train_titles.py

from __future__ import division
from __future__ import unicode_literals
from __future__ import print_function

import torch
import numpy as np
import random
import os
from metrics import compute_metrics, tensor_text_to_video_metrics, tensor_video_to_text_sim
import time
import datetime
import argparse
from modules.tokenization_clip import SimpleTokenizer as ClipTokenizer
from modules.file_utils import PYTORCH_PRETRAINED_BERT_CACHE
from modules.modeling_tv_titles_video import CLIP4Clip
from modules.optimization import BertAdam
from tqdm import tqdm

from sim_matrix.fusion_scores import fusion_scores
from util import parallel_apply, get_logger, AverageMeter
from dataloaders.data_dataloaders import DATALOADER_DICT
import torch.distributed as dist
global logger

def get_args(description='CLIP4Clip on Retrieval Task'):
    parser = argparse.ArgumentParser(description=description)
    parser.add_argument("--do_pretrain", action='store_true', help="Whether to run training.")
    parser.add_argument("--do_train", action='store_true', help="Whether to run training.")
    parser.add_argument("--do_eval", action='store_true', help="Whether to run eval on the dev set.")
    parser.add_argument("--do_eval_zero_shot", action='store_true', help="Whether to run eval on the dev set.")
    parser.add_argument("--do_test", action='store_true', help="Whether to run test on the dev set.")

    parser.add_argument('--train_csv', type=str, default='data/.train.csv', help='')
    parser.add_argument('--val_csv', type=str, default='data/.val.csv', help='')
    parser.add_argument('--data_path', type=str, default='data/caption.pickle', help='data pickle file path')
    parser.add_argument('--features_path', type=str, default='data/videos_feature.pickle', help='feature path')

    parser.add_argument('--num_thread_reader', type=int, default=1, help='')
    parser.add_argument('--lr', type=float, default=0.0001, help='initial learning rate')
    parser.add_argument('--epochs', type=int, default=20, help='upper epoch limit')
    parser.add_argument('--batch_size', type=int, default=256, help='batch size')
    parser.add_argument('--batch_size_val', type=int, default=3500, help='batch size eval')
    parser.add_argument('--lr_decay', type=float, default=0.9, help='Learning rate exp epoch decay')
    parser.add_argument('--n_display', type=int, default=100, help='Information display frequence')
    parser.add_argument('--video_dim', type=int, default=1024, help='video feature dimension')
    parser.add_argument('--seed', type=int, default=42, help='random seed')
    parser.add_argument('--max_words', type=int, default=20, help='')
    parser.add_argument('--max_frames', type=int, default=100, help='')
    parser.add_argument('--feature_framerate', type=int, default=1, help='')
    parser.add_argument('--margin', type=float, default=0.1, help='margin for loss')
    parser.add_argument('--hard_negative_rate', type=float, default=0.5, help='rate of intra negative sample')
    parser.add_argument('--negative_weighting', type=int, default=1, help='Weight the loss for intra negative')
    parser.add_argument('--n_pair', type=int, default=1, help='Num of pair to output from data loader')

    parser.add_argument("--output_dir", default=None, type=str, required=True,
                        help="The output directory where the model predictions and checkpoints will be written.")
    parser.add_argument("--cross_model", default="cross-base", type=str, required=False, help="Cross module")
    parser.add_argument("--init_model", default=None, type=str, required=False, help="Initial model.")
    parser.add_argument("--resume_model", default=None, type=str, required=False, help="Resume train model.")
    parser.add_argument("--do_lower_case", action='store_true', help="Set this flag if you are using an uncased model.")
    parser.add_argument("--warmup_proportion", default=0.1, type=float,
                        help="Proportion of training to perform linear learning rate warmup for. E.g., 0.1 = 10%% of training.")
    parser.add_argument('--gradient_accumulation_steps', type=int, default=1,
                        help="Number of updates steps to accumulate before performing a backward/update pass.")
    parser.add_argument('--n_gpu', type=int, default=1, help="Changed in the execute process.")

    parser.add_argument("--cache_dir", default="", type=str,
                        help="Where do you want to store the pre-trained models downloaded from s3")

    parser.add_argument('--fp16', action='store_true',
                        help="Whether to use 16-bit (mixed) precision (through NVIDIA apex) instead of 32-bit")
    parser.add_argument('--fp16_opt_level', type=str, default='O1',
                        help="For fp16: Apex AMP optimization level selected in ['O0', 'O1', 'O2', and 'O3']."
                             "See details at https://nvidia.github.io/apex/amp.html")

    parser.add_argument("--task_type", default="retrieval", type=str, help="Point the task `retrieval` to finetune.")
    parser.add_argument("--datatype", default="msrvtt", type=str, help="Point the dataset to finetune.")

    parser.add_argument("--world_size", default=0, type=int, help="distribted training")
    parser.add_argument("--local_rank", default=0, type=int, help="distribted training")
    parser.add_argument("--rank", default=0, type=int, help="distribted training")
    parser.add_argument('--coef_lr', type=float, default=1., help='coefficient for bert branch.')
    parser.add_argument('--use_mil', action='store_true', help="Whether use MIL as Miech et. al. (2020).")
    parser.add_argument('--sampled_use_mil', action='store_true', help="Whether MIL, has a high priority than use_mil.")

    parser.add_argument('--text_num_hidden_layers', type=int, default=12, help="Layer NO. of text.")
    parser.add_argument('--visual_num_hidden_layers', type=int, default=12, help="Layer NO. of visual.")
    parser.add_argument('--cross_num_hidden_layers', type=int, default=4, help="Layer NO. of cross.")

    parser.add_argument('--loose_type', action='store_true', help="Default using tight type for retrieval.")
    parser.add_argument('--expand_msrvtt_sentences', action='store_true', help="")

    parser.add_argument('--train_frame_order', type=int, default=0, choices=[0, 1, 2],
                        help="Frame order, 0: ordinary order; 1: reverse order; 2: random order.")
    parser.add_argument('--eval_frame_order', type=int, default=0, choices=[0, 1, 2],
                        help="Frame order, 0: ordinary order; 1: reverse order; 2: random order.")

    parser.add_argument('--freeze_layer_num', type=int, default=0, help="Layer NO. of CLIP need to freeze.")
    parser.add_argument('--slice_framepos', type=int, default=0, choices=[0, 1, 2],
                        help="0: cut from head frames; 1: cut from tail frames; 2: extract frames uniformly.")
    parser.add_argument('--linear_patch', type=str, default="2d", choices=["2d", "3d"],
                        help="linear projection of flattened patches.")
    parser.add_argument('--sim_header', type=str, default="meanP",
                        choices=["meanP", "seqLSTM", "seqTransf", "tightTransf", "seqTransf_topk"],
                        help="choice a similarity header.")
      
    parser.add_argument("--pretrained_clip_name", default="ViT-B/32", type=str, help="Choose a CLIP version")
    parser.add_argument("--strategy", default=2, type=int, help="Sampling strategies.")
    ### interaction
    parser.add_argument('--interaction', type=str, default='dp', help="interaction type for retrieval.")
    parser.add_argument('--wti_arch', type=int, default=2, help="select a architecture for weight branch")

    parser.add_argument('--text_pool_type', type=str, default='clip_top1')
    parser.add_argument("--k", default=1, type=int, help="topk caption.")
    parser.add_argument("--generate_images", default=None, type=str, help="generate images path")
    parser.add_argument("--freeze_text_encoder", action='store_true', help="whether freeze text encoder")
    args = parser.parse_args()

    if args.sim_header == "tightTransf":
        args.loose_type = False

    # Check paramenters
    if args.gradient_accumulation_steps < 1:
        raise ValueError("Invalid gradient_accumulation_steps parameter: {}, should be >= 1".format(
            args.gradient_accumulation_steps))
    if not args.do_train and not args.do_eval and not args.do_eval_zero_shot and not args.do_test:
        raise ValueError("At least one of `do_train` or `do_eval` or 'do_eval_zero_shot' 'do_test' must be True.")

    args.batch_size = int(args.batch_size / args.gradient_accumulation_steps)

    return args

def set_seed_logger(args):
    global logger
    # predefining random initial seeds
    random.seed(args.seed)
    os.environ['PYTHONHASHSEED'] = str(args.seed)
    np.random.seed(args.seed)
    torch.manual_seed(args.seed)
    torch.cuda.manual_seed(args.seed)
    torch.cuda.manual_seed_all(args.seed)  # if you are using multi-GPU.
    torch.backends.cudnn.benchmark = False
    torch.backends.cudnn.deterministic = True

    world_size = torch.distributed.get_world_size()
    torch.cuda.set_device(args.local_rank)
    args.world_size = world_size
    rank = torch.distributed.get_rank()
    args.rank = rank

    if not os.path.exists(args.output_dir):
        os.makedirs(args.output_dir, exist_ok=True)

    logger = get_logger(os.path.join(args.output_dir, "log.txt"))

    if args.local_rank == 0:
        logger.info("Effective parameters:")
        for key in sorted(args.__dict__):
            logger.info("  <<< {}: {}".format(key, args.__dict__[key]))

    return args

def init_device(args, local_rank):
    global logger

    device = torch.device("cuda" if torch.cuda.is_available() else "cpu", local_rank)

    n_gpu = torch.distributed.get_world_size()
    logger.info("device: {} n_gpu: {}".format(device, n_gpu))
    args.n_gpu = n_gpu

    if args.batch_size % args.n_gpu != 0:
        raise ValueError("Invalid batch_size/batch_size_val and n_gpu parameter: {}%{} and {}%{}, should be == 0".format(
            args.batch_size, args.n_gpu, args.batch_size_val, args.n_gpu))

    return device, n_gpu

def init_model(args, device, n_gpu, local_rank):

    if args.init_model:
        model_state_dict = torch.load(args.init_model, map_location='cpu')
    else:
        model_state_dict = None

    # Prepare model
    cache_dir = args.cache_dir if args.cache_dir else os.path.join(str(PYTORCH_PRETRAINED_BERT_CACHE), 'distributed')

    model = CLIP4Clip.from_pretrained(args.cross_model, cache_dir=cache_dir, state_dict=model_state_dict, task_config=args)
    model.to(device)
    if args.freeze_text_encoder == True:
        for name,param in model.named_parameters():
            if 'clip.transformer.' in name :
                param.requires_grad_(False)
    return model


def init_model_pre(args, device, n_gpu, local_rank, model_file):
    if args.init_model:
        model_state_dict = torch.load(args.init_model, map_location='cpu')
    else:
        model_state_dict = None

    # Prepare model
    cache_dir = args.cache_dir if args.cache_dir else os.path.join(str(PYTORCH_PRETRAINED_BERT_CACHE), 'distributed')

    model = CLIP4Clip.from_pretrained(args.cross_model, cache_dir=cache_dir, state_dict=model_state_dict, task_config=args)

    model_state_dict = torch.load(model_file, map_location='cpu')
    keys = model_state_dict.keys()
    for name,param in model.named_parameters():
        if 'clip.transformer.' in name and name in keys:
            param.requires_grad_(False)
            param.copy_(model_state_dict[name])

    model.to(device)
    return model

def prep_optimizer(args, model, num_train_optimization_steps, device, n_gpu, local_rank, coef_lr=1.):

    if hasattr(model, 'module'):
        model = model.module

    param_optimizer = list(model.named_parameters())
    no_decay = ['bias', 'LayerNorm.bias', 'LayerNorm.weight']

    decay_param_tp = [(n, p) for n, p in param_optimizer if not any(nd in n for nd in no_decay)]
    no_decay_param_tp = [(n, p) for n, p in param_optimizer if any(nd in n for nd in no_decay)]

    decay_clip_param_tp = [(n, p) for n, p in decay_param_tp if ("clip." in n )]
    decay_noclip_param_tp = [(n, p) for n, p in decay_param_tp if ("clip." not in n )]
    no_decay_clip_param_tp = [(n, p) for n, p in no_decay_param_tp if ("clip." in n )]
    no_decay_noclip_param_tp = [(n, p) for n, p in no_decay_param_tp if ("clip." not in n )]

    weight_decay = 0.2
    optimizer_grouped_parameters = [
        {'params': [p for n, p in decay_clip_param_tp], 'weight_decay': weight_decay, 'lr': args.lr * coef_lr},
        {'params': [p for n, p in decay_noclip_param_tp], 'weight_decay': weight_decay},
        {'params': [p for n, p in no_decay_clip_param_tp], 'weight_decay': 0.0, 'lr': args.lr * coef_lr},
        {'params': [p for n, p in no_decay_noclip_param_tp], 'weight_decay': 0.0}
    ]

    scheduler = None
    optimizer = BertAdam(optimizer_grouped_parameters, lr=args.lr, warmup=args.warmup_proportion,
                         schedule='warmup_cosine', b1=0.9, b2=0.98, e=1e-6,
                         t_total=num_train_optimization_steps, weight_decay=weight_decay,
                         max_grad_norm=1.0)

    model = torch.nn.parallel.DistributedDataParallel(model, device_ids=[local_rank],
                                                      output_device=local_rank, find_unused_parameters=True)

    return optimizer, scheduler, model

def save_model(epoch, args, model, optimizer, tr_loss, type_name=""):
    # Only save the model it-self
    model_to_save = model.module if hasattr(model, 'module') else model
    output_model_file = os.path.join(
        args.output_dir, "titles/pytorch_model.bin.{}{}".format("" if type_name=="" else type_name+".", epoch))
    optimizer_state_file = os.path.join(
        args.output_dir, "titles/pytorch_opt.bin.{}{}".format("" if type_name=="" else type_name+".", epoch))
    torch.save(model_to_save.state_dict(), output_model_file)
    torch.save({
            'epoch': epoch,
            'optimizer_state_dict': optimizer.state_dict(),
            'loss': tr_loss,
            }, optimizer_state_file)
    logger.info("Model saved to %s", output_model_file)
    logger.info("Optimizer saved to %s", optimizer_state_file)
    return output_model_file

def load_model(epoch, args, n_gpu, device, model_file=None):        # for evaluation and test
    if model_file is None or len(model_file) == 0:
        model_file = os.path.join(args.output_dir, "titles/pytorch_model.bin.{}".format(epoch))
    if os.path.exists(model_file):
        model_state_dict = torch.load(model_file, map_location='cpu')
        if args.local_rank == 0:
            logger.info("Model loaded from %s", model_file)
        # Prepare model
        cache_dir = args.cache_dir if args.cache_dir else os.path.join(str(PYTORCH_PRETRAINED_BERT_CACHE), 'distributed')
        model = CLIP4Clip.from_pretrained(args.cross_model, cache_dir=cache_dir, state_dict=model_state_dict, task_config=args)
        model.to(device)
    else:
        model = None
    return model


def test_load_model(args, n_gpu, device, model_file=None):
    if os.path.exists(model_file):
        model_state_dict = torch.load(model_file, map_location='cpu')
        if args.local_rank == 0:
            logger.info("Model loaded from %s", model_file)
        cache_dir = args.cache_dir if args.cache_dir else os.path.join(str(PYTORCH_PRETRAINED_BERT_CACHE), 'distributed')
        model = CLIP4Clip.from_pretrained(args.cross_model, cache_dir=cache_dir, state_dict=model_state_dict, task_config=args)
        model.to(device)
    else:
        model = None
    return model

def load_model_eval_zero_shot(epoch, args, n_gpu, device, model_file=None):
    if True:
        model_state_dict = None
        if args.local_rank == 0:
            logger.info("Model loaded from %s", model_file)
        # Prepare model
        cache_dir = args.cache_dir if args.cache_dir else os.path.join(str(PYTORCH_PRETRAINED_BERT_CACHE), 'distributed')
        model = CLIP4Clip.from_pretrained(args.cross_model, cache_dir=cache_dir, state_dict=model_state_dict, task_config=args)
        model.to(device)
    else:
        model = None
    return model

def train_epoch(epoch, args, model, train_dataloader, device, n_gpu, optimizer, scheduler, global_step, local_rank=0):
    global logger
    torch.cuda.empty_cache()
    model.train()
    log_step = args.n_display
    batch_time = AverageMeter()
    end = time.time()
    total_loss = 0
    
    logger.info('start dataloader !')
    
    for step, batch in enumerate(train_dataloader):
        if n_gpu == 1:
            # multi-gpu does scattering it-self, not consider
            batch = tuple(t.to(device=device, non_blocking=True) for t in batch)

        input_ids, input_mask, segment_ids, video, video_mask, title_ids, title_mask = batch
        loss = model(input_ids, input_mask, video, video_mask, title_ids, title_mask, train_video = False)

        if n_gpu > 1:
            loss = loss.mean()  # mean() to average on multi-gpu.
        if args.gradient_accumulation_steps > 1:
            loss = loss / args.gradient_accumulation_steps

        loss.backward()

        total_loss += float(loss)
        if (step + 1) % args.gradient_accumulation_steps == 0:

            torch.nn.utils.clip_grad_norm_(model.parameters(), 1.0)

            if scheduler is not None:
                scheduler.step()  # Update learning rate schedule

            optimizer.step()
            optimizer.zero_grad()

            # https://github.com/openai/CLIP/issues/46
            if hasattr(model, 'module'):
                torch.clamp_(model.module.clip.logit_scale.data, max=np.log(100))
            else:
                torch.clamp_(model.clip.logit_scale.data, max=np.log(100))

            batch_time.update(time.time() - end)
            end = time.time()
            cur_iter = epoch * len(train_dataloader) + step
            max_iter = args.epochs * len(train_dataloader)
            eta_sec = batch_time.avg * (max_iter - cur_iter + 1)
            eta_sec = str(datetime.timedelta(seconds=int(eta_sec)))      


            global_step += 1
            if global_step % log_step == 0 and local_rank == 0:
                logger.info(('Epoch: [{0}/{1}], step: [{2}/{3}], lr: {lr:.2e} eta: {4}\t'
                             'Loss: {loss:.4f}, Time/step: {batch_time.avg:.3f}').format(
                                epoch+1, args.epochs, step+1, len(train_dataloader),
                                eta_sec, loss=loss, batch_time=batch_time,
                                lr=optimizer.get_lr()[0]
                             ))

    total_loss = total_loss / len(train_dataloader)
    return total_loss, global_step

def _run_on_single_gpu(model, batch_list_t, batch_list_v, batch_list_title, batch_sequence_output_list, batch_visual_output_list, batch_title_output_list):
    sim_matrix = []
    for idx1, b1 in enumerate(batch_list_t):
        input_mask, *_tmp = b1
        sequence_output = batch_sequence_output_list[idx1]
        each_row = []
        for idx2, b2 in enumerate(batch_list_v):
            video_mask, *_tmp = b2
            title_mask, *_tmp = batch_list_title[idx2]
            visual_output = batch_visual_output_list[idx2]
            title_output = batch_title_output_list[idx2]

            b1b2_logits = model.get_titles_similarity_logits(sequence_output, visual_output, title_output,
                                                            input_mask, video_mask, title_mask, loose_type=model.loose_type)
            b1b2_logits = b1b2_logits.cpu().detach().numpy()
            each_row.append(b1b2_logits)
        each_row = np.concatenate(tuple(each_row), axis=-1)
        sim_matrix.append(each_row)
    return sim_matrix
  
def eval_epoch(args, model, test_dataloader, device, n_gpu):
    if hasattr(model, 'module'):
        model = model.module.to(device)
    else:
        model = model.to(device)

    multi_sentence_ = False
    cut_off_points_, sentence_num_, video_num_ = [], -1, -1
    if hasattr(test_dataloader.dataset, 'multi_sentence_per_video') \
            and test_dataloader.dataset.multi_sentence_per_video:
        multi_sentence_ = True
        cut_off_points_ = test_dataloader.dataset.cut_off_points
        sentence_num_ = test_dataloader.dataset.sentence_num
        video_num_ = test_dataloader.dataset.video_num
        cut_off_points_ = [itm - 1 for itm in cut_off_points_]

    if multi_sentence_:
        logger.warning("Eval under the multi-sentence per video clip setting.")
        logger.warning("sentence num: {}, video num: {}".format(sentence_num_, video_num_))

    model.eval()

    with torch.no_grad():
        batch_list_t = []
        batch_list_v = []
        batch_list_title = []
        batch_sequence_output_list, batch_visual_output_list, batch_title_output_list = [], [], []
        total_video_num = 0

        # ----------------------------
        # 1. cache the features
        # ----------------------------
        for bid, batch in enumerate(test_dataloader):
            batch = tuple(t.to(device) for t in batch)
            input_ids, input_mask, segment_ids, video, video_mask, title_ids, title_mask = batch
            if multi_sentence_:
                b, *_t = video.shape
                sequence_output = model.get_text_output(input_ids)
                batch_sequence_output_list.append(sequence_output)
                batch_list_t.append((input_mask,))

                s_, e_ = total_video_num, total_video_num + b
                filter_inds = [itm - s_ for itm in cut_off_points_ if itm >= s_ and itm < e_]

                if len(filter_inds) > 0:
                    video, video_mask = video[filter_inds, ...], video_mask[filter_inds, ...]
                    title_ids, title_mask = title_ids[filter_inds, ...], title_mask[filter_inds, ...]

                    visual_output = None
                    title_output = model.get_text_output(title_ids)

                    batch_visual_output_list.append(visual_output)
                    batch_list_v.append((video_mask,))

                    batch_title_output_list.append(title_output)
                    batch_list_title.append((title_mask,))
                total_video_num += b
            else:
                sequence_output, visual_output, title_output = model.get_text_video_title_output(input_ids, video, title_ids, title_mask=title_mask, text_mask=input_mask, train_video = False)

                batch_title_output_list.append(title_output)
                batch_list_title.append((title_mask,))

                batch_sequence_output_list.append(sequence_output)
                batch_list_t.append((input_mask,))

                batch_visual_output_list.append(visual_output)
                batch_list_v.append((video_mask,))

            print("{}/{}\r".format(bid, len(test_dataloader)), end="")


        # ----------------------------------
        # 2. calculate the similarity
        # ----------------------------------
        if n_gpu > 1:
            device_ids = list(range(n_gpu))
            batch_list_t_splits = []
            batch_list_v_splits = []
            batch_list_title_splits = []
            batch_t_output_splits = []
            batch_v_output_splits = []
            batch_title_output_splits = []
            bacth_len = len(batch_list_t)
            split_len = (bacth_len + n_gpu - 1) // n_gpu
            for dev_id in device_ids:
                s_, e_ = dev_id * split_len, (dev_id + 1) * split_len
                if dev_id == 0:
                    batch_list_t_splits.append(batch_list_t[s_:e_])
                    batch_list_v_splits.append(batch_list_v)
                    batch_list_title_splits.append(batch_list_title)

                    batch_t_output_splits.append(batch_sequence_output_list[s_:e_])
                    batch_v_output_splits.append(batch_visual_output_list)
                    batch_title_output_splits.append(batch_title_output_list)
                else:
                    devc = torch.device('cuda:{}'.format(str(dev_id)))
                    devc_batch_list = [tuple(t.to(devc) for t in b) for b in batch_list_t[s_:e_]]
                    batch_list_t_splits.append(devc_batch_list)
                    devc_batch_list = [tuple(t.to(devc) for t in b) for b in batch_list_v]
                    batch_list_v_splits.append(devc_batch_list)
                    devc_batch_list = [tuple(t.to(devc) for t in b) for b in batch_list_title]
                    batch_list_title_splits.append(devc_batch_list)

                    devc_batch_list = [b.to(devc) for b in batch_sequence_output_list[s_:e_]]
                    batch_t_output_splits.append(devc_batch_list)
                    devc_batch_list = [b.to(devc) for b in batch_visual_output_list]
                    batch_v_output_splits.append(devc_batch_list)
                    devc_batch_list = [b.to(devc) for b in batch_title_output_list]
                    batch_title_output_splits.append(devc_batch_list)

            parameters_tuple_list = [(batch_list_t_splits[dev_id], batch_list_v_splits[dev_id], batch_list_title_splits[dev_id],
                                      batch_t_output_splits[dev_id], batch_v_output_splits[dev_id], batch_title_output_splits[dev_id]) for dev_id in device_ids]
            parallel_outputs = parallel_apply(_run_on_single_gpu, model, parameters_tuple_list, device_ids)
            sim_matrix = []
            for idx in range(len(parallel_outputs)):
                sim_matrix += parallel_outputs[idx]
            sim_matrix = np.concatenate(tuple(sim_matrix), axis=0)

        else:
            print('########## Using one GPU for evaluation ##########')
            sim_matrix = _run_on_single_gpu(model, batch_list_t, batch_list_v, batch_list_title, batch_sequence_output_list, batch_visual_output_list, batch_title_output_list)
            sim_matrix = np.concatenate(tuple(sim_matrix), axis=0)
            np.save('sim_matrix/msrvtt_titles_matrix', sim_matrix)
    if multi_sentence_:
        logger.info("before reshape, sim matrix size: {} x {}".format(sim_matrix.shape[0], sim_matrix.shape[1]))
        cut_off_points2len_ = [itm + 1 for itm in cut_off_points_]
        max_length = max([e_-s_ for s_, e_ in zip([0]+cut_off_points2len_[:-1], cut_off_points2len_)])
        sim_matrix_new = []
        for s_, e_ in zip([0] + cut_off_points2len_[:-1], cut_off_points2len_):
            sim_matrix_new.append(np.concatenate((sim_matrix[s_:e_],
                                                  np.full((max_length-e_+s_, sim_matrix.shape[1]), -np.inf)), axis=0))
        sim_matrix = np.stack(tuple(sim_matrix_new), axis=0)
        np.save('sim_matrix/didemo_titles_matrix', sim_matrix)
        logger.info("after reshape, sim matrix size: {} x {} x {}".
                    format(sim_matrix.shape[0], sim_matrix.shape[1], sim_matrix.shape[2]))

        tv_metrics = tensor_text_to_video_metrics(sim_matrix)
        vt_metrics = compute_metrics(tensor_video_to_text_sim(sim_matrix))
    else:
        logger.info("sim matrix size: {}, {}".format(sim_matrix.shape[0], sim_matrix.shape[1]))
        tv_metrics = compute_metrics(sim_matrix)
        vt_metrics = compute_metrics(sim_matrix.T)
        logger.info('\t Length-T: {}, Length-V:{}'.format(len(sim_matrix), len(sim_matrix[0])))

    logger.info("Text-to-Video:")
    logger.info('\t>>>  R@1: {:.1f} - R@5: {:.1f} - R@10: {:.1f} - Median R: {:.1f} - Mean R: {:.1f}'.
                format(tv_metrics['R1'], tv_metrics['R5'], tv_metrics['R10'], tv_metrics['MR'], tv_metrics['MeanR']))
    logger.info("Video-to-Text:")
    logger.info('\t>>>  V2T$R@1: {:.1f} - V2T$R@5: {:.1f} - V2T$R@10: {:.1f} - V2T$Median R: {:.1f} - V2T$Mean R: {:.1f}'.
                format(vt_metrics['R1'], vt_metrics['R5'], vt_metrics['R10'], vt_metrics['MR'], vt_metrics['MeanR']))

    R1 = tv_metrics['R1']
    return R1



def train_titles(args, best_video_weight_path= None):
    global logger
    if args.world_size == 0:
        os.environ['MASTER_ADDR'] = 'localhost'
        os.environ['MASTER_PORT'] = '56220'
        os.environ["RANK"] = "0"
        os.environ['WORLD_SIZE'] = '1'
    if torch.distributed.is_initialized() != True:
        torch.distributed.init_process_group(backend="nccl")

    args = set_seed_logger(args)
    device, n_gpu = init_device(args, args.local_rank)

    tokenizer = ClipTokenizer()

    assert  args.task_type == "retrieval"

    model = init_model_pre(args, device, n_gpu, args.local_rank, best_video_weight_path)

    assert args.freeze_layer_num <= 12 and args.freeze_layer_num >= -1
    if hasattr(model, "clip") and args.freeze_layer_num > -1:
        for name, param in model.clip.named_parameters():
            # top layers always need to train
            if name.find("ln_final.") == 0 or name.find("text_projection") == 0 or name.find("logit_scale") == 0 \
                    or name.find("visual.ln_post.") == 0 or name.find("visual.proj") == 0:
                continue    # need to train
            elif name.find("visual.transformer.resblocks.") == 0 or name.find("transformer.resblocks.") == 0:
                layer_num = int(name.split(".resblocks.")[1].split(".")[0])
                if layer_num >= args.freeze_layer_num:
                    continue    # need to train

            if args.linear_patch == "3d" and name.find("conv2."):
                continue
            else:
                # paramenters which < freeze_layer_num will be freezed
                param.requires_grad = False

    assert args.datatype in DATALOADER_DICT

    assert DATALOADER_DICT[args.datatype]["test"] is not None \
           or DATALOADER_DICT[args.datatype]["val"] is not None

    test_dataloader, test_length = None, 0
    if DATALOADER_DICT[args.datatype]["test"] is not None:          # false pass
        test_dataloader, test_length = DATALOADER_DICT[args.datatype]["test"](args, tokenizer, load_video = False)

    if DATALOADER_DICT[args.datatype]["val"] is not None:
        val_dataloader, val_length = DATALOADER_DICT[args.datatype]["val"](args, tokenizer, load_video = False, subset="val")       # through this
    else:
        val_dataloader, val_length = test_dataloader, test_length

    if test_dataloader is None:     # false pass
        test_dataloader, test_length = val_dataloader, val_length

    if args.local_rank == 0:
        logger.info("***** Running test *****")
        logger.info("  Num examples = %d", test_length)
        logger.info("  Batch size = %d", args.batch_size_val)
        logger.info("  Num steps = %d", len(test_dataloader))
        logger.info("***** Running val *****")
        logger.info("  Num examples = %d", val_length)
        if args.train_frame_order == 0:
            logger.info("***** Video sequential order *****")
        elif args.train_frame_order == 1: 
            logger.info("***** Video reverse order *****")
        elif args.train_frame_order == 2:
            logger.info("***** Video shuffle order *****")

    if args.do_train:
        train_dataloader, train_length, train_sampler = DATALOADER_DICT[args.datatype]["train"](args, tokenizer, load_video = False, caption_aug = False)
        num_train_optimization_steps = (int(len(train_dataloader) + args.gradient_accumulation_steps - 1)
                                        / args.gradient_accumulation_steps) * args.epochs

        coef_lr = args.coef_lr
        # distribute model by torch.nn.distributing
        optimizer, scheduler, model = prep_optimizer(args, model, num_train_optimization_steps, device, n_gpu, args.local_rank, coef_lr=coef_lr)

        if args.local_rank == 0:
            logger.info("***** Running training *****")
            logger.info("  Num examples = %d", train_length)
            logger.info("  Batch size = %d", args.batch_size)
            logger.info("  Num steps = %d", num_train_optimization_steps * args.gradient_accumulation_steps)

        best_score = 0.00001
        best_output_model_file = "None"
        ## ##############################################################
        # resume optimizer state besides loss to continue train
        ## ##############################################################
        resumed_epoch = 0
        if args.resume_model:
            checkpoint = torch.load(args.resume_model, map_location='cpu')
            optimizer.load_state_dict(checkpoint['optimizer_state_dict'])
            resumed_epoch = checkpoint['epoch']+1
            resumed_loss = checkpoint['loss']
        
        global_step = 0

        logger.info("start training !")
        logger.info("resumed_epoch = %d", resumed_epoch) 

        state = True    # everything is correct

        for epoch in range(resumed_epoch, args.epochs):
            train_sampler.set_epoch(epoch)
            tr_loss, global_step = train_epoch(epoch, args, model, train_dataloader, device, n_gpu, optimizer,
                                               scheduler, global_step, local_rank=args.local_rank)

            logger.info('args.local_rank = %d', args.local_rank)
            
            try:
                if args.local_rank == 0:
                    logger.info("Epoch %d/%s Finished, Train Loss: %f", epoch + 1, args.epochs, tr_loss)
                    output_model_file = save_model(epoch, args, model, optimizer, tr_loss, type_name="")
                    # Run on val dataset, this process is *TIME-consuming*.
                    logger.info("Eval on val dataset")
                    R1 = eval_epoch(args, model, test_dataloader, device, n_gpu=1)
                    if best_score <= R1:
                        best_score = R1
                        best_output_model_file = output_model_file
                    logger.info("The best model is: {}, the R1 is: {:.4f}".format(best_output_model_file, best_score))
                    state = True
            except:
                logger.info('CUDA out of memory when evaluating. Testing model at the end!')
                state = False
                continue

        ## Uncomment if want to test on the best checkpoint
        if args.local_rank == 0 and state:
            model = load_model(-1, args, n_gpu=1, device=device, model_file=best_output_model_file)
            eval_epoch(args, model, test_dataloader, device, n_gpu=1)

    elif args.do_eval:
        if args.local_rank == 0:
            best_score = 0.00001
            for epoch in range(args.epochs):
                model = load_model(epoch, args, n_gpu=1, device=device)   # comment for zero-shot
                R1 = eval_epoch(args, model, test_dataloader, device, n_gpu=1)
                if best_score <= R1:
                    best_score = R1
                logger.info("The epoch of best model is: {}, the R1 is: {:.4f}".format(epoch, best_score))
    elif args.do_eval_zero_shot:
        if args.local_rank == 0:
            best_score = 0.00001
            epoch = 0
            model = load_model_eval_zero_shot(epoch, args, n_gpu=1, device=device)   # comment for zero-shot
            R1 = eval_epoch(args, model, test_dataloader, device, n_gpu=1)
            if best_score <= R1:
                best_score = R1
            logger.info("The epoch of best model is: {}, the R1 is: {:.4f}".format(epoch, best_score))

    elif args.do_test:
        if args.local_rank == 0:
            model_file = 'xxxxxx'
            best_score = 0.00001
            epoch = -1
            model = test_load_model(args, n_gpu=1, device=device, model_file=model_file)  # comment for zero-shot

            R1 = eval_epoch(args, model, test_dataloader, device, n_gpu=1)
            if best_score <= R1:
                best_score = R1
            logger.info(" the best model R1 is: {:.4f}".format(best_score))
if __name__ == "__main__":
    args = get_args()
    best_video_weight_path = os.path.join(args.output_dir, 'video/pytorch_best_model')
    train_titles(args, best_video_weight_path)
    if dist.get_rank() == 0:
        fusion_scores()