o }oi(@sTddlmZddlmZddlZddlZddlm Z ddl m Z ddl m Z mZmZddlmZmZddlmZddlmZd Zejjed d ejejfd d Zejje edddddddddddddddddddddddddde fdeed ed!ed"ed#eejd$eed%ed&e d'ed(ed)ed*ed+ed,e!d-ed.ed/ed0ed ej"f&d1d2Z#dS)3)OptionalN)pretrain)MockDataModule) default_logdefault_resumetensorboard_logger)nemotron_modelnemotron_trainer),distributed_fused_adam_with_cosine_annealing)TimingCallbacknemotron3_22b_16knamereturncCs ttdS)a Factory function to create a Nemotron3 22B model with 16k sequence length. Returns: run.Config[pl.LightningModule]: Configuration for the Nemotron3 22b and 16k sequence length model. Examples: CLI usage: $ nemo llm pretrain model=nemotron3_22b_16k ... Python API usage: >>> model_config = model() >>> print(model_config) )version)rNAMErrb/home/ubuntu/.local/lib/python3.10/site-packages/nemo/collections/llm/recipes/nemotron3_22b_16k.pymodels r)targetrdefaultTiz bf16-mixedg? ii@igh㈵>g-C6?dirrtensor_parallelismpipeline_parallelismpipeline_parallelism_typevirtual_pipeline_parallelismcontext_parallelismsequence_parallelism num_nodesnum_gpus_per_node max_steps precisionaccumulate_grad_batchesgradient_clip_vallimit_test_batcheslimit_val_batcheslog_every_n_stepsval_check_intervalcCstj|ttdid|d|d|d|d|d|d|d| d | d | d | d |d |d|d|dttgtjt|||dt||t|ddt | ||||| dt dS)a Create a pre-training recipe for Nemotron3 22B model with 16k sequence length. This function sets up a complete configuration for pre-training, including model, trainer, data, logging, optimization, and resumption settings. Args: dir (Optional[str]): Directory for saving logs and checkpoints. name (str): Name of the pre-training run. tensor_parallelism (int): Degree of tensor model parallelism. pipeline_parallelism (int): Degree of pipeline model parallelism. pipeline_parallelism_type (Optional[torch.dtype]): Data type for pipeline parallelism. virtual_pipeline_parallelism (Optional[int]): Size of virtual pipeline parallelism. context_parallelism (int): Degree of context parallelism. sequence_parallelism (bool): Whether to use sequence parallelism. num_nodes (int): Number of compute nodes to use. num_gpus_per_node (int): Number of GPUs per node. max_steps (int): Maximum number of training steps. precision (str): Precision configuration, one of fp32, 16-mixed or bf16-mixed. accumulate_grad_batches (int): Number of steps per gradient accumulation. gradient_clip_val (float): Value for gradient clipping. limit_test_batches (int): Limit the number of test batches. limit_val_batches (int): Limit the number of validation batches. log_every_n_steps (int): Log every n steps. val_check_interval (int): Run validation every N steps. global_batch_size (int): Global batch size. micro_batch_size (int): Micro batch size. seq_length (int): Sequence length. warmup_steps (int): Number of warmup steps. constant_steps (int): Number of constant steps. min_lr (float): Minimum learning rate. max_lr (float): Maximum learning rate. fn (Callable): The pre-training function to use. Returns: run.Partial: Partial configuration for pre-training. Examples: CLI usage: $ nemo llm pretrain --factory nemotron3_22b_16k $ nemo llm pretrain --factory "nemotron3_22b_16k(num_nodes=1, name='my_nemotron_pretrain')" Python API usage: >>> recipe = pretrain_recipe(name="nemotron_pretrain", num_nodes=1) >>> print(recipe) Note: This recipe uses a mock dataset, look for the finetune examples to see how to change the dataset. rrr r!r"r#r$r%r&r'r(r*r+r,r- callbacks) seq_lengthglobal_batch_sizemicro_batch_sizer )rrr)r' warmup_stepsconstant_stepsmin_lrmax_lr clip_grad)rtrainerdatalogoptimresumeNr) runPartialrr Configr rrrr r)rrrrr r!r"r#r$r%r&r'r(r)r*r+r,r-r0r1r/r2r3r4r5fnrrrpretrain_recipe3sjS      r@)$typingrlightning.pytorchpytorchplnemo_runr<torchnemo.collections.llm.apir"nemo.collections.llm.gpt.data.mockr(nemo.collections.llm.recipes.log.defaultrrr%nemo.collections.llm.recipes.nemotronrr 'nemo.collections.llm.recipes.optim.adamr nemo.utils.exp_managerr rclifactoryr>LightningModulerstrintdtypeboolfloatr=r@rrrrs