ConvNeXt V2 Image Classifier

convnext-v2-train.py

import os
import random
import math, warnings
from copy import deepcopy

import numpy as np
import lightning
import torch
import torch.nn as nn
import torch.nn.functional as F
from PIL import Image
from lightning.pytorch import seed_everything
from lightning.pytorch.callbacks import ModelCheckpoint

from torch.utils.data import DataLoader, Dataset, random_split
from torchvision import transforms
from torchvision.transforms import functional
from ml_collections import ConfigDict

def drop_path(
    x, drop_prob: float = 0.0, training: bool = False, scale_by_keep: bool = True
):
    """Drop paths (Stochastic Depth) per sample (when applied in main path of residual blocks).

    This is the same as the DropConnect impl I created for EfficientNet, etc networks, however,
    the original name is misleading as 'Drop Connect' is a different form of dropout in a separate paper...
    See discussion: https://github.com/tensorflow/tpu/issues/494#issuecomment-532968956 ... I've opted for
    changing the layer and argument names to 'drop path' rather than mix DropConnect as a layer name and use
    'survival rate' as the argument.

    """
    if drop_prob == 0.0 or not training:
        return x
    keep_prob = 1 - drop_prob
    shape = (x.shape[0],) + (1,) * (
        x.ndim - 1
    )  # work with diff dim tensors, not just 2D ConvNets
    random_tensor = x.new_empty(shape).bernoulli_(keep_prob)
    if keep_prob > 0.0 and scale_by_keep:
        random_tensor.div_(keep_prob)
    return x * random_tensor


def _trunc_normal_(tensor, mean, std, a, b):
    # Cut & paste from PyTorch official master until it's in a few official releases - RW
    # Method based on https://people.sc.fsu.edu/~jburkardt/presentations/truncated_normal.pdf
    def norm_cdf(x):
        # Computes standard normal cumulative distribution function
        return (1.0 + math.erf(x / math.sqrt(2.0))) / 2.0

    if (mean < a - 2 * std) or (mean > b + 2 * std):
        warnings.warn(
            "mean is more than 2 std from [a, b] in nn.init.trunc_normal_. "
            "The distribution of values may be incorrect.",
            stacklevel=2,
        )

    # Values are generated by using a truncated uniform distribution and
    # then using the inverse CDF for the normal distribution.
    # Get upper and lower cdf values
    l = norm_cdf((a - mean) / std)
    u = norm_cdf((b - mean) / std)

    # Uniformly fill tensor with values from [l, u], then translate to
    # [2l-1, 2u-1].
    tensor.uniform_(2 * l - 1, 2 * u - 1)

    # Use inverse cdf transform for normal distribution to get truncated
    # standard normal
    tensor.erfinv_()

    # Transform to proper mean, std
    tensor.mul_(std * math.sqrt(2.0))
    tensor.add_(mean)

    # Clamp to ensure it's in the proper range
    tensor.clamp_(min=a, max=b)
    return tensor


def trunc_normal_(tensor, mean=0.0, std=1.0, a=-2.0, b=2.0):
    with torch.no_grad():
        return _trunc_normal_(tensor, mean, std, a, b)


class DropPath(nn.Module):
    """Drop paths (Stochastic Depth) per sample  (when applied in main path of residual blocks)."""

    def __init__(self, drop_prob: float = 0.0, scale_by_keep: bool = True):
        super(DropPath, self).__init__()
        self.drop_prob = drop_prob
        self.scale_by_keep = scale_by_keep

    def forward(self, x):
        return drop_path(x, self.drop_prob, self.training, self.scale_by_keep)

    def extra_repr(self):
        return f"drop_prob={round(self.drop_prob,3):0.3f}"


class LayerNorm(nn.Module):
    """LayerNorm that supports two data formats: channels_last (default) or channels_first.
    The ordering of the dimensions in the inputs. channels_last corresponds to inputs with
    shape (batch_size, height, width, channels) while channels_first corresponds to inputs
    with shape (batch_size, channels, height, width).
    """

    def __init__(self, normalized_shape, eps=1e-6, data_format="channels_last"):
        super().__init__()
        self.weight = nn.Parameter(torch.ones(normalized_shape))
        self.bias = nn.Parameter(torch.zeros(normalized_shape))
        self.eps = eps
        self.data_format = data_format
        if self.data_format not in ["channels_last", "channels_first"]:
            raise NotImplementedError
        self.normalized_shape = (normalized_shape,)

    def forward(self, x):
        if self.data_format == "channels_last":
            return F.layer_norm(
                x, self.normalized_shape, self.weight, self.bias, self.eps
            )
        elif self.data_format == "channels_first":
            u = x.mean(1, keepdim=True)
            s = (x - u).pow(2).mean(1, keepdim=True)
            x = (x - u) / torch.sqrt(s + self.eps)
            x = self.weight[:, None, None] * x + self.bias[:, None, None]
            return x


class GRN(nn.Module):
    """GRN (Global Response Normalization) layer"""

    def __init__(self, dim):
        super().__init__()
        self.gamma = nn.Parameter(torch.zeros(1, 1, 1, dim))
        self.beta = nn.Parameter(torch.zeros(1, 1, 1, dim))

    def forward(self, x):
        Gx = torch.norm(x, p=2, dim=(1, 2), keepdim=True)
        Nx = Gx / (Gx.mean(dim=-1, keepdim=True) + 1e-6)
        return self.gamma * (x * Nx) + self.beta + x


class Block(nn.Module):
    """ConvNeXtV2 Block.

    Args:
        dim (int): Number of input channels.
        drop_path (float): Stochastic depth rate. Default: 0.0
    """

    def __init__(self, dim, drop_path=0.0):
        super().__init__()
        self.dwconv = nn.Conv2d(
            dim, dim, kernel_size=7, padding=3, groups=dim
        )  # depthwise conv
        self.norm = LayerNorm(dim, eps=1e-6)
        self.pwconv1 = nn.Linear(
            dim, 4 * dim
        )  # pointwise/1x1 convs, implemented with linear layers
        self.act = nn.GELU()
        self.grn = GRN(4 * dim)
        self.pwconv2 = nn.Linear(4 * dim, dim)
        self.drop_path = DropPath(drop_path) if drop_path > 0.0 else nn.Identity()

    def forward(self, x):
        input = x
        x = self.dwconv(x)
        x = x.permute(0, 2, 3, 1)  # (N, C, H, W) -> (N, H, W, C)
        x = self.norm(x)
        x = self.pwconv1(x)
        x = self.act(x)
        x = self.grn(x)
        x = self.pwconv2(x)
        x = x.permute(0, 3, 1, 2)  # (N, H, W, C) -> (N, C, H, W)

        x = input + self.drop_path(x)
        return x


class ConvNeXtV2(nn.Module):
    """ConvNeXt V2

    Args:
        in_chans (int): Number of input image channels. Default: 3
        num_classes (int): Number of classes for classification head. Default: 1000
        depths (tuple(int)): Number of blocks at each stage. Default: [3, 3, 9, 3]
        dims (int[]): Feature dimension at each stage. Default: [96, 192, 384, 768]
        drop_path_rate (float): Stochastic depth rate. Default: 0.
        head_init_scale (float): Init scaling value for classifier weights and biases. Default: 1.
    """

    def __init__(
        self,
        in_chans=3,
        num_classes=1000,
        depths=[3, 3, 9, 3],
        dims=[96, 192, 384, 768],
        drop_path_rate=0.0,
        head_init_scale=1.0,
    ):
        super().__init__()
        self.depths = depths
        self.downsample_layers = (
            nn.ModuleList()
        )  # stem and 3 intermediate downsampling conv layers
        stem = nn.Sequential(
            nn.Conv2d(in_chans, dims[0], kernel_size=4, stride=4),
            LayerNorm(dims[0], eps=1e-6, data_format="channels_first"),
        )
        self.downsample_layers.append(stem)
        for i in range(3):
            downsample_layer = nn.Sequential(
                LayerNorm(dims[i], eps=1e-6, data_format="channels_first"),
                nn.Conv2d(dims[i], dims[i + 1], kernel_size=2, stride=2),
            )
            self.downsample_layers.append(downsample_layer)

        self.stages = (
            nn.ModuleList()
        )  # 4 feature resolution stages, each consisting of multiple residual blocks
        dp_rates = [x.item() for x in torch.linspace(0, drop_path_rate, sum(depths))]
        cur = 0
        for i in range(4):
            stage = nn.Sequential(
                *[
                    Block(dim=dims[i], drop_path=dp_rates[cur + j])
                    for j in range(depths[i])
                ]
            )
            self.stages.append(stage)
            cur += depths[i]

        self.norm = nn.LayerNorm(dims[-1], eps=1e-6)  # final norm layer
        self.head = nn.Linear(dims[-1], num_classes)

        self.apply(self._init_weights)
        self.head.weight.data.mul_(head_init_scale)
        self.head.bias.data.mul_(head_init_scale)

    def _init_weights(self, m):
        if isinstance(m, (nn.Conv2d, nn.Linear)):
            trunc_normal_(m.weight, std=0.02)
            nn.init.constant_(m.bias, 0)

    def forward_features(self, x):
        for i in range(4):
            x = self.downsample_layers[i](x)
            x = self.stages[i](x)
        return self.norm(
            x.mean([-2, -1])
        )  # global average pooling, (N, C, H, W) -> (N, C)

    def forward(self, x):
        x = self.forward_features(x)
        x = self.head(x)
        return x


def convnextv2_atto(**kwargs):
    model = ConvNeXtV2(depths=[2, 2, 6, 2], dims=[40, 80, 160, 320], **kwargs)
    return model

def convnextv2_femto(**kwargs):
    model = ConvNeXtV2(depths=[2, 2, 6, 2], dims=[48, 96, 192, 384], **kwargs)
    return model

def convnext_pico(**kwargs):
    model = ConvNeXtV2(depths=[2, 2, 6, 2], dims=[64, 128, 256, 512], **kwargs)
    return model

def convnextv2_nano(**kwargs):
    model = ConvNeXtV2(depths=[2, 2, 8, 2], dims=[80, 160, 320, 640], **kwargs)
    return model

def convnextv2_tiny(**kwargs):
    model = ConvNeXtV2(depths=[3, 3, 9, 3], dims=[96, 192, 384, 768], **kwargs)
    return model

def convnextv2_base(**kwargs):
    model = ConvNeXtV2(depths=[3, 3, 27, 3], dims=[128, 256, 512, 1024], **kwargs)
    return model

def convnextv2_large(**kwargs):
    model = ConvNeXtV2(depths=[3, 3, 27, 3], dims=[192, 384, 768, 1536], **kwargs)
    return model

def convnextv2_huge(**kwargs):
    model = ConvNeXtV2(depths=[3, 3, 27, 3], dims=[352, 704, 1408, 2816], **kwargs)
    return model

# ========= Dataset =========

def rescale_pad(image, output_size, random_pad=False):
    h, w = image.shape[-2:]
    if h != output_size or w != output_size:
        r = min(output_size / h, output_size / w)
        new_h, new_w = int(h * r), int(w * r)
        if random_pad:
            r2 = random.uniform(0.9, 1)
            new_h, new_w = int(new_h * r2), int(new_w * r2)
        ph = output_size - new_h
        pw = output_size - new_w
        left = random.randint(0, pw) if random_pad else pw // 2
        right = pw - left
        top = random.randint(0, ph) if random_pad else ph // 2
        bottom = ph - top
        image = transforms.functional.resize(image, [new_h, new_w])
        image = transforms.functional.pad(
            image, [left, top, right, bottom], random.uniform(0, 1) if random_pad else 0
        )
    return image


def random_crop(image, min_rate=0.8):
    h, w = image.shape[-2:]
    new_h, new_w = int(h * random.uniform(min_rate, 1)), int(
        w * random.uniform(min_rate, 1)
    )
    top = np.random.randint(0, h - new_h)
    left = np.random.randint(0, w - new_w)
    image = image[:, top : top + new_h, left : left + new_w]
    return image


class ClassifierDataset(Dataset):
    def __init__(self, path, img_size):
        self.path = path
        self.img_size = img_size
        self.image_list = []
        self.label_list = []
        self.label_dict = {}

        label_folders = [d for d in os.listdir(path) if os.path.isdir(os.path.join(path, d))]
        for i, label_folder in enumerate(label_folders):
            self.label_dict[label_folder] = i
            folder_path = os.path.join(path, label_folder)
            for root, _dirs, files in os.walk(folder_path):
                for fname in files:
                    if os.path.splitext(fname)[1].lower() in [".png", ".jpg", ".jpeg", ".webp"]:
                        self.image_list.append(os.path.join(root, fname))
                        self.label_list.append(i)

        print("Label sequence:", self.label_dict)

    def __len__(self):
        length = len(self.image_list) * 2
        return length

    def __getitem__(self, index):
        real_len = len(self.image_list)
        fname = self.image_list[index % real_len]
        label = self.label_list[index % real_len]
        image = Image.open(os.path.join(self.path, fname)).convert("RGB")
        image = transforms.functional.to_tensor(image)
        image = random_crop(image, min_rate=0.8)
        image = rescale_pad(image, self.img_size, True)
        if index // real_len != 0:
            image = transforms.functional.hflip(image)
        
        label = torch.tensor([label], dtype=torch.float32)
        return image, label

#
class Classifier(lightning.LightningModule):
    def __init__(self, model, drop_path_rate=0.0, ema_decay=0):
        super().__init__()
        self.net = model(in_chans=3, num_classes=1, drop_path_rate=drop_path_rate)
        self.ema = deepcopy(self.net)
        self.ema_decay = 0.9999
        self.ema.requires_grad_(False)

    def configure_optimizers(self):
        optimizer = torch.optim.AdamW(self.net.parameters(), weight_decay=0.05)
        return optimizer

    def forward(self, x, use_ema=False):
        x = transforms.functional.normalize(x, 0.5, 0.5)
        return self.ema(x) if use_ema else self.net(x)

    def training_step(self, batch, batch_idx):
        images, labels = batch
        loss = F.mse_loss(self.forward(images), labels)
        self.log_dict({"train/loss": loss})
        return loss

    def validation_step(self, batch, batch_idx):
        images, labels = batch
        pred = self.forward(images)
        mae = F.l1_loss(pred, labels)
        pred_ema = self.forward(images, True)
        mae_ema = F.l1_loss(pred_ema, labels)
        logs = {
            "val/mae": mae,
            "val/mae_ema": mae_ema
        }
        self.log_dict(logs, sync_dist=True, prog_bar=True)

    def on_train_batch_end(self, outputs, batch, batch_idx):
        if self.ema is not None:
            with torch.no_grad():
                for ema_v, model_v in zip(self.ema.state_dict().values(), self.net.state_dict().values()):
                    ema_v.copy_(self.ema_decay * ema_v + (1.0 - self.ema_decay) * model_v)

def main(opt):
    print(opt)
    seed_everything(opt.seed)
    torch.set_float32_matmul_precision('high')
    generator = torch.Generator().manual_seed(opt.seed)
    
    full_dataset = ClassifierDataset(opt.dataset.path, opt.dataset.img_size)
    split_ratio = [opt.dataset.train_split, 1 - opt.dataset.train_split]
    train_dataset, val_dataset = random_split(full_dataset, split_ratio, generator)
    
    dataloader_args = {
        "num_workers": 8,
        "pin_memory": True,
        "persistent_workers": True,
    }
    train_dataloader = DataLoader(train_dataset, batch_size=opt.bsz_train, shuffle=True, **dataloader_args)
    val_dataloader = DataLoader(val_dataset, batch_size=opt.bsz_val, shuffle=False, **dataloader_args)
    
    classifier = Classifier(model=opt.model, drop_path_rate=opt.model_drop_path)
    callbacks = [
         ModelCheckpoint(
            monitor="val/mae",
            mode="min",
            save_top_k=1,
            save_last=True,
            auto_insert_metric_name=False,
            filename="epoch={epoch},mae={val/mae:.4f}",
        ),
        ModelCheckpoint(
            monitor="val/mae_ema",
            mode="min",
            save_top_k=1,
            save_last=False,
            auto_insert_metric_name=False,
            filename="epoch={epoch},mae-ema={val/mae_ema:.4f}",
        )
    ]
    trainer = lightning.Trainer(**opt.trainer, callbacks=callbacks)
    trainer.fit(classifier, train_dataloader, val_dataloader)
    
def get_config():
    config = ConfigDict()

    # model args
    config.model = convnextv2_tiny
    config.model_drop_path = 0.1
    config.resume = ""
    config.seed = 42
    config.bsz_train = 32
    config.bsz_val = 2

    # dataset args
    dataset = config.dataset = ConfigDict()
    dataset.path = "/tmp/classifier-dataset"
    dataset.train_split = 0.9999
    dataset.img_size = 768

    # training args
    trainer = config.trainer = ConfigDict()
    trainer.max_epochs = 100
    trainer.accumulate_grad_batches = 4
    trainer.accelerator = "gpu"
    trainer.precision = 32
    trainer.benchmark = True
    trainer.log_every_n_steps = 1
    trainer.val_check_interval = 0.025
    trainer.strategy = "auto" #"ddp_find_unused_parameters_false"

    return config

if __name__ == "__main__":
    main(get_config())