`bridge.recipes.qwen.qwen3`#

Module Contents#

Classes#

`Qwen3CommonKwargs`	Typed options accepted by Qwen3 recipe helper functions.
`Qwen3FinetuneKwargs`	Typed options accepted by Qwen3 finetuning recipe helper functions.

Functions#

`qwen3_600m_pretrain_config`	Return a pre-training config for Qwen3 0.6B.
`qwen3_1p7b_pretrain_config`	Return a pre-training config for Qwen3 1.7B.
`qwen3_4b_pretrain_config`	Return a pre-training config for Qwen3 4B.
`qwen3_8b_pretrain_config`	Return a pre-training config for Qwen3 8B.
`qwen3_14b_pretrain_config`	Return a pre-training config for Qwen3 14B.
`qwen3_32b_pretrain_config`	Return a pre-training config for Qwen3 32B.
`_qwen3_common`	Create a pre-training configuration for Qwen3 models using a given HuggingFace path.
`qwen3_600m_finetune_config`	Return a finetuning config for Qwen3 600M.
`qwen3_1p7b_finetune_config`	Return a finetuning config for Qwen3 1.7B.
`qwen3_4b_finetune_config`	Return a finetuning config for Qwen3 4B.
`qwen3_8b_finetune_config`	Return a finetuning config for Qwen3 8B.
`qwen3_14b_finetune_config`	Return a finetuning config for Qwen3 14B.
`qwen3_32b_finetune_config`	Return a finetuning config for Qwen3 32B.
`_qwen3_finetune_common`	Common finetuning configuration for all Qwen3 models.

API#

class bridge.recipes.qwen.qwen3.Qwen3CommonKwargs#

Bases: typing_extensions.TypedDict

Typed options accepted by Qwen3 recipe helper functions.

Initialization

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

hf_path: str#: 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_parallelism: int#: None

pipeline_parallelism: int#: None

pipeline_parallelism_dtype: torch.dtype | None#: None

virtual_pipeline_parallelism: int | None#: None

context_parallelism: int#: None

sequence_parallelism: bool#: None

use_megatron_fsdp: bool#: None

use_null_tokenizer: 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

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.qwen.qwen3.Qwen3FinetuneKwargs#

Bases: bridge.recipes.qwen.qwen3.Qwen3CommonKwargs

Typed options accepted by Qwen3 finetuning recipe helper functions.

Initialization

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

pretrained_checkpoint: str | None#: None

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

packed_sequence: bool#: None

finetune_lr: float#: None

wandb_project: str | None#: None

wandb_entity: str | None#: None

wandb_exp_name: str | None#: None

bridge.recipes.qwen.qwen3.qwen3_600m_pretrain_config(

**user_kwargs: typing_extensions.Unpack[bridge.recipes.qwen.qwen3.Qwen3CommonKwargs],

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

Return a pre-training config for Qwen3 0.6B.