Skip to content
New issue

Have a question about this project? Sign up for a free GitHub account to open an issue and contact its maintainers and the community.

Sign up for GitHub

By clicking “Sign up for GitHub”, you agree to our terms of service and privacy statement. We’ll occasionally send you account related emails.

Already on GitHub? Sign in to your account

Jump to bottom

Bump version to 0.27.0.dev0 #3670

Closed
wants to merge 1 commit into from
+12 −57
Closed

Bump version to 0.27.0.dev0 #3670

Show file tree
Hide file tree
Changes from all commits
Commits
Show all changes
1 commit
Select commit
File filter

Filter by extension

Filter by extension
Conversations
Failed to load comments.
Loading
Jump to
Jump to file
Failed to load files.
Loading
Diff view
Diff view
2 changes: 1 addition & 1 deletion composer/_version.py
View file
Open in desktop
Original file line number Diff line number Diff line change
Expand Up @@ -3,4 +3,4 @@

"""The Composer Version."""

__version__ = '0.26.0.dev0'
__version__ = '0.27.0.dev0'
45 changes: 0 additions & 45 deletions composer/trainer/trainer.py
View file
Open in desktop
Original file line number Diff line number Diff line change
Expand Up @@ -990,12 +990,6 @@ class Trainer:

To use DeepSpeed with default values, set to the empty dictionary ``{}``.
To disable DeepSpeed (the default), set to ``None``.
fsdp_config (dict[str, Any], optional): Configuration for FSDP.
See :doc:`FSDP Documentation </notes/distributed_training>` for more details.
To use FSDP with default values, set to the empty dictionary ``{}``. To
disable FSDP, set to ``None``. (default: ``None``)
fsdp_auto_wrap (bool, optional): option to let trainer wrap the module, or if
the module is already wrapped outside, allow the user to disable auto-wrapping.
parallelism_config (Union[dict[str, Any], ParallelismConfig], optional): Configuration for parallelism options.
Currently supports fsdp and tensor parallelism, whose respective configs are specified
as the keys ``fsdp`` and ``tp``. (default: ``None``)
Expand Down Expand Up @@ -1153,8 +1147,6 @@ def __init__(

# Parallelism
deepspeed_config: Optional[dict[str, Any]] = None,
fsdp_config: Optional[dict[str, Any]] = None,
fsdp_auto_wrap: bool = True,
parallelism_config: Optional[Union[dict[str, Any], ParallelismConfig]] = None,

# System/Numerics
Expand Down Expand Up @@ -1279,43 +1271,6 @@ def __init__(
assert not isinstance(device_train_microbatch_size, str)

# Distributed
if fsdp_config is not None:
warnings.warn(
VersionedDeprecationWarning(
"fsdp_config is deprecated. Please use parallelism_config['fsdp'] instead.",
remove_version='0.26.0',
),
)
if parallelism_config is None:
parallelism_config = {}
if isinstance(parallelism_config, ParallelismConfig):
raise ValueError(
'fsdp_config cannot be specified if parallelism_config is a ParallelismConfig object. '
'Please instead pass fsdp_config as a FSDPConfig object when constructing ParallelismConfig.',
)
elif parallelism_config.get('fsdp') is not None:
raise ValueError(
'fsdp_config is specified in both fsdp_config and parallelism_config. Please specify it in only in parallelism_config.',
)
parallelism_config['fsdp'] = fsdp_config
if not fsdp_auto_wrap:
warnings.warn(
VersionedDeprecationWarning(
"fsdp_auto_wrap=False is deprecated. Please use parallelism_config['fsdp']['auto_wrap'] instead.",
remove_version='0.26.0',
),
)
if parallelism_config is None:
parallelism_config = {}
if isinstance(parallelism_config, ParallelismConfig):
raise ValueError(
'fsdp_auto_wrap cannot be specified if parallelism_config is a ParallelismConfig object. '
'Please instead pass fsdp_auto_wrap to FSDPConfig as part of ParallelismConfig.',
)
else:
if parallelism_config.get('fsdp') is None:
parallelism_config['fsdp'] = {}
parallelism_config['fsdp']['auto_wrap'] = fsdp_auto_wrap
if parallelism_config is not None and not isinstance(parallelism_config, ParallelismConfig):
parallelism_config_args = {}
if 'fsdp' in parallelism_config and parallelism_config['fsdp'] is not None:
Expand Down
4 changes: 2 additions & 2 deletions docker/README.md
View file
Open in desktop
Original file line number Diff line number Diff line change
Expand Up @@ -15,8 +15,8 @@ all dependencies for both NLP and Vision models. They are built on top of the
<!-- BEGIN_COMPOSER_BUILD_MATRIX -->
| Composer Version | CUDA Support | Docker Tag |
|--------------------|----------------|------------------------------------------------------------------------------------------------------------------------------------------------------------------|
| 0.25.0 | Yes | `mosaicml/composer:latest`, `mosaicml/composer:0.25.0` |
| 0.25.0 | No | `mosaicml/composer:latest_cpu`, `mosaicml/composer:0.25.0_cpu` |
| 0.26.0 | Yes | `mosaicml/composer:latest`, `mosaicml/composer:0.26.0` |
| 0.26.0 | No | `mosaicml/composer:latest_cpu`, `mosaicml/composer:0.26.0_cpu` |
<!-- END_COMPOSER_BUILD_MATRIX -->

**Note**: For a lightweight installation, we recommended using a [MosaicML PyTorch Image](#pytorch-images) and manually
Expand Down
16 changes: 8 additions & 8 deletions docker/build_matrix.yaml
View file
Open in desktop
Original file line number Diff line number Diff line change
Expand Up @@ -168,36 +168,36 @@
TORCHVISION_VERSION: 0.18.1
- AWS_OFI_NCCL_VERSION: ''
BASE_IMAGE: nvidia/cuda:12.4.1-cudnn-devel-ubuntu20.04
COMPOSER_INSTALL_COMMAND: mosaicml[all]==0.25.0
COMPOSER_INSTALL_COMMAND: mosaicml[all]==0.26.0
CUDA_VERSION: 12.4.1
IMAGE_NAME: composer-0-25-0
IMAGE_NAME: composer-0-26-0
MOFED_VERSION: latest-23.10
NVIDIA_REQUIRE_CUDA_OVERRIDE: ''
PYTHON_VERSION: '3.11'
PYTORCH_NIGHTLY_URL: ''
PYTORCH_NIGHTLY_VERSION: ''
PYTORCH_VERSION: 2.5.0
TAGS:
- mosaicml/composer:0.25.0
- ghcr.io/databricks-mosaic/composer:0.25.0
- mosaicml/composer:0.26.0
- ghcr.io/databricks-mosaic/composer:0.26.0
- mosaicml/composer:latest
- ghcr.io/databricks-mosaic/composer:latest
TARGET: composer_stage
TORCHVISION_VERSION: 0.20.0
- AWS_OFI_NCCL_VERSION: ''
BASE_IMAGE: ubuntu:20.04
COMPOSER_INSTALL_COMMAND: mosaicml[all]==0.25.0
COMPOSER_INSTALL_COMMAND: mosaicml[all]==0.26.0
CUDA_VERSION: ''
IMAGE_NAME: composer-0-25-0-cpu
IMAGE_NAME: composer-0-26-0-cpu
MOFED_VERSION: latest-23.10
NVIDIA_REQUIRE_CUDA_OVERRIDE: ''
PYTHON_VERSION: '3.11'
PYTORCH_NIGHTLY_URL: ''
PYTORCH_NIGHTLY_VERSION: ''
PYTORCH_VERSION: 2.5.0
TAGS:
- mosaicml/composer:0.25.0_cpu
- ghcr.io/databricks-mosaic/composer:0.25.0_cpu
- mosaicml/composer:0.26.0_cpu
- ghcr.io/databricks-mosaic/composer:0.26.0_cpu
- mosaicml/composer:latest_cpu
- ghcr.io/databricks-mosaic/composer:latest_cpu
TARGET: composer_stage
Expand Down
2 changes: 1 addition & 1 deletion docker/generate_build_matrix.py
View file
Open in desktop
Original file line number Diff line number Diff line change
Expand Up @@ -244,7 +244,7 @@ def _main():
composer_entries = []

# The `GIT_COMMIT` is a placeholder and Jenkins will substitute it with the actual git commit for the `composer_staging` images
composer_versions = ['0.25.0'] # Only build images for the latest composer version
composer_versions = ['0.26.0'] # Only build images for the latest composer version
composer_python_versions = [PRODUCTION_PYTHON_VERSION] # just build composer against the latest

for product in itertools.product(composer_python_versions, composer_versions, cuda_options):
Expand Down
Loading