mgie_train.py

#
# For licensing see accompanying LICENSE file.
# Copyright (C) 2024 Apple Inc. All Rights Reserved.
#
# modified from https://github.com/haotian-liu/LLaVA/blob/7ace501183c4bdec6052ec1a30039cdc3242a67c/llava/train/train.py

import os
import copy
from dataclasses import dataclass, field
import json
import logging
import pathlib
from typing import Dict, Optional, Sequence, List

import torch

import transformers
from torch.utils.data import Dataset
from llava.train.llava_trainer import LLaVATrainer

from llava import conversation as conversation_lib
from llava.model import *

from PIL import Image
import torch.nn as nn

# TODO: import and use code from ../data/dataset.py

IGNORE_INDEX = -100
DEFAULT_PAD_TOKEN = "[PAD]"
DEFAULT_EOS_TOKEN = "</s>"
DEFAULT_BOS_TOKEN = "<s>"
DEFAULT_UNK_TOKEN = "<unk>"
DEFAULT_IMAGE_TOKEN = "<image>"
DEFAULT_IMAGE_PATCH_TOKEN = "<im_patch>"
DEFAULT_IM_START_TOKEN = "<im_start>"
DEFAULT_IM_END_TOKEN = "<im_end>"

import io, base64, pickle, random
from tqdm import tqdm
import numpy as np

def b2f(b): return Image.open(io.BytesIO(base64.b64decode(b))).convert('RGB')
def resize(f):
    w, h = f.size
    if w>h:
        p = (w-h)//2
        f = f.crop([p, 0, p+h, h])
    elif h>w:
        p = (h-w)//2
        f = f.crop([0, p, w, p+w])
    f = f.resize([512, 512])
    return f
def img2npy(f): return (2.0*np.array(f)/255.0-1.0).transpose((2, 0, 1)).astype(np.float32)

@dataclass
class ModelArguments:
    model_name_or_path: Optional[str] = field(default="facebook/opt-125m")
    version: Optional[str] = field(default="v0")
    freeze_backbone: bool = field(default=False)
    tune_mm_mlp_adapter: bool = field(default=False)
    vision_tower: Optional[str] = field(default=None)
    mm_vision_select_layer: Optional[int] = field(default=-1)   # default to the last layer
    pretrain_mm_mlp_adapter: Optional[str] = field(default=None)
    mm_use_im_start_end: bool = field(default=False)


@dataclass
class DataArguments:
    data_path: str = field(default=None,
                           metadata={"help": "Path to the training data."})
    lazy_preprocess: bool = False
    is_multimodal: bool = False
    sep_image_conv_front: bool = False
    image_token_len: int = 0
    image_folder: Optional[str] = field(default=None)
    image_aspect_ratio: str = 'square'


@dataclass
class TrainingArguments(transformers.TrainingArguments):
    cache_dir: Optional[str] = field(default=None)
    optim: str = field(default="adamw_torch")
    remove_unused_columns: bool = field(default=False)
    freeze_mm_mlp_adapter: bool = field(default=False)
    force_fsdp: bool = field(default=False)
    model_max_length: int = field(
        default=512,
        metadata={
            "help":
            "Maximum sequence length. Sequences will be right padded (and possibly truncated)."
        },
    )
    double_quant: bool = field(
        default=True,
        metadata={"help": "Compress the quantization statistics through double quantization."}
    )
    quant_type: str = field(
        default="nf4",
        metadata={"help": "Quantization data type to use. Should be one of `fp4` or `nf4`."}
    )
    bits: int = field(
        default=16,
        metadata={"help": "How many bits to use."}
    )
    lora_enable: bool = False
    lora_r: int = 64
    lora_alpha: int = 16
    lora_dropout: float = 0.05
    lora_weight_path: str = ""
    lora_bias: str = "none"


def maybe_zero_3(param, ignore_status=False, name=None):
    from deepspeed import zero
    from deepspeed.runtime.zero.partition_parameters import ZeroParamStatus
    if hasattr(param, "ds_id"):
        if param.ds_status == ZeroParamStatus.NOT_AVAILABLE:
            if not ignore_status:
                logging.warning(f"{name}: param.ds_status != ZeroParamStatus.NOT_AVAILABLE: {param.ds_status}")
        with zero.GatheredParameters([param]):
            param = param.data.detach().cpu().clone()
    else:
        param = param.detach().cpu().clone()
    return param


# Borrowed from peft.utils.get_peft_model_state_dict
def get_peft_state_maybe_zero_3(named_params, bias):
    if bias == "none":
        to_return = {k: t for k, t in named_params if "lora_" in k}
    elif bias == "all":
        to_return = {k: t for k, t in named_params if "lora_" in k or "bias" in k}
    elif bias == "lora_only":
        to_return = {}
        maybe_lora_bias = {}
        lora_bias_names = set()
        for k, t in named_params:
            if "lora_" in k:
                to_return[k] = t
                bias_name = k.split("lora_")[0] + "bias"
                lora_bias_names.add(bias_name)
            elif "bias" in k:
                maybe_lora_bias[k] = t
        for k, t in maybe_lora_bias:
            if bias_name in lora_bias_names:
                to_return[bias_name] = t
    else:
        raise NotImplementedError
    to_return = {k: maybe_zero_3(v, name=k) for k, v in to_return.items()}
    return to_return


def get_peft_state_non_lora_maybe_zero_3(named_params, require_grad_only=True):
    to_return = {k: t for k, t in named_params if "lora_" not in k}
    if require_grad_only:
        to_return = {k: t for k, t in to_return.items() if t.requires_grad}
    to_return = {k: maybe_zero_3(v, ignore_status=True).cpu() for k, v in to_return.items()}
    return to_return


def find_all_linear_names(model):
    cls = torch.nn.Linear
    lora_module_names = set()
    for name, module in model.named_modules():
        if isinstance(module, cls):
            names = name.split('.')
            lora_module_names.add(names[0] if len(names) == 1 else names[-1])


    if 'lm_head' in lora_module_names: # needed for 16-bit
        lora_module_names.remove('lm_head')
    return list(lora_module_names)


def safe_save_model_for_hf_trainer(trainer: transformers.Trainer,
                                   output_dir: str):
    """Collects the state dict and dump to disk."""
    if trainer.deepspeed:
        torch.cuda.synchronize()
        trainer.save_model(output_dir)
        return

    state_dict = trainer.model.state_dict()
    if trainer.args.should_save:
        cpu_state_dict = {
            key: value.cpu()
            for key, value in state_dict.items()
        }
        del state_dict
        trainer._save(output_dir, state_dict=cpu_state_dict)  # noqa


def smart_tokenizer_and_embedding_resize(
    special_tokens_dict: Dict,
    tokenizer: transformers.PreTrainedTokenizer,
    model: transformers.PreTrainedModel,
):
    """Resize tokenizer and embedding.

    Note: This is the unoptimized version that may make your embedding size not be divisible by 64.
    """
    num_new_tokens = tokenizer.add_special_tokens(special_tokens_dict)
    model.resize_token_embeddings(len(tokenizer))

    if num_new_tokens > 0:
        input_embeddings = model.get_input_embeddings().weight.data
        output_embeddings = model.get_output_embeddings().weight.data

        input_embeddings_avg = input_embeddings[:-num_new_tokens].mean(
            dim=0, keepdim=True)
        output_embeddings_avg = output_embeddings[:-num_new_tokens].mean(
            dim=0, keepdim=True)

        input_embeddings[-num_new_tokens:] = input_embeddings_avg
        output_embeddings[-num_new_tokens:] = output_embeddings_avg


def _tokenize_fn(strings: Sequence[str],
                 tokenizer: transformers.PreTrainedTokenizer) -> Dict:
    """Tokenize a list of strings."""
    tokenized_list = [
        tokenizer(
            text,
            return_tensors="pt",
            padding="longest",
            max_length=tokenizer.model_max_length,
            truncation=True,
        ) for text in strings
    ]
    input_ids = labels = [
        tokenized.input_ids[0] for tokenized in tokenized_list
    ]
    input_ids_lens = labels_lens = [
        tokenized.input_ids.ne(tokenizer.pad_token_id).sum().item()
        for tokenized in tokenized_list
    ]
    return dict(
        input_ids=input_ids,
        labels=labels,
        input_ids_lens=input_ids_lens,
        labels_lens=labels_lens,
    )


def _mask_targets(target, tokenized_lens, speakers):
    # cur_idx = 0
    cur_idx = tokenized_lens[0]
    tokenized_lens = tokenized_lens[1:]
    target[:cur_idx] = IGNORE_INDEX
    for tokenized_len, speaker in zip(tokenized_lens, speakers):
        if speaker == "human":
            target[cur_idx+2:cur_idx + tokenized_len] = IGNORE_INDEX
        cur_idx += tokenized_len


def _add_speaker_and_signal(header, source, get_conversation=True):
    """Add speaker and start/end signal on each round."""
    BEGIN_SIGNAL = "### "
    END_SIGNAL = "\n"
    conversation = header
    for sentence in source:
        from_str = sentence["from"]
        if from_str.lower() == "human":
            from_str = conversation_lib.default_conversation.roles[0]
        elif from_str.lower() == "gpt":
            from_str = conversation_lib.default_conversation.roles[1]
        else:
            from_str = 'unknown'
        sentence["value"] = (BEGIN_SIGNAL + from_str + ": " +
                             sentence["value"] + END_SIGNAL)
        if get_conversation:
            conversation += sentence["value"]
    conversation += BEGIN_SIGNAL
    return conversation


def preprocess_multimodal(
    sources: Sequence[str],
    multimodal_cfg: dict,
    cur_token_len: int,
) -> Dict:
    is_multimodal = multimodal_cfg['is_multimodal']
    # image_token_len = multimodal_cfg['image_token_len']
    image_token_len = cur_token_len
    if not is_multimodal:
        return sources

    for source in sources:
        if multimodal_cfg['sep_image_conv_front']:
            assert DEFAULT_IMAGE_TOKEN in source[0]['value']
            source[0]['value'] = source[0]['value'].replace(DEFAULT_IMAGE_TOKEN, '').strip()
            source[0]['value'] = DEFAULT_IMAGE_TOKEN + conversation_lib.default_conversation.sep + conversation_lib.default_conversation.roles[0] + ": " + source[0]['value']
        for sentence in source:
            replace_token = DEFAULT_IMAGE_PATCH_TOKEN * image_token_len
            if multimodal_cfg['use_im_start_end']:
                replace_token = DEFAULT_IM_START_TOKEN + replace_token + DEFAULT_IM_END_TOKEN
            sentence["value"] = sentence["value"].replace(DEFAULT_IMAGE_TOKEN, replace_token)

    return sources


def preprocess_v1(
    sources,
    tokenizer: transformers.PreTrainedTokenizer,
) -> Dict:
    conv = conversation_lib.default_conversation.copy()
    roles = {"human": conv.roles[0], "gpt": conv.roles[1]}

    # Apply prompt templates
    conversations = []
    for i, source in enumerate(sources):
        if roles[source[0]["from"]] != conv.roles[0]:
            # Skip the first one if it is not from human
            source = source[1:]

        conv.messages = []
        for j, sentence in enumerate(source):
            role = roles[sentence["from"]]
            assert role == conv.roles[j % 2], f"{i}"
            conv.append_message(role, sentence["value"])
        conversations.append(conv.get_prompt())

    # Tokenize conversations
    input_ids = tokenizer(
        conversations,
        return_tensors="pt",
        padding="longest",
        max_length=tokenizer.model_max_length,
        truncation=True,
    ).input_ids
    targets = input_ids.clone()

    assert conv.sep_style == conversation_lib.SeparatorStyle.TWO

    # Mask targets
    sep = conv.sep + conv.roles[1] + ": "
    for conversation, target in zip(conversations, targets):
        total_len = int(target.ne(tokenizer.pad_token_id).sum())

        rounds = conversation.split(conv.sep2)
        cur_len = 1
        target[:cur_len] = IGNORE_INDEX
        for i, rou in enumerate(rounds):
            if rou == "":
                break

            parts = rou.split(sep)
            if len(parts) != 2:
                break
            parts[0] += sep
            round_len = len(tokenizer(rou).input_ids)
            instruction_len = len(tokenizer(parts[0]).input_ids) - 2

            target[cur_len : cur_len + instruction_len] = IGNORE_INDEX

            cur_len += round_len
        target[cur_len:] = IGNORE_INDEX

        if cur_len < tokenizer.model_max_length:
            if cur_len != total_len:
                target[:] = IGNORE_INDEX
                print(
                    f"WARNING: tokenization mismatch: {cur_len} vs. {total_len}."
                    f" (ignored)"
                )

    return dict(
        input_ids=input_ids,
        labels=targets,
    )

def preprocess_mpt(
    sources,
    tokenizer: transformers.PreTrainedTokenizer,
) -> Dict:
    conv = conversation_lib.default_conversation.copy()
    roles = {"human": conv.roles[0], "gpt": conv.roles[1]}

    # Apply prompt templates
    conversations = []
    for i, source in enumerate(sources):
        if roles[source[0]["from"]] != conv.roles[0]:
            # Skip the first one if it is not from human
            source = source[1:]

        conv.messages = []
        for j, sentence in enumerate(source):
            role = roles[sentence["from"]]
            assert role == conv.roles[j % 2], f"{i}"
            conv.append_message(role, sentence["value"])
        conversations.append(conv.get_prompt())

    # Tokenize conversations
    input_ids = tokenizer(
        conversations,
        return_tensors="pt",
        padding="longest",
        max_length=tokenizer.model_max_length,
        truncation=True,
    ).input_ids
    targets = input_ids.clone()
    assert conv.sep_style == conversation_lib.SeparatorStyle.MPT

    # Mask targets
    sep = conv.sep + conv.roles[1]
    for conversation, target in zip(conversations, targets):
        total_len = int(target.ne(tokenizer.pad_token_id).sum())

        rounds = conversation.split(conv.sep)
        re_rounds = [conv.sep.join(rounds[:3])] # system + user + gpt
        for conv_idx in range(3, len(rounds), 2):
            re_rounds.append(conv.sep.join(rounds[conv_idx:conv_idx+2]))    # user + gpt
        cur_len = 0
        target[:cur_len] = IGNORE_INDEX
        for i, rou in enumerate(re_rounds):
            if rou == "":
                break

            parts = rou.split(sep)
            if len(parts) != 2:
                break
            parts[0] += sep
            round_len = len(tokenizer(rou).input_ids) + len(tokenizer(conv.sep).input_ids)
            instruction_len = len(tokenizer(parts[0]).input_ids)
            target[cur_len : cur_len + instruction_len] = IGNORE_INDEX

            cur_len += round_len
        target[cur_len:] = IGNORE_INDEX

        if cur_len < tokenizer.model_max_length:
            if cur_len != total_len:
                target[:] = IGNORE_INDEX
                print(
                    f"WARNING: tokenization mismatch: {cur_len} vs. {total_len}."
                    f" (ignored)"
                )

    return dict(
        input_ids=input_ids,
        labels=targets,
    )


def preprocess(
    sources: Sequence[str],
    tokenizer: transformers.PreTrainedTokenizer,
) -> Dict:
    """
    Given a list of sources, each is a conversation list. This transform:
    1. Add signal '### ' at the beginning each sentence, with end signal '\n';
    2. Concatenate conversations together;
    3. Tokenize the concatenated conversation;
    4. Make a deepcopy as the target. Mask human words with IGNORE_INDEX.
    """
    if conversation_lib.default_conversation.version == "v1":
        return preprocess_v1(sources, tokenizer)
    if conversation_lib.default_conversation.version == "mpt":
        return preprocess_mpt(sources, tokenizer)
    # add end signal and concatenate together
    conversations = []
    for source in sources:
        header = f"{conversation_lib.default_conversation.system}\n\n"
        conversation = _add_speaker_and_signal(header, source)
        conversations.append(conversation)
    # tokenize conversations
    conversations_tokenized = _tokenize_fn(conversations, tokenizer)
    input_ids = conversations_tokenized["input_ids"]
    targets = copy.deepcopy(input_ids)
    for target, source in zip(targets, sources):
        tokenized_lens = _tokenize_fn([header] + [s["value"] for s in source],
                                      tokenizer)["input_ids_lens"]
        speakers = [sentence["from"] for sentence in source]
        _mask_targets(target, tokenized_lens, speakers)

    return dict(input_ids=input_ids, labels=targets)


class SupervisedDataset(Dataset):
    """Dataset for supervised fine-tuning."""

    def __init__(self, data_path: str,
                 tokenizer: transformers.PreTrainedTokenizer):
        super(SupervisedDataset, self).__init__()
        logging.warning("Loading data...")
        list_data_dict = json.load(open(data_path, "r"))

        logging.warning("Formatting inputs...")
        sources = [example["conversations"] for example in list_data_dict]
        data_dict = preprocess(sources, tokenizer)

        self.input_ids = data_dict["input_ids"]
        self.labels = data_dict["labels"]

    def __len__(self):
        return len(self.input_ids)

    def __getitem__(self, i) -> Dict[str, torch.Tensor]:
        return dict(input_ids=self.input_ids[i], labels=self.labels[i])


class LazySupervisedDataset(Dataset):

    def __init__(self, data_path: str,
                 tokenizer: transformers.PreTrainedTokenizer,
                 multimodal_cfg: dict):
        super(LazySupervisedDataset, self).__init__()

        self.tokenizer, self.multimodal_cfg = tokenizer, multimodal_cfg

        self.pkl, self.prompt = pickle.load(open('./_data/ipr2pr.pkl', 'rb'))['task'], json.load(open('./_data/ipr2pr_expressive.json', 'r'))
        random.shuffle(self.pkl)
        print('--pkl: %d--'%(len(self.pkl)))

    def __len__(self):
        return len(self.pkl)

    def __getitem__(self, i) -> Dict[str, torch.Tensor]:
        item = self.pkl[i][0]

        tsv = open('./_data/ipr2pr.tsv', 'r')
        tsv.seek(item['lineidx'])
        b = tsv.readline().strip().split('\t')
        image = resize(b2f(b[0]))

        processor = self.multimodal_cfg['image_processor']
        image = processor.preprocess(image, return_tensors='pt')['pixel_values'][0]

        cur_token_len = (image.shape[1]//14)*(image.shape[2]//14)
        query = "what will this image be like if '%s'\n%s"%(item['instruction'], DEFAULT_IMAGE_TOKEN)
        ans = '%s [IMG0] [IMG1] [IMG2] [IMG3] [IMG4] [IMG5] [IMG6] [IMG7]'%(self.prompt[item['input']]['expressive'])
        sources = preprocess_multimodal(copy.deepcopy([[{'from': 'human', 'value': query}, {'from': 'gpt', 'value': ans}]]),
                                        self.multimodal_cfg, cur_token_len)

        data_dict = preprocess(sources, self.tokenizer)
        if isinstance(i, int): data_dict = dict(input_ids=data_dict['input_ids'][0],
                                                labels=data_dict['labels'][0])
        data_dict['image'] = image

        p2p_inp, p2p_ans = img2npy(resize(b2f(b[0])).resize([256, 256])), img2npy(resize(b2f(b[1])).resize([256, 256]))
        data_dict['p2p_inp'], data_dict['p2p_ans'] = p2p_inp, p2p_ans

        return data_dict


@dataclass
class DataCollatorForSupervisedDataset(object):
    """Collate examples for supervised fine-tuning."""

    tokenizer: transformers.PreTrainedTokenizer

    def __call__(self, instances: Sequence[Dict]) -> Dict[str, torch.Tensor]:
        input_ids, labels = tuple([instance[key] for instance in instances]
                                  for key in ("input_ids", "labels"))
        input_ids = torch.nn.utils.rnn.pad_sequence(
            input_ids,
            batch_first=True,
            padding_value=self.tokenizer.pad_token_id)
        labels = torch.nn.utils.rnn.pad_sequence(labels,
                                                 batch_first=True,
                                                 padding_value=IGNORE_INDEX)
        batch = dict(
            input_ids=input_ids,
            labels=labels,
            attention_mask=input_ids.ne(self.tokenizer.pad_token_id),
        )

        if 'image' in instances[0]:
            images = [instance['image'] for instance in instances]
            if all(x is not None and x.shape == images[0].shape for x in images):
                batch['images'] = torch.stack(images)
            else:
                batch['images'] = images

        batch['p2p_inp'], batch['p2p_ans'] = [torch.cat([torch.from_numpy(d['p2p_inp']).unsqueeze(dim=0) for d in instances], dim=0),
                                              torch.cat([torch.from_numpy(d['p2p_ans']).unsqueeze(dim=0) for d in instances], dim=0)]

        return batch


def make_supervised_data_module(tokenizer: transformers.PreTrainedTokenizer,
                                data_args) -> Dict:
    """Make dataset and collator for supervised fine-tuning."""
    dataset_cls = (LazySupervisedDataset
                   if data_args.lazy_preprocess else SupervisedDataset)
    train_dataset = dataset_cls(tokenizer=tokenizer,
                                data_path=data_args.data_path,
                                multimodal_cfg=dict(
                                    is_multimodal=data_args.is_multimodal,
                                    sep_image_conv_front=data_args.sep_image_conv_front,
                                    image_token_len=data_args.image_token_len,
                                    image_folder=data_args.image_folder,
                                    image_aspect_ratio=data_args.image_aspect_ratio,
                                    use_im_start_end=getattr(data_args, 'mm_use_im_start_end', False),
                                    image_processor=getattr(data_args, 'image_processor', None)))
    data_collator = DataCollatorForSupervisedDataset(tokenizer=tokenizer)
    return dict(train_dataset=train_dataset,
                eval_dataset=None,
                data_collator=data_collator)


def train():
    parser = transformers.HfArgumentParser((ModelArguments, DataArguments, TrainingArguments))
    model_args, data_args, training_args = parser.parse_args_into_dataclasses()
    compute_dtype = (torch.float16 if training_args.fp16 else (torch.bfloat16 if training_args.bf16 else torch.float32))

    bnb_model_from_pretrained_args = {}
    if training_args.bits in [4, 8]:
        from transformers import BitsAndBytesConfig
        from peft import prepare_model_for_int8_training
        bnb_model_from_pretrained_args.update(dict(
            device_map={"": training_args.device},
            load_in_4bit=training_args.bits == 4,
            load_in_8bit=training_args.bits == 8,
            quantization_config=BitsAndBytesConfig(
                load_in_4bit=training_args.bits == 4,
                load_in_8bit=training_args.bits == 8,
                llm_int8_threshold=6.0,
                llm_int8_has_fp16_weight=False,
                bnb_4bit_compute_dtype=compute_dtype,
                bnb_4bit_use_double_quant=training_args.double_quant,
                bnb_4bit_quant_type=training_args.quant_type # {'fp4', 'nf4'}
            )
        ))

    if model_args.vision_tower is not None:
        if 'mpt' in model_args.model_name_or_path:
            model = LlavaMPTForCausalLM.from_pretrained(
                model_args.model_name_or_path,
                cache_dir=training_args.cache_dir,
                **bnb_model_from_pretrained_args
            )
        else:
            model = LlavaLlamaForCausalLM.from_pretrained(
                model_args.model_name_or_path,
                cache_dir=training_args.cache_dir,
                **bnb_model_from_pretrained_args
            )
    else:
        model = transformers.LlamaForCausalLM.from_pretrained(
            model_args.model_name_or_path,
            cache_dir=training_args.cache_dir,
            **bnb_model_from_pretrained_args
        )
    model.config.use_cache = False

    if model_args.freeze_backbone:
        model.model.requires_grad_(False)

    if training_args.bits in [4, 8]:
        model.config.torch_dtype=(torch.float32 if training_args.fp16 else (torch.bfloat16 if training_args.bf16 else torch.float32))
        model = prepare_model_for_int8_training(model, use_gradient_checkpointing=training_args.gradient_checkpointing)

    if training_args.gradient_checkpointing and model_args.vision_tower is None:
        if hasattr(model, "enable_input_require_grads"):
            model.enable_input_require_grads()
        else:
            def make_inputs_require_grad(module, input, output):
                output.requires_grad_(True)
            model.get_input_embeddings().register_forward_hook(make_inputs_require_grad)

    if training_args.lora_enable:
        from peft import LoraConfig, get_peft_model
        lora_config = LoraConfig(
            r=training_args.lora_r,
            lora_alpha=training_args.lora_alpha,
            target_modules=find_all_linear_names(model),
            lora_dropout=training_args.lora_dropout,
            bias=training_args.lora_bias,
            task_type="CAUSAL_LM",
        )
        if training_args.bits == 16:
            if training_args.bf16:
                model.to(torch.bfloat16)
            if training_args.fp16:
                model.to(torch.float16)
        logging.warning("Adding LoRA adapters...")
        model = get_peft_model(model, lora_config)

    if 'mpt' in model_args.model_name_or_path:
        tokenizer = transformers.AutoTokenizer.from_pretrained(
            model_args.model_name_or_path,
            cache_dir=training_args.cache_dir,
            model_max_length=training_args.model_max_length,
            padding_side="right"
        )
    else:
        tokenizer = transformers.AutoTokenizer.from_pretrained(
            model_args.model_name_or_path,
            cache_dir=training_args.cache_dir,
            model_max_length=training_args.model_max_length,
            padding_side="right",
            use_fast=False,
        )

    if model_args.version == "v0":
        if tokenizer.pad_token is None:
            smart_tokenizer_and_embedding_resize(
                special_tokens_dict=dict(pad_token=DEFAULT_PAD_TOKEN),
                tokenizer=tokenizer,
                model=model,
            )
        if "llama" in model_args.model_name_or_path:
            tokenizer.add_special_tokens({
                "eos_token": DEFAULT_EOS_TOKEN,
                "bos_token": DEFAULT_BOS_TOKEN,
                "unk_token": DEFAULT_UNK_TOKEN,
            })
    else:
        tokenizer.pad_token = tokenizer.unk_token
        if "mpt" in model_args.model_name_or_path:
            conversation_lib.default_conversation = conversation_lib.conv_templates["mpt"]
        else:
            conversation_lib.default_conversation = conversation_lib.conv_templates["vicuna_v1_1"]

    if model_args.vision_tower is not None:
        model_vision_dict = model.get_model().initialize_vision_modules(
            vision_tower=model_args.vision_tower,
            mm_vision_select_layer=model_args.mm_vision_select_layer,
            pretrain_mm_mlp_adapter=model_args.pretrain_mm_mlp_adapter,
            fsdp=training_args.fsdp
        )
        model.get_vision_tower().to(dtype=torch.float16, device=training_args.device)
        vision_config = model_vision_dict['vision_config']

        data_args.image_token_len = model_vision_dict['image_token_len']
        data_args.image_processor = model_vision_dict['image_processor']
        data_args.is_multimodal = True

        model.config.tune_mm_mlp_adapter = training_args.tune_mm_mlp_adapter = model_args.tune_mm_mlp_adapter
        if model_args.tune_mm_mlp_adapter:
            model.requires_grad_(False)
            for p in model.get_model().mm_projector.parameters():
                p.requires_grad = True

        model.config.freeze_mm_mlp_adapter = training_args.freeze_mm_mlp_adapter
        if training_args.freeze_mm_mlp_adapter:
            for p in model.get_model().mm_projector.parameters():
                p.requires_grad = False

        if training_args.bits in [4, 8]:
            model.get_model().mm_projector.to(dtype=compute_dtype, device=training_args.device)

        model.config.mm_use_im_start_end = data_args.mm_use_im_start_end = model_args.mm_use_im_start_end
        vision_config.use_im_start_end = training_args.use_im_start_end = model_args.mm_use_im_start_end
        model.config.sep_image_conv_front = data_args.sep_image_conv_front
        model.initialize_vision_tokenizer(mm_use_im_start_end=model_args.mm_use_im_start_end, tokenizer=tokenizer, device=training_args.device,
                                          tune_mm_mlp_adapter=model_args.tune_mm_mlp_adapter, pretrain_mm_mlp_adapter=model_args.pretrain_mm_mlp_adapter)

        params_no_grad = [n for n, p in model.named_parameters() if not p.requires_grad]
        if len(params_no_grad) > 0:
            if training_args.fsdp is not None and len(training_args.fsdp) > 0:
                if len(params_no_grad) < 10:
                    print('[WARNING] Attempting to use FSDP while {} parameters do not require gradients: {}'. format(len(params_no_grad), params_no_grad))
                else:
                    print('[WARNING] Attempting to use FSDP while {} parameters do not require gradients: {}...(omitted)'. format(len(params_no_grad), ', '.join(params_no_grad[:10])))
                print("[WARNING] Attempting to use FSDP with partially frozen paramters, this is experimental.")
                print("[WARNING] As of 4/30/23, this feature requires PyTorch-nightly build.  See here for details: https://github.com/haotian-liu/LLaVA#experimental-use-fsdp-to-save-memory-in-pretraining")

                from torch.distributed.fsdp.fully_sharded_data_parallel import FullyShardedDataParallel as FSDP
                def patch_FSDP_use_orig_params(func):
                    def wrap_func(*args, **kwargs):
                        use_orig_params = kwargs.pop('use_orig_params', True)
                        return func(*args, **kwargs, use_orig_params=use_orig_params)
                    return wrap_func

                FSDP.__init__ = patch_FSDP_use_orig_params(FSDP.__init__)

    if training_args.bits in [4, 8]:
        from peft.tuners.lora import LoraLayer
        for name, module in model.named_modules():
            if isinstance(module, LoraLayer):
                if training_args.bf16:
                    module = module.to(torch.bfloat16)
            if 'norm' in name:
                module = module.to(torch.float32)
            if 'lm_head' in name or 'embed_tokens' in name:
                if hasattr(module, 'weight'):
                    if training_args.bf16 and module.weight.dtype == torch.float32:
                        module = module.to(torch.bfloat16)

    # start for MGIE
    os.makedirs('_log', exist_ok=True)

    pt = {}
    for i in tqdm(range(2)): pt.update(torch.load('./_ckpt/LLaVA-7B-v1/pytorch_model-0000%d-of-00002.bin'%(i+1), map_location='cpu'))
    miss, unexp = model.load_state_dict(pt, strict=False)
    print('miss:', miss), print('unexp:', unexp)

    tokenizer.add_tokens(['[IMG0]', '[IMG1]', '[IMG2]', '[IMG3]', '[IMG4]', '[IMG5]', '[IMG6]', '[IMG7]'], special_tokens=True)
    model.resize_token_embeddings(len(tokenizer))
    print(tokenizer), json.dump(tokenizer.get_vocab(), open('_log/vocabs.json', 'w'), indent=2)

    for n, p in model.named_parameters():
        if 'embed_tokens' in n or 'lm_head' in n or 'edit_head' in n or 'unet' in n: p.requires_grad = True
        else: p.requires_grad = False
    with open('_log/parameters.txt', 'w') as F:
        for n, p in model.named_parameters(): F.write('%s %s %s\n'%(n, str(p.shape), str(p.requires_grad)))

    with open('_log/args_train.txt', 'w') as F:
        for key in vars(training_args): F.write('%s: %s\n'%(str(key), str(vars(training_args)[key])))
    # end for MGIE

    data_module = make_supervised_data_module(tokenizer=tokenizer,
                                              data_args=data_args)
    trainer = LLaVATrainer(model=model,
                    tokenizer=tokenizer,
                    args=training_args,
                    **data_module)

    if list(pathlib.Path(training_args.output_dir).glob("checkpoint-*")):
        trainer.train(resume_from_checkpoint=True)
    else:
        trainer.train()
    trainer.save_state()

    if training_args.lora_enable:
        state_dict = get_peft_state_maybe_zero_3(
            model.named_parameters(), training_args.lora_bias
        )
        non_lora_state_dict = get_peft_state_non_lora_maybe_zero_3(
            model.named_parameters()
        )
        if training_args.local_rank == 0 or training_args.local_rank == -1:
            model.config.save_pretrained(training_args.output_dir)
            model.save_pretrained(training_args.output_dir, state_dict=state_dict)
            torch.save(non_lora_state_dict, os.path.join(training_args.output_dir, 'non_lora_trainables.bin'))
    else:
        safe_save_model_for_hf_trainer(trainer=trainer,
                                       output_dir=training_args.output_dir)


if __name__ == "__main__":
    train()