了解一下InternLM2

• 大模型的出现和发展得益于增长的数据量、计算能力的提升以及算法优化等因素。这些模型在各种任务中展现出惊人的性能，比如自然语言处理、计算机视觉、语音识别等。这种模型通常采用深度神经网络结构，如 Transformer、BERT、GPT（ Generative Pre-trained Transformer ）等。大模型的优势在于其能够捕捉和理解数据中更为复杂、抽象的特征和关系。通过大规模参数的学习，它们可以提高在各种任务上的泛化能力，并在未经过大量特定领域数据训练的情况下实现较好的表现。然而，大模型也面临着一些挑战，比如巨大的计算资源需求、高昂的训练成本、对大规模数据的依赖以及模型的可解释性等问题。因此，大模型的应用和发展也需要在性能、成本和道德等多个方面进行权衡和考量。InternLM-7B 包含了一个拥有 70 亿参数的基础模型和一个为实际场景量身定制的对话模型。该模型具有以下特点:1，利用数万亿的高质量 token 进行训练，建立了一个强大的知识库；2.支持 8k token 的上下文窗口长度，使得输入序列更长并增强了推理能力。基于 InternLM 训练框架，上海人工智能实验室已经发布了两个开源的预训练模型：InternLM-7B 和 InternLM-20B。

• InternLM 是一个开源的轻量级训练框架，旨在支持大模型训练而无需大量的依赖。通过单一的代码库，它支持在拥有数千个 GPU 的大型集群上进行预训练，并在单个 GPU 上进行微调，同时实现了卓越的性能优化。在 1024 个 GPU 上训练时，InternLM 可以实现近 90% 的加速效率。基于 InternLM 训练框架，上海人工智能实验室已经发布了两个开源的预训练模型：InternLM-7B 和 InternLM-20B。Lagent 是一个轻量级、开源的基于大语言模型的智能体（agent）框架，支持用户快速地将一个大语言模型转变为多种类型的智能体，并提供了一些典型工具为大语言模型赋能。通过 Lagent 框架可以更好的发挥 InternLM 的全部性能。

• 7B demo 的训练配置文件样例如下:

  • JOB_NAME = "7b_train"
    SEQ_LEN = 2048
    HIDDEN_SIZE = 4096
    NUM_ATTENTION_HEAD = 32
    MLP_RATIO = 8 / 3
    NUM_LAYER = 32
    VOCAB_SIZE = 103168
    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.
        # load_ckpt_folder=LOAD_CKPT_FOLDER, # Ckpt path to resume training(load weights and scheduler/context states).
        # load_model_only_folder=MODEL_ONLY_FOLDER, # Path to initialize with given model weights.
        load_optimizer=True,  # Wheter to load optimizer states when continuing training.
        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.
        snapshot_ckpt_folder="/".join([SAVE_CKPT_FOLDER, "snapshot"]),  # directory for snapshot ckpt storage path.
        oss_snapshot_freq=int(CHECKPOINT_EVERY / 2),  # snapshot ckpt save frequency.
    )
    TRAIN_FOLDER = "/path/to/dataset"
    VALID_FOLDER = "/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,
        # packed_length = micro_bsz * SEQ_LEN
        micro_bsz=2,
        # defaults to the value of micro_num
        valid_micro_num=4,
        # defaults to 0, means disable evaluate
        valid_every=50,
        pack_sample_into_one=False,
        total_steps=50000,
        skip_batches="",
        rampup_batch_size="",
        # Datasets with less than 50 rows will be discarded
        min_length=50,
        # train_folder=TRAIN_FOLDER,
        # valid_folder=VALID_FOLDER,
    )
    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=True,
        overlap_sync_param=True,
        # 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,
    )
    
    model = dict(
        checkpoint=False,  # The proportion of layers for activation aheckpointing, the optional value are True/False/[0-1]
        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,
        mlp_ratio=MLP_RATIO,
        apply_post_layer_norm=False,
        dtype="torch.float16",  # Support: "torch.float16", "torch.half", "torch.bfloat16", "torch.float32", "torch.tf32"
        norm_type="rmsnorm",
        layer_norm_epsilon=1e-5,
        use_flash_attn=True,
        num_chunks=1,  # if num_chunks > 1, interleaved pipeline scheduler is used.
    )
    """
    zero1 parallel:
        1. if zero1 <= 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.
        2. if zero1 == 1, zero is not used, and all dp groups retain the full amount of model parameters.
        3. zero1 > 1 and zero1 <= 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.
    pipeline parallel (dict):
        1. size: int, the size of pipeline parallel.
        2. interleaved_overlap: bool, enable/disable communication overlap when using interleaved pipeline scheduler.
    tensor parallel: tensor parallel size, usually the number of GPUs per node.
    """
    parallel = dict(
        zero1=8,
        pipeline=dict(size=1, interleaved_overlap=True),
        sequence_parallel=False,
    )
    
    cudnn_deterministic = False
    cudnn_benchmark = False
    

• 30B demo 训练配置文件样例如下:

  • JOB_NAME = "30b_train"
    SEQ_LEN = 2048
    HIDDEN_SIZE = 6144
    NUM_ATTENTION_HEAD = 48
    MLP_RATIO = 8 / 3
    NUM_LAYER = 60
    VOCAB_SIZE = 103168
    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.
        # load_ckpt_folder=LOAD_CKPT_FOLDER, # Ckpt path to resume training(load weights and scheduler/context states).
        # load_model_only_folder=MODEL_ONLY_FOLDER, # Path to initialize with given model weights.
        load_optimizer=True,  # Wheter to load optimizer states when continuing training.
        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.
        snapshot_ckpt_folder="/".join([SAVE_CKPT_FOLDER, "snapshot"]),  # directory for snapshot ckpt storage path.
        oss_snapshot_freq=int(CHECKPOINT_EVERY / 2),  # snapshot ckpt save frequency.
    )
    TRAIN_FOLDER = "/path/to/dataset"
    VALID_FOLDER = "/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,
        # packed_length = micro_bsz * SEQ_LEN
        micro_bsz=2,
        # defaults to the value of micro_num
        valid_micro_num=4,
        # defaults to 0, means disable evaluate
        valid_every=50,
        pack_sample_into_one=False,
        total_steps=50000,
        skip_batches="",
        rampup_batch_size="",
        # Datasets with less than 50 rows will be discarded
        min_length=50,
        # train_folder=TRAIN_FOLDER,
        # valid_folder=VALID_FOLDER,
    )
    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=True,
        overlap_sync_param=True,
        # 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,
    )
    
    model = dict(
        checkpoint=False,  # The proportion of layers for activation aheckpointing, the optional value are True/False/[0-1]
        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,
        mlp_ratio=MLP_RATIO,
        apply_post_layer_norm=False,
        dtype="torch.float16",  # Support: "torch.float16", "torch.half", "torch.bfloat16", "torch.float32", "torch.tf32"
        norm_type="rmsnorm",
        layer_norm_epsilon=1e-5,
        use_flash_attn=True,
        num_chunks=1,  # if num_chunks > 1, interleaved pipeline scheduler is used.
    )
    """
    zero1 parallel:
        1. if zero1 <= 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.
        2. if zero1 == 1, zero is not used, and all dp groups retain the full amount of model parameters.
        3. zero1 > 1 and zero1 <= 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.
    pipeline parallel (dict):
        1. size: int, the size of pipeline parallel.
        2. interleaved_overlap: bool, enable/disable communication overlap when using interleaved pipeline scheduler.
    tensor parallel: tensor parallel size, usually the number of GPUs per node.
    """
    parallel = dict(
        zero1=-1,
        tensor=4,
        pipeline=dict(size=1, interleaved_overlap=True),
        sequence_parallel=False,
    )
    cudnn_deterministic = False
    cudnn_benchmark = False
    

30B Demo — InternLM 0.2.0 文档