llava_llama3_8b_full_CLIP_lora训练显存占用异常过大

[Yanllan]

torch.cuda.OutOfMemoryError: CUDA out of memory. Tried to allocate 30.10 GiB. GPU 0 has a total capacty of 79.14 GiB of which 1.72 GiB is free. Including non-PyTorch memory, this process has 77.39 GiB memory in use. Of the allocated memory 76.34 GiB is allocated by PyTorch, and 332.09 MiB is reserved by PyTorch but unallocated. If reserved but unallocated memory is large try setting max_split_size_mb to avoid fragmentation. See documentation for Memory Management and PYTORCH_CUDA_ALLOC_CONF

I successfully ran llava_llama3_8b_full_CLIP_lora training about two months ago. Now when I re-ran the code at that time, the video memory was out of memory, and it seems to require more than 110GB of video memory. What might be the cause? How to solve it? The configuration code is as follows:

Copyright (c) OpenMMLab. All rights reserved.

from mmengine.hooks import (CheckpointHook, DistSamplerSeedHook, IterTimerHook, LoggerHook, ParamSchedulerHook) from mmengine.optim import AmpOptimWrapper, CosineAnnealingLR, LinearLR from peft import LoraConfig from torch.optim import AdamW from transformers import (AutoModelForCausalLM, AutoTokenizer, CLIPImageProcessor, CLIPVisionModel)

from xtuner.dataset import ConcatDataset, LLaVADataset from xtuner.dataset.collate_fns import default_collate_fn from xtuner.dataset.map_fns import llava_map_fn, template_map_fn_factory from xtuner.dataset.samplers import LengthGroupedSampler from xtuner.engine.hooks import DatasetInfoHook, EvaluateChatHook from xtuner.engine.runner import TrainLoop from xtuner.model import LLaVAModel from xtuner.utils import PROMPT_TEMPLATE import torch #######################################################################

PART 1 Settings

#######################################################################

Model

llm_name_or_path = './llama3.1-8b-instruct' visual_encoder_name_or_path = './clip-vit-large-patch14-336'

Specify the pretrained pth

pretrained_pth = './work_dirs/llava_llama3_8b_instruct_clip_vit_large_p14_336_e1_gpu8_sharegpt4v_pretrain/iter_24000.pth' # noqa: E501

Data

data_root = ''

sharegpt4v_caption_data_path = data_root + './finetune.json' # noqa: E501 sharegpt4v_caption_image_folder = ''

prompt_template = PROMPT_TEMPLATE.llama3_chat max_length = int(4096 - (336 / 14)**2)

Scheduler & Optimizer

batch_size = 1 # per_device accumulative_counts = 1 dataloader_num_workers = 4 max_epochs = 1 optim_type = AdamW lr = 5e-6 betas = (0.9, 0.999) weight_decay = 0 max_norm = 1 # grad clip warmup_ratio = 0.03

Save

save_steps = 300 save_total_limit = 3 # Maximum checkpoints to keep (-1 means unlimited)

Evaluate the generation performance during the training

evaluation_freq = 300 SYSTEM = '' evaluation_images = './synpic21776.jpg' evaluation_inputs = ['照片中是什么症状？', 'Please describe this picture']

#######################################################################

PART 2 Model & Tokenizer & Image Processor

####################################################################### tokenizer = dict( type=AutoTokenizer.from_pretrained, pretrained_model_name_or_path=llm_name_or_path, trust_remote_code=True, padding_side='right')

image_processor = dict( type=CLIPImageProcessor.from_pretrained, pretrained_model_name_or_path=visual_encoder_name_or_path, trust_remote_code=True)

model = dict( type=LLaVAModel, freeze_llm=False, freeze_visual_encoder=True, pretrained_pth=pretrained_pth, llm=dict( type=AutoModelForCausalLM.from_pretrained, pretrained_model_name_or_path=llm_name_or_path,

torch_dtype=torch.float16,

    trust_remote_code=True),
visual_encoder=dict(

    type=CLIPVisionModel.from_pretrained,
    pretrained_model_name_or_path=visual_encoder_name_or_path),
visual_encoder_lora=dict(
    type=LoraConfig, r=64, lora_alpha=16, lora_dropout=0.05, bias='none'))

#######################################################################

PART 3 Dataset & Dataloader

####################################################################### sharegpt4v_caption_dataset = dict( type=LLaVADataset, data_path=sharegpt4v_caption_data_path, image_folder=sharegpt4v_caption_image_folder, tokenizer=tokenizer, image_processor=image_processor, dataset_map_fn=llava_map_fn, template_map_fn=dict( type=template_map_fn_factory, template=prompt_template), max_length=max_length, pad_image_to_square=True)

train_dataset = dict( type=ConcatDataset, datasets=[ sharegpt4v_caption_dataset, ])

train_dataloader = dict( batch_size=batch_size, num_workers=dataloader_num_workers, pin_memory=True, dataset=train_dataset, sampler=dict( type=LengthGroupedSampler, length_property='modality_length', per_device_batch_size=batch_size * accumulative_counts), collate_fn=dict(type=default_collate_fn))

#######################################################################

PART 4 Scheduler & Optimizer

#######################################################################

optimizer

optim_wrapper = dict( type=AmpOptimWrapper, optimizer=dict( type=optim_type, lr=lr, betas=betas, weight_decay=weight_decay), clip_grad=dict(max_norm=max_norm, error_if_nonfinite=False), accumulative_counts=accumulative_counts, loss_scale='dynamic', dtype='float16')

learning policy

More information: https://github.com/open-mmlab/mmengine/blob/main/docs/en/tutorials/param_scheduler.md # noqa: E501

param_scheduler = [ dict( type=LinearLR, start_factor=1e-5, by_epoch=True, begin=0, end=warmup_ratio max_epochs, convert_to_iter_based=True), dict( type=CosineAnnealingLR, eta_min=0.0, by_epoch=True, begin=warmup_ratio max_epochs, end=max_epochs, convert_to_iter_based=True) ]

train, val, test setting

train_cfg = dict(type=TrainLoop, max_epochs=max_epochs)

#######################################################################

PART 5 Runtime

#######################################################################

Log the dialogue periodically during the training process, optional

custom_hooks = [ dict(type=DatasetInfoHook, tokenizer=tokenizer), dict( type=EvaluateChatHook, tokenizer=tokenizer, image_processor=image_processor, every_n_iters=evaluation_freq, evaluation_inputs=evaluation_inputs, evaluation_images=evaluation_images, system=SYSTEM, prompt_template=prompt_template) ]

configure default hooks

default_hooks = dict(

record the time of every iteration.

timer=dict(type=IterTimerHook),
# print log every 10 iterations.
logger=dict(type=LoggerHook, log_metric_by_epoch=False, interval=10),
# enable the parameter scheduler.
param_scheduler=dict(type=ParamSchedulerHook),
# save checkpoint per `save_steps`.
checkpoint=dict(
    type=CheckpointHook,
    by_epoch=False,
    interval=save_steps,
    max_keep_ckpts=save_total_limit),
# set sampler seed in distributed evrionment.
sampler_seed=dict(type=DistSamplerSeedHook),

)

configure environment

env_cfg = dict(

whether to enable cudnn benchmark

cudnn_benchmark=False,
# set multi process parameters
mp_cfg=dict(mp_start_method='fork', opencv_num_threads=0),
# set distributed parameters
dist_cfg=dict(backend='nccl'),

)

set visualizer

visualizer = None

set log level

log_level = 'INFO'

load from which checkpoint

load_from = None

whether to resume training from the loaded checkpoint

resume = False

Defaults to use random seed and disable deterministic

randomness = dict(seed=None, deterministic=False)

set log processor

log_processor = dict(by_epoch=False)