releases.shpreview
Hugging Face/Accelerate

Accelerate

Mon
Wed
Fri
JunJulAugSepOctNovDecJanFebMarAprMay
Less
More
Releases1Avg0/wkVersionsv1.13.0
HighlightsAll Releases

v1.13.0: Neuron support, IPEX removal, and distributed training fixes

↗
v1.13.0

AWS Neuron support

We now have support for AWS Neuron (Trainium/Inferentia) devices. Thanks @michaelbenayoun for adding this.

XPU Improvements

We've removed IPEX dependency and improved device-agnostic code for XPU.

FSDP2 Improvements

We've added a bunch of important fixes for FSDP2 users: upcasting only grad-requiring params, better tied embedding errors, DCP optimizer loading, bf16 optimizer step crash fix, and torch < 2.7.0 compatibility.

DeepSpeed Sequence Parallelism

We've added several fixes to the DeepSpeed + Sequence Parallelism integration introduced in v1.12.0, including evaluation support during SP training and proper process group handling.

FP8

We've enhanced FP8 training. Thanks @shimizust for fixing torchao support.

Performance

Accelerate now imports faster by deferring heavy dependencies, and torch.compile hooks are disabled lazily.

Minor fixes

v1.12.0: Deepspeed Ulysses/ALST

↗
v1.12.0

Deepspeed Ulysses/ALST integration

Deepspeed Ulysses/ALST is an efficient way of training on long sequences by employing sequence parallelism and attention head parallelism. You can learn more about this technology in this paper https://arxiv.org/abs/2506.13996 or this deepspeed tutorial https://www.deepspeed.ai/tutorials/ulysses-alst-sequence-parallelism/.

<img width="2368" height="1250" alt="0d8bd9e0" src="https://github.com/user-attachments/assets/b94e90c9-4368-4711-ad57-58de3c714ebc" />

To enable Deepspeed Ulysses, you first need to create ParallelismConfig and setting sp related args:

parallelism_config = ParallelismConfig(
    sp_backend="deepspeed",
    sp_size=2,
    sp_handler=DeepSpeedSequenceParallelConfig(...),
)

Then, you need to make sure to compute the correct loss as described on our docs

        ...
        losses_per_rank = torch.distributed.nn.functional.all_gather(loss, group=sp_group)
        good_tokens = (shift_labels != -100).view(-1).sum()
        good_tokens_per_rank = torch.distributed.nn.functional.all_gather(good_tokens, group=sp_group)
        total_loss = sum(
            losses_per_rank[rank] * good_tokens_per_rank[rank]
            for rank in range(sp_world_size)
            if good_tokens_per_rank[rank] > 0
        )
        total_good_tokens = sum(good_tokens_per_rank)
        loss = total_loss / max(total_good_tokens, 1)

Thanks @S1ro1 for starting this work and for @stas00 for finishing this work. Also thanks @kashif for adding docs and reviewing/testing this PR !

This feature will also be available in HF Trainer thanks for this PR from @stas00: https://github.com/huggingface/transformers/pull/41832

Minor changes

New Contributors

Full Changelog: https://github.com/huggingface/accelerate/compare/v1.11.0...v1.12.0

v1.11.0: TE MXFP8, FP16/BF16 with MPS, Python 3.10

↗
v1.11.0

TE MXFP8 support

We've added support for MXFP8 in our TransformerEngine integration. To use that, you need to set use_mxfp8_block_scaling in fp8_config. See nvidia docs [here]. (https://docs.nvidia.com/deeplearning/transformer-engine/user-guide/examples/fp8_primer.html#MXFP8-and-block-scaling)

FP16/BF16 Training for MPS devices

BF16 and FP16 support for MPS devices is finally here. You can now pass mixed_precision = "fp16" or "bf16" when training on a mac (fp16 requires torch 2.8 and bf16 requires torch 2.6)

FSDP updates

The following PRs add respectively support to ignored_params and no_sync() for FSDPv2:

Mixed precision can now be passed as a dtype string from accelerate cli flag or fsdp_config in accelerate config file:

Nd-parallel updates

Some minor updates concerning nd-parallelism.

Bump to Python 3.10

We've dropped support for python 3.9 as it reached EOL in October.

Lots of minor fixes:

New Contributors

Full Changelog: https://github.com/huggingface/accelerate/compare/v1.10.1...v1.11.0

v1.10.0: N-D Parallelism

↗
v1.10.0

N-D Parallelism

Training large models across multiple GPUs can be complex, especially when combining different parallelism strategies (e.g TP, CP, DP). To simplify this process, we've collaborated with Axolotl to introduce an easy-to-use integration that allows you to apply any combination of parallelism strategies directly in your training script. Just pass a ParallelismConfig specifying the size of each parallelism type—it's that simple. Learn more about how it works in our latest blogpost.

parallelism_config = ParallelismConfig(
    dp_shard_size=2,
    dp_replicate_size=2,
    cp_size=2,
    tp_size=2,
)
accelerator = Accelerator(
    parallelism_config=parallelism_config,
   ...
)
model = AutoModelForCausalLM.from_pretrained("your-model-name", device_mesh=accelerator.torch_device_mesh)
model = accelerator.prepare(model)

FSDP improvements

We've fixed ignored modules attribute. With this, it is now possible to train PEFT model that moe layers that contrains q_proj and v_proj parameters. This is especially important for fine-tuning gpt-oss model.

Minor improvements

New Contributors

Full Changelog: https://github.com/huggingface/accelerate/compare/v1.9.0...v1.10.0

v1.9.0: Trackio support, Model loading speedup, Minor distributed improvements

↗
v1.9.0

Trackio tracker support

We've added support for a trackio, lightweight, 💯 free experiment tracking Python library built on top of 🤗 Datasets and Spaces.

Main features are:

  • Local-first design: dashboard runs locally by default. You can also host it on Spaces by specifying a space_id.
  • Persists logs locally (or in a private Hugging Face Dataset)
  • Visualize experiments with a Gradio dashboard locally (or on Hugging Face Spaces)
  • Everything here, including hosting on Hugging Faces, is free!

To use it with accelerate, you need to set log_with and initialize the trackers

accelerator = Accelerator(log_with="trackio")
config={"learning_rate": 0.001, "batch_size": 32}
# init_kwargs in order to host the dashboard on spaces
init_kwargs = {"trackio": {"space_id": "hf_username/space_name"}
accelerator.init_trackers("example_project", config=config, init_kwargs=init_kwargs})

Thanks @pcuenca for the integration !

Model loading speedup when relying set_module_tensor_to_device

Setting tensor while clearing cache is very slow, so we added clear_device option to disable it. Another small optimization is using non_blocking everywhere and syncing just before returning control to the user. This makes the loading slightly faster.

FDSP, Deepspeed, FP8 minor improvements

🚨🚨🚨 Breaking changes 🚨🚨🚨

find_executable_batch_size() will no longer halves the batch after every OOM. Instead, we will multiply the batch size by 0.9. This should help user not waste gpu capacity.

What's Changed

New Contributors

Full Changelog: https://github.com/huggingface/accelerate/compare/v1.8.1...v1.9.0

v1.8.0: FSDPv2 + FP8, Regional Compilation for DeepSpeed, Faster Distributed Training on Intel CPUs, ipex.optimize deprecation

↗
v1.8.0

FSDPv2 refactor + FP8 support

We've simplified how to prepare FSDPv2 models, as there were too many ways to compose FSDP2 with other features (e.g., FP8, torch.compile, activation checkpointing, etc.). Although the setup is now more restrictive, it leads to fewer errors and a more performant user experience. We’ve also added support for FP8. You can read about the results here. Thanks to @S1ro1 for this contribution!

Faster Distributed Training on Intel CPUs

We updated the CCL_WORKER_COUNT variable and added KMP parameters for Intel CPU users. This significantly improves distributed training performance (e.g., Tensor Parallelism), with up to a 40% speed-up on Intel 4th Gen Xeon when training transformer TP models.

Regional Compilation for DeepSpeed

We added support for regional compilation with the DeepSpeed engine. DeepSpeed’s .compile() modifies models in-place using torch.nn.Module.compile(...), rather than the out-of-place torch.compile(...), so we had to account for that. Thanks @IlyasMoutawwakil for this feature!

ipex.optimize deprecation

ipex.optimize is being deprecated. Most optimizations have been upstreamed to PyTorch, and future improvements will land there directly. For users without PyTorch 2.8, we’ll continue to rely on IPEX for now.

Better XPU Support

We've greatly expanded and stabilized support for Intel XPUs:

Trackers

We've added support for SwanLab as an experiment tracking backend. Huge thanks to @ShaohonChen for this contribution ! We also deferred all tracker initializations to prevent premature setup of distributed environments.

What's Changed

New Contributors

Full Changelog: https://github.com/huggingface/accelerate/compare/v1.7.0...v1.8.0

v1.7.0 : Regional compilation, Layerwise casting hook, FSDPv2 + QLoRA

↗
v1.7.0

Regional compilation

Instead of compiling the entire model at once, regional compilation targets repeated blocks (such as decoder layers) first. This allows the compiler to cache and reuse optimized code for subsequent blocks, significantly reducing the cold start compilation time typically seen during the first inference. Thanks @IlyasMoutawwakil for the feature ! You can view the full benchmark here, and check out our updated compilation guide for more details!

To enable this feature, set use_regional_compilation=True in the TorchDynamoPlugin configuration.

# Configure the compilation backend
dynamo_plugin = TorchDynamoPlugin(
    use_regional_compilation=True,
    ... # other parameters
)
# Initialize accelerator with the plugin
accelerator = Accelerator(dynamo_plugin=dynamo_plugin)
# This will apply compile_regions to your model
model = accelerator.prepare(model)

Layerwise casting hook

We've introduced a new hook that enables per-layer upcasting and downcasting (e.g., for Linear layers) during inference. This allows users to run models with separate storage and compute dtypes, resulting in memory savings. The concept was first implemented in diffusers, where downcasting models to FP8 proved effective without major quality degradation. Contributed by @sayakpaul in https://github.com/huggingface/accelerate/pull/3427

model = ....
storage_dtype = torch.float8_e4m3fn
compute_dtype = torch.bfloat16
attach_layerwise_casting_hooks(
            model,
            storage_dtype=storage_dtype,
            compute_dtype=compute_dtype,
        )

Better FSDP2 support

This release includes numerous new features and bug fixes. Notably, we’ve added support for FULL_STATE_DICT, a widely used option in FSDP, now enabling .save_pretrained() in transformers to work with FSDP2 wrapped models. QLoRA training is now supported as well but more testing is needed. We have also resolved a backend issue related to parameter offloading to CPU. Additionally, a significant memory spike that occurred when cpu_ram_efficient_loading=True was enabled has been fixed. Several other minor improvements and fixes are also included—see the What’s Changed section for full details.

Better HPU support:

We have added a documentation for Intel Gaudi hardware ! The support is already available since v1.5.0 through this PR.

Torch.compile breaking change for dynamic argument

We've updated the logic for setting self.dynamic to explicitly preserve None rather than defaulting to False when the USE_DYNAMIC environment variable is unset. This change aligns the behavior with the PyTorch documentation for torch.compile. Thanks to @yafshar for contributing this improvement in #3567.

What's Changed

New Contributors

Full Changelog: https://github.com/huggingface/accelerate/compare/v1.6.0...v1.7.0

v1.6.0: FSDPv2, DeepSpeed TP and XCCL backend support

↗
v1.6.0

FSDPv2 support

This release introduces the support for FSDPv2 thanks to @S1ro1.

If you are using python code, you need to set fsdp_version=2 in FullyShardedDataParallelPlugin:

from accelerate import FullyShardedDataParallelPlugin, Accelerator

fsdp_plugin = FullyShardedDataParallelPlugin(
    fsdp_version=2
    # other options...
)
accelerator = Accelerator(fsdp_plugin=fsdp_plugin)

If want to convert a YAML config that contains the FSDPv1 config to FSDPv2 one , use our conversion tool:

accelerate to-fsdp2 --config_file config.yaml --output_file new_config.yaml`

To learn more about the difference between FSDPv1 and FSDPv2, read the following documentation.

DeepSpeed TP support

We have added initial support for DeepSpeed + TP. Not many changes were required as the DeepSpeed APIs was already compatible. We only needed to make sure that the dataloader was compatible with TP and that we were able to save the TP weights. Thanks @inkcherry for the work ! https://github.com/huggingface/accelerate/pull/3390.

To use TP with deepspeed, you need to update the setting in the deepspeed config file by including tensor_parallel key:

    ....
    "tensor_parallel":{
      "autotp_size": ${autotp_size}
    },
   ...

More details in this deepspeed PR.

Support for XCCL distributed backend

We've added support for XCCL which is an Intel distributed backend which can be used with XPU devices. More details in this torch PR. Thanks @dvrogozh for the integration !

What's Changed

New Contributors

Full Changelog: https://github.com/huggingface/accelerate/compare/v1.5.2...v1.6.0

v1.5.0: HPU support

↗
v1.5.0

HPU Support

  • Adds in HPU accelerator support for 🤗 Accelerate

What's Changed

New Contributors

Full Changelog: https://github.com/huggingface/accelerate/compare/v1.4.0...v1.5.0

v1.4.0: `torchao` FP8, TP & dataLoader support, fix memory leak

↗
v1.4.0

torchao FP8, initial Tensor Parallel support, and memory leak fixes

torchao FP8

This release introduces a new FP8 API and brings in a new backend: torchao. To use, pass in AORecipeKwargs to the Accelerator while setting mixed_precision="fp8". This is initial support, as it matures we will incorporate more into it (such as accelerate config/yaml) in future releases. See our benchmark examples here

TensorParallel

We have intial support for an in-house solution to TP when working with accelerate dataloaders. check out the PR here

Bug fixes

What's Changed

New Contributors

Full Changelog: https://github.com/huggingface/accelerate/compare/v1.3.0...v1.4.0

v1.3.0 Bug fixes + Require torch 2.0

↗
v1.3.0

Torch 2.0

As it's been ~2 years since torch 2.0 was first released, we are now requiring this as the minimum version for Accelerate, which similarly was done in transformers as of its last release.

Core

Big Modeling

Examples

Full Changelog

What's Changed

New Contributors

Full Changelog: https://github.com/huggingface/accelerate/compare/v1.2.1...v1.3.0

v1.2.0: Bug Squashing & Fixes across the board

↗
v1.2.0

Core

Big Modeling

DeepSpeed

Documentation

New Contributors

Full Changelog

Code Diff

Release diff: https://github.com/huggingface/accelerate/compare/v1.1.1...v1.2.0

v1.1.0: Python 3.9 minimum, torch dynamo deepspeed support, and bug fixes

↗
v1.1.0

Internals:

DeepSpeed

Megatron

Big Model Inference

Examples

Full Changelog

New Contributors

Full Changelog: https://github.com/huggingface/accelerate/compare/v1.0.1...v1.1.0

Accelerate 1.0.0 is here!

↗
v1.0.0

🚀 Accelerate 1.0 🚀

With accelerate 1.0, we are officially stating that the core parts of the API are now "stable" and ready for the future of what the world of distributed training and PyTorch has to handle. With these release notes, we will focus first on the major breaking changes to get your code fixed, followed by what is new specifically between 0.34.0 and 1.0.

To read more, check out our official blog here

Migration assistance

  • Passing in dispatch_batches, split_batches, even_batches, and use_seedable_sampler to the Accelerator() should now be handled by creating an accelerate.utils.DataLoaderConfiguration() and passing this to the Accelerator() instead (Accelerator(dataloader_config=DataLoaderConfiguration(...)))
  • Accelerator().use_fp16 and AcceleratorState().use_fp16 have been removed; this should be replaced by checking accelerator.mixed_precision == "fp16"
  • Accelerator().autocast() no longer accepts a cache_enabled argument. Instead, an AutocastKwargs() instance should be used which handles this flag (among others) passing it to the Accelerator (Accelerator(kwargs_handlers=[AutocastKwargs(cache_enabled=True)]))
  • accelerate.utils.is_tpu_available should be replaced with accelerate.utils.is_torch_xla_available
  • accelerate.utils.modeling.shard_checkpoint should be replaced with split_torch_state_dict_into_shards from the huggingface_hub library
  • accelerate.tqdm.tqdm() no longer accepts True/False as the first argument, and instead, main_process_only should be passed in as a named argument

Multiple Model DeepSpeed Support

After long request, we finally have multiple model DeepSpeed support in Accelerate! (though it is quite early still). Read the full tutorial here, however essentially:

When using multiple models, a DeepSpeed plugin should be created for each model (and as a result, a separate config). a few examples are below:

Knowledge distillation

(Where we train only one model, zero3, and another is used for inference, zero2)

from accelerate import Accelerator
from accelerate.utils import DeepSpeedPlugin

zero2_plugin = DeepSpeedPlugin(hf_ds_config="zero2_config.json")
zero3_plugin = DeepSpeedPlugin(hf_ds_config="zero3_config.json")

deepspeed_plugins = {"student": zero2_plugin, "teacher": zero3_plugin}


accelerator = Accelerator(deepspeed_plugins=deepspeed_plugins)

To then select which plugin to be used at a certain time (aka when calling prepare), we call `accelerator.state.select_deepspeed_plugin("name"), where the first plugin is active by default:

accelerator.state.select_deepspeed_plugin("student")
student_model, optimizer, scheduler = ...
student_model, optimizer, scheduler, train_dataloader = accelerator.prepare(student_model, optimizer, scheduler, train_dataloader)

accelerator.state.select_deepspeed_plugin("teacher") # This will automatically enable zero init
teacher_model = AutoModel.from_pretrained(...)
teacher_model = accelerator.prepare(teacher_model)
Multiple disjoint models

For disjoint models, separate accelerators should be used for each model, and their own .backward() should be called later:

for batch in dl:
    outputs1 = first_model(**batch)
    first_accelerator.backward(outputs1.loss)
    first_optimizer.step()
    first_scheduler.step()
    first_optimizer.zero_grad()
    
    outputs2 = model2(**batch)
    second_accelerator.backward(outputs2.loss)
    second_optimizer.step()
    second_scheduler.step()
    second_optimizer.zero_grad()

FP8

We've enabled MS-AMP support up to FSDP. At this time we are not going forward with implementing FSDP support with MS-AMP, due to design issues between both libraries that don't make them inter-op easily.

FSDP

  • Fixed FSDP auto_wrap using characters instead of full str for layers
  • Re-enable setting state dict type manually

Big Modeling

  • Removed cpu restriction for bnb training

What's Changed

New Contributors

Full Changelog: https://github.com/huggingface/accelerate/compare/v0.34.2...v1.0.0

Last Checked
14h ago
Latest
v1.13.0
Source
@huggingface/accelerate9.7k
Tracking since Mar 5, 2021
.json·.md·.atom