Upload folder using huggingface_hub
9855f47 verified 7 months ago
28.3 kB
"""Console script for clip_benchmark."""

import argparse
import ast
import csv
import json
import os
import pathlib
import random
import sys
from copy import copy
from itertools import product
from functools import partial

# Determine the parent directory of the clip_benchmark package
parent_dir = os.path.abspath(os.path.join(os.path.dirname(__file__), ".."))
# Add the parent directory to sys.path
sys.path.insert(0, parent_dir)

import torch
from clip_benchmark.datasets.builder import (build_dataset, dataset_collection,
                                             get_dataset_collate_fn,
                                             get_dataset_collection_from_file,
                                             get_dataset_default_task,
                                             is_video_dataset, is_audio_dataset)
from clip_benchmark.metrics import (linear_probe, multiclass_retrieval,
                                    visualization, zeroshot_classification,
                                    zeroshot_retrieval)
from clip_benchmark.model_collection import (get_model_collection_from_file,
                                             model_collection)


class ParseKwargs(argparse.Action):
    def __call__(self, parser, namespace, values, option_string=None):
        # kw = {}
        kw = getattr(namespace, self.dest, {})
        for value in values:
            key, value = value.split("=")
            try:
                kw[key] = ast.literal_eval(value)
            except ValueError:
                kw[key] = str(
                    value
                )  # fallback to string (avoid need to escape on command line)
        setattr(namespace, self.dest, kw)


from dataclasses import dataclass
import core.vision_encoder.pe as pe
import core.vision_encoder.transforms as transforms
from core.audio_visual_encoder import PEAudioVisual, PEAudioVisualTransform

@dataclass
class Visualization:
    enabled: bool = False
    delete_post_ln: bool = False
    extract_layer: int = None
    attn_pooling: str = None  # "K", "V"
    attn_rollout: bool = False


def get_parser_args():
    parser = argparse.ArgumentParser()
    subparsers = parser.add_subparsers()
    parser_eval = subparsers.add_parser("eval", help="Evaluate")
    parser_eval.add_argument(
        "--dataset",
        type=str,
        default="cifar10",
        nargs="+",
        help="Dataset(s) to use for the benchmark. Can be the name of a dataset, or a collection name ('vtab', 'vtab+', 'imagenet_robustness', 'retrieval') or path of a text file where each line is a dataset name",
    )
    parser_eval.add_argument(
        "--dataset_root",
        default="root",
        type=str,
        help="dataset root folder where the datasets are downloaded. Can be in the form of a template depending on dataset name, e.g., --dataset_root='datasets/{dataset}'. This is useful if you evaluate on multiple datasets.",
    )
    parser_eval.add_argument(
        "--split", type=str, default="test", help="Dataset split to use"
    )
    parser_eval.add_argument(
        "--test_split",
        dest="split",
        action="store",
        type=str,
        default="test",
        help="Dataset split to use",
    )
    parser_eval.add_argument(
        "--train_split",
        type=str,
        nargs="+",
        default="train",
        help="Dataset(s) train split names",
    )
    mutually_exclusive = parser_eval.add_mutually_exclusive_group()
    mutually_exclusive.add_argument(
        "--val_split",
        default=None,
        type=str,
        nargs="+",
        help="Dataset(s) validation split names. Mutually exclusive with val_proportion.",
    )
    mutually_exclusive.add_argument(
        "--val_proportion",
        default=None,
        type=float,
        nargs="+",
        help="what is the share of the train dataset will be used for validation part, if it doesn't predefined. Mutually exclusive with val_split",
    )
    parser_eval.add_argument(
        "--model",
        type=str,
        nargs="+",
        default=["ViT-B-32-quickgelu"],
        help="Model architecture to use from OpenCLIP",
    )
    parser_eval.add_argument(
        "--pretrained",
        type=str,
        nargs="+",
        default=["laion400m_e32"],
        help="Model checkpoint name to use from OpenCLIP",
    )
    parser_eval.add_argument(
        "--pretrained_model",
        type=str,
        default="",
        nargs="+",
        help="Pre-trained model(s) to use. Can be the full model name where `model` and `pretrained` are comma separated (e.g., --pretrained_model='ViT-B-32-quickgelu,laion400m_e32'), a model collection name ('openai' or 'openclip_base' or 'openclip_multilingual' or 'openclip_all'), or path of a text file where each line is a model fullname where model and pretrained are comma separated (e.g., ViT-B-32-quickgelu,laion400m_e32). --model and --pretrained are ignored if --pretrained_model is used.",
    )
    parser_eval.add_argument(
        "--task",
        type=str,
        default="auto",
        choices=[
            "zeroshot_classification",
            "zeroshot_retrieval",
            "multiclass_retreival",
            "linear_probe",
            "auto",
        ],
        help="Task to evaluate on. With --task=auto, the task is automatically inferred from the dataset.",
    )
    parser_eval.add_argument(
        "--no_amp",
        action="store_false",
        dest="amp",
        default=True,
        help="whether to use mixed precision",
    )
    parser_eval.add_argument("--num_workers", default=4, type=int)
    parser_eval.add_argument(
        "--recall_k",
        default=[1, 5, 10],
        type=int,
        help="for retrieval, select the k for Recall@K metric. ",
        nargs="+",
    )
    parser_eval.add_argument(
        "--fewshot_k",
        default=-1,
        type=int,
        help="for linear probe, how many shots. -1 = whole dataset.",
    )
    parser_eval.add_argument(
        "--fewshot_epochs",
        default=10,
        type=int,
        help="for linear probe, how many epochs.",
    )
    parser_eval.add_argument(
        "--fewshot_lr",
        default=0.1,
        type=float,
        help="for linear probe, what is the learning rate.",
    )
    parser_eval.add_argument(
        "--skip_load",
        action="store_true",
        help="for linear probes, when everything is cached, no need to load model.",
    )
    parser_eval.add_argument(
        "--distributed", action="store_true", help="evaluation in parallel"
    )
    parser_eval.add_argument("--seed", default=0, type=int, help="random seed.")
    parser_eval.add_argument("--batch_size", default=64, type=int)
    parser_eval.add_argument(
        "--normalize", default=True, type=bool, help="features normalization"
    )
    parser_eval.add_argument(
        "--model_cache_dir",
        default=None,
        type=str,
        help="directory to where downloaded models are cached",
    )
    parser_eval.add_argument(
        "--feature_root",
        default="features",
        type=str,
        help="feature root folder where the features are stored.",
    )
    parser_eval.add_argument(
        "--annotation_file",
        default="",
        type=str,
        help="text annotation file for retrieval datasets. Only needed  for when `--task` is `zeroshot_retrieval`.",
    )
    parser_eval.add_argument(
        "--custom_classname_file",
        default=None,
        type=str,
        help="use custom json file with classnames for each dataset, where keys are dataset names and values are list of classnames.",
    )
    parser_eval.add_argument(
        "--custom_template_file",
        default=None,
        type=str,
        help="use custom json file with prompts for each dataset, where keys are dataset names and values are list of prompts. For instance, to use CuPL prompts, use --custom_template_file='cupl_prompts.json'",
    )
    parser_eval.add_argument(
        "--dump_classnames",
        default=False,
        action="store_true",
        help="dump classnames to the results json file.",
    )
    parser_eval.add_argument(
        "--dump_templates",
        default=False,
        action="store_true",
        help="dump templates to the results json file.",
    )

    parser_eval.add_argument(
        "--name", default=None, type=str, help="Overwrite the name used."
    )

    parser_eval.add_argument(
        "--image-mean",
        type=float,
        nargs="+",
        default=None,
        metavar="MEAN",
        help="Override default image mean value of dataset",
    )
    parser_eval.add_argument(
        "--image-std",
        type=float,
        nargs="+",
        default=None,
        metavar="STD",
        help="Override default image std deviation of of dataset",
    )
    parser_eval.add_argument(
        "--force-preprocess-cfg", nargs="*", default={}, action=ParseKwargs
    )
    parser_eval.add_argument(
        "--force-vision-cfg",
        nargs="*",
        default={},
        action=ParseKwargs,
        help="Overwrite fields of the vision cfg with the args. Only specified kwdargs will be updated.",
    )
    parser_eval.add_argument(
        "--num-frames",
        default=8,
        type=int,
        help="number of frames to use for video datasets",
    )
    parser_eval.add_argument(
        "--reweight-retrieval",
        default=True,
        action=argparse.BooleanOptionalAction,
        help="use the softmax trick to reweight the retrieval scores",
    )
    parser_eval.add_argument(
        "--reweight-scale",
        default=1.0,
        type=float,
        help="Scale the scores prior to doing the softmax trick to reweight"
    )
    parser_eval.add_argument(
        "--visualize",
        nargs="*",
        default={},
        action=ParseKwargs,
        help="Visualization tool. Spits out visualizations as just bytes in the console. To be read by bento.",
    )

    parser_eval.add_argument(
        "--language",
        default="en",
        type=str,
        nargs="+",
        help="language(s) of classname and prompts to use for zeroshot classification.",
    )
    parser_eval.add_argument(
        "--output",
        default="result.json",
        type=str,
        help="output file where to dump the metrics. Can be in form of a template, e.g., --output='{dataset}_{pretrained}_{model}_{language}_{task}.json'",
    )
    parser_eval.add_argument(
        "--quiet",
        dest="verbose",
        action="store_false",
        help="suppress verbose messages",
    )
    parser_eval.add_argument(
        "--save_clf",
        default=None,
        type=str,
        help="optionally save the classification layer output by the text tower",
    )
    parser_eval.add_argument(
        "--load_clfs",
        nargs="+",
        default=[],
        type=str,
        help="optionally load and average mutliple layers output by text towers.",
    )
    parser_eval.add_argument(
        "--skip_existing",
        default=False,
        action="store_true",
        help="whether to skip an evaluation if the output file exists.",
    )
    parser_eval.add_argument(
        "--model_type", default="open_clip", type=str, help="clip model type"
    )
    parser_eval.add_argument(
        "--wds_cache_dir",
        default=None,
        type=str,
        help="optional cache directory for webdataset only",
    )
    parser_eval.set_defaults(which="eval")

    parser_build = subparsers.add_parser("build", help="Build CSV from evaluations")
    parser_build.add_argument(
        "files", type=str, nargs="+", help="path(s) of JSON result files"
    )
    parser_build.add_argument(
        "--output", type=str, default="benchmark.csv", help="CSV output file"
    )
    parser_build.set_defaults(which="build")

    args = parser.parse_args()
    return parser, args


def main():
    parser, base = get_parser_args()
    if not hasattr(base, "which"):
        parser.print_help()
        return
    if base.which == "eval":
        main_eval(base)
    elif base.which == "build":
        main_build(base)


def main_build(base):
    # Build a benchmark single CSV file from a set of evaluations (JSON files)
    rows = []
    fieldnames = set()

    def process_file(path: str):
        data = json.load(open(path))
        row = {}
        row.update(data["metrics"])
        row.update(data)
        del row["metrics"]
        row["model_fullname"] = row["model"] + " " + row["pretrained"]
        for field in row.keys():
            fieldnames.add(field)
        rows.append(row)

    for path in base.files:
        if os.path.isdir(path):
            files = [
                os.path.join(path, f) for f in os.listdir(path) if f.endswith(".json")
            ]
            for file in files:
                process_file(file)
        else:
            process_file(path)
    with open(base.output, "w") as csvfile:
        writer = csv.DictWriter(csvfile, fieldnames=fieldnames)
        writer.writeheader()
        for row in rows:
            writer.writerow(row)


def main_eval(base):
    # Get list of pre-trained models to evaluate
    pretrained_model = _as_list(base.pretrained_model)
    if pretrained_model:
        models = []
        for name in pretrained_model:
            if os.path.isfile(name):
                # if path, read file, each line is a pre-trained model
                models.extend(get_model_collection_from_file(name))
            elif name in model_collection:
                # if part of `model_collection`, retrieve from it
                models.extend(model_collection[name])
            else:
                # if not, assume it is in the form of `model,pretrained`
                model, pretrained = name.split(",")
                models.append((model, pretrained))
    else:
        models = list(product(base.model, base.pretrained))

    # Get list of datasets to evaluate on
    datasets = []
    for name in _as_list(base.dataset):
        if os.path.isfile(name):
            # If path, read file, each line is a dataset name
            datasets.extend(get_dataset_collection_from_file(name))
        elif name in dataset_collection:
            # if part of `dataset_collection`, retrieve from it
            datasets.extend(dataset_collection[name])
        else:
            # if not, assume it is simply the name of the dataset
            datasets.append(name)
    train_splits = _as_list(base.train_split)
    train_splits = _single_option_to_multiple_datasets(
        train_splits, datasets, "train_split"
    )
    proportions, val_splits = None, None
    if base.val_split is not None:
        val_splits = _as_list(base.val_split)
        val_splits = _single_option_to_multiple_datasets(
            val_splits, datasets, "val_split"
        )
    if base.val_proportion is not None:
        proportions = _as_list(base.val_proportion)
        proportions = _single_option_to_multiple_datasets(
            proportions, datasets, "val_proportion"
        )

    dataset_info = {}
    for i in range(len(datasets)):
        dataset_info[datasets[i]] = {
            "train_split": train_splits[i],
            "val_split": val_splits[i] if val_splits is not None else None,
            "proportion": proportions[i] if proportions is not None else None,
        }
    # Get list of languages to evaluate on
    languages = _as_list(base.language)

    if base.verbose:
        print(f"Models: {models}")
        print(f"Datasets: {datasets}")
        print(f"Languages: {languages}")
    runs = product(models, datasets, languages)
    if base.distributed:
        local_rank, rank, world_size = world_info_from_env()
        runs = list(runs)
        # randomize runs so that runs are balanced across gpus
        random.seed(base.seed)
        random.shuffle(runs)
        runs = [r for i, r in enumerate(runs) if i % world_size == rank]
    for (model, pretrained), (dataset), (language) in runs:
        # We iterative over all possible model/dataset/languages
        args = copy(base)
        args.model = model
        args.pretrained = pretrained
        args.dataset = dataset
        args.language = language
        args.train_split = dataset_info[dataset]["train_split"]
        args.val_split = dataset_info[dataset]["val_split"]
        args.val_proportion = dataset_info[dataset]["proportion"]
        run(args)


def _as_list(l):
    if not l:
        return []
    return [l] if type(l) != list else l


def _single_option_to_multiple_datasets(cur_option, datasets, name):
    cur_len = len(cur_option)
    ds_len = len(datasets)
    if cur_len != ds_len:
        # If user wants to use same value for all datasets
        if cur_len == 1:
            return [cur_option[0]] * ds_len
        else:
            raise ValueError(f"The incommensurable number of {name}")
    else:
        return cur_option


def get_basename_and_parent_folder(path):
    """
    Returns the basename and the folder two parents above.
    Args:
        path (str): The input path.
    Returns:
        str: The basename and the folder two parents above.
    """
    p = pathlib.Path(path)
    parent_folder = (
        p.parents[1].name if len(p.parents) >= 2 else ""
    )  # Get the name of the parent folder two levels up
    basename = p.stem  # Get the basename
    return f"{parent_folder}-{basename}"


def run(args):
    """Console script for clip_benchmark."""
    if torch.cuda.is_available():
        if args.distributed:
            local_rank, rank, world_size = world_info_from_env()
            device = "cuda:%d" % local_rank
            torch.cuda.set_device(device)
        else:
            device = "cuda"
        args.device = device
    else:
        args.device = "cpu"
    # set seed.
    torch.manual_seed(args.seed)
    task = args.task
    if args.dataset.startswith("wds/"):
        dataset_name = args.dataset.replace("wds/", "", 1)
    else:
        dataset_name = args.dataset
    if task == "auto":
        task = get_dataset_default_task(dataset_name)
    local_model_flag = os.path.isfile(args.pretrained) or os.path.isdir(args.pretrained)
    pretrained_slug = (
        get_basename_and_parent_folder(args.pretrained)
        if local_model_flag
        else args.pretrained
    )
    pretrained_slug_full_path = (
        args.pretrained.replace("/", "_")
        if os.path.isfile(args.pretrained)
        else args.pretrained
    )
    dataset_slug = dataset_name.replace("/", "_")
    output = args.output.format(
        model=args.model,
        pretrained=(
            pretrained_slug if args.name is None else f"{pretrained_slug}-{args.name}"
        ),
        pretrained_full_path=pretrained_slug_full_path,
        task=task,
        dataset=dataset_slug,
        language=args.language,
        num_frames=args.num_frames,
    )
    # hack to replace timm/hf model
    output = output.replace("timm/", "timm_")

    if os.path.exists(output) and args.skip_existing:
        if args.verbose:
            print(f"Skip {output}, exists already.")
        return
    if args.verbose:
        print(
            f"Running '{task}' on '{dataset_name}' with the model '{args.pretrained}' on language '{args.language}'"
        )
    dataset_root = args.dataset_root.format(
        dataset=dataset_name, dataset_cleaned=dataset_name.replace("/", "-")
    )
    if args.skip_load:
        model, transform, collate_fn, dataloader = None, None, None, None
    else:
        if args.model.startswith("pe-av"):
            # Load PE-AV model
            model = PEAudioVisual.from_config(args.model, pretrained=True).cuda()
            transform = PEAudioVisualTransform.from_config(args.model)
            tokenizer = partial(transform.tokenizer, padding=True, return_tensors="pt")
        else:
            model_name = args.model
            model = pe.CLIP.from_config(model_name, pretrained=True)  # Downloads from HF
            model = model.cuda()
            transform = transforms.get_image_transform(model.image_size)
            tokenizer = transforms.get_text_tokenizer(model.context_length)

        model.eval()

        dataset = build_dataset(
            dataset_name=args.dataset,
            root=dataset_root,
            transform=transform,
            split=args.split,
            annotation_file=args.annotation_file,
            download=True,
            language=args.language,
            task=task,
            custom_template_file=args.custom_template_file,
            custom_classname_file=args.custom_classname_file,
            wds_cache_dir=args.wds_cache_dir,
            num_frames=args.num_frames,
        )
        if hasattr(dataset, "collate_fn"):
            collate_fn = dataset.collate_fn
        else:
            collate_fn = get_dataset_collate_fn(args.dataset)

        if hasattr(transform, "collate_fn"):
            # Union the dataloader's collate fn (which deals with images) with clipbench's collate fn (which deals with text)
            def collate_union(collate_img, collate_text):
                def collate(batch):
                    img_batch = collate_img([(x[0], torch.zeros(1)) for x in batch])
                    text_batch = collate_text(
                        [[torch.zeros(3, 224, 224)] + list(x[1:]) for x in batch]
                    )
                    return tuple(list(img_batch[:1]) + list(text_batch[1:]))

                return collate

            collate_fn = collate_union(transform.collate_fn, collate_fn)

            def test(*args, **kwargs):
                breakpoint()
                return collate_fn(*args, **kwargs)

        if args.verbose:
            try:
                print(f"Dataset size: {len(dataset)}")
            except TypeError:
                print("IterableDataset has no len()")
            print(f"Dataset split: {args.split}")
            if hasattr(dataset, "classes") and dataset.classes:
                try:
                    print(f"Dataset classes: {dataset.classes}")
                    print(f"Dataset number of classes: {len(dataset.classes)}")
                except AttributeError:
                    print("Dataset has no classes.")

        if args.dataset.startswith("wds/"):
            dataloader = torch.utils.data.DataLoader(
                dataset.batched(args.batch_size, collation_fn=collate_fn),
                batch_size=None,
                shuffle=False,
                num_workers=args.num_workers,
            )
        else:
            dataloader = torch.utils.data.DataLoader(
                dataset,
                batch_size=args.batch_size,
                shuffle=False,
                num_workers=args.num_workers,
                collate_fn=collate_fn,
            )


    if task == "zeroshot_classification":
        zeroshot_templates = (
            dataset.templates if hasattr(dataset, "templates") else None
        )
        if args.verbose:
            print(f"Zero-shot templates: {zeroshot_templates}")
        classnames = dataset.classes if hasattr(dataset, "classes") else None
        assert (
            zeroshot_templates is not None and classnames is not None
        ), "Dataset does not support classification"
        metrics = zeroshot_classification.evaluate(
            model,
            dataloader,
            tokenizer,
            classnames,
            zeroshot_templates,
            video_dataset=is_video_dataset(args.dataset),
            device=args.device,
            amp=args.amp,
            verbose=args.verbose,
            save_clf=args.save_clf,
            load_clfs=args.load_clfs,
            args=args,
        )
    elif task == "zeroshot_retrieval":
        metrics = zeroshot_retrieval.evaluate(
            model,
            dataloader,
            tokenizer,
            video_dataset=is_video_dataset(args.dataset),
            recall_k_list=args.recall_k,
            device=args.device,
            amp=args.amp,
            args=args,
            audio_dataset=is_audio_dataset(args.dataset),
            transform=transform,
        )
    elif task == "multiclass_retrieval":
        metrics = multiclass_retrieval.evaluate(
            model,
            dataloader,
            tokenizer,
            device=args.device,
            amp=args.amp,
            args=args,
            retrieval_template=dataset.retrieval_template,
        )
    elif task == "linear_probe":
        # we also need the train and validation splits for linear probing.
        train_dataset = None
        train_dataset = build_dataset(
            dataset_name=args.dataset,
            root=dataset_root,
            transform=transform,
            split=args.train_split,
            annotation_file=args.annotation_file,
            download=True,
        )
        if args.val_split is not None:
            val_dataset = build_dataset(
                dataset_name=args.dataset,
                root=dataset_root,
                transform=transform,
                split=args.val_split,
                annotation_file=args.annotation_file,
                download=True,
            )
        elif args.val_proportion is not None:
            train_dataset, val_dataset = torch.utils.data.random_split(
                train_dataset, [1 - args.val_proportion, args.val_proportion]
            )
        else:
            val_dataset = None
        train_dataloader = torch.utils.data.DataLoader(
            train_dataset,
            batch_size=args.batch_size,
            shuffle=False,
            num_workers=args.num_workers,
            collate_fn=collate_fn,
            pin_memory=True,
        )
        if val_dataset is not None:
            val_dataloader = torch.utils.data.DataLoader(
                val_dataset,
                batch_size=args.batch_size,
                shuffle=False,
                num_workers=args.num_workers,
                collate_fn=collate_fn,
                pin_memory=True,
            )
        else:
            val_dataloader = None
        metrics = linear_probe.evaluate(
            model,
            train_dataloader,
            dataloader,
            args.fewshot_k,
            args.batch_size,
            args.num_workers,
            args.fewshot_lr,
            args.fewshot_epochs,
            (args.model + "-" + args.pretrained + "-" + args.dataset).replace("/", "_"),
            args.seed,
            args.feature_root,
            val_dataloader=val_dataloader,
            device=args.device,
            normalize=args.normalize,
            amp=args.amp,
            verbose=args.verbose,
        )
    else:
        raise ValueError(
            "Unsupported task: {}. task should be `zeroshot_classification`, `zeroshot_retrieval`, `multiclass_retrieval`, `linear_probe`, or `captioning`".format(
                task
            )
        )
    dump = {
        "dataset": args.dataset,
        "model": args.model,
        "pretrained": args.pretrained,
        "task": task,
        "metrics": metrics,
        "language": args.language,
        "name": args.name if args.name is not None else "",
    }
    if hasattr(dataset, "classes") and dataset.classes and args.dump_classnames:
        dump["classnames"] = dataset.classes
    if hasattr(dataset, "templates") and dataset.templates and args.dump_templates:
        dump["templates"] = dataset.templates
    if args.verbose:
        print(f"Dump results to: {output}")
    os.makedirs(os.path.dirname(output), exist_ok=True)
    with open(output, "w") as f:
        json.dump(dump, f)
    return 0


def world_info_from_env():
    # from openclip
    local_rank = 0
    for v in (
        "LOCAL_RANK",
        "MPI_LOCALRANKID",
        "SLURM_LOCALID",
        "OMPI_COMM_WORLD_LOCAL_RANK",
    ):
        if v in os.environ:
            local_rank = int(os.environ[v])
            break
    global_rank = 0
    for v in ("RANK", "PMI_RANK", "SLURM_PROCID", "OMPI_COMM_WORLD_RANK"):
        if v in os.environ:
            global_rank = int(os.environ[v])
            break
    world_size = 1
    for v in ("WORLD_SIZE", "PMI_SIZE", "SLURM_NTASKS", "OMPI_COMM_WORLD_SIZE"):
        if v in os.environ:
            world_size = int(os.environ[v])
            break
    return local_rank, global_rank, world_size


if __name__ == "__main__":
    sys.exit(main())  # pragma: no cover