o wi"(@sddlmZddlmZddlZddlZddlm Z m Z ddl m Z ddl mZddlmZddlmZmZmZddlmZdd lmZmZdd lmZd Zejjed d ejej fddZ!ejje edddddddddddddddddddddddddde fdee"d e"d!e#d"e#d#eej$d$ee#d%e#d&e%d'e#d(e#d)e#d*e"d+e#d,e&d-e#d.e#d/e#d0e#d ej'f&d1d2Z(ejje ed     3 d8dee"d e"d'e#d(e#d4ee"d5e%d ej'fd6d7Z)dS)9)OptionalN)finetunepretrain)MockDataModule) PEFT_STR2CLS)default_finetune_recipe) default_logdefault_resumetensorboard_logger),distributed_fused_adam_with_cosine_annealing) qwen2_model qwen2_trainer)TimingCallback qwen2_1p5bnamereturncCs ttdS)aa Factory function to create a Qwen2 1.5b model configuration. Returns: run.Config[pl.LightningModule]: Configuration for the Qwen2 1.5b model. Examples: CLI usage: $ nemo llm pretrain model=qwen2_1p5b ... Python API usage: >>> model_config = model() >>> print(model_config) )version)r NAMErrd/home/ubuntu/sommelier/.venv/lib/python3.10/site-packages/nemo/collections/llm/recipes/qwen2_1p5b.pymodel!s r)targetrdefaultFiz bf16-mixedg? iigiUMu>ga2U0*3?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)aY Create a pre-training recipe for Qwen2 1.5b model. 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 qwen2_1p5b $ nemo llm pretrain --factory "qwen2_1p5b(num_nodes=1, name='my_qwen2_pretrain')" Python API usage: >>> recipe = pretrain_recipe(name="qwen2_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. r r!r"r#r$r%r&r'r(r)r*r,r-r.r/ callbacks) seq_lengthglobal_batch_sizemicro_batch_sizer)rrr )r) warmup_stepsconstant_stepsmin_lrmax_lr clip_grad)rtrainerdatalogoptimresumeNr) runPartialrr Configrrrr r r )rrr r!r"r#r$r%r&r'r(r)r*r+r,r-r.r/r2r3r1r4r5r6r7fnrrrpretrain_recipe5sjS      rBlora peft_schemepacked_sequencecCstttd|||||}|dus|dkrd|jj_|S|dvr3tt||_ d|jj_|St d|)a Create a fine-tuning recipe for Qwen2 1.5b model. This function sets up a complete configuration for fine-tuning, including model, trainer, data, logging, optimization, and resumption settings. The recipe uses LoRA (Low-Rank Adaptation) for efficient fine-tuning, unless peft_scheme is set to None. Args: dir (Optional[str]): Directory for saving logs and checkpoints. name (str): Name of the fine-tuning run. num_nodes (int): Number of compute nodes to use. num_gpus_per_node (int): Number of GPUs per node. peft_scheme (Optional[str]): Name of the peft scheme to use for fine-tuning. Allowed values: 'lora'/'dora'/'none'/None. packed_sequence (Optional[bool]): Packing multiple training sequences into one long sequence for training efficiency. Default sequence length is 2048. Returns: run.Partial: Partial configuration for fine-tuning. Examples: CLI usage: $ nemo llm finetune --factory qwen2_1p5b Python API usage: >>> recipe = finetune_recipe(name="qwen2_1p5b_finetune", num_nodes=2) >>> print(recipe) Note: This recipe uses the SQuAD dataset for fine-tuning. zQwen/Qwen2-1.5BNnonegh㈵>)rCdorag-C6?zUnrecognized peft scheme: ) rrlowerr<configlrr>r@rpeft ValueError)rrr&r'rDrEreciperrrfinetune_recipes(   rN)NrrrrCF)*typingrlightning.pytorchpytorchplnemo_runr>torchnemo.collections.llm.apirr"nemo.collections.llm.gpt.data.mockrnemo.collections.llm.peftr-nemo.collections.llm.recipes.finetune_defaultr(nemo.collections.llm.recipes.log.defaultrr r 'nemo.collections.llm.recipes.optim.adamr "nemo.collections.llm.recipes.qwen2r r nemo.utils.exp_managerrrclifactoryr@LightningModulerstrintdtypeboolfloatr?rBrNrrrrs              z