# Copyright (c) 2024, NVIDIA CORPORATION. All rights reserved.
import torch

from megatron.core.fusions.fused_bias_dropout import get_bias_dropout_add
from megatron.core.tensor_parallel.layers import ColumnParallelLinear, RowParallelLinear
from megatron.core.transformer.attention import SelfAttention, SelfAttentionSubmodules
from megatron.core.extensions.transformer_engine import (
    TELayerNormColumnParallelLinear,
    TENorm,
    TERowParallelLinear,
)
from megatron.core.transformer.dot_product_attention import DotProductAttention
from megatron.core.transformer.enums import AttnMaskType
from megatron.core.transformer.identity_op import IdentityOp
from megatron.core.transformer.mlp import MLP, MLPSubmodules
from megatron.core.transformer.spec_utils import ModuleSpec
from megatron.core.transformer.transformer_layer import (
    TransformerLayer,
    TransformerLayerSubmodules,
)


def get_layer_spec(config=None, is_vit=False, *args, **kwargs) -> ModuleSpec:
    attn_mask_type = AttnMaskType.no_mask if is_vit else AttnMaskType.causal

    mlp = get_mlp_module_spec(use_te=False)
    return ModuleSpec(
        module=TransformerLayer,
        submodules=TransformerLayerSubmodules(
            input_layernorm=TENorm,
            self_attention=ModuleSpec(
                module=SelfAttention,
                params={
                    "attn_mask_type": attn_mask_type
                },
                submodules=SelfAttentionSubmodules(
                    linear_qkv=ColumnParallelLinear,
                    core_attention=DotProductAttention,
                    linear_proj=RowParallelLinear,
                    q_layernorm=IdentityOp,
                    k_layernorm=IdentityOp,
                ),
            ),
            self_attn_bda=get_bias_dropout_add,
            pre_mlp_layernorm=TENorm,
            mlp=mlp,
            mlp_bda=get_bias_dropout_add,
        ),
    )


def get_mlp_module_spec(config=None, use_te=False, *args, **kwargs) -> ModuleSpec:
    return ModuleSpec(
        module=MLP,
        submodules=MLPSubmodules(
            linear_fc1=TELayerNormColumnParallelLinear if use_te else ColumnParallelLinear,
            linear_fc2=TERowParallelLinear if use_te else RowParallelLinear,
        ),
    )