main_dialog.py

import torch
import torch.backends.cudnn as cudnn
from torch.utils.tensorboard import SummaryWriter
from torch.utils.data import Dataset
from PIL import Image
import cv2
import util.misc as misc
from util.misc import NativeScalerWithGradNormCount as NativeScaler
from llama.llama_mmdiffuser_dialog import LLaMA_mmdiffuser
from data.MMDialog_dataset import mmdialog_Finetune_Dataset, transform_train, format_mmDialog_sprompt

import argparse
import datetime
import json
import numpy as np
import os
import time
from pathlib import Path

from engine_dialog import train_one_epoch


def get_args_parser():
    parser = argparse.ArgumentParser('llama_adapterV2 pre-training', add_help=False)
    parser.add_argument('--batch_size', default=64, type=int,
                        help='Batch size per GPU (effective batch size is batch_size * accum_iter * # gpus')
    parser.add_argument('--epochs', default=5, type=int)
    parser.add_argument('--accum_iter', default=1, type=int,
                        help='Accumulate gradient iterations (for increasing the effective batch size under memory constraints)')

    # Model parameters
    parser.add_argument('--llama_type', default='7B', type=str,
                        help='Type of LLaMA model') #

    parser.add_argument('--llama_path', default='./share_ckpts/llama1_weight', type=str,
                        help='path to LLaMA pretrained checkpoint')
    parser.add_argument('--llama_bias_path', default="./share_ckpts/llama-adapter/7fa55208379faf2dd862565284101b0e4a2a72114d6490a95e432cf9d9b6c813_BIAS-7B.pth", type=str,
                        help='path to LLaMA-adapter BIAS-7B pretrained checkpoint')
    parser.add_argument('--diffuser_path', default="./share_ckpts/stabilityai/stable-diffusion-xl-base-1.0/", type=str,
                        help='path to Diffuser pretrained checkpoint')
    parser.add_argument('--querry_path', default='./checkpoints/20231106/epoch2_queryblock.pkl', type=str,
                        help='path to querry model of the M2Chat')

    parser.add_argument('--max_words', default=512, type=int,
                        help='max number of input words')

    # Optimizer parameters
    parser.add_argument('--weight_decay', type=float, default=0.05,
                        help='weight decay (default: 0.05)')

    parser.add_argument('--lr', type=float, default=None, metavar='LR',
                        help='learning rate (absolute lr)')
    parser.add_argument('--blr', type=float, default=1e-3, metavar='LR',
                        help='base learning rate: absolute_lr = base_lr * total_batch_size / 256')
    parser.add_argument('--min_lr', type=float, default=0., metavar='LR',
                        help='lower lr bound for cyclic schedulers that hit 0')

    parser.add_argument('--warmup_epochs', type=int, default=40, metavar='N',
                        help='epochs to warmup LR')

    # Dataset parameters
    parser.add_argument('--data_config', default='./configs/t2i_pretrain.yaml', type=str,
                        help='dataset config path')
    parser.add_argument('--num_workers', default=10, type=int)
    parser.add_argument('--pin_mem', action='store_true',
                        help='Pin CPU memory in DataLoader for more efficient (sometimes) transfer to GPU.')
    parser.add_argument('--no_pin_mem', action='store_false', dest='pin_mem')
    parser.set_defaults(pin_mem=True)


    parser.add_argument('--output_dir', default='./output',
                        help='path where to save, empty for no saving')
    parser.add_argument('--log_dir', default='./output',
                        help='path where to tensorboard log')
    parser.add_argument('--device', default='cuda',
                        help='device to use for training / testing')
    parser.add_argument('--seed', default=0, type=int)

    parser.add_argument('--start_epoch', default=0, type=int, metavar='N',
                        help='start epoch')

    # distributed training parameters
    parser.add_argument('--world_size', default=1, type=int,
                        help='number of distributed processes')
    parser.add_argument('--local_rank', default=-1, type=int)
    parser.add_argument('--dist_on_itp', action='store_true')
    parser.add_argument('--dist_url', default='env://',
                        help='url used to set up distributed training')

    parser.add_argument('--split_epoch', type=int, default=50)

    return parser


def main(args):
    misc.init_distributed_mode(args)

    print('job dir: {}'.format(os.path.dirname(os.path.realpath(__file__))))
    print("{}".format(args).replace(', ', ',\n'))

    device = torch.device(args.device)

    # fix the seed for reproducibility
    seed = args.seed + misc.get_rank()
    torch.manual_seed(seed)
    np.random.seed(seed)
    cudnn.benchmark = True

    # define the model
    name = "BIAS-7B"
    phase="finetune"


    # BIAS-7B or https://xxx/sha256_BIAS-7B.pth -> 7B
    llama_type = name.split('.')[0].split('-')[-1]
    llama_ckpt_dir = os.path.join(args.llama_path, llama_type)
    llama_tokenzier_path = os.path.join(args.llama_path, 'tokenizer.model')

    print(f'Loading LLaMA-mmdiffuser from {args.llama_bias_path}')
    ckpt = torch.load(args.llama_bias_path, map_location='cpu')
    model_cfg = ckpt.get('config', {})

    model = LLaMA_mmdiffuser(
        llama_ckpt_dir, llama_tokenzier_path, args.diffuser_path,
        max_seq_len=512, max_batch_size=1,
        clip_model='ViT-L/14',
        v_embed_dim=768, v_depth=8,
        v_num_heads=16, v_mlp_ratio=4.0,
        query_len=10, query_layer=31,
        w_bias=model_cfg.get('w_bias', False), 
        w_lora=model_cfg.get('w_lora', False), 
        lora_rank=model_cfg.get('lora_rank', 16),
        w_new_gate=model_cfg.get('w_lora', False), # for compatibility
        phase=phase)

    load_result = model.load_state_dict(ckpt['model'], strict=False)
    query_ckpt = torch.load(args.querry_path, map_location='cpu')
    query_load_result1 = model.query_block.load_state_dict(query_ckpt['query_block'], strict=False)
    query_load_result2 = model.sd_query.load_state_dict(query_ckpt['sd_query'], strict=False)

    assert len(load_result.unexpected_keys) == 0, f"Unexpected keys: {load_result.unexpected_keys}"
    model.to(device)
    model.get_trainable_params(phase='finetune') # reset learnable paramaters

    train_param_name = ['sd_query', 'query_block']
    for name, para in model.named_parameters():
        for train_name in train_param_name:
            if train_name in name:
                para.data = para.data.float()
                para.requires_grad = True
                # print(name)
            else:
                para.requires_grad = False
    
    model_without_ddp = model
    print("Model = %s" % str(model_without_ddp))

    print("Trainable Params:")
    print([(key, val.shape) for key, val in model.named_parameters() if val.requires_grad])

    if args.distributed:
        model = torch.nn.parallel.DistributedDataParallel(model, device_ids=[args.gpu], find_unused_parameters=True)
        model_without_ddp = model.module

    # training detail
    eff_batch_size = args.batch_size * args.accum_iter * misc.get_world_size()

    if args.lr is None:  # only base_lr is specified
        args.lr = args.blr * eff_batch_size / 256

    print("base lr: %.2e" % (args.lr * 256 / eff_batch_size))
    print("actual lr: %.2e" % args.lr)

    print("accumulate grad iterations: %d" % args.accum_iter)
    print("effective batch size: %d" % eff_batch_size)

    # following timm: set wd as 0 for bias and norm layers
    param_groups = misc.add_weight_decay(model_without_ddp, args.weight_decay)
    optimizer = torch.optim.AdamW(param_groups, lr=args.lr, betas=(0.9, 0.95))
    print(optimizer)
    loss_scaler = NativeScaler()



    dataset_train = mmdialog_Finetune_Dataset(args.data_config, transform=transform_train,
                                max_words=args.max_words, tokenizer_path=llama_tokenzier_path)
    print(dataset_train)
    num_tasks = misc.get_world_size()
    global_rank = misc.get_rank()
    sampler_train = torch.utils.data.DistributedSampler(
        dataset_train, num_replicas=num_tasks, rank=global_rank, shuffle=True
    )
    print("Sampler_train = %s" % str(sampler_train))

    data_loader_train = torch.utils.data.DataLoader(
        dataset_train, sampler=sampler_train,
        batch_size=args.batch_size,
        num_workers=args.num_workers,
        pin_memory=args.pin_mem,
        drop_last=True,
    )

    # SummaryWrite
    if global_rank == 0 and args.log_dir is not None:
        os.makedirs(args.log_dir, exist_ok=True)
        log_writer = SummaryWriter(log_dir=args.log_dir)
    else:
        log_writer = None


    print(f"Start training for {args.epochs} epochs")
    start_time = time.time()
    for epoch in range(args.start_epoch, args.epochs):
        if args.distributed:
            data_loader_train.sampler.set_epoch(epoch)

        train_stats = train_one_epoch(
            model, data_loader_train,
            optimizer, device, epoch, loss_scaler,
            log_writer=log_writer,
            args=args
        )

        if args.output_dir and (epoch % 1 == 0 or epoch + 1 == args.epochs):
            with torch.cuda.amp.autocast():
                # prompt = "A door with a sticker of a cat door on it"
                context1 = "Kobe's jersey hangs next to LeBron's locker tonight. (via"
                caption1 = "a man in a blue shirt and a red shirt is in front of a closet . "
                # caption1 = "a piece of cake is sitting on a plate . "
                context2 = "Dear Kobe, I created this masterpiece for you. I wish you were here in person to see the MambaMentality flowing within this painting, but I know you're smiling down from heaven! We all love you and miss you! KobeBryant"
                caption2 = "a painting of a bald man with blue hair ."
                # answer = context2+"\n###Image:"+caption2
                formprompts = format_mmDialog_sprompt('answer based on the input dialog', context1, caption1)
                # formprompts = formprompts + answer
                img = Image.fromarray(cv2.imread('/aifs4su/mmdata/rawdata/videogen/MMDialogDataset/MMDialogDataset/train/-8621082798323348566.jpg'))
                img = model_without_ddp.clip_transform(img).unsqueeze(0).to(device)# img = model_without_ddp.t2i_generate(prompts=[formprompts])
                # print(img)
                print([formprompts])
                print("Reference Answer:"+context2+" </IC>" + caption2 + "</IC>" + "<|img|>"*128)
                model_without_ddp.eval()
                sdxl_img = model_without_ddp.t2i_generate(prompts=[caption2], use_origin=True, seed=epoch)
                gen_text, img = model_without_ddp.ti2ti_generate(img, [formprompts])
                model_without_ddp.train()
            img_resize = [img[0].resize((256, 256))]
            # print(img_resize)
            np_image = np.asarray(img_resize)
            
            sdxl_img_resize = [sdxl_img[0].resize((256, 256))]
            sdxl_np_image = np.asarray(sdxl_img_resize)
            if misc.is_main_process():
                log_writer.add_images("validation", np_image, epoch, dataformats="NHWC")
                log_writer.add_images("sdxl", sdxl_np_image, epoch, dataformats="NHWC")
                aligner_to_save = {     #'sd_query', 'sd_query_block', 'text_unpooled_proj', 'text_pooled_proj'
                    "sd_query":model_without_ddp.sd_query.state_dict(),
                    "query_block":model_without_ddp.query_block.state_dict(),
                    "epoch":epoch
                }
                print(gen_text)
                log_writer.add_text("generate prompts", gen_text[0], global_step=epoch)
                ckpt_dir = os.path.join(args.log_dir,"checkpoints/")
                if not os.path.exists(ckpt_dir):
                    os.makedirs(ckpt_dir)
                save_dir = os.path.join(ckpt_dir,f"epoch{epoch}_queryblock.pkl")
                torch.save(aligner_to_save,save_dir)
            if args.output_dir and (epoch % 100 == 0 or epoch + 1 == args.epochs):    
                misc.save_model(
                    args=args, model=model, model_without_ddp=model_without_ddp, optimizer=optimizer,
                    loss_scaler=loss_scaler, epoch=epoch)

        log_stats = {**{f'train_{k}': v for k, v in train_stats.items()},
                     'epoch': epoch,
                     **{f'val_{k}': v for k, v in train_stats.items()}}

        if args.output_dir and misc.is_main_process():
            if log_writer is not None:
                log_writer.flush()
            with open(os.path.join(args.output_dir, "log.txt"), mode="a", encoding="utf-8") as f:
                f.write(json.dumps(log_stats) + "\n")

    total_time = time.time() - start_time
    total_time_str = str(datetime.timedelta(seconds=int(total_time)))
    print('Training time {}'.format(total_time_str))


if __name__ == '__main__':
    args = get_args_parser()
    args = args.parse_args()
    if args.output_dir:
        Path(args.output_dir).mkdir(parents=True, exist_ok=True)
    main(args)