# Copyright (c) Facebook, Inc. and its affiliates. All Rights Reserved.
# Copyright (c) 2018-2019 NVIDIA CORPORATION. All rights reserved.
import os

from yacs.config import CfgNode as CN

# -----------------------------------------------------------------------------
# Convention about Training / Test specific parameters
# -----------------------------------------------------------------------------
# Whenever an argument can be either used for training or for testing, the
# corresponding name will be post-fixed by a _TRAIN for a training parameter,
# or _TEST for a test-specific parameter.
# For example, the number of images during training will be
# IMAGES_PER_BATCH_TRAIN, while the number of images for testing will be
# IMAGES_PER_BATCH_TEST

# -----------------------------------------------------------------------------
# Config definition
# -----------------------------------------------------------------------------

_C = CN()
# AMP otimization level
# O0 is fully FP32
# O1 weights are stored in FP32, but some ops are performed in FP16
# O1 is the recommended setting for stability
# O2 Mostly FP16, model cast to FP16 except batch norm
# O2 is faster by introduces some stability issues
# O3 Pure FP16, faster still, but less stable
# O4 FP16 and NHWC uses custom ops to go even faster
# O4 is MLPerf settings, crazy fast, but hard to tune
# and keep stable
_C.OPT_LEVEL = "O1"
# Distribution type 
# torch = Pytorch DDP
# smd = Sagemaker Distributed Data Parallel
# smd accelerates training when running distributed on multiple 
# p3.16, p3dn, and p4d instances. It is only available on these
# instance types.
# auto = Automatically enable smd when available, otherwise use
# pytorch DDP
_C.DISTRIBUTION = "auto"
_C.OUTPUT_DIR = "/opt/ml/checkpoints"

_C.MODEL = CN()
_C.MODEL.RPN_ONLY = False
_C.MODEL.MASK_ON = False
_C.MODEL.KEYPOINT_ON = False
_C.MODEL.DEVICE = "cuda"
_C.MODEL.META_ARCHITECTURE = "GeneralizedRCNN"
_C.MODEL.CLS_AGNOSTIC_BBOX_REG = False

# If the WEIGHT starts with a catalog://, like :R-50, the code will look for
# the path in paths_catalog. Else, it will use it as the specified absolute
# path
_C.MODEL.WEIGHT = '/opt/ml/input/data/weights/pytorch/R-50.pkl'

# -----------------------------------------------------------------------------
# INPUT
# -----------------------------------------------------------------------------
_C.INPUT = CN()
# Dataloader type
_C.INPUT.DATALOADER = 'COCO'
# Train input dir
_C.INPUT.TRAIN_INPUT_DIR = '/opt/ml/input/data/coco/train'
# Validation input dir
_C.INPUT.VAL_INPUT_DIR = '/opt/ml/input/data/coco/val'
# Train annotations
_C.INPUT.TRAIN_ANNO_DIR = '/opt/ml/input/data/coco/annotations/instances_train2017.json'
# Validation annotations
_C.INPUT.VAL_ANNO_DIR = '/opt/ml/input/data/coco/annotations/instances_val2017.json'
# Size of the smallest side of the image during training
_C.INPUT.MIN_SIZE_TRAIN = (800,)  # (800,)
# Maximum size of the side of the image during training
_C.INPUT.MAX_SIZE_TRAIN = 1333
# Size of the smallest side of the image during testing
_C.INPUT.MIN_SIZE_TEST = 800
# Maximum size of the side of the image during testing
_C.INPUT.MAX_SIZE_TEST = 1333
# Values to be used for image normalization
_C.INPUT.PIXEL_MEAN = [102.9801, 115.9465, 122.7717]
# Values to be used for image normalization
_C.INPUT.PIXEL_STD = [1., 1., 1.]
# Convert image to BGR format (for Caffe2 models), in range 0-255
_C.INPUT.TO_BGR255 = True
# Add multiplicative gaussian noise
_C.INPUT.ADD_NOISE = False

# -----------------------------------------------------------------------------
# Dataset
# -----------------------------------------------------------------------------
_C.DATASETS = CN()
# List of the dataset names for training, as present in paths_catalog.py
_C.DATASETS.TRAIN = ()
# List of the dataset names for testing, as present in paths_catalog.py
_C.DATASETS.TEST = ()

# -----------------------------------------------------------------------------
# DataLoader
# -----------------------------------------------------------------------------
_C.DATALOADER = CN()
# Number of data loading threads
_C.DATALOADER.NUM_WORKERS = 4
# If > 0, this enforces that each collated batch should have a size divisible
# by SIZE_DIVISIBILITY
_C.DATALOADER.SIZE_DIVISIBILITY = 0
# If True, each batch should contain only images for which the aspect ratio
# is compatible. This groups portrait images together, and landscape images
# are not batched with portrait images.
_C.DATALOADER.ASPECT_RATIO_GROUPING = True

# ---------------------------------------------------------------------------- #
# Backbone options
# ---------------------------------------------------------------------------- #
_C.MODEL.BACKBONE = CN()

# The backbone conv body to use
# The string must match a function that is imported in modeling.model_builder
# (e.g., 'FPN.add_fpn_ResNet101_conv5_body' to specify a ResNet-101-FPN
# backbone)
_C.MODEL.BACKBONE.CONV_BODY = "R-50-FPN"

# Add StopGrad at a specified stage so the bottom layers are frozen
_C.MODEL.BACKBONE.FREEZE_CONV_BODY_AT = 2
_C.MODEL.BACKBONE.OUT_CHANNELS = 256 * 4
# GN for backbone
_C.MODEL.BACKBONE.USE_GN = False


# ---------------------------------------------------------------------------- #
# FPN options
# ---------------------------------------------------------------------------- #
_C.MODEL.FPN = CN()
_C.MODEL.FPN.USE_GN = False
_C.MODEL.FPN.USE_RELU = False

# ---------------------------------------------------------------------------- #
# Group Norm options
# ---------------------------------------------------------------------------- #
_C.MODEL.GROUP_NORM = CN()
# Number of dimensions per group in GroupNorm (-1 if using NUM_GROUPS)
_C.MODEL.GROUP_NORM.DIM_PER_GP = -1
# Number of groups in GroupNorm (-1 if using DIM_PER_GP)
_C.MODEL.GROUP_NORM.NUM_GROUPS = 32
# GroupNorm's small constant in the denominator
_C.MODEL.GROUP_NORM.EPSILON = 1e-5


# ---------------------------------------------------------------------------- #
# RPN options
# ---------------------------------------------------------------------------- #
_C.MODEL.RPN = CN()
_C.MODEL.RPN.USE_FPN = False
# Base RPN anchor sizes given in absolute pixels w.r.t. the scaled network input
_C.MODEL.RPN.ANCHOR_SIZES = (32, 64, 128, 256, 512)
# Stride of the feature map that RPN is attached.
# For FPN, number of strides should match number of scales
_C.MODEL.RPN.ANCHOR_STRIDE = (16,)
# RPN anchor aspect ratios
_C.MODEL.RPN.ASPECT_RATIOS = (0.5, 1.0, 2.0)
# Remove RPN anchors that go outside the image by RPN_STRADDLE_THRESH pixels
# Set to -1 or a large value, e.g. 100000, to disable pruning anchors
_C.MODEL.RPN.STRADDLE_THRESH = 0
# Minimum overlap required between an anchor and ground-truth box for the
# (anchor, gt box) pair to be a positive example (IoU >= FG_IOU_THRESHOLD
# ==> positive RPN example)
_C.MODEL.RPN.FG_IOU_THRESHOLD = 0.7
# Maximum overlap allowed between an anchor and ground-truth box for the
# (anchor, gt box) pair to be a negative examples (IoU < BG_IOU_THRESHOLD
# ==> negative RPN example)
_C.MODEL.RPN.BG_IOU_THRESHOLD = 0.3
# Total number of RPN examples per image
_C.MODEL.RPN.BATCH_SIZE_PER_IMAGE = 256
# Target fraction of foreground (positive) examples per RPN minibatch
_C.MODEL.RPN.POSITIVE_FRACTION = 0.5
# Number of top scoring RPN proposals to keep before applying NMS
# When FPN is used, this is *per FPN level* (not total)
_C.MODEL.RPN.PRE_NMS_TOP_N_TRAIN = 12000
_C.MODEL.RPN.PRE_NMS_TOP_N_TEST = 6000
# Number of top scoring RPN proposals to keep after applying NMS
_C.MODEL.RPN.POST_NMS_TOP_N_TRAIN = 2000
_C.MODEL.RPN.POST_NMS_TOP_N_TEST = 1000
# NMS threshold used on RPN proposals
_C.MODEL.RPN.NMS_THRESH = 0.7
# Proposal height and width both need to be greater than RPN_MIN_SIZE
# (a the scale used during training or inference)
_C.MODEL.RPN.MIN_SIZE = 0
# Number of top scoring RPN proposals to keep after combining proposals from
# all FPN levels
_C.MODEL.RPN.FPN_POST_NMS_TOP_N_TRAIN = 2000
_C.MODEL.RPN.FPN_POST_NMS_TOP_N_TEST = 2000
_C.MODEL.RPN.FPN_POST_NMS_TOP_N_PER_IMAGE = True
# Custom rpn head, empty to use default conv or separable conv
_C.MODEL.RPN.RPN_HEAD = "SingleConvRPNHead"
_C.MODEL.RPN.LS = 0.0

# ---------------------------------------------------------------------------- #
# ROI HEADS options
# ---------------------------------------------------------------------------- #
_C.MODEL.ROI_HEADS = CN()
_C.MODEL.ROI_HEADS.USE_FPN = False
# Overlap threshold for an RoI to be considered foreground (if >= FG_IOU_THRESHOLD)
_C.MODEL.ROI_HEADS.FG_IOU_THRESHOLD = 0.5
# Overlap threshold for an RoI to be considered background
# (class = 0 if overlap in [0, BG_IOU_THRESHOLD))
_C.MODEL.ROI_HEADS.BG_IOU_THRESHOLD = 0.5
# Default weights on (dx, dy, dw, dh) for normalizing bbox regression targets
# These are empirically chosen to approximately lead to unit variance targets
_C.MODEL.ROI_HEADS.BBOX_REG_WEIGHTS = (10., 10., 5., 5.)
# RoI minibatch size *per image* (number of regions of interest [ROIs])
# Total number of RoIs per training minibatch =
#   TRAIN.BATCH_SIZE_PER_IM * TRAIN.IMS_PER_BATCH
# E.g., a common configuration is: 512 * 2 * 8 = 8192
_C.MODEL.ROI_HEADS.BATCH_SIZE_PER_IMAGE = 512
# Target fraction of RoI minibatch that is labeled foreground (i.e. class > 0)
_C.MODEL.ROI_HEADS.POSITIVE_FRACTION = 0.25
# Only used on test mode

# Minimum score threshold (assuming scores in a [0, 1] range); a value chosen to
# balance obtaining high recall with not having too many low precision
# detections that will slow down inference post processing steps (like NMS)
_C.MODEL.ROI_HEADS.SCORE_THRESH = 0.05
# Overlap threshold used for non-maximum suppression (suppress boxes with
# IoU >= this threshold)
_C.MODEL.ROI_HEADS.NMS = 0.5
# Maximum number of detections to return per image (100 is based on the limit
# established for the COCO dataset)
_C.MODEL.ROI_HEADS.DETECTIONS_PER_IMG = 100

# ---------------------------------------------------------------------------- #
# CASCADE ROI HEADS options
# ---------------------------------------------------------------------------- #
_C.MODEL.ROI_HEADS.CASCADE = CN()
_C.MODEL.ROI_HEADS.CASCADE.FG_IOU_THRESHOLD = (0.5, 0.6, 0.7)
_C.MODEL.ROI_HEADS.CASCADE.BG_IOU_THRESHOLD = (0.5, 0.6, 0.7)
_C.MODEL.ROI_HEADS.CASCADE.BBOX_REG_WEIGHTS = ((10., 10., 5., 5.), (20., 20., 10., 10.), (30., 30., 15., 15.))
_C.MODEL.ROI_HEADS.CASCADE.STAGES = 3
_C.MODEL.ROI_HEADS.CASCADE.STAGE_WEIGHTS = (1., 1., 1.)

# ---------------------------------------------------------------------------- #
# BOX ROI HEADS options
# ---------------------------------------------------------------------------- #
_C.MODEL.ROI_BOX_HEAD = CN()
_C.MODEL.ROI_BOX_HEAD.TYPE = "StandardBoxHead"
# whether to decode prediction bbox
_C.MODEL.ROI_BOX_HEAD.DECODE = False
_C.MODEL.ROI_BOX_HEAD.LOSS = "SmoothL1Loss"
_C.MODEL.ROI_BOX_HEAD.CARL = False
_C.MODEL.ROI_BOX_HEAD.ISR_P = False
_C.MODEL.ROI_BOX_HEAD.ISR_N = False
_C.MODEL.ROI_BOX_HEAD.FEATURE_EXTRACTOR = "ResNet50Conv5ROIFeatureExtractor"
_C.MODEL.ROI_BOX_HEAD.PREDICTOR = "FastRCNNPredictor"
_C.MODEL.ROI_BOX_HEAD.POOLER_RESOLUTION = 14
_C.MODEL.ROI_BOX_HEAD.POOLER_SAMPLING_RATIO = 0
_C.MODEL.ROI_BOX_HEAD.POOLER_SCALES = (1.0 / 16,)
_C.MODEL.ROI_BOX_HEAD.NUM_CLASSES = 81
# Hidden layer dimension when using an MLP for the RoI box head
_C.MODEL.ROI_BOX_HEAD.MLP_HEAD_DIM = 1024
# GN
_C.MODEL.ROI_BOX_HEAD.USE_GN = False
# Dilation
_C.MODEL.ROI_BOX_HEAD.DILATION = 1
_C.MODEL.ROI_BOX_HEAD.CONV_HEAD_DIM = 256
_C.MODEL.ROI_BOX_HEAD.NUM_STACKED_CONVS = 4

# ---------------------------------------------------------------------------- #
# MASK ROI HEADS options
# ---------------------------------------------------------------------------- #
_C.MODEL.ROI_MASK_HEAD = CN()
_C.MODEL.ROI_MASK_HEAD.FEATURE_EXTRACTOR = "ResNet50Conv5ROIFeatureExtractor"
_C.MODEL.ROI_MASK_HEAD.PREDICTOR = "MaskRCNNC4Predictor"
_C.MODEL.ROI_MASK_HEAD.POOLER_RESOLUTION = 14
_C.MODEL.ROI_MASK_HEAD.POOLER_SAMPLING_RATIO = 0
_C.MODEL.ROI_MASK_HEAD.POOLER_SCALES = (1.0 / 16,)
_C.MODEL.ROI_MASK_HEAD.MLP_HEAD_DIM = 1024
_C.MODEL.ROI_MASK_HEAD.CONV_LAYERS = (256, 256, 256, 256)
_C.MODEL.ROI_MASK_HEAD.RESOLUTION = 14
_C.MODEL.ROI_MASK_HEAD.SHARE_BOX_FEATURE_EXTRACTOR = True
# Whether or not resize and translate masks to the input image.
_C.MODEL.ROI_MASK_HEAD.POSTPROCESS_MASKS = False
_C.MODEL.ROI_MASK_HEAD.POSTPROCESS_MASKS_THRESHOLD = 0.5
# Dilation
_C.MODEL.ROI_MASK_HEAD.DILATION = 1
# GN
_C.MODEL.ROI_MASK_HEAD.USE_GN = False
#Label smoothing
_C.MODEL.ROI_MASK_HEAD.LS = 0.0

_C.MODEL.ROI_KEYPOINT_HEAD = CN()
_C.MODEL.ROI_KEYPOINT_HEAD.FEATURE_EXTRACTOR = "KeypointRCNNFeatureExtractor"
_C.MODEL.ROI_KEYPOINT_HEAD.PREDICTOR = "KeypointRCNNPredictor"
_C.MODEL.ROI_KEYPOINT_HEAD.POOLER_RESOLUTION = 14
_C.MODEL.ROI_KEYPOINT_HEAD.POOLER_SAMPLING_RATIO = 0
_C.MODEL.ROI_KEYPOINT_HEAD.POOLER_SCALES = (1.0 / 16,)
_C.MODEL.ROI_KEYPOINT_HEAD.MLP_HEAD_DIM = 1024
_C.MODEL.ROI_KEYPOINT_HEAD.CONV_LAYERS = tuple(512 for _ in range(8))
_C.MODEL.ROI_KEYPOINT_HEAD.RESOLUTION = 14
_C.MODEL.ROI_KEYPOINT_HEAD.NUM_CLASSES = 17
_C.MODEL.ROI_KEYPOINT_HEAD.SHARE_BOX_FEATURE_EXTRACTOR = True

# ---------------------------------------------------------------------------- #
# ResNe[X]t options (ResNets = {ResNet, ResNeXt}
# Note that parts of a resnet may be used for both the backbone and the head
# These options apply to both
# ---------------------------------------------------------------------------- #
_C.MODEL.RESNETS = CN()

# Number of groups to use; 1 ==> ResNet; > 1 ==> ResNeXt
_C.MODEL.RESNETS.NUM_GROUPS = 1

# Baseline width of each group
_C.MODEL.RESNETS.WIDTH_PER_GROUP = 64

# Place the stride 2 conv on the 1x1 filter
# Use True only for the original MSRA ResNet; use False for C2 and Torch models
_C.MODEL.RESNETS.STRIDE_IN_1X1 = True

# Residual transformation function
_C.MODEL.RESNETS.TRANS_FUNC = "BottleneckWithFixedBatchNorm"
# ResNet's stem function (conv1 and pool1)
_C.MODEL.RESNETS.STEM_FUNC = "StemWithFixedBatchNorm"

# Apply dilation in stage "res5"
_C.MODEL.RESNETS.RES5_DILATION = 1

_C.MODEL.RESNETS.RES2_OUT_CHANNELS = 256
_C.MODEL.RESNETS.STEM_OUT_CHANNELS = 64

# ---------------------------------------------------------------------------- #
# Cascade Options
# ---------------------------------------------------------------------------- #
_C.MODEL.ROI_HEADS.CASCADE_STAGES = 3
_C.MODEL.ROI_HEADS.CASCADE_STAGE_WEIGHTS = (1., 1., 1.)

# ---------------------------------------------------------------------------- #
# Solver
# ---------------------------------------------------------------------------- #
_C.SOLVER = CN()
_C.SOLVER.MAX_ITER = 40000

_C.SOLVER.BASE_LR = 0.001
_C.SOLVER.BIAS_LR_FACTOR = 2

_C.SOLVER.MOMENTUM = 0.9

_C.SOLVER.WEIGHT_DECAY = 0.0005
_C.SOLVER.WEIGHT_DECAY_BIAS = 0

_C.SOLVER.GAMMA = 0.1
_C.SOLVER.ALPHA = 0.001
_C.SOLVER.STEPS = (30000,)

_C.SOLVER.WARMUP_FACTOR = 1.0 / 3
_C.SOLVER.WARMUP_ITERS = 500
_C.SOLVER.WARMUP_METHOD = "linear"

_C.SOLVER.CHECKPOINT_PERIOD = 2500

_C.SOLVER.OPTIMIZER = "SGD"
_C.SOLVER.BETA1 = 0.9
_C.SOLVER.BETA2 = 0.5
_C.SOLVER.LR_SCHEDULE = "MULTISTEP"

_C.SOLVER.GRADIENT_CLIPPING = 0.0

# Number of images per batch
# This is global, so if we have 8 GPUs and IMS_PER_BATCH = 16, each GPU will
# see 2 images per batch
_C.SOLVER.IMS_PER_BATCH = 16

# ---------------------------------------------------------------------------- #
# Specific test options
# ---------------------------------------------------------------------------- #
_C.TEST = CN()
_C.TEST.EXPECTED_RESULTS = []
_C.TEST.EXPECTED_RESULTS_SIGMA_TOL = 4
# Number of images per batch
# This is global, so if we have 8 GPUs and IMS_PER_BATCH = 16, each GPU will
# see 2 images per batch
_C.TEST.IMS_PER_BATCH = 8
# Number of detections per image
_C.TEST.DETECTIONS_PER_IMG = 100
# Run eval each epoch
_C.TEST.PER_EPOCH_EVAL = True

# ---------------------------------------------------------------------------- #
# Misc options
# ---------------------------------------------------------------------------- #
_C.SAVE_CHECKPOINTS = True
_C.SAVE_INTERVAL = 1
_C.LOG_INTERVAL = 50
# Disable reduced logging
_C.DISABLE_REDUCED_LOGGING = False
# Runner hooks
_C.HOOKS = []