvpatel2000 in https://github.com/mosaicml/composer/pull/3188
* Update `NeptuneLogger` by AleksanderWWW in https://github.com/mosaicml/composer/pull/3165
* Filter neptune warning in doctests by mvpatel2000 in https://github.com/mosaicml/composer/pull/3195
* Removal of metrics deepcopy before computing the metrics by gregjauvion in https://github.com/mosaicml/composer/pull/3180
* Fix MLFlow Tag Name for Resumption by KuuCi in https://github.com/mosaicml/composer/pull/3194
* Fix mistral gating by dakinggg in https://github.com/mosaicml/composer/pull/3199
* Bump version to 0.21.3 by mvpatel2000 in https://github.com/mosaicml/composer/pull/3198

New Contributors
* gregjauvion made their first contribution in https://github.com/mosaicml/composer/pull/3180

**Full Changelog**: https://github.com/mosaicml/composer/compare/v0.21.2...v0.21.3

Bug Fixes

**1. Enable torch 2.2.2 (3161)**

Composer currently monkeypatches PyTorch for nightly versions in order to fix upstream bugs. With the release of torch 2.2.2, these monkeypatches were mistakenly applied to the stable release due to incorrect gating on imports. This release fixes the gating, enabling torch 2.2.2.

**2. MPS Metric Computation on CPU (3105)**

Due to bugs in computing torchmetrics on Mac devices, we move metric computation onto CPU. This previously had issues with data not properly moving to CPU.

Thank you to hyenal for this contribution!

**3. Batch Sampler Support (3105)**

Composer now supports batch sampler, which previously resulted in an error if specified in the dataloader.

Thank you to Ghelfi for this contribution!

What's Changed
* Make codequality callable by mvpatel2000 in https://github.com/mosaicml/composer/pull/3133
* Explicitly print checkpoint downloading exception by bigning in https://github.com/mosaicml/composer/pull/3131
* Change release actions by mvpatel2000 in https://github.com/mosaicml/composer/pull/3136
* Passing rank and num_replicas to dist.get_sampler by ShashankMosaicML in https://github.com/mosaicml/composer/pull/3137
* Fix broadcast by mvpatel2000 in https://github.com/mosaicml/composer/pull/3138
* Compressor fixes by mbway in https://github.com/mosaicml/composer/pull/3142
* In case of MPS device also copy batch to CPU by hyenal in https://github.com/mosaicml/composer/pull/3105
* Composer object store download retry by bigning in https://github.com/mosaicml/composer/pull/3140
* Bump databricks-sdk from 0.22.0 to 0.23.0 by dependabot in https://github.com/mosaicml/composer/pull/3144
* Update transformers requirement from !=4.34.0,<4.39,>=4.11 to >=4.11,!=4.34.0,<4.40 by dependabot in https://github.com/mosaicml/composer/pull/3148
* Update protobuf requirement from <3.21 to <5.27 by dependabot in https://github.com/mosaicml/composer/pull/3147
* Bump traitlets from 5.14.1 to 5.14.2 by dependabot in https://github.com/mosaicml/composer/pull/3145
* Bump to 0.21 by mvpatel2000 in https://github.com/mosaicml/composer/pull/3150
* Fixing sequence parallel error conditions and adding type float for microbatch_size in typehints by ShashankMosaicML in https://github.com/mosaicml/composer/pull/3139
* Fix torch monkeypatch version check by dakinggg in https://github.com/mosaicml/composer/pull/3155
* Update torchmetrics requirement from <1.3.2,>=0.10.0 to >=0.10.0,<1.3.3 by dependabot in https://github.com/mosaicml/composer/pull/3157
* Bump gitpython from 3.1.42 to 3.1.43 by dependabot in https://github.com/mosaicml/composer/pull/3160
* Prevent crash if signal handler cannot be set by mbway in https://github.com/mosaicml/composer/pull/3152
* Pin pillow for code quality workflow by dakinggg in https://github.com/mosaicml/composer/pull/3162
* Fix torch version check by dakinggg in https://github.com/mosaicml/composer/pull/3161
* add more retry to checkpoint downloading by bigning in https://github.com/mosaicml/composer/pull/3164
* Append to gpu rank log files instead of throwing error by jjanezhang in https://github.com/mosaicml/composer/pull/3166
* Call `set_epoch` on `Dataloader.batch_sampler` if defined by Ghelfi in https://github.com/mosaicml/composer/pull/3124
* Bump version to 0.21.2 by mvpatel2000 in https://github.com/mosaicml/composer/pull/3168

New Contributors
* hyenal made their first contribution in https://github.com/mosaicml/composer/pull/3105
* Ghelfi made their first contribution in https://github.com/mosaicml/composer/pull/3124

**Full Changelog**: https://github.com/mosaicml/composer/compare/v0.21.1...v0.21.2

**Bug Fixes**

**1. Fix to HSDP checkpoint loading**

The previous release broke checkpoint loading when using HSDP with mutliple replicas. This patch release fixes checkpoint loading.

What's Changed
* Fix broadcast by mvpatel2000 in https://github.com/mosaicml/composer/pull/3138

**Full Changelog**: https://github.com/mosaicml/composer/compare/v0.21.0...v0.21.1