# Ultralytics 🚀 AGPL-3.0 License - https://ultralytics.com/license
"""PyTorch utils."""

import math
import os
import platform
import subprocess
import time
import warnings
from contextlib import contextmanager
from copy import deepcopy
from pathlib import Path

import torch
import torch.distributed as dist
import torch.nn as nn
import torch.nn.functional as F
from torch.nn.parallel import DistributedDataParallel as DDP

from utils.general import LOGGER, check_version, colorstr, file_date, git_describe

LOCAL_RANK = int(os.getenv("LOCAL_RANK", -1))  # https://pytorch.org/docs/stable/elastic/run.html
RANK = int(os.getenv("RANK", -1))
WORLD_SIZE = int(os.getenv("WORLD_SIZE", 1))

try:
    import thop  # for FLOPs computation
except ImportError:
    thop = None

# Suppress PyTorch warnings
warnings.filterwarnings("ignore", message="User provided device_type of 'cuda', but CUDA is not available. Disabling")
warnings.filterwarnings("ignore", category=UserWarning)


def smart_inference_mode(torch_1_9=check_version(torch.__version__, "1.9.0")):
    """Applies torch.inference_mode() if torch>=1.9.0 or torch.no_grad() otherwise as a decorator to functions."""

    def decorate(fn):
        """Applies torch.inference_mode() if torch>=1.9.0, otherwise torch.no_grad(), as a decorator to functions."""
        return (torch.inference_mode if torch_1_9 else torch.no_grad)()(fn)

    return decorate


def smartCrossEntropyLoss(label_smoothing=0.0):
    """Returns CrossEntropyLoss with optional label smoothing for torch>=1.10.0; warns if label smoothing used with
    older versions.
    """
    if check_version(torch.__version__, "1.10.0"):
        return nn.CrossEntropyLoss(label_smoothing=label_smoothing)
    if label_smoothing > 0:
        LOGGER.warning(f"WARNING ⚠️ label smoothing {label_smoothing} requires torch>=1.10.0")
    return nn.CrossEntropyLoss()


def smart_DDP(model):
    """
    Initializes DDP for a model with version checks; fails for torch==1.12.0 due to known issues.

    See https://github.com/ultralytics/yolov5/issues/8395.
    """
    assert not check_version(torch.__version__, "1.12.0", pinned=True), (
        "torch==1.12.0 torchvision==0.13.0 DDP training is not supported due to a known issue. "
        "Please upgrade or downgrade torch to use DDP. See https://github.com/ultralytics/yolov5/issues/8395"
    )
    if check_version(torch.__version__, "1.11.0"):
        return DDP(model, device_ids=[LOCAL_RANK], output_device=LOCAL_RANK, static_graph=True)
    else:
        return DDP(model, device_ids=[LOCAL_RANK], output_device=LOCAL_RANK)


def reshape_classifier_output(model, n=1000):
    """Reshapes the last layer of a model to have 'n' outputs; supports YOLOv3, ResNet, EfficientNet, adjusting Linear
    and Conv2d layers.
    """
    from models.common import Classify

    name, m = list((model.model if hasattr(model, "model") else model).named_children())[-1]  # last module
    if isinstance(m, Classify):  # YOLOv3 Classify() head
        if m.linear.out_features != n:
            m.linear = nn.Linear(m.linear.in_features, n)
    elif isinstance(m, nn.Linear):  # ResNet, EfficientNet
        if m.out_features != n:
            setattr(model, name, nn.Linear(m.in_features, n))
    elif isinstance(m, nn.Sequential):
        types = [type(x) for x in m]
        if nn.Linear in types:
            i = types.index(nn.Linear)  # nn.Linear index
            if m[i].out_features != n:
                m[i] = nn.Linear(m[i].in_features, n)
        elif nn.Conv2d in types:
            i = types.index(nn.Conv2d)  # nn.Conv2d index
            if m[i].out_channels != n:
                m[i] = nn.Conv2d(m[i].in_channels, n, m[i].kernel_size, m[i].stride, bias=m[i].bias is not None)


@contextmanager
def torch_distributed_zero_first(local_rank: int):
    """Context manager ensuring ordered execution in distributed training by synchronizing local masters first."""
    if local_rank not in [-1, 0]:
        dist.barrier(device_ids=[local_rank])
    yield
    if local_rank == 0:
        dist.barrier(device_ids=[0])


def device_count():
    """Returns the count of available CUDA devices; supports Linux and Windows, using nvidia-smi."""
    assert platform.system() in ("Linux", "Windows"), "device_count() only supported on Linux or Windows"
    try:
        cmd = "nvidia-smi -L | wc -l" if platform.system() == "Linux" else 'nvidia-smi -L | find /c /v ""'  # Windows
        return int(subprocess.run(cmd, shell=True, capture_output=True, check=True).stdout.decode().split()[-1])
    except Exception:
        return 0


def select_device(device="", batch_size=0, newline=True):
    """Selects the device for running models, handling CPU, GPU, and MPS with optional batch size divisibility check."""
    s = f"YOLOv3 🚀 {git_describe() or file_date()} Python-{platform.python_version()} torch-{torch.__version__} "
    device = str(device).strip().lower().replace("cuda:", "").replace("none", "")  # to string, 'cuda:0' to '0'
    cpu = device == "cpu"
    mps = device == "mps"  # Apple Metal Performance Shaders (MPS)
    if cpu or mps:
        os.environ["CUDA_VISIBLE_DEVICES"] = "-1"  # force torch.cuda.is_available() = False
    elif device:  # non-cpu device requested
        os.environ["CUDA_VISIBLE_DEVICES"] = device  # set environment variable - must be before assert is_available()
        assert torch.cuda.is_available() and torch.cuda.device_count() >= len(device.replace(",", "")), (
            f"Invalid CUDA '--device {device}' requested, use '--device cpu' or pass valid CUDA device(s)"
        )

    if not cpu and not mps and torch.cuda.is_available():  # prefer GPU if available
        devices = device.split(",") if device else "0"  # range(torch.cuda.device_count())  # i.e. 0,1,6,7
        n = len(devices)  # device count
        if n > 1 and batch_size > 0:  # check batch_size is divisible by device_count
            assert batch_size % n == 0, f"batch-size {batch_size} not multiple of GPU count {n}"
        space = " " * (len(s) + 1)
        for i, d in enumerate(devices):
            p = torch.cuda.get_device_properties(i)
            s += f"{'' if i == 0 else space}CUDA:{d} ({p.name}, {p.total_memory / (1 << 20):.0f}MiB)\n"  # bytes to MB
        arg = "cuda:0"
    elif mps and getattr(torch, "has_mps", False) and torch.backends.mps.is_available():  # prefer MPS if available
        s += "MPS\n"
        arg = "mps"
    else:  # revert to CPU
        s += "CPU\n"
        arg = "cpu"

    if not newline:
        s = s.rstrip()
    LOGGER.info(s)
    return torch.device(arg)


def time_sync():
    """Synchronizes PyTorch across available CUDA devices and returns current time in seconds."""
    if torch.cuda.is_available():
        torch.cuda.synchronize()
    return time.time()


def profile(input, ops, n=10, device=None):
    """YOLOv3 speed/memory/FLOPs profiler
    Usage:
        input = torch.randn(16, 3, 640, 640)
        m1 = lambda x: x * torch.sigmoid(x)
        m2 = nn.SiLU()
        profile(input, [m1, m2], n=100)  # profile over 100 iterations.
    """
    results = []
    if not isinstance(device, torch.device):
        device = select_device(device)
    print(
        f"{'Params':>12s}{'GFLOPs':>12s}{'GPU_mem (GB)':>14s}{'forward (ms)':>14s}{'backward (ms)':>14s}"
        f"{'input':>24s}{'output':>24s}"
    )

    for x in input if isinstance(input, list) else [input]:
        x = x.to(device)
        x.requires_grad = True
        for m in ops if isinstance(ops, list) else [ops]:
            m = m.to(device) if hasattr(m, "to") else m  # device
            m = m.half() if hasattr(m, "half") and isinstance(x, torch.Tensor) and x.dtype is torch.float16 else m
            tf, tb, t = 0, 0, [0, 0, 0]  # dt forward, backward
            try:
                flops = thop.profile(m, inputs=(x,), verbose=False)[0] / 1e9 * 2  # GFLOPs
            except Exception:
                flops = 0

            try:
                for _ in range(n):
                    t[0] = time_sync()
                    y = m(x)
                    t[1] = time_sync()
                    try:
                        _ = (sum(yi.sum() for yi in y) if isinstance(y, list) else y).sum().backward()
                        t[2] = time_sync()
                    except Exception:  # no backward method
                        # print(e)  # for debug
                        t[2] = float("nan")
                    tf += (t[1] - t[0]) * 1000 / n  # ms per op forward
                    tb += (t[2] - t[1]) * 1000 / n  # ms per op backward
                mem = torch.cuda.memory_reserved() / 1e9 if torch.cuda.is_available() else 0  # (GB)
                s_in, s_out = (tuple(x.shape) if isinstance(x, torch.Tensor) else "list" for x in (x, y))  # shapes
                p = sum(x.numel() for x in m.parameters()) if isinstance(m, nn.Module) else 0  # parameters
                print(f"{p:12}{flops:12.4g}{mem:>14.3f}{tf:14.4g}{tb:14.4g}{str(s_in):>24s}{str(s_out):>24s}")
                results.append([p, flops, mem, tf, tb, s_in, s_out])
            except Exception as e:
                print(e)
                results.append(None)
            torch.cuda.empty_cache()
    return results


def is_parallel(model):
    """Checks if a model is using DataParallel (DP) or DistributedDataParallel (DDP)."""
    return type(model) in (nn.parallel.DataParallel, nn.parallel.DistributedDataParallel)


def de_parallel(model):
    """Returns a single-GPU model if input model is using DataParallel (DP) or DistributedDataParallel (DDP)."""
    return model.module if is_parallel(model) else model


def initialize_weights(model):
    """Initializes weights for Conv2D, BatchNorm2d, and activation layers (Hardswish, LeakyReLU, ReLU, ReLU6, SiLU) in a
    model.
    """
    for m in model.modules():
        t = type(m)
        if t is nn.Conv2d:
            pass  # nn.init.kaiming_normal_(m.weight, mode='fan_out', nonlinearity='relu')
        elif t is nn.BatchNorm2d:
            m.eps = 1e-3
            m.momentum = 0.03
        elif t in [nn.Hardswish, nn.LeakyReLU, nn.ReLU, nn.ReLU6, nn.SiLU]:
            m.inplace = True


def find_modules(model, mclass=nn.Conv2d):
    """Finds indices of layers in 'model' matching 'mclass'; default searches for 'nn.Conv2d'."""
    return [i for i, m in enumerate(model.module_list) if isinstance(m, mclass)]


def sparsity(model):
    """Calculates and returns the global sparsity of a model as the ratio of zero-valued parameters to total
    parameters.
    """
    a, b = 0, 0
    for p in model.parameters():
        a += p.numel()
        b += (p == 0).sum()
    return b / a


def prune(model, amount=0.3):
    """Prunes Conv2d layers in a model to a specified global sparsity using l1 unstructured pruning."""
    import torch.nn.utils.prune as prune

    for name, m in model.named_modules():
        if isinstance(m, nn.Conv2d):
            prune.l1_unstructured(m, name="weight", amount=amount)  # prune
            prune.remove(m, "weight")  # make permanent
    LOGGER.info(f"Model pruned to {sparsity(model):.3g} global sparsity")


def fuse_conv_and_bn(conv, bn):
    """Fuses Conv2d and BatchNorm2d layers for efficiency; see https://tehnokv.com/posts/fusing-batchnorm-and-conv/."""
    fusedconv = (
        nn.Conv2d(
            conv.in_channels,
            conv.out_channels,
            kernel_size=conv.kernel_size,
            stride=conv.stride,
            padding=conv.padding,
            dilation=conv.dilation,
            groups=conv.groups,
            bias=True,
        )
        .requires_grad_(False)
        .to(conv.weight.device)
    )

    # Prepare filters
    w_conv = conv.weight.clone().view(conv.out_channels, -1)
    w_bn = torch.diag(bn.weight.div(torch.sqrt(bn.eps + bn.running_var)))
    fusedconv.weight.copy_(torch.mm(w_bn, w_conv).view(fusedconv.weight.shape))

    # Prepare spatial bias
    b_conv = torch.zeros(conv.weight.size(0), device=conv.weight.device) if conv.bias is None else conv.bias
    b_bn = bn.bias - bn.weight.mul(bn.running_mean).div(torch.sqrt(bn.running_var + bn.eps))
    fusedconv.bias.copy_(torch.mm(w_bn, b_conv.reshape(-1, 1)).reshape(-1) + b_bn)

    return fusedconv


def model_info(model, verbose=False, imgsz=640):
    """
    Prints model layers, parameters, gradients, and GFLOPs if verbose; handles various `imgsz`.

    Usage: model_info(model).
    """
    n_p = sum(x.numel() for x in model.parameters())  # number parameters
    n_g = sum(x.numel() for x in model.parameters() if x.requires_grad)  # number gradients
    if verbose:
        print(f"{'layer':>5} {'name':>40} {'gradient':>9} {'parameters':>12} {'shape':>20} {'mu':>10} {'sigma':>10}")
        for i, (name, p) in enumerate(model.named_parameters()):
            name = name.replace("module_list.", "")
            print(
                "%5g %40s %9s %12g %20s %10.3g %10.3g"
                % (i, name, p.requires_grad, p.numel(), list(p.shape), p.mean(), p.std())
            )

    try:  # FLOPs
        p = next(model.parameters())
        stride = max(int(model.stride.max()), 32) if hasattr(model, "stride") else 32  # max stride
        im = torch.empty((1, p.shape[1], stride, stride), device=p.device)  # input image in BCHW format
        flops = thop.profile(deepcopy(model), inputs=(im,), verbose=False)[0] / 1e9 * 2  # stride GFLOPs
        imgsz = imgsz if isinstance(imgsz, list) else [imgsz, imgsz]  # expand if int/float
        fs = f", {flops * imgsz[0] / stride * imgsz[1] / stride:.1f} GFLOPs"  # 640x640 GFLOPs
    except Exception:
        fs = ""

    name = Path(model.yaml_file).stem.replace("yolov5", "YOLOv3") if hasattr(model, "yaml_file") else "Model"
    LOGGER.info(f"{name} summary: {len(list(model.modules()))} layers, {n_p} parameters, {n_g} gradients{fs}")


def scale_img(img, ratio=1.0, same_shape=False, gs=32):  # img(16,3,256,416)
    """Scales and optionally pads an image tensor to a specified ratio, maintaining its aspect ratio constrained by
    `gs`.
    """
    if ratio == 1.0:
        return img
    h, w = img.shape[2:]
    s = (int(h * ratio), int(w * ratio))  # new size
    img = F.interpolate(img, size=s, mode="bilinear", align_corners=False)  # resize
    if not same_shape:  # pad/crop img
        h, w = (math.ceil(x * ratio / gs) * gs for x in (h, w))
    return F.pad(img, [0, w - s[1], 0, h - s[0]], value=0.447)  # value = imagenet mean


def copy_attr(a, b, include=(), exclude=()):
    """Copies attributes from object b to a, with options to include or exclude specific attributes."""
    for k, v in b.__dict__.items():
        if (len(include) and k not in include) or k.startswith("_") or k in exclude:
            continue
        else:
            setattr(a, k, v)


def smart_optimizer(model, name="Adam", lr=0.001, momentum=0.9, decay=1e-5):
    """Initializes a smart optimizer for YOLOv3 with custom parameter groups for different weight decays and biases."""
    g = [], [], []  # optimizer parameter groups
    bn = tuple(v for k, v in nn.__dict__.items() if "Norm" in k)  # normalization layers, i.e. BatchNorm2d()
    for v in model.modules():
        for p_name, p in v.named_parameters(recurse=0):
            if p_name == "bias":  # bias (no decay)
                g[2].append(p)
            elif p_name == "weight" and isinstance(v, bn):  # weight (no decay)
                g[1].append(p)
            else:
                g[0].append(p)  # weight (with decay)

    if name == "Adam":
        optimizer = torch.optim.Adam(g[2], lr=lr, betas=(momentum, 0.999))  # adjust beta1 to momentum
    elif name == "AdamW":
        optimizer = torch.optim.AdamW(g[2], lr=lr, betas=(momentum, 0.999), weight_decay=0.0)
    elif name == "RMSProp":
        optimizer = torch.optim.RMSprop(g[2], lr=lr, momentum=momentum)
    elif name == "SGD":
        optimizer = torch.optim.SGD(g[2], lr=lr, momentum=momentum, nesterov=True)
    else:
        raise NotImplementedError(f"Optimizer {name} not implemented.")

    optimizer.add_param_group({"params": g[0], "weight_decay": decay})  # add g0 with weight_decay
    optimizer.add_param_group({"params": g[1], "weight_decay": 0.0})  # add g1 (BatchNorm2d weights)
    LOGGER.info(
        f"{colorstr('optimizer:')} {type(optimizer).__name__}(lr={lr}) with parameter groups "
        f"{len(g[1])} weight(decay=0.0), {len(g[0])} weight(decay={decay}), {len(g[2])} bias"
    )
    return optimizer


def smart_hub_load(repo="ultralytics/yolov5", model="yolov5s", **kwargs):
    """
    Loads YOLO model from Ultralytics repo with smart error handling, supports `force_reload` on failure.

    See https://github.com/ultralytics/yolov5
    """
    if check_version(torch.__version__, "1.9.1"):
        kwargs["skip_validation"] = True  # validation causes GitHub API rate limit errors
    if check_version(torch.__version__, "1.12.0"):
        kwargs["trust_repo"] = True  # argument required starting in torch 0.12
    try:
        return torch.hub.load(repo, model, **kwargs)
    except Exception:
        return torch.hub.load(repo, model, force_reload=True, **kwargs)


def smart_resume(ckpt, optimizer, ema=None, weights="yolov5s.pt", epochs=300, resume=True):
    """Resumes or fine-tunes training from a checkpoint with optimizer and EMA support; updates epochs based on
    progress.
    """
    best_fitness = 0.0
    start_epoch = ckpt["epoch"] + 1
    if ckpt["optimizer"] is not None:
        optimizer.load_state_dict(ckpt["optimizer"])  # optimizer
        best_fitness = ckpt["best_fitness"]
    if ema and ckpt.get("ema"):
        ema.ema.load_state_dict(ckpt["ema"].float().state_dict())  # EMA
        ema.updates = ckpt["updates"]
    if resume:
        assert start_epoch > 0, (
            f"{weights} training to {epochs} epochs is finished, nothing to resume.\n"
            f"Start a new training without --resume, i.e. 'python train.py --weights {weights}'"
        )
        LOGGER.info(f"Resuming training from {weights} from epoch {start_epoch} to {epochs} total epochs")
    if epochs < start_epoch:
        LOGGER.info(f"{weights} has been trained for {ckpt['epoch']} epochs. Fine-tuning for {epochs} more epochs.")
        epochs += ckpt["epoch"]  # finetune additional epochs
    return best_fitness, start_epoch, epochs


class EarlyStopping:
    """Monitors training to halt if no improvement in fitness metric is observed for a specified number of epochs."""

    def __init__(self, patience=30):
        """Initializes EarlyStopping to monitor training, halting if no improvement in 'patience' epochs, defaulting to
        30.
        """
        self.best_fitness = 0.0  # i.e. mAP
        self.best_epoch = 0
        self.patience = patience or float("inf")  # epochs to wait after fitness stops improving to stop
        self.possible_stop = False  # possible stop may occur next epoch

    def __call__(self, epoch, fitness):
        """Updates stopping criteria based on fitness; returns True to stop if no improvement in 'patience' epochs."""
        if fitness >= self.best_fitness:  # >= 0 to allow for early zero-fitness stage of training
            self.best_epoch = epoch
            self.best_fitness = fitness
        delta = epoch - self.best_epoch  # epochs without improvement
        self.possible_stop = delta >= (self.patience - 1)  # possible stop may occur next epoch
        stop = delta >= self.patience  # stop training if patience exceeded
        if stop:
            LOGGER.info(
                f"Stopping training early as no improvement observed in last {self.patience} epochs. "
                f"Best results observed at epoch {self.best_epoch}, best model saved as best.pt.\n"
                f"To update EarlyStopping(patience={self.patience}) pass a new patience value, "
                f"i.e. `python train.py --patience 300` or use `--patience 0` to disable EarlyStopping."
            )
        return stop


class ModelEMA:
    """Updated Exponential Moving Average (EMA) from https://github.com/rwightman/pytorch-image-models
    Keeps a moving average of everything in the model state_dict (parameters and buffers)
    For EMA details see https://www.tensorflow.org/api_docs/python/tf/train/ExponentialMovingAverage.
    """

    def __init__(self, model, decay=0.9999, tau=2000, updates=0):
        """Initializes EMA with model, optional decay (default 0.9999), tau (2000), and updates count, setting model to
        eval mode.
        """
        self.ema = deepcopy(de_parallel(model)).eval()  # FP32 EMA
        self.updates = updates  # number of EMA updates
        self.decay = lambda x: decay * (1 - math.exp(-x / tau))  # decay exponential ramp (to help early epochs)
        for p in self.ema.parameters():
            p.requires_grad_(False)

    def update(self, model):
        """Updates EMA parameters based on model weights, decay factor, and increment update count."""
        self.updates += 1
        d = self.decay(self.updates)

        msd = de_parallel(model).state_dict()  # model state_dict
        for k, v in self.ema.state_dict().items():
            if v.dtype.is_floating_point:  # true for FP16 and FP32
                v *= d
                v += (1 - d) * msd[k].detach()
        # assert v.dtype == msd[k].dtype == torch.float32, f'{k}: EMA {v.dtype} and model {msd[k].dtype} must be FP32'

    def update_attr(self, model, include=(), exclude=("process_group", "reducer")):
        """Updates EMA attributes by copying from model, excluding 'process_group' and 'reducer' by default."""
        copy_attr(self.ema, model, include, exclude)