`bridge.recipes.gemma.gemma3`#

Module Contents#

Classes#

`Gemma3CommonKwargs`	Typed options accepted by Gemma3 family recipe helpers.
`Gemma3FinetuneKwargs`	Typed options accepted by Gemma3 finetuning recipe helper functions.

Functions#

`gemma3_1b_pretrain_config`	Return a pre-training config for Gemma3 1B.
`_gemma3_common`	Create a pre-training configuration for Gemma3 family models.
`gemma3_1b_finetune_config`	Return a finetuning config for Gemma3 1B.
`_gemma3_finetune_common`	Create a finetuning configuration for Gemma3 models.

Data#

`SEQUENCE_LENGTH_32K`
`SEQUENCE_LENGTH_128K`

API#

class bridge.recipes.gemma.gemma3.Gemma3CommonKwargs#

Bases: typing_extensions.TypedDict

Typed options accepted by Gemma3 family recipe helpers.

Initialization

Initialize self. See help(type(self)) for accurate signature.

provider_class: type#: None

hf_path: str | None#: None

dir: str | None#: None

name: str#: None

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#: None

tensor_model_parallel_size: int#: None

pipeline_model_parallel_size: int#: None

pipeline_dtype: torch.dtype | None#: None

virtual_pipeline_model_parallel_size: int | None#: None

context_parallel_size: int#: None

sequence_parallel: bool#: None

use_megatron_fsdp: bool#: None

account_for_embedding_in_pipeline_split: bool#: None

account_for_loss_in_pipeline_split: bool#: None

train_iters: int#: None

global_batch_size: int#: None

micro_batch_size: int#: None

seq_length: int#: None

lr: float#: None

min_lr: float#: None

lr_warmup_iters: int#: None

lr_decay_iters: int | None#: None

eval_interval: int#: None

save_interval: int#: None

use_null_tokenizer: bool#: None

precision_config: megatron.bridge.training.mixed_precision.MixedPrecisionConfig | str | None#: None

comm_overlap_config: megatron.bridge.training.comm_overlap.CommOverlapConfig | None#: None

class bridge.recipes.gemma.gemma3.Gemma3FinetuneKwargs#

Bases: typing_extensions.TypedDict

Typed options accepted by Gemma3 finetuning recipe helper functions.

This is separate from Gemma3CommonKwargs to avoid confusion - finetuning uses SQuAD dataset by default, not the data path fields.

Initialization

Initialize self. See help(type(self)) for accurate signature.

dir: str | None#: None

name: str#: None

pretrained_checkpoint: str | None#: None

peft: str | megatron.bridge.peft.base.PEFT | None#: None

packed_sequence: bool#: None

train_iters: int#: None

global_batch_size: int | None#: None

micro_batch_size: int#: None

seq_length: int | None#: None

eval_interval: int#: None

save_interval: int#: None

finetune_lr: float | None#: None

min_lr: float#: None

lr_warmup_iters: int#: None

lr_decay_iters: int | None#: None

wandb_project: str | None#: None

wandb_entity: str | None#: None

wandb_exp_name: str | None#: None

precision_config: megatron.bridge.training.mixed_precision.MixedPrecisionConfig | str | None#: None

bridge.recipes.gemma.gemma3.SEQUENCE_LENGTH_32K: int#: 32768

bridge.recipes.gemma.gemma3.SEQUENCE_LENGTH_128K: int#: 131072

bridge.recipes.gemma.gemma3.gemma3_1b_pretrain_config(

**user_kwargs: typing_extensions.Unpack[bridge.recipes.gemma.gemma3.Gemma3CommonKwargs],

) → megatron.bridge.training.config.ConfigContainer#

Return a pre-training config for Gemma3 1B.

See _gemma3_common for the full list of parameters.

bridge.recipes.gemma.gemma3._gemma3_common( provider_class: type, hf_path: str | None = None, dir: str | None = None, name: str = 'default', 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, tensor_model_parallel_size: int = 1, pipeline_model_parallel_size: int = 1, pipeline_dtype: torch.dtype | None = None, virtual_pipeline_model_parallel_size: int | None = None, context_parallel_size: int = 1, sequence_parallel: bool = False, use_megatron_fsdp: bool = False, account_for_embedding_in_pipeline_split: bool = False, account_for_loss_in_pipeline_split: bool = False, train_iters: int = 1168251, global_batch_size: int = 512, micro_batch_size: int = 1, seq_length: int = 131072, lr: float = 0.0003, min_lr: float = 3e-05, lr_warmup_iters: int = 2000, lr_decay_iters: int | None = None, eval_interval: int = 2000, save_interval: int = 500, use_null_tokenizer: bool = True, precision_config: megatron.bridge.training.mixed_precision.MixedPrecisionConfig | str | None = 'bf16_mixed', comm_overlap_config: megatron.bridge.training.comm_overlap.CommOverlapConfig | None = None, ) → megatron.bridge.training.config.ConfigContainer#

Create a pre-training configuration for Gemma3 family models.

Parameters:

provider_class (type) – Gemma3 model provider class (e.g., Gemma3ModelProvider1B).
hf_path (str | None) – HuggingFace model path (e.g., “google/gemma-3-1b-pt”).
dir (str | None) – Base directory for saving logs and checkpoints.
name (str) – Name of the pre-training run.
data_paths (list[str] | None) – List of paths to dataset files. If None, mock data will be used.
data_args_path (str | None) – Path to file containing data arguments.
train_data_path (list[str] | None) – List of training data paths.
valid_data_path (list[str] | None) – List of validation data paths.
test_data_path (list[str] | None) – List of test data paths.
per_split_data_args_path (str | None) – Path to JSON file with per-split data configuration.
mock (bool) – Whether to use mock data. If True, ignores data_paths.
tensor_model_parallel_size (int) – Degree of tensor model parallelism.
pipeline_model_parallel_size (int) – Degree of pipeline model parallelism.
pipeline_dtype (torch.dtype | None) – Data type for pipeline parallelism.
virtual_pipeline_model_parallel_size (int | None) – Size of virtual pipeline parallelism.
context_parallel_size (int) – Degree of context parallelism.
sequence_parallel (bool) – Whether to use sequence parallelism.
use_megatron_fsdp (bool) – Whether to use Megatron FSDP.
account_for_embedding_in_pipeline_split (bool) – Whether to account for embedding in pipeline split.
account_for_loss_in_pipeline_split (bool) – Whether to account for loss in pipeline split.
train_iters (int) – Total number of training iterations.
global_batch_size (int) – Global batch size for training.
micro_batch_size (int) – Micro batch size for training.
seq_length (int) – Sequence length for training data.
lr (float) – Learning rate.
min_lr (float) – Minimum learning rate for cosine decay.
lr_warmup_iters (int) – Number of warmup iterations for the learning rate.
lr_decay_iters (int | None) – Number of iterations over which to decay the LR.
eval_interval (int) – Evaluation interval.
save_interval (int) – Checkpoint save interval.
use_null_tokenizer (bool) – Whether to use null tokenizer for synthetic data.
precision_config (MixedPrecisionConfig | str | None) – Precision configuration for the model.
comm_overlap_config (CommOverlapConfig | None) – Communication overlap configuration.

Returns:

Configuration for pre-training.

Return type:

ConfigContainer

bridge.recipes.gemma.gemma3.gemma3_1b_finetune_config(

**user_kwargs: typing_extensions.Unpack[bridge.recipes.gemma.gemma3.Gemma3FinetuneKwargs],

) → megatron.bridge.training.config.ConfigContainer#

Return a finetuning config for Gemma3 1B.

Default configuration: 1 node, 8 GPUs, LoRA

LoRA (default): TP=1, PP=1, LR=1e-4, dim=8, alpha=16
DoRA: TP=1, PP=1, LR=1e-4, dim=8, alpha=16
Full SFT (peft=None): TP=1, PP=1, LR=5e-6

Matches NeMo2 recipe at nemo/collections/llm/recipes/gemma3_1b.py

bridge.recipes.gemma.gemma3._gemma3_finetune_common( hf_path: str, dir: str | None = None, name: str = 'default', pretrained_checkpoint: str | None = None, packed_sequence: bool = False, train_iters: int = 100, global_batch_size: int | None = None, micro_batch_size: int = 1, seq_length: int | None = None, eval_interval: int = 50, save_interval: int = 100, finetune_lr: float | None = None, min_lr: float = 0.0, lr_warmup_iters: int = 10, lr_decay_iters: int | None = None, wandb_project: str | None = None, wandb_entity: str | None = None, wandb_exp_name: str | None = None, precision_config: megatron.bridge.training.mixed_precision.MixedPrecisionConfig | str | None = None, ) → megatron.bridge.training.config.ConfigContainer#

Create a finetuning configuration for Gemma3 models.