Skip to content

Make mamba2 recipe imports consistent with codebase #1038

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

Merged
ananthsub merged 1 commit into from
Oct 22, 2025
Merged

Make mamba2 recipe imports consistent with codebase #1038

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
34 changes: 34 additions & 0 deletions src/megatron/bridge/recipes/mamba/__init__.py
View file
Open in desktop
Original file line number Diff line number Diff line change
@@ -0,0 +1,34 @@
# Copyright (c) 2025, NVIDIA CORPORATION. All rights reserved.
#
# Licensed under the Apache License, Version 2.0 (the "License");
# you may not use this file except in compliance with the License.
# You may obtain a copy of the License at
#
# http://www.apache.org/licenses/LICENSE-2.0
#
# Unless required by applicable law or agreed to in writing, software
# distributed under the License is distributed on an "AS IS" BASIS,
# WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied.
# See the License for the specific language governing permissions and
# limitations under the License.

from .mamba2 import (
mamba2_1p3b_pretrain_config,
mamba2_2p7b_pretrain_config,
mamba2_8b_pretrain_config,
mamba2_130m_pretrain_config,
mamba2_370m_pretrain_config,
mamba2_780m_pretrain_config,
mamba2_hybrid_8b_pretrain_config,
)


__all__ = [
"mamba2_130m_pretrain_config",
"mamba2_370m_pretrain_config",
"mamba2_780m_pretrain_config",
"mamba2_1p3b_pretrain_config",
"mamba2_2p7b_pretrain_config",
"mamba2_8b_pretrain_config",
"mamba2_hybrid_8b_pretrain_config",
]
340 changes: 340 additions & 0 deletions src/megatron/bridge/recipes/mamba/mamba2.py
View file
Open in desktop
Original file line number Diff line number Diff line change
@@ -0,0 +1,340 @@
# Copyright (c) 2025, NVIDIA CORPORATION. All rights reserved.
#
# Licensed under the Apache License, Version 2.0 (the "License");
# you may not use this file except in compliance with the License.
# You may obtain a copy of the License at
#
# http://www.apache.org/licenses/LICENSE-2.0
#
# Unless required by applicable law or agreed to in writing, software
# distributed under the License is distributed on an "AS IS" BASIS,
# WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied.
# See the License for the specific language governing permissions and
# limitations under the License.

import os

import torch
from typing_extensions import TypedDict, Unpack

from megatron.bridge.models.mamba import (
MambaModelProvider1P3B,
MambaModelProvider2P7B,
MambaModelProvider130M,
MambaModelProvider370M,
MambaModelProvider780M,
NVIDIAMambaHybridProvider8B,
NVIDIAMambaModelProvider8B,
)
from megatron.bridge.recipes.utils.dataset_utils import get_blend_fields_from_data_paths
from megatron.bridge.recipes.utils.optimizer_utils import distributed_fused_adam_with_cosine_annealing
from megatron.bridge.recipes.utils.tokenizer_utils import DEFAULT_NULL_TOKENIZER_VOCAB_SIZE
from megatron.bridge.training.comm_overlap import CommOverlapConfig
from megatron.bridge.training.config import (
CheckpointConfig,
ConfigContainer,
DistributedDataParallelConfig,
GPTDatasetConfig,
LoggerConfig,
RNGConfig,
TokenizerConfig,
TrainingConfig,
)
from megatron.bridge.training.mixed_precision import MixedPrecisionConfig


class Mamba2CommonKwargs(TypedDict, total=False):
"""Typed options accepted by Mamba2 recipe helper functions."""

# Core identifiers
model_provider: (
type[MambaModelProvider130M]
| type[MambaModelProvider370M]
| type[MambaModelProvider780M]
| type[MambaModelProvider1P3B]
| type[MambaModelProvider2P7B]
| type[NVIDIAMambaModelProvider8B]
| type[NVIDIAMambaHybridProvider8B]
)
tokenizer_model: str | None
dir: str | None
name: str
# Dataset configuration
data_paths: list[str] | None
data_args_path: str | None
train_data_path: list[str] | None
valid_data_path: list[str] | None
test_data_path: list[str] | None
per_split_data_args_path: str | None
mock: bool
# Model configuration
tensor_parallelism: int
pipeline_parallelism: int
pipeline_parallelism_dtype: torch.dtype | None
virtual_pipeline_parallelism: int | None
context_parallelism: int
sequence_parallelism: bool
# Training hyperparameters
train_iters: int
global_batch_size: int
micro_batch_size: int
seq_length: int
lr: float
min_lr: float
lr_warmup_iters: int
lr_decay_iters: int | None
# Tokenizer selection
use_null_tokenizer: bool
# Precision / overlap configs
precision_config: MixedPrecisionConfig | str | None
comm_overlap_config: CommOverlapConfig | None


def mamba2_130m_pretrain_config(**user_kwargs: Unpack[Mamba2CommonKwargs]) -> ConfigContainer:
"""Return a pre-training config for Mamba2 130M."""
recommended: Mamba2CommonKwargs = {
"model_provider": MambaModelProvider130M,
"tensor_parallelism": 1,
"pipeline_parallelism": 1,
"sequence_parallelism": False,
"precision_config": "bf16_mixed",
"use_null_tokenizer": False,
}
kwargs: Mamba2CommonKwargs = {**recommended, **user_kwargs}
return _mamba2_common(tokenizer_model=kwargs.get("tokenizer_model"), **kwargs)


def mamba2_370m_pretrain_config(**user_kwargs: Unpack[Mamba2CommonKwargs]) -> ConfigContainer:
"""Return a pre-training config for Mamba2 370M."""
recommended: Mamba2CommonKwargs = {
"model_provider": MambaModelProvider370M,
"tensor_parallelism": 1,
"pipeline_parallelism": 1,
"sequence_parallelism": False,
"precision_config": "bf16_mixed",
"use_null_tokenizer": False,
}
kwargs: Mamba2CommonKwargs = {**recommended, **user_kwargs}
return _mamba2_common(tokenizer_model=kwargs.get("tokenizer_model"), **kwargs)


def mamba2_780m_pretrain_config(**user_kwargs: Unpack[Mamba2CommonKwargs]) -> ConfigContainer:
"""Return a pre-training config for Mamba2 780M."""
recommended: Mamba2CommonKwargs = {
"model_provider": MambaModelProvider780M,
"tensor_parallelism": 1,
"pipeline_parallelism": 1,
"sequence_parallelism": False,
"precision_config": "bf16_mixed",
"use_null_tokenizer": False,
}
kwargs: Mamba2CommonKwargs = {**recommended, **user_kwargs}
return _mamba2_common(tokenizer_model=kwargs.get("tokenizer_model"), **kwargs)


def mamba2_1p3b_pretrain_config(**user_kwargs: Unpack[Mamba2CommonKwargs]) -> ConfigContainer:
"""Return a pre-training config for Mamba2 1.3B."""
recommended: Mamba2CommonKwargs = {
"model_provider": MambaModelProvider1P3B,
"tensor_parallelism": 1,
"pipeline_parallelism": 1,
"sequence_parallelism": False,
"precision_config": "bf16_mixed",
"use_null_tokenizer": False,
}
kwargs: Mamba2CommonKwargs = {**recommended, **user_kwargs}
return _mamba2_common(tokenizer_model=kwargs.get("tokenizer_model"), **kwargs)


def mamba2_2p7b_pretrain_config(**user_kwargs: Unpack[Mamba2CommonKwargs]) -> ConfigContainer:
"""Return a pre-training config for Mamba2 2.7B."""
recommended: Mamba2CommonKwargs = {
"model_provider": MambaModelProvider2P7B,
"tensor_parallelism": 1,
"pipeline_parallelism": 1,
"sequence_parallelism": False,
"precision_config": "bf16_mixed",
"use_null_tokenizer": False,
}
kwargs: Mamba2CommonKwargs = {**recommended, **user_kwargs}
return _mamba2_common(tokenizer_model=kwargs.get("tokenizer_model"), **kwargs)


def mamba2_8b_pretrain_config(**user_kwargs: Unpack[Mamba2CommonKwargs]) -> ConfigContainer:
"""Return a pre-training config for Mamba2 8B."""
recommended: Mamba2CommonKwargs = {
"model_provider": NVIDIAMambaModelProvider8B,
"tensor_parallelism": 8,
"pipeline_parallelism": 1,
"sequence_parallelism": False,
"precision_config": "bf16_mixed",
"use_null_tokenizer": True,
}
kwargs: Mamba2CommonKwargs = {**recommended, **user_kwargs}
return _mamba2_common(tokenizer_model=None, **kwargs)


def mamba2_hybrid_8b_pretrain_config(**user_kwargs: Unpack[Mamba2CommonKwargs]) -> ConfigContainer:
"""Return a pre-training config for Mamba2 Hybrid 8B."""
recommended: Mamba2CommonKwargs = {
"model_provider": NVIDIAMambaHybridProvider8B,
"tensor_parallelism": 8,
"pipeline_parallelism": 1,
"sequence_parallelism": False,
"precision_config": "bf16_mixed",
"use_null_tokenizer": True,
}
kwargs: Mamba2CommonKwargs = {**recommended, **user_kwargs}
return _mamba2_common(tokenizer_model=None, **kwargs)


def _mamba2_common(
model_provider: (
type[MambaModelProvider130M]
| type[MambaModelProvider370M]
| type[MambaModelProvider780M]
| type[MambaModelProvider1P3B]
| type[MambaModelProvider2P7B]
| type[NVIDIAMambaModelProvider8B]
| type[NVIDIAMambaHybridProvider8B]
),
tokenizer_model: str | None = None,
dir: str | None = None,
name: str = "default",
# Dataset configuration
data_paths: list[str] | None = None,
data_args_path: str | None = None,
train_data_path: list[str] | None = None,
valid_data_path: list[str] | None = None,
test_data_path: list[str] | None = None,
per_split_data_args_path: str | None = None,
mock: bool = False,
# Model configuration
tensor_parallelism: int = 1,
pipeline_parallelism: int = 1,
pipeline_parallelism_dtype: torch.dtype | None = None,
virtual_pipeline_parallelism: int | None = None,
context_parallelism: int = 1,
sequence_parallelism: bool = False,
# Training hyperparameters
train_iters: int = 1_168_251,
global_batch_size: int = 8,
micro_batch_size: int = 1,
seq_length: int = 4096,
lr: float = 3e-4,
min_lr: float = 3e-5,
lr_warmup_iters: int = 2000,
lr_decay_iters: int | None = None,
# Tokenizer selection
use_null_tokenizer: bool = False,
# Precision recipe
precision_config: MixedPrecisionConfig | str | None = "bf16_mixed",
comm_overlap_config: CommOverlapConfig | None = None,
) -> ConfigContainer:
"""
Create a pre-training configuration for Mamba 2.x models.

Args mirror the individual recipe helpers; see those functions for recommended defaults.
"""
base_output_dir = dir if dir is not None else os.path.join(os.getcwd(), "nemo_experiments")
run_output_dir = os.path.join(base_output_dir, name)
checkpoint_dir = os.path.join(run_output_dir, "checkpoints")
tensorboard_dir = os.path.join(run_output_dir, "tb_logs")

blend, blend_per_split, split = get_blend_fields_from_data_paths(
data_paths, data_args_path, train_data_path, valid_data_path, test_data_path, per_split_data_args_path, mock
)

model_cfg = model_provider(
tensor_model_parallel_size=tensor_parallelism,
pipeline_model_parallel_size=pipeline_parallelism,
pipeline_dtype=pipeline_parallelism_dtype,
virtual_pipeline_model_parallel_size=virtual_pipeline_parallelism,
context_parallel_size=context_parallelism,
sequence_parallel=sequence_parallelism,
)

opt_config, scheduler = distributed_fused_adam_with_cosine_annealing(
lr_warmup_iters=lr_warmup_iters,
lr_decay_iters=lr_decay_iters,
adam_beta1=0.9,
adam_beta2=0.95,
adam_eps=1e-5,
weight_decay=0.1,
max_lr=lr,
min_lr=min_lr,
)

cfg = ConfigContainer(
model=model_cfg,
train=TrainingConfig(
train_iters=train_iters,
eval_interval=100,
eval_iters=32,
global_batch_size=global_batch_size,
micro_batch_size=micro_batch_size,
),
optimizer=opt_config,
scheduler=scheduler,
ddp=DistributedDataParallelConfig(
check_for_nan_in_grad=True,
grad_reduce_in_fp32=True,
overlap_grad_reduce=True,
overlap_param_gather=True,
use_distributed_optimizer=True,
),
dataset=GPTDatasetConfig(
random_seed=1234,
reset_attention_mask=False,
reset_position_ids=False,
eod_mask_loss=False,
sequence_length=seq_length,
num_dataset_builder_threads=1,
blend=blend,
blend_per_split=blend_per_split,
split=split,
data_sharding=True,
dataloader_type="single",
num_workers=8,
skip_getting_attention_mask_from_dataset=True,
),
logger=LoggerConfig(
log_interval=10,
tensorboard_dir=tensorboard_dir,
),
tokenizer=(
TokenizerConfig(
tokenizer_type="NullTokenizer",
tokenizer_model=None,
vocab_size=DEFAULT_NULL_TOKENIZER_VOCAB_SIZE,
)
if use_null_tokenizer
else TokenizerConfig(
tokenizer_type="HuggingFaceTokenizer",
tokenizer_model=tokenizer_model or "EleutherAI/gpt-neox-20b",
)
),
checkpoint=CheckpointConfig(
save_interval=2000,
save=checkpoint_dir,
load=checkpoint_dir,
ckpt_format="torch_dist",
fully_parallel_load=True,
),
rng=RNGConfig(seed=1234),
comm_overlap=comm_overlap_config,
mixed_precision=precision_config,
)

return cfg


__all__ = [
"mamba2_130m_pretrain_config",
"mamba2_370m_pretrain_config",
"mamba2_780m_pretrain_config",
"mamba2_1p3b_pretrain_config",
"mamba2_2p7b_pretrain_config",
"mamba2_8b_pretrain_config",
"mamba2_hybrid_8b_pretrain_config",
]
Loading