Feat/add zeropp

configs/7B_llama2_zeropp.py

@@ -0,0 +1,211 @@
+JOB_NAME = "7b_llama2_train"
+model_type = "LLAMA2"
+DO_ALERT = False
+
+VOCAB_SIZE = 32000
+SEQ_LEN = 2048
+HIDDEN_SIZE = 4096
+NUM_ATTENTION_HEAD = 32
+NUM_KV_ATTENTION_HEAD = 8
+MLP_RATIO = 3.5
+NUM_LAYER = 32
+
+
+MODEL_ONLY_FOLDER = "local:llm_ckpts/xxxx"
+# Ckpt folder format:
+# fs: 'local:/mnt/nfs/XXX'
+SAVE_CKPT_FOLDER = "local:llm_ckpts"
+LOAD_CKPT_FOLDER = "local:llm_ckpts/49"
+
+# boto3 Ckpt folder format:
+# import os
+# BOTO3_IP = os.environ["BOTO3_IP"] # boto3 bucket endpoint
+# SAVE_CKPT_FOLDER = f"boto3:s3://model_weights.{BOTO3_IP}/internlm"
+# LOAD_CKPT_FOLDER = f"boto3:s3://model_weights.{BOTO3_IP}/internlm/snapshot/1/"
+CHECKPOINT_EVERY = 50
+ckpt = dict(
+    enable_save_ckpt=False,  # enable ckpt save.
+    save_ckpt_folder=SAVE_CKPT_FOLDER,  # Path to save training ckpt.
+    # 'auto_resume' is designed to automatically load the latest checkpoint from 'save_ckpt_folder' when encountering
+    # training interruptions/hangs caused by hardware failures, using a scheduling system (such as k8s/slurm)
+    # with an automatic restart mechanism upon training reboot.
+    # Please be aware that if `auto_resume` is not set (its default value is True), it will not load the checkpoint
+    # path specified in `load_ckpt_info` by default.
+    # If you want to initialize your model weights from another model, you must set `auto_resume` to False.
+    # If you want to train from scratch, please set `auto_resume` to False and 'load_ckpt_info' to None.
+    auto_resume=False,
+    checkpoint_every=CHECKPOINT_EVERY,
+    async_upload=True,  # async ckpt upload. (only work for boto3 ckpt)
+    async_upload_tmp_folder="/dev/shm/internlm_tmp_ckpt/",  # path for temporarily files during asynchronous upload.
+    oss_snapshot_freq=int(CHECKPOINT_EVERY / 2),  # snapshot ckpt save frequency.
+)
+
+TRAIN_FOLDER = None
+VALID_FOLDER = None  # "/path/to/dataset"
+data = dict(
+    seq_len=SEQ_LEN,
+    # micro_num means the number of micro_batch contained in one gradient update
+    micro_num=4,
+    # unit_schedule_size is used with ZeroPP and means the number of micro_batch contained in one scheduling unit.
+    # If unit_schedule_size is not set, ZeroPP will use Interleaved 1F1B schedule. Otherwise it will use interleaved GPipe.
+    unit_schedule_size=4,
+    # packed_length = micro_bsz * SEQ_LEN
+    micro_bsz=1,
+    # defaults to the value of micro_num
+    valid_micro_num=4,
+    # defaults to 0, means disable evaluate
+    valid_every=0,
+    pack_sample_into_one=False,
+    total_steps=20,
+    skip_batches="",
+    # rampup_batch_size (str): A string with three space-separated integers representing the
+    #       starting batch size, the increment, and the number of steps between
+    #       each increment. For example, "192 24 8" means that the batch size (micro_num)
+    #       starts at 192 and increases by 24 every 8 steps. Defaults to None.
+    #       (IMPORTANT): The interval step size is 'micro_bsz'.
+    rampup_batch_size="",
+    # Datasets with less than 50 rows will be discarded
+    min_length=50,
+    train_folder=TRAIN_FOLDER,
+    valid_folder=VALID_FOLDER,
+    empty_cache_and_diag_interval=200,
+    diag_outlier_ratio=1.1,
+)
+
+grad_scaler = dict(
+    fp16=dict(
+        # the initial loss scale, defaults to 2**16
+        initial_scale=2**16,
+        # the minimum loss scale, defaults to None
+        min_scale=1,
+        # the number of steps to increase loss scale when no overflow occurs
+        growth_interval=1000,
+    ),
+    # the multiplication factor for increasing loss scale, defaults to 2
+    growth_factor=2,
+    # the multiplication factor for decreasing loss scale, defaults to 0.5
+    backoff_factor=0.5,
+    # the maximum loss scale, defaults to None
+    max_scale=2**24,
+    # the number of overflows before decreasing loss scale, defaults to 2
+    hysteresis=2,
+)
+
+hybrid_zero_optimizer = dict(
+    # Enable low_level_optimzer overlap_communication
+    overlap_sync_grad=False,
+    overlap_sync_param=False,
+    # bucket size for nccl communication params
+    reduce_bucket_size=512 * 1024 * 1024,
+    # grad clipping
+    clip_grad_norm=1.0,
+)
+
+loss = dict(
+    label_smoothing=0,
+)
+
+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,
+)
+
+lr_scheduler = dict(
+    total_steps=data["total_steps"],
+    init_steps=0,  # optimizer_warmup_step
+    warmup_ratio=0.01,
+    eta_min=1e-5,
+    last_epoch=-1,
+)
+
+beta2_scheduler = dict(
+    init_beta2=adam["adam_beta2"],
+    c=adam["adam_beta2_c"],
+    cur_iter=-1,
+)
+
+use_fp32_norm = False
+model = dict(
+    checkpoint=False,
+    num_chunks=4,
+    num_attention_heads=NUM_ATTENTION_HEAD,
+    embed_split_hidden=True,
+    vocab_size=VOCAB_SIZE,
+    embed_grad_scale=1,
+    parallel_output=True,
+    hidden_size=HIDDEN_SIZE,
+    num_layers=NUM_LAYER,
+    no_bias=True,
+    mlp_ratio=MLP_RATIO,
+    apply_post_layer_norm=False,
+    dtype="torch.bfloat16",
+    norm_type="rmsnorm",
+    layer_norm_epsilon=1e-5,
+    num_kv_attention_heads=NUM_KV_ATTENTION_HEAD,
+    use_flash_attn=True,
+    # Whether the odd and even columns of the query and key in the model are normally interleaved.
+    # If it's True, the model's odd and even columns are normally ordered; if it's False,
+    # it means that the model has prematurely concatenated all odd columns and even columns in front
+    # and back, in order to improve the RoPE's computational efficiency.
+    # Example:
+    # qk_interleaved = True: q[-1] = [q1,q2,q3,q4,q5,q6,...], k[-1] = [k1,k2,k3,k4,k5,k6,...]
+    # qk_interleaved = False: q[-1] = [q1,q3,q5,...,q2,q4,q6,...], k[-1] = [k1,k3,k5,...,k2,k4,k6,...]
+    qk_interleaved=False,
+)
+
+"""
+zero1 parallel (dict):
+    1. size: int
+        * if size <= 0, the size of the zero process group is equal to the size of the dp process group,
+            so parameters will be divided within the range of dp.
+        * if size == 1, zero is not used, and all dp groups retain the full amount of model parameters.
+        * if size > 1 and size <= dp world size, the world size of zero is a subset of dp world size.
+        For smaller models, it is usually a better choice to split the parameters within nodes with a setting <= 8.
+    2. fsdp: bool, enable/disable torch's fully sharded data parallel, defaults to False.
+tensor parallel (dict):
+    1. size: int, the size of tensor parallel.
+    2. mode: str, the tensor parallel mode, should be in ['mtp', 'msp', 'fsp', 'isp'],
+        defaults to 'mtp', means the pure megatron tensor parallel without sequence parallel.
+        msp: megatron tensor parallel with sequence parallel, sequence parallel size = tensor parallel size.
+        fsp: tensor parallel by flash-attn with sequence parallel, sequence parallel size = tensor parallel size.
+        isp: customed intern sequence parallel without tensor parallel, can be used with weight parallel.
+pipeline parallel (dict):
+    1. size: int, the size of pipeline parallel.
+    2. interleaved_overlap: bool, enable/disable communication overlap when using interleaved pipeline scheduler,
+        defaults to False.
+weight parallel (dict):
+    1. size: int, the size of weight parallel.
+    2. overlap: bool, enable/disable all_gather/reduce_scatter communication overlap, defaults to False.
+    3. memory_pool: bool, enable/disable memory pool, defaults to False.
+"""
+parallel = dict(
+    zero1=dict(size=-1),
+    tensor=dict(size=1, mode="isp"),
+    #pipeline=dict(size=4, interleaved_overlap=True, use_zeropp=True, decouple_grad=True),
+    pipeline=dict(size=4, interleaved_overlap=True, use_zeropp=True, decouple_grad=True),
+    weight=dict(size=8, overlap=True, memory_pool=False),
+)
+
+cudnn_deterministic = False
+cudnn_benchmark = False
+
+monitor = dict(
+    # feishu alert configs
+    alert=dict(
+        enable_feishu_alert=DO_ALERT,
+        feishu_alert_address=None,  # feishu webhook to send alert message
+        light_monitor_address=None,  # light_monitor address to send heartbeat
+        alert_file_path=f"llm_alter/{JOB_NAME}_alert.log",
+    ),
+    tensorboard=dict(
+        queue_max_length=10,
+    ),
+)
+
+# metric_dtype can be "fp32" or other string
+# only when set to "fp32" will use fp32 to calc in metrics
+# metric_dtype = "fp32"

internlm/core/context/parallel_context.py

@@ -19,6 +19,7 @@
 from internlm.utils.common import SingletonMeta
 from internlm.utils.logger import get_logger
 from internlm.utils.timeout import LLM_NCCL_TIMEOUT
+from internlm.utils.zeropp_manager import ZeroppManager
 
 from . import process_group_initializer as pgroup_initializer
 from .process_group_initializer import ParallelMode
@@ -666,7 +667,10 @@ def set_seed(self, seed: int, dpseed_with_tpoffset: bool = False):
     def set_virtual_pipeline_parallel_size(self, size):
         self.virtual_pipeline_parallel_size = size
 
-    def set_virtual_pipeline_parallel_rank(self, rank):
+    def set_virtual_pipeline_parallel_rank(self, rank, clear_zeropp_cache=True):
+        if rank != self.virtual_pipeline_parallel_rank:
+            if clear_zeropp_cache:
+                ZeroppManager.clear_cached_wp_parameters()
         self.virtual_pipeline_parallel_rank = rank
 
 

internlm/core/parallel/shard.py

@@ -95,14 +95,19 @@ def pipeline_parallel_sharding_wrapper(
 
     models = []
 
-    for start, end in parts:
+    use_checkpoint = kwargs["checkpoint"]
+    for part_idx, (start, end) in enumerate(parts):
         kwargs["num_layers"] = end - start
         kwargs["first"] = start == 0
         # If there is no content in the final layer, assign the last layer.
         kwargs["last"] = end == num_layers and len(all_parts[-1]) != 0
         kwargs["device"] = device
         kwargs["start_layer_idx"] = start
-
+        if hasattr(gpc.config.parallel.pipeline, "use_zeropp") and gpc.config.parallel.pipeline.use_zeropp:
+            if part_idx == num_chunks - 1 and gpc.is_pipeline_last_stage(ignore_virtual=True):
+                kwargs["checkpoint"] = False
+            else:
+                kwargs["checkpoint"] = use_checkpoint
         chunk = model_builder(**kwargs).to(device)
         setattr(chunk, "first_layer", start)
         setattr(chunk, "last_layer", end)

internlm/core/scheduler/__init__.py

@@ -1,10 +1,15 @@
 from .base_scheduler import BaseScheduler
 from .no_pipeline_scheduler import NonPipelineScheduler
-from .pipeline_scheduler import InterleavedPipelineScheduler, PipelineScheduler
+from .pipeline_scheduler import (
+    InterleavedPipelineScheduler,
+    PipelineScheduler,
+    ZeroPPScheduler,
+)
 
 __all__ = [
     "BaseScheduler",
     "NonPipelineScheduler",
     "InterleavedPipelineScheduler",
     "PipelineScheduler",
+    "ZeroPPScheduler",
 ]