import copy import multiprocessing as mp import random import numpy as np import pytest import torch from torch import nn from torch.nn.parallel import DistributedDataParallel as DDP from torch.testing import assert_close import internlm from internlm.core.context.parallel_context import Config from internlm.core.trainer import Trainer from internlm.core.scheduler import ( InterleavedPipelineScheduler, NonPipelineScheduler, PipelineScheduler, SchedulerHook, ) from internlm.data.utils import unpack_data from internlm.core.scheduler.pipeline_scheduler import get_tensor_shape from internlm.core.context import global_context as gpc from internlm.core.context import ParallelMode from internlm.core.scheduler import SchedulerMetricHook from internlm.model.metrics import AccPerplex from internlm.train import ( get_train_data_loader, get_validation_data_loader, initialize_llm_profile, initialize_model, initialize_optimizer, load_new_batch, record_current_batch_training_metrics, ) from internlm.core.engine import Engine from internlm.model.loss import FlashGPTLMLoss from internlm.core.gradient_handler import PipelineSharedModuleGradientHandler from internlm.core.trainer import TrainState class MlpModel(nn.Module): def __init__(self): super(MlpModel, self).__init__() self.linear1 = nn.Linear(4, 8) self.linear2 = nn.Linear(8, 8) self.linear3 = nn.Linear(8, 8) self.linear4 = nn.Linear(8, 8) self.linear5 = nn.Linear(8, 8) self.linear6 = nn.Linear(8, 8) self.linear7 = nn.Linear(8, 8) self.linear8 = nn.Linear(8, 8) self.linear9 = nn.Linear(8, 8) self.linear10 = nn.Linear(8, 8) self.linear11 = nn.Linear(8, 8) self.linear12 = nn.Linear(8, 8) self.linear13 = nn.Linear(8, 8) self.linear14 = nn.Linear(8, 8) self.linear15 = nn.Linear(8, 8) self.linear16 = nn.Linear(8, 4) def forward(self, hidden_states=None, cu_seqlens=None, input_ids=None, indexes=None, inference_params=None): print('MLP:', input_ids, input_ids.dtype, flush=True) input_ids = self.linear1(input_ids) input_ids = self.linear2(input_ids) input_ids = self.linear3(input_ids) input_ids = self.linear4(input_ids) input_ids = self.linear5(input_ids) input_ids = self.linear6(input_ids) input_ids = self.linear7(input_ids) input_ids = self.linear8(input_ids) input_ids = self.linear9(input_ids) input_ids = self.linear10(input_ids) input_ids = self.linear11(input_ids) input_ids = self.linear12(input_ids) input_ids = self.linear13(input_ids) input_ids = self.linear14(input_ids) input_ids = self.linear15(input_ids) input_ids = self.linear16(input_ids) return input_ids config = Config( dict( HIDDEN_SIZE=4, parallel=dict(zero1=1, pipeline=dict(size=8, interleaved_overlap=False), sequence_parallel=False, tensor=1), model_type="INTERNLM", data=dict(seq_len=4, micro_num=4, micro_bsz=1, pack_sample_into_one=False, min_length=0, total_steps=9999), model=dict( dtype=torch.bfloat16, ), resume_tb_folder="", tensorboard_folder="", alert_address=None, monitor=dict(alert=dict(enable_feishu_alert=False, feishu_alert_address=None, light_monitor_address=None)), grad_scaler=dict( fp16=dict( initial_scale=1, min_scale=1, growth_interval=1, ), growth_factor=1.1, backoff_factor=0.9, max_scale=1, hysteresis=1, ), adam=dict( lr=1e-4, adam_beta1=0.9, adam_beta2=0.95, adam_beta2_c=0, adam_eps=1e-8, weight_decay=0.01, ), hybrid_zero_optimizer=dict( overlap_sync_grad=False, overlap_sync_param=False, reduce_bucket_size=512 * 1024 * 1024, clip_grad_norm=1.0, ), beta2_scheduler = dict( init_beta2=0.95, c=0, cur_iter=-1, ), lr_scheduler = dict( total_steps=100, init_steps=0, # optimizer_warmup_step warmup_ratio=0.01, eta_min=1e-5, last_epoch=-1, ) ) ) def build_environment(rank, world_size): import os os.environ["RANK"] = str(rank) os.environ["LOCAL_RANK"] = str(rank) os.environ["WORLD_SIZE"] = str(world_size) os.environ["MASTER_ADDR"] = "127.0.0.1" os.environ["MASTER_PORT"] = "44444" torch.cuda.empty_cache() # launcher="torch" internlm.launch_from_torch(config=config, seed=1024) def loose_close(a, b, dtype: torch.dtype = torch.float32): if dtype is torch.float32: rtol = 1.3e-6 atol = 1e-5 elif dtype is torch.bfloat16: rtol = 2e-2 atol = 2e-2 if isinstance(a, torch.Tensor): a = a.detach().to(dtype) b = b.detach().to(dtype) assert_close(a, b, rtol=rtol, atol=atol) def seed_all(seed, cuda_deterministic=False): random.seed(seed) np.random.seed(seed) torch.manual_seed(seed) if torch.cuda.is_available(): torch.cuda.manual_seed(seed) torch.cuda.manual_seed_all(seed) if cuda_deterministic: # slower, more reproducible torch.backends.cudnn.deterministic = True torch.backends.cudnn.benchmark = False else: torch.backends.cudnn.deterministic = False torch.backends.cudnn.benchmark = True def exam_pipeline_parallel(args): import os rank, world_size = args dtype = torch.float32 build_environment(rank, world_size) local_rank = int(os.environ["LOCAL_RANK"]) print('rank_com:', rank, local_rank) device = torch.device(f"cuda:{local_rank}") # print('device_id:', device) # torch.cuda.set_device(device) seed_all(1024) torch_model = MlpModel().to(device) pp_model = copy.deepcopy(torch_model).to(dtype) tensor_shape = get_tensor_shape() tensor_shape = ( 4, 4, ) # print('tensor_shape:', tensor_shape) scatter_gather = gpc.is_initialized(ParallelMode.TENSOR) if gpc.is_first_rank(ParallelMode.PIPELINE): print(rank, 'is first pp') scheduler_hooks = [ SchedulerMetricHook( skip=False ), ] gpc.config.NUM_MICRO_BATCHES = gpc.config.data.micro_num scheduler = PipelineScheduler( data_process_func=None, num_microbatches=gpc.config.data.micro_num, dtype=gpc.config.model["dtype"], tensor_shape=tensor_shape, scatter_gather_tensors=scatter_gather, scheduler_hooks=scheduler_hooks, ) optimizer, beta2_scheduler, lr_scheduler = initialize_optimizer(model=pp_model) criterion = FlashGPTLMLoss(parallel_output=True, label_smoothing=0) engine = Engine( model=pp_model, optimizer=optimizer, lr_scheduler=lr_scheduler, beta2_scheduler=beta2_scheduler, criterion=criterion, gradient_handlers= [dict(type="PipelineSharedModuleGradientHandler")], clip_grad_norm=gpc.config.hybrid_zero_optimizer.get("clip_grad_norm", 0.0), ) scheduler.pre_processing(engine) engine.train() engine.zero_grad() input_list = [{'input_ids':torch.tensor([[0,1,2,3],[0,1,2,3],[0,1,2,3],[0,1,2,3]]).to(device).to(dtype)}, torch.tensor([[1],[1],[1],[1]]).to(device).to(torch.int64)] torch_input = torch.tensor([[0,1,2,3]]).to(device).to(torch.float32) torch_label = torch.tensor([[1]]).to(device).to(torch.int64) # print('label_shape:', input_list[1].shape) # input_list = [{'input_ids':torch.rand(1, 4).cuda()}, torch.rand(1, 4).cuda()] # input = input_list[0] # print(input) # output = torch_model(input) # print(output) print('local_rank:', gpc.get_local_rank(ParallelMode.PIPELINE), 'start schedule') _, _, loss = scheduler.forward_backward_step(engine, input_list, forward_only=False, return_loss=True, return_output_label=False) engine.step() print('local_rank:', gpc.get_local_rank(ParallelMode.PIPELINE), 'end schedule') torch_output = torch_model(input_ids=torch_input) torch_loss = criterion(torch_output, torch_label).unsqueeze(0) # if rank == 0: # print('loss:', loss) # print('torch_loss:', torch_loss) #loose_close(loss, torch_loss, dtype=dtype) torch_loss.backward() print('local_rank:', gpc.get_local_rank(ParallelMode.PIPELINE), 'everything3') def test_pipeline_parallel(): ctx = mp.get_context("spawn") with ctx.Pool(processes=8) as pool: pool.map( exam_pipeline_parallel, [[rank, 8] for rank in range(8)], ) pool.close() pool.join() if __name__ == "__main__": pytest.main(["-s", "-q", "test_pipeline.py"])