cap_strange / 3_8b_image /xtuner_config.py
wchai's picture
Upload 3_8b_image/xtuner_config.py with huggingface_hub
c7b2255 verified
SYSTEM = ''
accumulative_counts = 96
allava_cl_data_path = '/data/wenhao/projects/xtuner/data/ALLaVA-4V/ALLaVA-Caption-LAION-4V.jsonl'
allava_cl_dataset = dict(
dataset_map_fn='xtuner.dataset.map_fns.cambrian_map_fn',
image_folder='/data/wenhao/projects/xtuner/data/ALLaVA-4V',
image_processor=dict(
crop_size=378,
pretrained_model_name_or_path=
'laion/CLIP-ViT-bigG-14-laion2B-39B-b160k',
size=378,
trust_remote_code=True,
type='transformers.CLIPImageProcessor.from_pretrained'),
max_length=4096,
offline_processed_text_folder=
'/data/wenhao/projects/xtuner/data/ALLaVA-4V/pre_token_cl_llama31',
pad_image_to_square=True,
template_map_fn=dict(
template='xtuner.utils.PROMPT_TEMPLATE.llama3_chat',
type='xtuner.dataset.map_fns.template_map_fn_factory'),
type='xtuner.dataset.CambrianDataset')
allava_cl_image_folder = '/data/wenhao/projects/xtuner/data/ALLaVA-4V'
allava_cv_data_path = '/data/wenhao/projects/xtuner/data/ALLaVA-4V/ALLaVA-Caption-VFLAN-4V.jsonl'
allava_cv_dataset = dict(
dataset_map_fn='xtuner.dataset.map_fns.cambrian_map_fn',
image_folder='/data/wenhao/projects/xtuner/data/ALLaVA-4V',
image_processor=dict(
crop_size=378,
pretrained_model_name_or_path=
'laion/CLIP-ViT-bigG-14-laion2B-39B-b160k',
size=378,
trust_remote_code=True,
type='transformers.CLIPImageProcessor.from_pretrained'),
max_length=4096,
offline_processed_text_folder=
'/data/wenhao/projects/xtuner/data/ALLaVA-4V/pre_token_cv_llama31',
pad_image_to_square=True,
template_map_fn=dict(
template='xtuner.utils.PROMPT_TEMPLATE.llama3_chat',
type='xtuner.dataset.map_fns.template_map_fn_factory'),
type='xtuner.dataset.CambrianDataset')
allava_data_root = '/data/wenhao/projects/xtuner/data/ALLaVA-4V'
allava_il_data_path = '/data/wenhao/projects/xtuner/data/ALLaVA-4V/ALLaVA-Instruct-LAION-4V.jsonl'
allava_il_dataset = dict(
dataset_map_fn='xtuner.dataset.map_fns.cambrian_map_fn',
image_folder='/data/wenhao/projects/xtuner/data/ALLaVA-4V',
image_processor=dict(
crop_size=378,
pretrained_model_name_or_path=
'laion/CLIP-ViT-bigG-14-laion2B-39B-b160k',
size=378,
trust_remote_code=True,
type='transformers.CLIPImageProcessor.from_pretrained'),
max_length=4096,
offline_processed_text_folder=
'/data/wenhao/projects/xtuner/data/ALLaVA-4V/pre_token_il_llama31',
pad_image_to_square=True,
template_map_fn=dict(
template='xtuner.utils.PROMPT_TEMPLATE.llama3_chat',
type='xtuner.dataset.map_fns.template_map_fn_factory'),
type='xtuner.dataset.CambrianDataset')
allava_il_image_folder = '/data/wenhao/projects/xtuner/data/ALLaVA-4V'
allava_image_folder = '/data/wenhao/projects/xtuner/data/ALLaVA-4V'
allava_iv_data_path = '/data/wenhao/projects/xtuner/data/ALLaVA-4V/ALLaVA-Instruct-VFLAN-4V.jsonl'
allava_iv_dataset = dict(
dataset_map_fn='xtuner.dataset.map_fns.cambrian_map_fn',
image_folder='/data/wenhao/projects/xtuner/data/ALLaVA-4V',
image_processor=dict(
crop_size=378,
pretrained_model_name_or_path=
'laion/CLIP-ViT-bigG-14-laion2B-39B-b160k',
size=378,
trust_remote_code=True,
type='transformers.CLIPImageProcessor.from_pretrained'),
max_length=4096,
offline_processed_text_folder=
'/data/wenhao/projects/xtuner/data/ALLaVA-4V/pre_token_iv_llama31',
pad_image_to_square=True,
template_map_fn=dict(
template='xtuner.utils.PROMPT_TEMPLATE.llama3_chat',
type='xtuner.dataset.map_fns.template_map_fn_factory'),
type='xtuner.dataset.CambrianDataset')
batch_size = 1
betas = (
0.9,
0.999,
)
cambrian_data_path = '/data/wenhao/projects/xtuner/data/Cambrian-10M/jsons/Cambrian7M_withsystemprompt.jsonl'
cambrian_data_root = '/data/wenhao/projects/xtuner/data/Cambrian-10M/'
cambrian_dataset = dict(
dataset_map_fn='xtuner.dataset.map_fns.cambrian_map_fn',
image_folder='/data/wenhao/projects/xtuner/data/Cambrian-10M/',
image_processor=dict(
crop_size=378,
pretrained_model_name_or_path=
'laion/CLIP-ViT-bigG-14-laion2B-39B-b160k',
size=378,
trust_remote_code=True,
type='transformers.CLIPImageProcessor.from_pretrained'),
max_length=4096,
offline_processed_text_folder=
'/data/wenhao/projects/xtuner/data/Cambrian-10M/pre_token_llama3',
pad_image_to_square=True,
template_map_fn=dict(
template='xtuner.utils.PROMPT_TEMPLATE.llama3_chat',
type='xtuner.dataset.map_fns.template_map_fn_factory'),
type='xtuner.dataset.CambrianDataset')
cambrian_image_folder = '/data/wenhao/projects/xtuner/data/Cambrian-10M/'
cambrian_processed_text_folder = '/data/wenhao/projects/xtuner/data/Cambrian-10M/pre_token_llama3'
custom_hooks = [
dict(
tokenizer=dict(
padding_side='right',
pretrained_model_name_or_path='meta-llama/Meta-Llama-3-8B-Instruct',
trust_remote_code=True,
type='transformers.AutoTokenizer.from_pretrained'),
type='xtuner.engine.DatasetInfoHook'),
dict(
evaluation_images='https://llava-vl.github.io/static/images/view.jpg',
evaluation_inputs=[
'请描述一下这张照片',
'Please describe this picture',
],
every_n_iters=100,
image_processor=dict(
crop_size=378,
pretrained_model_name_or_path=
'laion/CLIP-ViT-bigG-14-laion2B-39B-b160k',
size=378,
trust_remote_code=True,
type='transformers.CLIPImageProcessor.from_pretrained'),
prompt_template='xtuner.utils.PROMPT_TEMPLATE.llama3_chat',
system='',
tokenizer=dict(
padding_side='right',
pretrained_model_name_or_path='meta-llama/Meta-Llama-3-8B-Instruct',
trust_remote_code=True,
type='transformers.AutoTokenizer.from_pretrained'),
type='xtuner.engine.EvaluateChatHook'),
]
dataloader_num_workers = 0
default_hooks = dict(
checkpoint=dict(
by_epoch=False,
interval=200,
max_keep_ckpts=2,
type='mmengine.hooks.CheckpointHook'),
logger=dict(interval=10, type='mmengine.hooks.LoggerHook'),
param_scheduler=dict(type='mmengine.hooks.ParamSchedulerHook'),
sampler_seed=dict(type='mmengine.hooks.DistSamplerSeedHook'),
timer=dict(type='mmengine.hooks.IterTimerHook'))
dense_data_path = '/data/wenhao/projects/xtuner/data/DenseFusion-1M/DenseFusion-1M/DenseFusion-1M-instruct.jsonl'
dense_data_root = '/data/wenhao/projects/xtuner/data/DenseFusion-1M/'
dense_dataset = dict(
dataset_map_fn='xtuner.dataset.map_fns.cambrian_map_fn',
image_folder='/data/wenhao/projects/xtuner/data/DenseFusion-1M/1M_data',
image_processor=dict(
crop_size=378,
pretrained_model_name_or_path=
'laion/CLIP-ViT-bigG-14-laion2B-39B-b160k',
size=378,
trust_remote_code=True,
type='transformers.CLIPImageProcessor.from_pretrained'),
max_length=4096,
offline_processed_text_folder=
'/data/wenhao/projects/xtuner/data/DenseFusion-1M/pre_token_llama3',
pad_image_to_square=True,
template_map_fn=dict(
template='xtuner.utils.PROMPT_TEMPLATE.llama3_chat',
type='xtuner.dataset.map_fns.template_map_fn_factory'),
type='xtuner.dataset.CambrianDataset')
dense_image_folder = '/data/wenhao/projects/xtuner/data/DenseFusion-1M/1M_data'
dense_processed_text_folder = '/data/wenhao/projects/xtuner/data/DenseFusion-1M/pre_token_llama3'
env_cfg = dict(
cudnn_benchmark=False,
dist_cfg=dict(backend='nccl'),
mp_cfg=dict(mp_start_method='fork', opencv_num_threads=0))
evaluation_freq = 100
evaluation_images = 'https://llava-vl.github.io/static/images/view.jpg'
evaluation_inputs = [
'请描述一下这张照片',
'Please describe this picture',
]
evol_data_path = '/data/wenhao/projects/xtuner/data/ALLaVA-4V/allava_text/Evol-Instruct-GPT4-Turbo-143K.jsonl'
evol_data_root = '/data/wenhao/projects/xtuner/data/ALLaVA-4V/allava_text/'
evol_dataset = dict(
dataset_map_fn='xtuner.dataset.map_fns.cambrian_map_fn',
image_folder=
'/data/wenhao/projects/xtuner/data/ALLaVA-4V/allava_text/images',
image_processor=dict(
crop_size=378,
pretrained_model_name_or_path=
'laion/CLIP-ViT-bigG-14-laion2B-39B-b160k',
size=378,
trust_remote_code=True,
type='transformers.CLIPImageProcessor.from_pretrained'),
max_length=4096,
offline_processed_text_folder=
'/data/wenhao/projects/xtuner/data/ALLaVA-4V/pre_token_evol_llama31',
pad_image_to_square=True,
template_map_fn=dict(
template='xtuner.utils.PROMPT_TEMPLATE.llama3_chat',
type='xtuner.dataset.map_fns.template_map_fn_factory'),
type='xtuner.dataset.CambrianDataset')
evol_image_folder = '/data/wenhao/projects/xtuner/data/ALLaVA-4V/allava_text/images'
face_data_path = '/data/wenhao/projects/xtuner/data/FaceCaption-15M/FaceCaption-100K.jsonl'
face_data_root = '/data/wenhao/projects/xtuner/data/FaceCaption-15M/'
face_dataset = dict(
dataset_map_fn='xtuner.dataset.map_fns.cambrian_map_fn',
image_folder='/data/wenhao/projects/xtuner/data/FaceCaption-15M/full_data',
image_processor=dict(
crop_size=378,
pretrained_model_name_or_path=
'laion/CLIP-ViT-bigG-14-laion2B-39B-b160k',
size=378,
trust_remote_code=True,
type='transformers.CLIPImageProcessor.from_pretrained'),
max_length=4096,
offline_processed_text_folder=
'/data/wenhao/projects/xtuner/data/FaceCaption-15M/pre_token_llama3',
pad_image_to_square=True,
template_map_fn=dict(
template='xtuner.utils.PROMPT_TEMPLATE.llama3_chat',
type='xtuner.dataset.map_fns.template_map_fn_factory'),
type='xtuner.dataset.CambrianDataset')
face_image_folder = '/data/wenhao/projects/xtuner/data/FaceCaption-15M/full_data'
face_processed_text_folder = '/data/wenhao/projects/xtuner/data/FaceCaption-15M/pre_token_llama3'
image_processor = dict(
crop_size=378,
pretrained_model_name_or_path='laion/CLIP-ViT-bigG-14-laion2B-39B-b160k',
size=378,
trust_remote_code=True,
type='transformers.CLIPImageProcessor.from_pretrained')
launcher = 'slurm'
llava_mix_data_path = '/data/wenhao/projects/xtuner/data/LLaVA-Instruct-150K/llava_v1_5_mix665k.jsonl'
llava_mix_data_root = '/data/wenhao/projects/xtuner/data/LLaVA-Instruct-150K/'
llava_mix_dataset = dict(
dataset_map_fn='xtuner.dataset.map_fns.cambrian_map_fn',
image_folder='/data/wenhao/projects/xtuner/data/LLaVA-Instruct-150K/images',
image_processor=dict(
crop_size=378,
pretrained_model_name_or_path=
'laion/CLIP-ViT-bigG-14-laion2B-39B-b160k',
size=378,
trust_remote_code=True,
type='transformers.CLIPImageProcessor.from_pretrained'),
max_length=4096,
offline_processed_text_folder=
'/data/wenhao/projects/xtuner/data/LLaVA-Instruct-150K/pre_token_llama31',
pad_image_to_square=True,
template_map_fn=dict(
template='xtuner.utils.PROMPT_TEMPLATE.llama3_chat',
type='xtuner.dataset.map_fns.template_map_fn_factory'),
type='xtuner.dataset.CambrianDataset')
llava_mix_image_folder = '/data/wenhao/projects/xtuner/data/LLaVA-Instruct-150K/images'
llavanext_data_path = '/data/wenhao/projects/xtuner/data/LLaVA-NeXT-Data/llava_next.jsonl'
llavanext_data_root = '/data/wenhao/projects/xtuner/data/LLaVA-NeXT-Data/'
llavanext_dataset = dict(
dataset_map_fn='xtuner.dataset.map_fns.cambrian_map_fn',
image_folder='/data/wenhao/projects/xtuner/data/LLaVA-NeXT-Data/images',
image_processor=dict(
crop_size=378,
pretrained_model_name_or_path=
'laion/CLIP-ViT-bigG-14-laion2B-39B-b160k',
size=378,
trust_remote_code=True,
type='transformers.CLIPImageProcessor.from_pretrained'),
max_length=4096,
offline_processed_text_folder=
'/data/wenhao/projects/xtuner/data/LLaVA-NeXT-Data/pre_token_llama31',
pad_image_to_square=True,
template_map_fn=dict(
template='xtuner.utils.PROMPT_TEMPLATE.llama3_chat',
type='xtuner.dataset.map_fns.template_map_fn_factory'),
type='xtuner.dataset.CambrianDataset')
llavanext_image_folder = '/data/wenhao/projects/xtuner/data/LLaVA-NeXT-Data/images'
llm_name_or_path = 'meta-llama/Meta-Llama-3-8B-Instruct'
load_from = 'work_dirs/new_image/iter_449600.pth'
log_level = 'INFO'
lr = 1e-05
m4_data_path = '/data/wenhao/projects/xtuner/data/M4-Instruct-Data/m4_instruct_image.jsonl'
m4_data_root = '/data/wenhao/projects/xtuner/data/M4-Instruct-Data/'
m4_dataset = dict(
dataset_map_fn='xtuner.dataset.map_fns.cambrian_map_fn',
image_folder='/data/wenhao/projects/xtuner/data/M4-Instruct-Data/',
image_processor=dict(
crop_size=378,
pretrained_model_name_or_path=
'laion/CLIP-ViT-bigG-14-laion2B-39B-b160k',
size=378,
trust_remote_code=True,
type='transformers.CLIPImageProcessor.from_pretrained'),
max_length=4096,
offline_processed_text_folder=
'/data/wenhao/projects/xtuner/data/M4-Instruct-Data/pre_token_llama31',
pad_image_to_square=True,
template_map_fn=dict(
template='xtuner.utils.PROMPT_TEMPLATE.llama3_chat',
type='xtuner.dataset.map_fns.template_map_fn_factory'),
type='xtuner.dataset.CambrianDataset')
m4_image_folder = '/data/wenhao/projects/xtuner/data/M4-Instruct-Data/'
max_epochs = 1
max_length = 4096
max_norm = 1
model = dict(
freeze_llm=False,
freeze_visual_encoder=False,
llm=dict(
pretrained_model_name_or_path='meta-llama/Meta-Llama-3-8B-Instruct',
torch_dtype='torch.float16',
trust_remote_code=True,
type='transformers.AutoModelForCausalLM.from_pretrained'),
llm_lora=dict(
bias='none',
lora_alpha=256,
lora_dropout=0.05,
r=512,
task_type='CAUSAL_LM',
type='peft.LoraConfig'),
pretrained_pth=
'/data/wenhao/projects/xtuner/work_dirs/final_new_v/projector',
type='xtuner.model.PikaModel',
visual_encoder=dict(
pretrained_model_name_or_path=
'/data/wenhao/projects/xtuner/work_dirs/final_new_v/visual_encoder',
type='xtuner.model.pika.PikaSigVidEncoder.from_pretrained',
visual_token_merge_ratio=0.1))
optim_type = 'torch.optim.AdamW'
optim_wrapper = dict(
optimizer=dict(
betas=(
0.9,
0.999,
),
lr=1e-05,
type='torch.optim.AdamW',
weight_decay=0),
type='DeepSpeedOptimWrapper')
param_scheduler = [
dict(
begin=0,
by_epoch=True,
convert_to_iter_based=True,
end=0.03,
start_factor=1e-05,
type='mmengine.optim.LinearLR'),
dict(
T_max=1,
begin=0.03,
by_epoch=True,
convert_to_iter_based=True,
eta_min=0.0,
type='mmengine.optim.CosineAnnealingLR'),
]
pretrained_pth = '/data/wenhao/projects/xtuner/work_dirs/final_new_v/projector'
prompt_template = 'xtuner.utils.PROMPT_TEMPLATE.llama3_chat'
randomness = dict(deterministic=False, seed=1416244085)
resume = True
runner_type = 'FlexibleRunner'
save_steps = 200
save_total_limit = 2
sharegpt4v_data_path = '/data/wenhao/projects/xtuner/data/ShareGPT4V/sharegpt4v_instruct_gpt4-vision_cap100k.jsonl'
sharegpt4v_data_root = '/data/wenhao/projects/xtuner/data/ShareGPT4V'
sharegpt4v_dataset = dict(
dataset_map_fn='xtuner.dataset.map_fns.cambrian_map_fn',
image_folder='/data/wenhao/projects/xtuner/data',
image_processor=dict(
crop_size=378,
pretrained_model_name_or_path=
'laion/CLIP-ViT-bigG-14-laion2B-39B-b160k',
size=378,
trust_remote_code=True,
type='transformers.CLIPImageProcessor.from_pretrained'),
max_length=4096,
offline_processed_text_folder=
'/data/wenhao/projects/xtuner/data/ShareGPT4V/pre_token_sharegpt4v_llama31',
pad_image_to_square=True,
template_map_fn=dict(
template='xtuner.utils.PROMPT_TEMPLATE.llama3_chat',
type='xtuner.dataset.map_fns.template_map_fn_factory'),
type='xtuner.dataset.CambrianDataset')
sharegpt4v_image_folder = '/data/wenhao/projects/xtuner/data'
size = 378
strategy = dict(
config=dict(
bf16=dict(enabled=True),
fp16=dict(enabled=False, initial_scale_power=16),
gradient_accumulation_steps='auto',
gradient_clipping='auto',
train_micro_batch_size_per_gpu='auto',
zero_allow_untested_optimizer=True,
zero_force_ds_cpu_optimizer=False,
zero_optimization=dict(overlap_comm=True, stage=2)),
exclude_frozen_parameters=True,
gradient_accumulation_steps=96,
gradient_clipping=1,
sequence_parallel_size=1,
train_micro_batch_size_per_gpu=1,
type='xtuner.engine.DeepSpeedStrategy')
tokenizer = dict(
padding_side='right',
pretrained_model_name_or_path='meta-llama/Meta-Llama-3-8B-Instruct',
trust_remote_code=True,
type='transformers.AutoTokenizer.from_pretrained')
train_cfg = dict(by_epoch=True, max_epochs=1, val_interval=1)
train_dataloader = dict(
batch_size=1,
collate_fn=dict(type='xtuner.dataset.collate_fns.default_collate_fn'),
dataset=dict(
datasets=[
dict(
dataset_map_fn='xtuner.dataset.map_fns.cambrian_map_fn',
image_folder=
'/data/wenhao/projects/xtuner/data/M4-Instruct-Data/',
image_processor=dict(
crop_size=378,
pretrained_model_name_or_path=
'laion/CLIP-ViT-bigG-14-laion2B-39B-b160k',
size=378,
trust_remote_code=True,
type='transformers.CLIPImageProcessor.from_pretrained'),
max_length=4096,
offline_processed_text_folder=
'/data/wenhao/projects/xtuner/data/M4-Instruct-Data/pre_token_llama31',
pad_image_to_square=True,
template_map_fn=dict(
template='xtuner.utils.PROMPT_TEMPLATE.llama3_chat',
type='xtuner.dataset.map_fns.template_map_fn_factory'),
type='xtuner.dataset.CambrianDataset'),
dict(
dataset_map_fn='xtuner.dataset.map_fns.cambrian_map_fn',
image_folder=
'/data/wenhao/projects/xtuner/data/LLaVA-NeXT-Data/images',
image_processor=dict(
crop_size=378,
pretrained_model_name_or_path=
'laion/CLIP-ViT-bigG-14-laion2B-39B-b160k',
size=378,
trust_remote_code=True,
type='transformers.CLIPImageProcessor.from_pretrained'),
max_length=4096,
offline_processed_text_folder=
'/data/wenhao/projects/xtuner/data/LLaVA-NeXT-Data/pre_token_llama31',
pad_image_to_square=True,
template_map_fn=dict(
template='xtuner.utils.PROMPT_TEMPLATE.llama3_chat',
type='xtuner.dataset.map_fns.template_map_fn_factory'),
type='xtuner.dataset.CambrianDataset'),
dict(
dataset_map_fn='xtuner.dataset.map_fns.cambrian_map_fn',
image_folder='/data/wenhao/projects/xtuner/data/ALLaVA-4V',
image_processor=dict(
crop_size=378,
pretrained_model_name_or_path=
'laion/CLIP-ViT-bigG-14-laion2B-39B-b160k',
size=378,
trust_remote_code=True,
type='transformers.CLIPImageProcessor.from_pretrained'),
max_length=4096,
offline_processed_text_folder=
'/data/wenhao/projects/xtuner/data/ALLaVA-4V/pre_token_cl_llama31',
pad_image_to_square=True,
template_map_fn=dict(
template='xtuner.utils.PROMPT_TEMPLATE.llama3_chat',
type='xtuner.dataset.map_fns.template_map_fn_factory'),
type='xtuner.dataset.CambrianDataset'),
dict(
dataset_map_fn='xtuner.dataset.map_fns.cambrian_map_fn',
image_folder='/data/wenhao/projects/xtuner/data/ALLaVA-4V',
image_processor=dict(
crop_size=378,
pretrained_model_name_or_path=
'laion/CLIP-ViT-bigG-14-laion2B-39B-b160k',
size=378,
trust_remote_code=True,
type='transformers.CLIPImageProcessor.from_pretrained'),
max_length=4096,
offline_processed_text_folder=
'/data/wenhao/projects/xtuner/data/ALLaVA-4V/pre_token_cv_llama31',
pad_image_to_square=True,
template_map_fn=dict(
template='xtuner.utils.PROMPT_TEMPLATE.llama3_chat',
type='xtuner.dataset.map_fns.template_map_fn_factory'),
type='xtuner.dataset.CambrianDataset'),
dict(
dataset_map_fn='xtuner.dataset.map_fns.cambrian_map_fn',
image_folder='/data/wenhao/projects/xtuner/data/ALLaVA-4V',
image_processor=dict(
crop_size=378,
pretrained_model_name_or_path=
'laion/CLIP-ViT-bigG-14-laion2B-39B-b160k',
size=378,
trust_remote_code=True,
type='transformers.CLIPImageProcessor.from_pretrained'),
max_length=4096,
offline_processed_text_folder=
'/data/wenhao/projects/xtuner/data/ALLaVA-4V/pre_token_il_llama31',
pad_image_to_square=True,
template_map_fn=dict(
template='xtuner.utils.PROMPT_TEMPLATE.llama3_chat',
type='xtuner.dataset.map_fns.template_map_fn_factory'),
type='xtuner.dataset.CambrianDataset'),
dict(
dataset_map_fn='xtuner.dataset.map_fns.cambrian_map_fn',
image_folder='/data/wenhao/projects/xtuner/data/ALLaVA-4V',
image_processor=dict(
crop_size=378,
pretrained_model_name_or_path=
'laion/CLIP-ViT-bigG-14-laion2B-39B-b160k',
size=378,
trust_remote_code=True,
type='transformers.CLIPImageProcessor.from_pretrained'),
max_length=4096,
offline_processed_text_folder=
'/data/wenhao/projects/xtuner/data/ALLaVA-4V/pre_token_iv_llama31',
pad_image_to_square=True,
template_map_fn=dict(
template='xtuner.utils.PROMPT_TEMPLATE.llama3_chat',
type='xtuner.dataset.map_fns.template_map_fn_factory'),
type='xtuner.dataset.CambrianDataset'),
dict(
dataset_map_fn='xtuner.dataset.map_fns.cambrian_map_fn',
image_folder=
'/data/wenhao/projects/xtuner/data/LLaVA-Instruct-150K/images',
image_processor=dict(
crop_size=378,
pretrained_model_name_or_path=
'laion/CLIP-ViT-bigG-14-laion2B-39B-b160k',
size=378,
trust_remote_code=True,
type='transformers.CLIPImageProcessor.from_pretrained'),
max_length=4096,
offline_processed_text_folder=
'/data/wenhao/projects/xtuner/data/LLaVA-Instruct-150K/pre_token_llama31',
pad_image_to_square=True,
template_map_fn=dict(
template='xtuner.utils.PROMPT_TEMPLATE.llama3_chat',
type='xtuner.dataset.map_fns.template_map_fn_factory'),
type='xtuner.dataset.CambrianDataset'),
dict(
dataset_map_fn='xtuner.dataset.map_fns.cambrian_map_fn',
image_folder=
'/data/wenhao/projects/xtuner/data/ALLaVA-4V/allava_text/images',
image_processor=dict(
crop_size=378,
pretrained_model_name_or_path=
'laion/CLIP-ViT-bigG-14-laion2B-39B-b160k',
size=378,
trust_remote_code=True,
type='transformers.CLIPImageProcessor.from_pretrained'),
max_length=4096,
offline_processed_text_folder=
'/data/wenhao/projects/xtuner/data/ALLaVA-4V/pre_token_evol_llama31',
pad_image_to_square=True,
template_map_fn=dict(
template='xtuner.utils.PROMPT_TEMPLATE.llama3_chat',
type='xtuner.dataset.map_fns.template_map_fn_factory'),
type='xtuner.dataset.CambrianDataset'),
dict(
dataset_map_fn='xtuner.dataset.map_fns.cambrian_map_fn',
image_folder='/data/wenhao/projects/xtuner/data',
image_processor=dict(
crop_size=378,
pretrained_model_name_or_path=
'laion/CLIP-ViT-bigG-14-laion2B-39B-b160k',
size=378,
trust_remote_code=True,
type='transformers.CLIPImageProcessor.from_pretrained'),
max_length=4096,
offline_processed_text_folder=
'/data/wenhao/projects/xtuner/data/ShareGPT4V/pre_token_sharegpt4v_llama31',
pad_image_to_square=True,
template_map_fn=dict(
template='xtuner.utils.PROMPT_TEMPLATE.llama3_chat',
type='xtuner.dataset.map_fns.template_map_fn_factory'),
type='xtuner.dataset.CambrianDataset'),
dict(
dataset_map_fn='xtuner.dataset.map_fns.cambrian_map_fn',
image_folder=
'/data/wenhao/projects/xtuner/data/FaceCaption-15M/full_data',
image_processor=dict(
crop_size=378,
pretrained_model_name_or_path=
'laion/CLIP-ViT-bigG-14-laion2B-39B-b160k',
size=378,
trust_remote_code=True,
type='transformers.CLIPImageProcessor.from_pretrained'),
max_length=4096,
offline_processed_text_folder=
'/data/wenhao/projects/xtuner/data/FaceCaption-15M/pre_token_llama3',
pad_image_to_square=True,
template_map_fn=dict(
template='xtuner.utils.PROMPT_TEMPLATE.llama3_chat',
type='xtuner.dataset.map_fns.template_map_fn_factory'),
type='xtuner.dataset.CambrianDataset'),
],
type='xtuner.dataset.ConcatDataset'),
num_workers=0,
sampler=dict(shuffle=True, type='mmengine.dataset.DefaultSampler'))
train_dataset = dict(
datasets=[
dict(
dataset_map_fn='xtuner.dataset.map_fns.cambrian_map_fn',
image_folder='/data/wenhao/projects/xtuner/data/M4-Instruct-Data/',
image_processor=dict(
crop_size=378,
pretrained_model_name_or_path=
'laion/CLIP-ViT-bigG-14-laion2B-39B-b160k',
size=378,
trust_remote_code=True,
type='transformers.CLIPImageProcessor.from_pretrained'),
max_length=4096,
offline_processed_text_folder=
'/data/wenhao/projects/xtuner/data/M4-Instruct-Data/pre_token_llama31',
pad_image_to_square=True,
template_map_fn=dict(
template='xtuner.utils.PROMPT_TEMPLATE.llama3_chat',
type='xtuner.dataset.map_fns.template_map_fn_factory'),
type='xtuner.dataset.CambrianDataset'),
dict(
dataset_map_fn='xtuner.dataset.map_fns.cambrian_map_fn',
image_folder=
'/data/wenhao/projects/xtuner/data/LLaVA-NeXT-Data/images',
image_processor=dict(
crop_size=378,
pretrained_model_name_or_path=
'laion/CLIP-ViT-bigG-14-laion2B-39B-b160k',
size=378,
trust_remote_code=True,
type='transformers.CLIPImageProcessor.from_pretrained'),
max_length=4096,
offline_processed_text_folder=
'/data/wenhao/projects/xtuner/data/LLaVA-NeXT-Data/pre_token_llama31',
pad_image_to_square=True,
template_map_fn=dict(
template='xtuner.utils.PROMPT_TEMPLATE.llama3_chat',
type='xtuner.dataset.map_fns.template_map_fn_factory'),
type='xtuner.dataset.CambrianDataset'),
dict(
dataset_map_fn='xtuner.dataset.map_fns.cambrian_map_fn',
image_folder='/data/wenhao/projects/xtuner/data/ALLaVA-4V',
image_processor=dict(
crop_size=378,
pretrained_model_name_or_path=
'laion/CLIP-ViT-bigG-14-laion2B-39B-b160k',
size=378,
trust_remote_code=True,
type='transformers.CLIPImageProcessor.from_pretrained'),
max_length=4096,
offline_processed_text_folder=
'/data/wenhao/projects/xtuner/data/ALLaVA-4V/pre_token_cl_llama31',
pad_image_to_square=True,
template_map_fn=dict(
template='xtuner.utils.PROMPT_TEMPLATE.llama3_chat',
type='xtuner.dataset.map_fns.template_map_fn_factory'),
type='xtuner.dataset.CambrianDataset'),
dict(
dataset_map_fn='xtuner.dataset.map_fns.cambrian_map_fn',
image_folder='/data/wenhao/projects/xtuner/data/ALLaVA-4V',
image_processor=dict(
crop_size=378,
pretrained_model_name_or_path=
'laion/CLIP-ViT-bigG-14-laion2B-39B-b160k',
size=378,
trust_remote_code=True,
type='transformers.CLIPImageProcessor.from_pretrained'),
max_length=4096,
offline_processed_text_folder=
'/data/wenhao/projects/xtuner/data/ALLaVA-4V/pre_token_cv_llama31',
pad_image_to_square=True,
template_map_fn=dict(
template='xtuner.utils.PROMPT_TEMPLATE.llama3_chat',
type='xtuner.dataset.map_fns.template_map_fn_factory'),
type='xtuner.dataset.CambrianDataset'),
dict(
dataset_map_fn='xtuner.dataset.map_fns.cambrian_map_fn',
image_folder='/data/wenhao/projects/xtuner/data/ALLaVA-4V',
image_processor=dict(
crop_size=378,
pretrained_model_name_or_path=
'laion/CLIP-ViT-bigG-14-laion2B-39B-b160k',
size=378,
trust_remote_code=True,
type='transformers.CLIPImageProcessor.from_pretrained'),
max_length=4096,
offline_processed_text_folder=
'/data/wenhao/projects/xtuner/data/ALLaVA-4V/pre_token_il_llama31',
pad_image_to_square=True,
template_map_fn=dict(
template='xtuner.utils.PROMPT_TEMPLATE.llama3_chat',
type='xtuner.dataset.map_fns.template_map_fn_factory'),
type='xtuner.dataset.CambrianDataset'),
dict(
dataset_map_fn='xtuner.dataset.map_fns.cambrian_map_fn',
image_folder='/data/wenhao/projects/xtuner/data/ALLaVA-4V',
image_processor=dict(
crop_size=378,
pretrained_model_name_or_path=
'laion/CLIP-ViT-bigG-14-laion2B-39B-b160k',
size=378,
trust_remote_code=True,
type='transformers.CLIPImageProcessor.from_pretrained'),
max_length=4096,
offline_processed_text_folder=
'/data/wenhao/projects/xtuner/data/ALLaVA-4V/pre_token_iv_llama31',
pad_image_to_square=True,
template_map_fn=dict(
template='xtuner.utils.PROMPT_TEMPLATE.llama3_chat',
type='xtuner.dataset.map_fns.template_map_fn_factory'),
type='xtuner.dataset.CambrianDataset'),
dict(
dataset_map_fn='xtuner.dataset.map_fns.cambrian_map_fn',
image_folder=
'/data/wenhao/projects/xtuner/data/LLaVA-Instruct-150K/images',
image_processor=dict(
crop_size=378,
pretrained_model_name_or_path=
'laion/CLIP-ViT-bigG-14-laion2B-39B-b160k',
size=378,
trust_remote_code=True,
type='transformers.CLIPImageProcessor.from_pretrained'),
max_length=4096,
offline_processed_text_folder=
'/data/wenhao/projects/xtuner/data/LLaVA-Instruct-150K/pre_token_llama31',
pad_image_to_square=True,
template_map_fn=dict(
template='xtuner.utils.PROMPT_TEMPLATE.llama3_chat',
type='xtuner.dataset.map_fns.template_map_fn_factory'),
type='xtuner.dataset.CambrianDataset'),
dict(
dataset_map_fn='xtuner.dataset.map_fns.cambrian_map_fn',
image_folder=
'/data/wenhao/projects/xtuner/data/ALLaVA-4V/allava_text/images',
image_processor=dict(
crop_size=378,
pretrained_model_name_or_path=
'laion/CLIP-ViT-bigG-14-laion2B-39B-b160k',
size=378,
trust_remote_code=True,
type='transformers.CLIPImageProcessor.from_pretrained'),
max_length=4096,
offline_processed_text_folder=
'/data/wenhao/projects/xtuner/data/ALLaVA-4V/pre_token_evol_llama31',
pad_image_to_square=True,
template_map_fn=dict(
template='xtuner.utils.PROMPT_TEMPLATE.llama3_chat',
type='xtuner.dataset.map_fns.template_map_fn_factory'),
type='xtuner.dataset.CambrianDataset'),
dict(
dataset_map_fn='xtuner.dataset.map_fns.cambrian_map_fn',
image_folder='/data/wenhao/projects/xtuner/data',
image_processor=dict(
crop_size=378,
pretrained_model_name_or_path=
'laion/CLIP-ViT-bigG-14-laion2B-39B-b160k',
size=378,
trust_remote_code=True,
type='transformers.CLIPImageProcessor.from_pretrained'),
max_length=4096,
offline_processed_text_folder=
'/data/wenhao/projects/xtuner/data/ShareGPT4V/pre_token_sharegpt4v_llama31',
pad_image_to_square=True,
template_map_fn=dict(
template='xtuner.utils.PROMPT_TEMPLATE.llama3_chat',
type='xtuner.dataset.map_fns.template_map_fn_factory'),
type='xtuner.dataset.CambrianDataset'),
dict(
dataset_map_fn='xtuner.dataset.map_fns.cambrian_map_fn',
image_folder=
'/data/wenhao/projects/xtuner/data/FaceCaption-15M/full_data',
image_processor=dict(
crop_size=378,
pretrained_model_name_or_path=
'laion/CLIP-ViT-bigG-14-laion2B-39B-b160k',
size=378,
trust_remote_code=True,
type='transformers.CLIPImageProcessor.from_pretrained'),
max_length=4096,
offline_processed_text_folder=
'/data/wenhao/projects/xtuner/data/FaceCaption-15M/pre_token_llama3',
pad_image_to_square=True,
template_map_fn=dict(
template='xtuner.utils.PROMPT_TEMPLATE.llama3_chat',
type='xtuner.dataset.map_fns.template_map_fn_factory'),
type='xtuner.dataset.CambrianDataset'),
],
type='xtuner.dataset.ConcatDataset')
visual_encoder_name_or_path = '/data/wenhao/projects/xtuner/work_dirs/final_new_v/visual_encoder'
visual_token_merge_ratio = 0.1
visualizer = None
warmup_ratio = 0.03
weight_decay = 0
work_dir = 'work_dirs/new_image'