SYSTEM = '' accumulative_counts = 256 batch_size = 1 betas = ( 0.9, 0.999, ) custom_hooks = [ dict( tokenizer=dict( padding_side='right', pretrained_model_name_or_path= '/mnt/program/wlx_data/experiments/language-1022/rwkv-1b6/iter_745524.hf', trust_remote_code=True, type='transformers.AutoTokenizer.from_pretrained'), type='xtuner.engine.DatasetInfoHook'), dict( evaluation_images= '/mnt/program/wlx_data/projects/aurora-rwkv/assets/view.jpg', evaluation_inputs=[ 'Please describe this picture in detail', 'What should I be aware of when I visit this place?', ], every_n_iters=10000, image_processor=dict( crop_size=384, pretrained_model_name_or_path= 'laion/CLIP-ViT-bigG-14-laion2B-39B-b160k', size=384, trust_remote_code=True, type='transformers.CLIPImageProcessor.from_pretrained'), prompt_template='xtuner.utils.PROMPT_TEMPLATE.vicuna', system='', tokenizer=dict( padding_side='right', pretrained_model_name_or_path= '/mnt/program/wlx_data/experiments/language-1022/rwkv-1b6/iter_745524.hf', trust_remote_code=True, type='transformers.AutoTokenizer.from_pretrained'), type='xtuner.engine.EvaluateChatHook'), ] data_path = '/mnt/program/wlx_data/datasets/lmms-lab/LLaVA-Video-178K/llava_video_frames/30_60_s_youtube_v0_1/converted_jsons/filtered_converted_30_60_s_youtube_v0_1_cap_processed.json' data_root = '/mnt/program/wlx_data/datasets/lmms-lab/LLaVA-Video-178K/llava_video_frames/30_60_s_youtube_v0_1/' dataloader_num_workers = 8 default_hooks = dict( checkpoint=dict( by_epoch=False, interval=10000, 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')) env_cfg = dict( cudnn_benchmark=False, dist_cfg=dict(backend='nccl'), mp_cfg=dict(mp_start_method='fork', opencv_num_threads=0)) evaluation_freq = 10000 evaluation_images = '/mnt/program/wlx_data/projects/aurora-rwkv/assets/view.jpg' evaluation_inputs = [ 'Please describe this picture in detail', 'What should I be aware of when I visit this place?', ] image_folder = '/mnt/program/wlx_data/datasets/lmms-lab/LLaVA-Video-178K/llava_video_frames/30_60_s_youtube_v0_1/' image_processor = dict( crop_size=384, pretrained_model_name_or_path='laion/CLIP-ViT-bigG-14-laion2B-39B-b160k', size=384, trust_remote_code=True, type='transformers.CLIPImageProcessor.from_pretrained') language_prefix = '/mnt/program/wlx_data/datasets/lmms-lab/LLaVA-Video-178K/llava_video_frames/' launcher = 'pytorch' llava_dataset_1 = dict( data_path= '/mnt/program/wlx_data/datasets/lmms-lab/LLaVA-Video-178K/llava_video_frames/0_30_s_academic_v0_1/converted_jsons/filtered_converted_0_30_s_academic_mc_v0_1_qa_processed.json', dataset_map_fn='xtuner.dataset.map_fns.aurora_map_fn', image_folder= '/mnt/program/wlx_data/datasets/lmms-lab/LLaVA-Video-178K/llava_video_frames/0_30_s_academic_v0_1/', image_processor=dict( crop_size=384, pretrained_model_name_or_path= 'laion/CLIP-ViT-bigG-14-laion2B-39B-b160k', size=384, trust_remote_code=True, type='transformers.CLIPImageProcessor.from_pretrained'), max_length=4096, offline_processed_text_folder= '/mnt/program/wlx_data/datasets/lmms-lab/LLaVA-Video-178K/llava_video_frames/0_30_s_academic_v0_1/tokenized-by-rwkv-v6/mc/tokenized_sandwich_data', pad_image_to_square=True, template_map_fn=dict( template='xtuner.utils.PROMPT_TEMPLATE.vicuna', type='xtuner.dataset.map_fns.template_map_fn_factory'), tokenizer=dict( padding_side='right', pretrained_model_name_or_path= '/mnt/program/wlx_data/experiments/language-1022/rwkv-1b6/iter_745524.hf', trust_remote_code=True, type='transformers.AutoTokenizer.from_pretrained'), type='xtuner.dataset.AuroraDataset') llava_dataset_10 = dict( data_path= '/mnt/program/wlx_data/datasets/lmms-lab/LLaVA-Video-178K/llava_video_frames/2_3_m_youtube_v0_1/converted_jsons/converted_2_3_m_youtube_mc_v0_1_qa_processed.json', dataset_map_fn='xtuner.dataset.map_fns.aurora_map_fn', image_folder= '/mnt/program/wlx_data/datasets/lmms-lab/LLaVA-Video-178K/llava_video_frames/2_3_m_youtube_v0_1/', image_processor=dict( crop_size=384, pretrained_model_name_or_path= 'laion/CLIP-ViT-bigG-14-laion2B-39B-b160k', size=384, trust_remote_code=True, type='transformers.CLIPImageProcessor.from_pretrained'), max_length=4096, offline_processed_text_folder= '/mnt/program/wlx_data/datasets/lmms-lab/LLaVA-Video-178K/llava_video_frames/2_3_m_youtube_v0_1/tokenized-by-rwkv-v6/mc/tokenized_sandwich_data', pad_image_to_square=True, template_map_fn=dict( template='xtuner.utils.PROMPT_TEMPLATE.vicuna', type='xtuner.dataset.map_fns.template_map_fn_factory'), tokenizer=dict( padding_side='right', pretrained_model_name_or_path= '/mnt/program/wlx_data/experiments/language-1022/rwkv-1b6/iter_745524.hf', trust_remote_code=True, type='transformers.AutoTokenizer.from_pretrained'), type='xtuner.dataset.AuroraDataset') llava_dataset_11 = dict( data_path= '/mnt/program/wlx_data/datasets/lmms-lab/LLaVA-Video-178K/llava_video_frames/30_60_s_academic_v0_1/converted_jsons/converted_30_60_s_academic_mc_v0_1_qa_processed.json', dataset_map_fn='xtuner.dataset.map_fns.aurora_map_fn', image_folder= '/mnt/program/wlx_data/datasets/lmms-lab/LLaVA-Video-178K/llava_video_frames/30_60_s_academic_v0_1/', image_processor=dict( crop_size=384, pretrained_model_name_or_path= 'laion/CLIP-ViT-bigG-14-laion2B-39B-b160k', size=384, trust_remote_code=True, type='transformers.CLIPImageProcessor.from_pretrained'), max_length=4096, offline_processed_text_folder= '/mnt/program/wlx_data/datasets/lmms-lab/LLaVA-Video-178K/llava_video_frames/30_60_s_academic_v0_1/tokenized-by-rwkv-v6/mc/tokenized_sandwich_data', pad_image_to_square=True, template_map_fn=dict( template='xtuner.utils.PROMPT_TEMPLATE.vicuna', type='xtuner.dataset.map_fns.template_map_fn_factory'), tokenizer=dict( padding_side='right', pretrained_model_name_or_path= '/mnt/program/wlx_data/experiments/language-1022/rwkv-1b6/iter_745524.hf', trust_remote_code=True, type='transformers.AutoTokenizer.from_pretrained'), type='xtuner.dataset.AuroraDataset') llava_dataset_12 = dict( data_path= '/mnt/program/wlx_data/datasets/lmms-lab/LLaVA-Video-178K/llava_video_frames/30_60_s_nextqa/converted_jsons/converted_30_60_s_nextqa_mc_qa_processed.json', dataset_map_fn='xtuner.dataset.map_fns.aurora_map_fn', image_folder= '/mnt/program/wlx_data/datasets/lmms-lab/LLaVA-Video-178K/llava_video_frames/30_60_s_nextqa/', image_processor=dict( crop_size=384, pretrained_model_name_or_path= 'laion/CLIP-ViT-bigG-14-laion2B-39B-b160k', size=384, trust_remote_code=True, type='transformers.CLIPImageProcessor.from_pretrained'), max_length=4096, offline_processed_text_folder= '/mnt/program/wlx_data/datasets/lmms-lab/LLaVA-Video-178K/llava_video_frames/30_60_s_nextqa/tokenized-by-rwkv-v6/mc/tokenized_sandwich_data', pad_image_to_square=True, template_map_fn=dict( template='xtuner.utils.PROMPT_TEMPLATE.vicuna', type='xtuner.dataset.map_fns.template_map_fn_factory'), tokenizer=dict( padding_side='right', pretrained_model_name_or_path= '/mnt/program/wlx_data/experiments/language-1022/rwkv-1b6/iter_745524.hf', trust_remote_code=True, type='transformers.AutoTokenizer.from_pretrained'), type='xtuner.dataset.AuroraDataset') llava_dataset_13 = dict( data_path= '/mnt/program/wlx_data/datasets/lmms-lab/LLaVA-Video-178K/llava_video_frames/30_60_s_perceptiontest/converted_jsons/converted_30_60_s_perceptiontest_mc_qa_processed.json', dataset_map_fn='xtuner.dataset.map_fns.aurora_map_fn', image_folder= '/mnt/program/wlx_data/datasets/lmms-lab/LLaVA-Video-178K/llava_video_frames/30_60_s_perceptiontest/', image_processor=dict( crop_size=384, pretrained_model_name_or_path= 'laion/CLIP-ViT-bigG-14-laion2B-39B-b160k', size=384, trust_remote_code=True, type='transformers.CLIPImageProcessor.from_pretrained'), max_length=4096, offline_processed_text_folder= '/mnt/program/wlx_data/datasets/lmms-lab/LLaVA-Video-178K/llava_video_frames/30_60_s_perceptiontest/tokenized-by-rwkv-v6/mc/tokenized_sandwich_data', pad_image_to_square=True, template_map_fn=dict( template='xtuner.utils.PROMPT_TEMPLATE.vicuna', type='xtuner.dataset.map_fns.template_map_fn_factory'), tokenizer=dict( padding_side='right', pretrained_model_name_or_path= '/mnt/program/wlx_data/experiments/language-1022/rwkv-1b6/iter_745524.hf', trust_remote_code=True, type='transformers.AutoTokenizer.from_pretrained'), type='xtuner.dataset.AuroraDataset') llava_dataset_14 = dict( data_path= '/mnt/program/wlx_data/datasets/lmms-lab/LLaVA-Video-178K/llava_video_frames/30_60_s_youtube_v0_1/converted_jsons/converted_30_60_s_youtube_mc_v0_1_qa_processed.json', dataset_map_fn='xtuner.dataset.map_fns.aurora_map_fn', image_folder= '/mnt/program/wlx_data/datasets/lmms-lab/LLaVA-Video-178K/llava_video_frames/30_60_s_youtube_v0_1/', image_processor=dict( crop_size=384, pretrained_model_name_or_path= 'laion/CLIP-ViT-bigG-14-laion2B-39B-b160k', size=384, trust_remote_code=True, type='transformers.CLIPImageProcessor.from_pretrained'), max_length=4096, offline_processed_text_folder= '/mnt/program/wlx_data/datasets/lmms-lab/LLaVA-Video-178K/llava_video_frames/30_60_s_youtube_v0_1/tokenized-by-rwkv-v6/mc/tokenized_sandwich_data', pad_image_to_square=True, template_map_fn=dict( template='xtuner.utils.PROMPT_TEMPLATE.vicuna', type='xtuner.dataset.map_fns.template_map_fn_factory'), tokenizer=dict( padding_side='right', pretrained_model_name_or_path= '/mnt/program/wlx_data/experiments/language-1022/rwkv-1b6/iter_745524.hf', trust_remote_code=True, type='transformers.AutoTokenizer.from_pretrained'), type='xtuner.dataset.AuroraDataset') llava_dataset_2 = dict( data_path= '/mnt/program/wlx_data/datasets/lmms-lab/LLaVA-Video-178K/llava_video_frames/0_30_s_nextqa/converted_jsons/converted_0_30_s_nextqa_mc_qa_processed.json', dataset_map_fn='xtuner.dataset.map_fns.aurora_map_fn', image_folder= '/mnt/program/wlx_data/datasets/lmms-lab/LLaVA-Video-178K/llava_video_frames/0_30_s_nextqa/', image_processor=dict( crop_size=384, pretrained_model_name_or_path= 'laion/CLIP-ViT-bigG-14-laion2B-39B-b160k', size=384, trust_remote_code=True, type='transformers.CLIPImageProcessor.from_pretrained'), max_length=4096, offline_processed_text_folder= '/mnt/program/wlx_data/datasets/lmms-lab/LLaVA-Video-178K/llava_video_frames/0_30_s_nextqa/tokenized-by-rwkv-v6/mc/tokenized_sandwich_data', pad_image_to_square=True, template_map_fn=dict( template='xtuner.utils.PROMPT_TEMPLATE.vicuna', type='xtuner.dataset.map_fns.template_map_fn_factory'), tokenizer=dict( padding_side='right', pretrained_model_name_or_path= '/mnt/program/wlx_data/experiments/language-1022/rwkv-1b6/iter_745524.hf', trust_remote_code=True, type='transformers.AutoTokenizer.from_pretrained'), type='xtuner.dataset.AuroraDataset') llava_dataset_20 = dict( data_path= '/mnt/program/wlx_data/datasets/lmms-lab/LLaVA-Video-178K/llava_video_frames/0_30_s_academic_v0_1/converted_jsons/filtered_converted_0_30_s_academic_oe_v0_1_qa_processed.json', dataset_map_fn='xtuner.dataset.map_fns.aurora_map_fn', image_folder= '/mnt/program/wlx_data/datasets/lmms-lab/LLaVA-Video-178K/llava_video_frames/0_30_s_academic_v0_1/', image_processor=dict( crop_size=384, pretrained_model_name_or_path= 'laion/CLIP-ViT-bigG-14-laion2B-39B-b160k', size=384, trust_remote_code=True, type='transformers.CLIPImageProcessor.from_pretrained'), max_length=4096, offline_processed_text_folder= '/mnt/program/wlx_data/datasets/lmms-lab/LLaVA-Video-178K/llava_video_frames/0_30_s_academic_v0_1/tokenized-by-rwkv-v6/oe/tokenized_sandwich_data', pad_image_to_square=True, template_map_fn=dict( template='xtuner.utils.PROMPT_TEMPLATE.vicuna', type='xtuner.dataset.map_fns.template_map_fn_factory'), tokenizer=dict( padding_side='right', pretrained_model_name_or_path= '/mnt/program/wlx_data/experiments/language-1022/rwkv-1b6/iter_745524.hf', trust_remote_code=True, type='transformers.AutoTokenizer.from_pretrained'), type='xtuner.dataset.AuroraDataset') llava_dataset_21 = dict( data_path= '/mnt/program/wlx_data/datasets/lmms-lab/LLaVA-Video-178K/llava_video_frames/0_30_s_activitynetqa/converted_jsons/converted_0_30_s_activitynetqa_oe_qa_processed.json', dataset_map_fn='xtuner.dataset.map_fns.aurora_map_fn', image_folder= '/mnt/program/wlx_data/datasets/lmms-lab/LLaVA-Video-178K/llava_video_frames/0_30_s_activitynetqa/', image_processor=dict( crop_size=384, pretrained_model_name_or_path= 'laion/CLIP-ViT-bigG-14-laion2B-39B-b160k', size=384, trust_remote_code=True, type='transformers.CLIPImageProcessor.from_pretrained'), max_length=4096, offline_processed_text_folder= '/mnt/program/wlx_data/datasets/lmms-lab/LLaVA-Video-178K/llava_video_frames/0_30_s_activitynetqa/tokenized-by-rwkv-v6/oe/tokenized_sandwich_data', pad_image_to_square=True, template_map_fn=dict( template='xtuner.utils.PROMPT_TEMPLATE.vicuna', type='xtuner.dataset.map_fns.template_map_fn_factory'), tokenizer=dict( padding_side='right', pretrained_model_name_or_path= '/mnt/program/wlx_data/experiments/language-1022/rwkv-1b6/iter_745524.hf', trust_remote_code=True, type='transformers.AutoTokenizer.from_pretrained'), type='xtuner.dataset.AuroraDataset') llava_dataset_22 = dict( data_path= '/mnt/program/wlx_data/datasets/lmms-lab/LLaVA-Video-178K/llava_video_frames/0_30_s_nextqa/converted_jsons/converted_0_30_s_nextqa_oe_qa_processed.json', dataset_map_fn='xtuner.dataset.map_fns.aurora_map_fn', image_folder= '/mnt/program/wlx_data/datasets/lmms-lab/LLaVA-Video-178K/llava_video_frames/0_30_s_nextqa/', image_processor=dict( crop_size=384, pretrained_model_name_or_path= 'laion/CLIP-ViT-bigG-14-laion2B-39B-b160k', size=384, trust_remote_code=True, type='transformers.CLIPImageProcessor.from_pretrained'), max_length=4096, offline_processed_text_folder= '/mnt/program/wlx_data/datasets/lmms-lab/LLaVA-Video-178K/llava_video_frames/0_30_s_nextqa/tokenized-by-rwkv-v6/oe/tokenized_sandwich_data', pad_image_to_square=True, template_map_fn=dict( template='xtuner.utils.PROMPT_TEMPLATE.vicuna', type='xtuner.dataset.map_fns.template_map_fn_factory'), tokenizer=dict( padding_side='right', pretrained_model_name_or_path= '/mnt/program/wlx_data/experiments/language-1022/rwkv-1b6/iter_745524.hf', trust_remote_code=True, type='transformers.AutoTokenizer.from_pretrained'), type='xtuner.dataset.AuroraDataset') llava_dataset_23 = dict( data_path= '/mnt/program/wlx_data/datasets/lmms-lab/LLaVA-Video-178K/llava_video_frames/0_30_s_youtube_v0_1/converted_jsons/filtered_converted_0_30_s_youtube_oe_v0_1_qa_processed.json', dataset_map_fn='xtuner.dataset.map_fns.aurora_map_fn', image_folder= '/mnt/program/wlx_data/datasets/lmms-lab/LLaVA-Video-178K/llava_video_frames/0_30_s_youtube_v0_1/', image_processor=dict( crop_size=384, pretrained_model_name_or_path= 'laion/CLIP-ViT-bigG-14-laion2B-39B-b160k', size=384, trust_remote_code=True, type='transformers.CLIPImageProcessor.from_pretrained'), max_length=4096, offline_processed_text_folder= '/mnt/program/wlx_data/datasets/lmms-lab/LLaVA-Video-178K/llava_video_frames/0_30_s_youtube_v0_1/tokenized-by-rwkv-v6/oe/tokenized_sandwich_data', pad_image_to_square=True, template_map_fn=dict( template='xtuner.utils.PROMPT_TEMPLATE.vicuna', type='xtuner.dataset.map_fns.template_map_fn_factory'), tokenizer=dict( padding_side='right', pretrained_model_name_or_path= '/mnt/program/wlx_data/experiments/language-1022/rwkv-1b6/iter_745524.hf', trust_remote_code=True, type='transformers.AutoTokenizer.from_pretrained'), type='xtuner.dataset.AuroraDataset') llava_dataset_24 = dict( data_path= '/mnt/program/wlx_data/datasets/lmms-lab/LLaVA-Video-178K/llava_video_frames/1_2_m_academic_v0_1/converted_jsons/converted_1_2_m_academic_oe_v0_1_qa_processed.json', dataset_map_fn='xtuner.dataset.map_fns.aurora_map_fn', image_folder= '/mnt/program/wlx_data/datasets/lmms-lab/LLaVA-Video-178K/llava_video_frames/1_2_m_academic_v0_1/', image_processor=dict( crop_size=384, pretrained_model_name_or_path= 'laion/CLIP-ViT-bigG-14-laion2B-39B-b160k', size=384, trust_remote_code=True, type='transformers.CLIPImageProcessor.from_pretrained'), max_length=4096, offline_processed_text_folder= '/mnt/program/wlx_data/datasets/lmms-lab/LLaVA-Video-178K/llava_video_frames/1_2_m_academic_v0_1/tokenized-by-rwkv-v6/oe/tokenized_sandwich_data', pad_image_to_square=True, template_map_fn=dict( template='xtuner.utils.PROMPT_TEMPLATE.vicuna', type='xtuner.dataset.map_fns.template_map_fn_factory'), tokenizer=dict( padding_side='right', pretrained_model_name_or_path= '/mnt/program/wlx_data/experiments/language-1022/rwkv-1b6/iter_745524.hf', trust_remote_code=True, type='transformers.AutoTokenizer.from_pretrained'), type='xtuner.dataset.AuroraDataset') llava_dataset_25 = dict( data_path= '/mnt/program/wlx_data/datasets/lmms-lab/LLaVA-Video-178K/llava_video_frames/1_2_m_activitynetqa/converted_jsons/converted_1_2_m_activitynetqa_oe_qa_processed.json', dataset_map_fn='xtuner.dataset.map_fns.aurora_map_fn', image_folder= '/mnt/program/wlx_data/datasets/lmms-lab/LLaVA-Video-178K/llava_video_frames/1_2_m_activitynetqa/', image_processor=dict( crop_size=384, pretrained_model_name_or_path= 'laion/CLIP-ViT-bigG-14-laion2B-39B-b160k', size=384, trust_remote_code=True, type='transformers.CLIPImageProcessor.from_pretrained'), max_length=4096, offline_processed_text_folder= '/mnt/program/wlx_data/datasets/lmms-lab/LLaVA-Video-178K/llava_video_frames/1_2_m_activitynetqa/tokenized-by-rwkv-v6/oe/tokenized_sandwich_data', pad_image_to_square=True, template_map_fn=dict( template='xtuner.utils.PROMPT_TEMPLATE.vicuna', type='xtuner.dataset.map_fns.template_map_fn_factory'), tokenizer=dict( padding_side='right', pretrained_model_name_or_path= '/mnt/program/wlx_data/experiments/language-1022/rwkv-1b6/iter_745524.hf', trust_remote_code=True, type='transformers.AutoTokenizer.from_pretrained'), type='xtuner.dataset.AuroraDataset') llava_dataset_26 = dict( data_path= '/mnt/program/wlx_data/datasets/lmms-lab/LLaVA-Video-178K/llava_video_frames/1_2_m_nextqa/converted_jsons/converted_1_2_m_nextqa_oe_qa_processed.json', dataset_map_fn='xtuner.dataset.map_fns.aurora_map_fn', image_folder= '/mnt/program/wlx_data/datasets/lmms-lab/LLaVA-Video-178K/llava_video_frames/1_2_m_nextqa/', image_processor=dict( crop_size=384, pretrained_model_name_or_path= 'laion/CLIP-ViT-bigG-14-laion2B-39B-b160k', size=384, trust_remote_code=True, type='transformers.CLIPImageProcessor.from_pretrained'), max_length=4096, offline_processed_text_folder= '/mnt/program/wlx_data/datasets/lmms-lab/LLaVA-Video-178K/llava_video_frames/1_2_m_nextqa/tokenized-by-rwkv-v6/oe/tokenized_sandwich_data', pad_image_to_square=True, template_map_fn=dict( template='xtuner.utils.PROMPT_TEMPLATE.vicuna', type='xtuner.dataset.map_fns.template_map_fn_factory'), tokenizer=dict( padding_side='right', pretrained_model_name_or_path= '/mnt/program/wlx_data/experiments/language-1022/rwkv-1b6/iter_745524.hf', trust_remote_code=True, type='transformers.AutoTokenizer.from_pretrained'), type='xtuner.dataset.AuroraDataset') llava_dataset_27 = dict( data_path= '/mnt/program/wlx_data/datasets/lmms-lab/LLaVA-Video-178K/llava_video_frames/1_2_m_youtube_v0_1/converted_jsons/converted_1_2_m_youtube_oe_v0_1_qa_processed.json', dataset_map_fn='xtuner.dataset.map_fns.aurora_map_fn', image_folder= '/mnt/program/wlx_data/datasets/lmms-lab/LLaVA-Video-178K/llava_video_frames/1_2_m_youtube_v0_1/', image_processor=dict( crop_size=384, pretrained_model_name_or_path= 'laion/CLIP-ViT-bigG-14-laion2B-39B-b160k', size=384, trust_remote_code=True, type='transformers.CLIPImageProcessor.from_pretrained'), max_length=4096, offline_processed_text_folder= '/mnt/program/wlx_data/datasets/lmms-lab/LLaVA-Video-178K/llava_video_frames/1_2_m_youtube_v0_1/tokenized-by-rwkv-v6/oe/tokenized_sandwich_data', pad_image_to_square=True, template_map_fn=dict( template='xtuner.utils.PROMPT_TEMPLATE.vicuna', type='xtuner.dataset.map_fns.template_map_fn_factory'), tokenizer=dict( padding_side='right', pretrained_model_name_or_path= '/mnt/program/wlx_data/experiments/language-1022/rwkv-1b6/iter_745524.hf', trust_remote_code=True, type='transformers.AutoTokenizer.from_pretrained'), type='xtuner.dataset.AuroraDataset') llava_dataset_28 = dict( data_path= '/mnt/program/wlx_data/datasets/lmms-lab/LLaVA-Video-178K/llava_video_frames/2_3_m_academic_v0_1/converted_jsons/converted_2_3_m_academic_oe_v0_1_qa_processed.json', dataset_map_fn='xtuner.dataset.map_fns.aurora_map_fn', image_folder= '/mnt/program/wlx_data/datasets/lmms-lab/LLaVA-Video-178K/llava_video_frames/2_3_m_academic_v0_1/', image_processor=dict( crop_size=384, pretrained_model_name_or_path= 'laion/CLIP-ViT-bigG-14-laion2B-39B-b160k', size=384, trust_remote_code=True, type='transformers.CLIPImageProcessor.from_pretrained'), max_length=4096, offline_processed_text_folder= '/mnt/program/wlx_data/datasets/lmms-lab/LLaVA-Video-178K/llava_video_frames/2_3_m_academic_v0_1/tokenized-by-rwkv-v6/oe/tokenized_sandwich_data', pad_image_to_square=True, template_map_fn=dict( template='xtuner.utils.PROMPT_TEMPLATE.vicuna', type='xtuner.dataset.map_fns.template_map_fn_factory'), tokenizer=dict( padding_side='right', pretrained_model_name_or_path= '/mnt/program/wlx_data/experiments/language-1022/rwkv-1b6/iter_745524.hf', trust_remote_code=True, type='transformers.AutoTokenizer.from_pretrained'), type='xtuner.dataset.AuroraDataset') llava_dataset_29 = dict( data_path= '/mnt/program/wlx_data/datasets/lmms-lab/LLaVA-Video-178K/llava_video_frames/2_3_m_activitynetqa/converted_jsons/converted_2_3_m_activitynetqa_oe_qa_processed.json', dataset_map_fn='xtuner.dataset.map_fns.aurora_map_fn', image_folder= '/mnt/program/wlx_data/datasets/lmms-lab/LLaVA-Video-178K/llava_video_frames/2_3_m_activitynetqa/', image_processor=dict( crop_size=384, pretrained_model_name_or_path= 'laion/CLIP-ViT-bigG-14-laion2B-39B-b160k', size=384, trust_remote_code=True, type='transformers.CLIPImageProcessor.from_pretrained'), max_length=4096, offline_processed_text_folder= '/mnt/program/wlx_data/datasets/lmms-lab/LLaVA-Video-178K/llava_video_frames/2_3_m_activitynetqa/tokenized-by-rwkv-v6/oe/tokenized_sandwich_data', pad_image_to_square=True, template_map_fn=dict( template='xtuner.utils.PROMPT_TEMPLATE.vicuna', type='xtuner.dataset.map_fns.template_map_fn_factory'), tokenizer=dict( padding_side='right', pretrained_model_name_or_path= '/mnt/program/wlx_data/experiments/language-1022/rwkv-1b6/iter_745524.hf', trust_remote_code=True, type='transformers.AutoTokenizer.from_pretrained'), type='xtuner.dataset.AuroraDataset') llava_dataset_3 = dict( data_path= '/mnt/program/wlx_data/datasets/lmms-lab/LLaVA-Video-178K/llava_video_frames/0_30_s_perceptiontest/converted_jsons/converted_0_30_s_perceptiontest_mc_qa_processed.json', dataset_map_fn='xtuner.dataset.map_fns.aurora_map_fn', image_folder= '/mnt/program/wlx_data/datasets/lmms-lab/LLaVA-Video-178K/llava_video_frames/0_30_s_perceptiontest/', image_processor=dict( crop_size=384, pretrained_model_name_or_path= 'laion/CLIP-ViT-bigG-14-laion2B-39B-b160k', size=384, trust_remote_code=True, type='transformers.CLIPImageProcessor.from_pretrained'), max_length=4096, offline_processed_text_folder= '/mnt/program/wlx_data/datasets/lmms-lab/LLaVA-Video-178K/llava_video_frames/0_30_s_perceptiontest/tokenized-by-rwkv-v6/mc/tokenized_sandwich_data', pad_image_to_square=True, template_map_fn=dict( template='xtuner.utils.PROMPT_TEMPLATE.vicuna', type='xtuner.dataset.map_fns.template_map_fn_factory'), tokenizer=dict( padding_side='right', pretrained_model_name_or_path= '/mnt/program/wlx_data/experiments/language-1022/rwkv-1b6/iter_745524.hf', trust_remote_code=True, type='transformers.AutoTokenizer.from_pretrained'), type='xtuner.dataset.AuroraDataset') llava_dataset_30 = dict( data_path= '/mnt/program/wlx_data/datasets/lmms-lab/LLaVA-Video-178K/llava_video_frames/2_3_m_nextqa/converted_jsons/converted_2_3_m_nextqa_oe_qa_processed.json', dataset_map_fn='xtuner.dataset.map_fns.aurora_map_fn', image_folder= '/mnt/program/wlx_data/datasets/lmms-lab/LLaVA-Video-178K/llava_video_frames/2_3_m_nextqa/', image_processor=dict( crop_size=384, pretrained_model_name_or_path= 'laion/CLIP-ViT-bigG-14-laion2B-39B-b160k', size=384, trust_remote_code=True, type='transformers.CLIPImageProcessor.from_pretrained'), max_length=4096, offline_processed_text_folder= '/mnt/program/wlx_data/datasets/lmms-lab/LLaVA-Video-178K/llava_video_frames/2_3_m_nextqa/tokenized-by-rwkv-v6/oe/tokenized_sandwich_data', pad_image_to_square=True, template_map_fn=dict( template='xtuner.utils.PROMPT_TEMPLATE.vicuna', type='xtuner.dataset.map_fns.template_map_fn_factory'), tokenizer=dict( padding_side='right', pretrained_model_name_or_path= '/mnt/program/wlx_data/experiments/language-1022/rwkv-1b6/iter_745524.hf', trust_remote_code=True, type='transformers.AutoTokenizer.from_pretrained'), type='xtuner.dataset.AuroraDataset') llava_dataset_31 = dict( data_path= '/mnt/program/wlx_data/datasets/lmms-lab/LLaVA-Video-178K/llava_video_frames/2_3_m_youtube_v0_1/converted_jsons/converted_2_3_m_youtube_oe_v0_1_qa_processed.json', dataset_map_fn='xtuner.dataset.map_fns.aurora_map_fn', image_folder= '/mnt/program/wlx_data/datasets/lmms-lab/LLaVA-Video-178K/llava_video_frames/2_3_m_youtube_v0_1/', image_processor=dict( crop_size=384, pretrained_model_name_or_path= 'laion/CLIP-ViT-bigG-14-laion2B-39B-b160k', size=384, trust_remote_code=True, type='transformers.CLIPImageProcessor.from_pretrained'), max_length=4096, offline_processed_text_folder= '/mnt/program/wlx_data/datasets/lmms-lab/LLaVA-Video-178K/llava_video_frames/2_3_m_youtube_v0_1/tokenized-by-rwkv-v6/oe/tokenized_sandwich_data', pad_image_to_square=True, template_map_fn=dict( template='xtuner.utils.PROMPT_TEMPLATE.vicuna', type='xtuner.dataset.map_fns.template_map_fn_factory'), tokenizer=dict( padding_side='right', pretrained_model_name_or_path= '/mnt/program/wlx_data/experiments/language-1022/rwkv-1b6/iter_745524.hf', trust_remote_code=True, type='transformers.AutoTokenizer.from_pretrained'), type='xtuner.dataset.AuroraDataset') llava_dataset_32 = dict( data_path= '/mnt/program/wlx_data/datasets/lmms-lab/LLaVA-Video-178K/llava_video_frames/30_60_s_academic_v0_1/converted_jsons/converted_30_60_s_academic_oe_v0_1_qa_processed.json', dataset_map_fn='xtuner.dataset.map_fns.aurora_map_fn', image_folder= '/mnt/program/wlx_data/datasets/lmms-lab/LLaVA-Video-178K/llava_video_frames/30_60_s_academic_v0_1/', image_processor=dict( crop_size=384, pretrained_model_name_or_path= 'laion/CLIP-ViT-bigG-14-laion2B-39B-b160k', size=384, trust_remote_code=True, type='transformers.CLIPImageProcessor.from_pretrained'), max_length=4096, offline_processed_text_folder= '/mnt/program/wlx_data/datasets/lmms-lab/LLaVA-Video-178K/llava_video_frames/30_60_s_academic_v0_1/tokenized-by-rwkv-v6/oe/tokenized_sandwich_data', pad_image_to_square=True, template_map_fn=dict( template='xtuner.utils.PROMPT_TEMPLATE.vicuna', type='xtuner.dataset.map_fns.template_map_fn_factory'), tokenizer=dict( padding_side='right', pretrained_model_name_or_path= '/mnt/program/wlx_data/experiments/language-1022/rwkv-1b6/iter_745524.hf', trust_remote_code=True, type='transformers.AutoTokenizer.from_pretrained'), type='xtuner.dataset.AuroraDataset') llava_dataset_33 = dict( data_path= '/mnt/program/wlx_data/datasets/lmms-lab/LLaVA-Video-178K/llava_video_frames/30_60_s_activitynetqa/converted_jsons/converted_30_60_s_activitynetqa_oe_qa_processed.json', dataset_map_fn='xtuner.dataset.map_fns.aurora_map_fn', image_folder= '/mnt/program/wlx_data/datasets/lmms-lab/LLaVA-Video-178K/llava_video_frames/30_60_s_activitynetqa/', image_processor=dict( crop_size=384, pretrained_model_name_or_path= 'laion/CLIP-ViT-bigG-14-laion2B-39B-b160k', size=384, trust_remote_code=True, type='transformers.CLIPImageProcessor.from_pretrained'), max_length=4096, offline_processed_text_folder= '/mnt/program/wlx_data/datasets/lmms-lab/LLaVA-Video-178K/llava_video_frames/30_60_s_activitynetqa/tokenized-by-rwkv-v6/oe/tokenized_sandwich_data', pad_image_to_square=True, template_map_fn=dict( template='xtuner.utils.PROMPT_TEMPLATE.vicuna', type='xtuner.dataset.map_fns.template_map_fn_factory'), tokenizer=dict( padding_side='right', pretrained_model_name_or_path= '/mnt/program/wlx_data/experiments/language-1022/rwkv-1b6/iter_745524.hf', trust_remote_code=True, type='transformers.AutoTokenizer.from_pretrained'), type='xtuner.dataset.AuroraDataset') llava_dataset_34 = dict( data_path= '/mnt/program/wlx_data/datasets/lmms-lab/LLaVA-Video-178K/llava_video_frames/30_60_s_nextqa/converted_jsons/converted_30_60_s_nextqa_oe_qa_processed.json', dataset_map_fn='xtuner.dataset.map_fns.aurora_map_fn', image_folder= '/mnt/program/wlx_data/datasets/lmms-lab/LLaVA-Video-178K/llava_video_frames/30_60_s_nextqa/', image_processor=dict( crop_size=384, pretrained_model_name_or_path= 'laion/CLIP-ViT-bigG-14-laion2B-39B-b160k', size=384, trust_remote_code=True, type='transformers.CLIPImageProcessor.from_pretrained'), max_length=4096, offline_processed_text_folder= '/mnt/program/wlx_data/datasets/lmms-lab/LLaVA-Video-178K/llava_video_frames/30_60_s_nextqa/tokenized-by-rwkv-v6/oe/tokenized_sandwich_data', pad_image_to_square=True, template_map_fn=dict( template='xtuner.utils.PROMPT_TEMPLATE.vicuna', type='xtuner.dataset.map_fns.template_map_fn_factory'), tokenizer=dict( padding_side='right', pretrained_model_name_or_path= '/mnt/program/wlx_data/experiments/language-1022/rwkv-1b6/iter_745524.hf', trust_remote_code=True, type='transformers.AutoTokenizer.from_pretrained'), type='xtuner.dataset.AuroraDataset') llava_dataset_35 = dict( data_path= '/mnt/program/wlx_data/datasets/lmms-lab/LLaVA-Video-178K/llava_video_frames/30_60_s_youtube_v0_1/converted_jsons/filtered_converted_30_60_s_youtube_oe_v0_1_qa_processed.json', dataset_map_fn='xtuner.dataset.map_fns.aurora_map_fn', image_folder= '/mnt/program/wlx_data/datasets/lmms-lab/LLaVA-Video-178K/llava_video_frames/30_60_s_youtube_v0_1/', image_processor=dict( crop_size=384, pretrained_model_name_or_path= 'laion/CLIP-ViT-bigG-14-laion2B-39B-b160k', size=384, trust_remote_code=True, type='transformers.CLIPImageProcessor.from_pretrained'), max_length=4096, offline_processed_text_folder= '/mnt/program/wlx_data/datasets/lmms-lab/LLaVA-Video-178K/llava_video_frames/30_60_s_youtube_v0_1/tokenized-by-rwkv-v6/oe/tokenized_sandwich_data', pad_image_to_square=True, template_map_fn=dict( template='xtuner.utils.PROMPT_TEMPLATE.vicuna', type='xtuner.dataset.map_fns.template_map_fn_factory'), tokenizer=dict( padding_side='right', pretrained_model_name_or_path= '/mnt/program/wlx_data/experiments/language-1022/rwkv-1b6/iter_745524.hf', trust_remote_code=True, type='transformers.AutoTokenizer.from_pretrained'), type='xtuner.dataset.AuroraDataset') llava_dataset_4 = dict( data_path= '/mnt/program/wlx_data/datasets/lmms-lab/LLaVA-Video-178K/llava_video_frames/0_30_s_youtube_v0_1/converted_jsons/filtered_converted_0_30_s_youtube_mc_v0_1_qa_processed.json', dataset_map_fn='xtuner.dataset.map_fns.aurora_map_fn', image_folder= '/mnt/program/wlx_data/datasets/lmms-lab/LLaVA-Video-178K/llava_video_frames/0_30_s_youtube_v0_1/', image_processor=dict( crop_size=384, pretrained_model_name_or_path= 'laion/CLIP-ViT-bigG-14-laion2B-39B-b160k', size=384, trust_remote_code=True, type='transformers.CLIPImageProcessor.from_pretrained'), max_length=4096, offline_processed_text_folder= '/mnt/program/wlx_data/datasets/lmms-lab/LLaVA-Video-178K/llava_video_frames/0_30_s_youtube_v0_1/tokenized-by-rwkv-v6/mc/tokenized_sandwich_data', pad_image_to_square=True, template_map_fn=dict( template='xtuner.utils.PROMPT_TEMPLATE.vicuna', type='xtuner.dataset.map_fns.template_map_fn_factory'), tokenizer=dict( padding_side='right', pretrained_model_name_or_path= '/mnt/program/wlx_data/experiments/language-1022/rwkv-1b6/iter_745524.hf', trust_remote_code=True, type='transformers.AutoTokenizer.from_pretrained'), type='xtuner.dataset.AuroraDataset') llava_dataset_40 = dict( data_path= '/mnt/program/wlx_data/datasets/lmms-lab/LLaVA-Video-178K/llava_video_frames/0_30_s_academic_v0_1/converted_jsons/filtered_converted_0_30_s_academic_v0_1_cap_processed.json', dataset_map_fn='xtuner.dataset.map_fns.aurora_map_fn', image_folder= '/mnt/program/wlx_data/datasets/lmms-lab/LLaVA-Video-178K/llava_video_frames/0_30_s_academic_v0_1/', image_processor=dict( crop_size=384, pretrained_model_name_or_path= 'laion/CLIP-ViT-bigG-14-laion2B-39B-b160k', size=384, trust_remote_code=True, type='transformers.CLIPImageProcessor.from_pretrained'), max_length=4096, offline_processed_text_folder= '/mnt/program/wlx_data/datasets/lmms-lab/LLaVA-Video-178K/llava_video_frames/0_30_s_academic_v0_1/tokenized-by-rwkv-v6/cap/tokenized_sandwich_data', pad_image_to_square=True, template_map_fn=dict( template='xtuner.utils.PROMPT_TEMPLATE.vicuna', type='xtuner.dataset.map_fns.template_map_fn_factory'), tokenizer=dict( padding_side='right', pretrained_model_name_or_path= '/mnt/program/wlx_data/experiments/language-1022/rwkv-1b6/iter_745524.hf', trust_remote_code=True, type='transformers.AutoTokenizer.from_pretrained'), type='xtuner.dataset.AuroraDataset') llava_dataset_41 = dict( data_path= '/mnt/program/wlx_data/datasets/lmms-lab/LLaVA-Video-178K/llava_video_frames/0_30_s_youtube_v0_1/converted_jsons/filtered_converted_0_30_s_youtube_v0_1_cap_processed.json', dataset_map_fn='xtuner.dataset.map_fns.aurora_map_fn', image_folder= '/mnt/program/wlx_data/datasets/lmms-lab/LLaVA-Video-178K/llava_video_frames/0_30_s_youtube_v0_1/', image_processor=dict( crop_size=384, pretrained_model_name_or_path= 'laion/CLIP-ViT-bigG-14-laion2B-39B-b160k', size=384, trust_remote_code=True, type='transformers.CLIPImageProcessor.from_pretrained'), max_length=4096, offline_processed_text_folder= '/mnt/program/wlx_data/datasets/lmms-lab/LLaVA-Video-178K/llava_video_frames/0_30_s_youtube_v0_1/tokenized-by-rwkv-v6/cap/tokenized_sandwich_data', pad_image_to_square=True, template_map_fn=dict( template='xtuner.utils.PROMPT_TEMPLATE.vicuna', type='xtuner.dataset.map_fns.template_map_fn_factory'), tokenizer=dict( padding_side='right', pretrained_model_name_or_path= '/mnt/program/wlx_data/experiments/language-1022/rwkv-1b6/iter_745524.hf', trust_remote_code=True, type='transformers.AutoTokenizer.from_pretrained'), type='xtuner.dataset.AuroraDataset') llava_dataset_42 = dict( data_path= '/mnt/program/wlx_data/datasets/lmms-lab/LLaVA-Video-178K/llava_video_frames/1_2_m_academic_v0_1/converted_jsons/converted_1_2_m_academic_v0_1_cap_processed.json', dataset_map_fn='xtuner.dataset.map_fns.aurora_map_fn', image_folder= '/mnt/program/wlx_data/datasets/lmms-lab/LLaVA-Video-178K/llava_video_frames/1_2_m_academic_v0_1/', image_processor=dict( crop_size=384, pretrained_model_name_or_path= 'laion/CLIP-ViT-bigG-14-laion2B-39B-b160k', size=384, trust_remote_code=True, type='transformers.CLIPImageProcessor.from_pretrained'), max_length=4096, offline_processed_text_folder= '/mnt/program/wlx_data/datasets/lmms-lab/LLaVA-Video-178K/llava_video_frames/1_2_m_academic_v0_1/tokenized-by-rwkv-v6/cap/tokenized_sandwich_data', pad_image_to_square=True, template_map_fn=dict( template='xtuner.utils.PROMPT_TEMPLATE.vicuna', type='xtuner.dataset.map_fns.template_map_fn_factory'), tokenizer=dict( padding_side='right', pretrained_model_name_or_path= '/mnt/program/wlx_data/experiments/language-1022/rwkv-1b6/iter_745524.hf', trust_remote_code=True, type='transformers.AutoTokenizer.from_pretrained'), type='xtuner.dataset.AuroraDataset') llava_dataset_43 = dict( data_path= '/mnt/program/wlx_data/datasets/lmms-lab/LLaVA-Video-178K/llava_video_frames/1_2_m_youtube_v0_1/converted_jsons/converted_1_2_m_youtube_v0_1_cap_processed.json', dataset_map_fn='xtuner.dataset.map_fns.aurora_map_fn', image_folder= '/mnt/program/wlx_data/datasets/lmms-lab/LLaVA-Video-178K/llava_video_frames/1_2_m_youtube_v0_1/', image_processor=dict( crop_size=384, pretrained_model_name_or_path= 'laion/CLIP-ViT-bigG-14-laion2B-39B-b160k', size=384, trust_remote_code=True, type='transformers.CLIPImageProcessor.from_pretrained'), max_length=4096, offline_processed_text_folder= '/mnt/program/wlx_data/datasets/lmms-lab/LLaVA-Video-178K/llava_video_frames/1_2_m_youtube_v0_1/tokenized-by-rwkv-v6/cap/tokenized_sandwich_data', pad_image_to_square=True, template_map_fn=dict( template='xtuner.utils.PROMPT_TEMPLATE.vicuna', type='xtuner.dataset.map_fns.template_map_fn_factory'), tokenizer=dict( padding_side='right', pretrained_model_name_or_path= '/mnt/program/wlx_data/experiments/language-1022/rwkv-1b6/iter_745524.hf', trust_remote_code=True, type='transformers.AutoTokenizer.from_pretrained'), type='xtuner.dataset.AuroraDataset') llava_dataset_44 = dict( data_path= '/mnt/program/wlx_data/datasets/lmms-lab/LLaVA-Video-178K/llava_video_frames/2_3_m_academic_v0_1/converted_jsons/converted_2_3_m_academic_v0_1_cap_processed.json', dataset_map_fn='xtuner.dataset.map_fns.aurora_map_fn', image_folder= '/mnt/program/wlx_data/datasets/lmms-lab/LLaVA-Video-178K/llava_video_frames/2_3_m_academic_v0_1/', image_processor=dict( crop_size=384, pretrained_model_name_or_path= 'laion/CLIP-ViT-bigG-14-laion2B-39B-b160k', size=384, trust_remote_code=True, type='transformers.CLIPImageProcessor.from_pretrained'), max_length=4096, offline_processed_text_folder= '/mnt/program/wlx_data/datasets/lmms-lab/LLaVA-Video-178K/llava_video_frames/2_3_m_academic_v0_1/tokenized-by-rwkv-v6/cap/tokenized_sandwich_data', pad_image_to_square=True, template_map_fn=dict( template='xtuner.utils.PROMPT_TEMPLATE.vicuna', type='xtuner.dataset.map_fns.template_map_fn_factory'), tokenizer=dict( padding_side='right', pretrained_model_name_or_path= '/mnt/program/wlx_data/experiments/language-1022/rwkv-1b6/iter_745524.hf', trust_remote_code=True, type='transformers.AutoTokenizer.from_pretrained'), type='xtuner.dataset.AuroraDataset') llava_dataset_45 = dict( data_path= '/mnt/program/wlx_data/datasets/lmms-lab/LLaVA-Video-178K/llava_video_frames/2_3_m_youtube_v0_1/converted_jsons/converted_2_3_m_youtube_v0_1_cap_processed.json', dataset_map_fn='xtuner.dataset.map_fns.aurora_map_fn', image_folder= '/mnt/program/wlx_data/datasets/lmms-lab/LLaVA-Video-178K/llava_video_frames/2_3_m_youtube_v0_1/', image_processor=dict( crop_size=384, pretrained_model_name_or_path= 'laion/CLIP-ViT-bigG-14-laion2B-39B-b160k', size=384, trust_remote_code=True, type='transformers.CLIPImageProcessor.from_pretrained'), max_length=4096, offline_processed_text_folder= '/mnt/program/wlx_data/datasets/lmms-lab/LLaVA-Video-178K/llava_video_frames/2_3_m_youtube_v0_1/tokenized-by-rwkv-v6/cap/tokenized_sandwich_data', pad_image_to_square=True, template_map_fn=dict( template='xtuner.utils.PROMPT_TEMPLATE.vicuna', type='xtuner.dataset.map_fns.template_map_fn_factory'), tokenizer=dict( padding_side='right', pretrained_model_name_or_path= '/mnt/program/wlx_data/experiments/language-1022/rwkv-1b6/iter_745524.hf', trust_remote_code=True, type='transformers.AutoTokenizer.from_pretrained'), type='xtuner.dataset.AuroraDataset') llava_dataset_46 = dict( data_path= '/mnt/program/wlx_data/datasets/lmms-lab/LLaVA-Video-178K/llava_video_frames/30_60_s_academic_v0_1/converted_jsons/converted_30_60_s_academic_v0_1_cap_processed.json', dataset_map_fn='xtuner.dataset.map_fns.aurora_map_fn', image_folder= '/mnt/program/wlx_data/datasets/lmms-lab/LLaVA-Video-178K/llava_video_frames/30_60_s_academic_v0_1/', image_processor=dict( crop_size=384, pretrained_model_name_or_path= 'laion/CLIP-ViT-bigG-14-laion2B-39B-b160k', size=384, trust_remote_code=True, type='transformers.CLIPImageProcessor.from_pretrained'), max_length=4096, offline_processed_text_folder= '/mnt/program/wlx_data/datasets/lmms-lab/LLaVA-Video-178K/llava_video_frames/30_60_s_academic_v0_1/tokenized-by-rwkv-v6/cap/tokenized_sandwich_data', pad_image_to_square=True, template_map_fn=dict( template='xtuner.utils.PROMPT_TEMPLATE.vicuna', type='xtuner.dataset.map_fns.template_map_fn_factory'), tokenizer=dict( padding_side='right', pretrained_model_name_or_path= '/mnt/program/wlx_data/experiments/language-1022/rwkv-1b6/iter_745524.hf', trust_remote_code=True, type='transformers.AutoTokenizer.from_pretrained'), type='xtuner.dataset.AuroraDataset') llava_dataset_47 = dict( data_path= '/mnt/program/wlx_data/datasets/lmms-lab/LLaVA-Video-178K/llava_video_frames/30_60_s_youtube_v0_1/converted_jsons/filtered_converted_30_60_s_youtube_v0_1_cap_processed.json', dataset_map_fn='xtuner.dataset.map_fns.aurora_map_fn', image_folder= '/mnt/program/wlx_data/datasets/lmms-lab/LLaVA-Video-178K/llava_video_frames/30_60_s_youtube_v0_1/', image_processor=dict( crop_size=384, pretrained_model_name_or_path= 'laion/CLIP-ViT-bigG-14-laion2B-39B-b160k', size=384, trust_remote_code=True, type='transformers.CLIPImageProcessor.from_pretrained'), max_length=4096, offline_processed_text_folder= '/mnt/program/wlx_data/datasets/lmms-lab/LLaVA-Video-178K/llava_video_frames/30_60_s_youtube_v0_1/tokenized-by-rwkv-v6/cap/tokenized_sandwich_data', pad_image_to_square=True, template_map_fn=dict( template='xtuner.utils.PROMPT_TEMPLATE.vicuna', type='xtuner.dataset.map_fns.template_map_fn_factory'), tokenizer=dict( padding_side='right', pretrained_model_name_or_path= '/mnt/program/wlx_data/experiments/language-1022/rwkv-1b6/iter_745524.hf', trust_remote_code=True, type='transformers.AutoTokenizer.from_pretrained'), type='xtuner.dataset.AuroraDataset') llava_dataset_5 = dict( data_path= '/mnt/program/wlx_data/datasets/lmms-lab/LLaVA-Video-178K/llava_video_frames/1_2_m_academic_v0_1/converted_jsons/converted_1_2_m_academic_mc_v0_1_qa_processed.json', dataset_map_fn='xtuner.dataset.map_fns.aurora_map_fn', image_folder= '/mnt/program/wlx_data/datasets/lmms-lab/LLaVA-Video-178K/llava_video_frames/1_2_m_academic_v0_1/', image_processor=dict( crop_size=384, pretrained_model_name_or_path= 'laion/CLIP-ViT-bigG-14-laion2B-39B-b160k', size=384, trust_remote_code=True, type='transformers.CLIPImageProcessor.from_pretrained'), max_length=4096, offline_processed_text_folder= '/mnt/program/wlx_data/datasets/lmms-lab/LLaVA-Video-178K/llava_video_frames/1_2_m_academic_v0_1/tokenized-by-rwkv-v6/mc/tokenized_sandwich_data', pad_image_to_square=True, template_map_fn=dict( template='xtuner.utils.PROMPT_TEMPLATE.vicuna', type='xtuner.dataset.map_fns.template_map_fn_factory'), tokenizer=dict( padding_side='right', pretrained_model_name_or_path= '/mnt/program/wlx_data/experiments/language-1022/rwkv-1b6/iter_745524.hf', trust_remote_code=True, type='transformers.AutoTokenizer.from_pretrained'), type='xtuner.dataset.AuroraDataset') llava_dataset_6 = dict( data_path= '/mnt/program/wlx_data/datasets/lmms-lab/LLaVA-Video-178K/llava_video_frames/1_2_m_nextqa/converted_jsons/converted_1_2_m_nextqa_mc_qa_processed.json', dataset_map_fn='xtuner.dataset.map_fns.aurora_map_fn', image_folder= '/mnt/program/wlx_data/datasets/lmms-lab/LLaVA-Video-178K/llava_video_frames/1_2_m_nextqa/', image_processor=dict( crop_size=384, pretrained_model_name_or_path= 'laion/CLIP-ViT-bigG-14-laion2B-39B-b160k', size=384, trust_remote_code=True, type='transformers.CLIPImageProcessor.from_pretrained'), max_length=4096, offline_processed_text_folder= '/mnt/program/wlx_data/datasets/lmms-lab/LLaVA-Video-178K/llava_video_frames/1_2_m_nextqa/tokenized-by-rwkv-v6/mc/tokenized_sandwich_data', pad_image_to_square=True, template_map_fn=dict( template='xtuner.utils.PROMPT_TEMPLATE.vicuna', type='xtuner.dataset.map_fns.template_map_fn_factory'), tokenizer=dict( padding_side='right', pretrained_model_name_or_path= '/mnt/program/wlx_data/experiments/language-1022/rwkv-1b6/iter_745524.hf', trust_remote_code=True, type='transformers.AutoTokenizer.from_pretrained'), type='xtuner.dataset.AuroraDataset') llava_dataset_7 = dict( data_path= '/mnt/program/wlx_data/datasets/lmms-lab/LLaVA-Video-178K/llava_video_frames/1_2_m_youtube_v0_1/converted_jsons/converted_1_2_m_youtube_mc_v0_1_qa_processed.json', dataset_map_fn='xtuner.dataset.map_fns.aurora_map_fn', image_folder= '/mnt/program/wlx_data/datasets/lmms-lab/LLaVA-Video-178K/llava_video_frames/1_2_m_youtube_v0_1/', image_processor=dict( crop_size=384, pretrained_model_name_or_path= 'laion/CLIP-ViT-bigG-14-laion2B-39B-b160k', size=384, trust_remote_code=True, type='transformers.CLIPImageProcessor.from_pretrained'), max_length=4096, offline_processed_text_folder= '/mnt/program/wlx_data/datasets/lmms-lab/LLaVA-Video-178K/llava_video_frames/1_2_m_youtube_v0_1/tokenized-by-rwkv-v6/mc/tokenized_sandwich_data', pad_image_to_square=True, template_map_fn=dict( template='xtuner.utils.PROMPT_TEMPLATE.vicuna', type='xtuner.dataset.map_fns.template_map_fn_factory'), tokenizer=dict( padding_side='right', pretrained_model_name_or_path= '/mnt/program/wlx_data/experiments/language-1022/rwkv-1b6/iter_745524.hf', trust_remote_code=True, type='transformers.AutoTokenizer.from_pretrained'), type='xtuner.dataset.AuroraDataset') llava_dataset_8 = dict( data_path= '/mnt/program/wlx_data/datasets/lmms-lab/LLaVA-Video-178K/llava_video_frames/2_3_m_academic_v0_1/converted_jsons/converted_2_3_m_academic_mc_v0_1_qa_processed.json', dataset_map_fn='xtuner.dataset.map_fns.aurora_map_fn', image_folder= '/mnt/program/wlx_data/datasets/lmms-lab/LLaVA-Video-178K/llava_video_frames/2_3_m_academic_v0_1/', image_processor=dict( crop_size=384, pretrained_model_name_or_path= 'laion/CLIP-ViT-bigG-14-laion2B-39B-b160k', size=384, trust_remote_code=True, type='transformers.CLIPImageProcessor.from_pretrained'), max_length=4096, offline_processed_text_folder= '/mnt/program/wlx_data/datasets/lmms-lab/LLaVA-Video-178K/llava_video_frames/2_3_m_academic_v0_1/tokenized-by-rwkv-v6/mc/tokenized_sandwich_data', pad_image_to_square=True, template_map_fn=dict( template='xtuner.utils.PROMPT_TEMPLATE.vicuna', type='xtuner.dataset.map_fns.template_map_fn_factory'), tokenizer=dict( padding_side='right', pretrained_model_name_or_path= '/mnt/program/wlx_data/experiments/language-1022/rwkv-1b6/iter_745524.hf', trust_remote_code=True, type='transformers.AutoTokenizer.from_pretrained'), type='xtuner.dataset.AuroraDataset') llava_dataset_9 = dict( data_path= '/mnt/program/wlx_data/datasets/lmms-lab/LLaVA-Video-178K/llava_video_frames/2_3_m_nextqa/converted_jsons/converted_2_3_m_nextqa_mc_qa_processed.json', dataset_map_fn='xtuner.dataset.map_fns.aurora_map_fn', image_folder= '/mnt/program/wlx_data/datasets/lmms-lab/LLaVA-Video-178K/llava_video_frames/2_3_m_nextqa/', image_processor=dict( crop_size=384, pretrained_model_name_or_path= 'laion/CLIP-ViT-bigG-14-laion2B-39B-b160k', size=384, trust_remote_code=True, type='transformers.CLIPImageProcessor.from_pretrained'), max_length=4096, offline_processed_text_folder= '/mnt/program/wlx_data/datasets/lmms-lab/LLaVA-Video-178K/llava_video_frames/2_3_m_nextqa/tokenized-by-rwkv-v6/mc/tokenized_sandwich_data', pad_image_to_square=True, template_map_fn=dict( template='xtuner.utils.PROMPT_TEMPLATE.vicuna', type='xtuner.dataset.map_fns.template_map_fn_factory'), tokenizer=dict( padding_side='right', pretrained_model_name_or_path= '/mnt/program/wlx_data/experiments/language-1022/rwkv-1b6/iter_745524.hf', trust_remote_code=True, type='transformers.AutoTokenizer.from_pretrained'), type='xtuner.dataset.AuroraDataset') llava_ov_strategy = False llm_name_or_path = '/mnt/program/wlx_data/experiments/language-1022/rwkv-1b6/iter_745524.hf' load_from = '/mnt/program/wlx_data/experiments/language-llava-video-178k-no-sort-unfreeze-llm-1112/iter_90000.pth' log_level = 'INFO' lr = 1e-05 max_epochs = 1 max_length = 4096 max_norm = 1 model = dict( freeze_llm=False, freeze_proj=False, freeze_visual_encoder=False, llava_ov_strategy=False, llm=dict( pretrained_model_name_or_path= '/mnt/program/wlx_data/experiments/language-1022/rwkv-1b6/iter_745524.hf', quantization_config=None, torch_dtype='torch.float16', trust_remote_code=True, type='transformers.AutoModelForCausalLM.from_pretrained'), pretrained_pth= '/mnt/program/wlx_data/experiments/language-1022/rwkv-1b6/iter_745524.hf/projector/', slowfast=False, type='xtuner.model.AuroraModel', visual_encoder=dict( order='ascending', pretrained_model_name_or_path= '/mnt/program/wlx_data/experiments/language-1022/rwkv-1b6/iter_745524.hf/visual_encoder/', sort_after_merge=True, type='xtuner.model.aurora.AuroraSigEncoder.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 = '/mnt/program/wlx_data/experiments/language-1022/rwkv-1b6/iter_745524.hf/projector/' prompt_template = 'xtuner.utils.PROMPT_TEMPLATE.vicuna' randomness = dict(deterministic=False, seed=None) resume = True runner_type = 'FlexibleRunner' save_steps = 10000 save_total_limit = 2 size = 384 slowfast = False 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=256, 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= '/mnt/program/wlx_data/experiments/language-1022/rwkv-1b6/iter_745524.hf', 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( data_path= '/mnt/program/wlx_data/datasets/lmms-lab/LLaVA-Video-178K/llava_video_frames/0_30_s_academic_v0_1/converted_jsons/filtered_converted_0_30_s_academic_mc_v0_1_qa_processed.json', dataset_map_fn='xtuner.dataset.map_fns.aurora_map_fn', image_folder= '/mnt/program/wlx_data/datasets/lmms-lab/LLaVA-Video-178K/llava_video_frames/0_30_s_academic_v0_1/', image_processor=dict( crop_size=384, pretrained_model_name_or_path= 'laion/CLIP-ViT-bigG-14-laion2B-39B-b160k', size=384, trust_remote_code=True, type='transformers.CLIPImageProcessor.from_pretrained'), max_length=4096, offline_processed_text_folder= '/mnt/program/wlx_data/datasets/lmms-lab/LLaVA-Video-178K/llava_video_frames/0_30_s_academic_v0_1/tokenized-by-rwkv-v6/mc/tokenized_sandwich_data', pad_image_to_square=True, template_map_fn=dict( template='xtuner.utils.PROMPT_TEMPLATE.vicuna', type='xtuner.dataset.map_fns.template_map_fn_factory'), tokenizer=dict( padding_side='right', pretrained_model_name_or_path= '/mnt/program/wlx_data/experiments/language-1022/rwkv-1b6/iter_745524.hf', trust_remote_code=True, type='transformers.AutoTokenizer.from_pretrained'), type='xtuner.dataset.AuroraDataset'), dict( data_path= '/mnt/program/wlx_data/datasets/lmms-lab/LLaVA-Video-178K/llava_video_frames/0_30_s_nextqa/converted_jsons/converted_0_30_s_nextqa_mc_qa_processed.json', dataset_map_fn='xtuner.dataset.map_fns.aurora_map_fn', image_folder= '/mnt/program/wlx_data/datasets/lmms-lab/LLaVA-Video-178K/llava_video_frames/0_30_s_nextqa/', image_processor=dict( crop_size=384, pretrained_model_name_or_path= 'laion/CLIP-ViT-bigG-14-laion2B-39B-b160k', size=384, trust_remote_code=True, type='transformers.CLIPImageProcessor.from_pretrained'), max_length=4096, offline_processed_text_folder= '/mnt/program/wlx_data/datasets/lmms-lab/LLaVA-Video-178K/llava_video_frames/0_30_s_nextqa/tokenized-by-rwkv-v6/mc/tokenized_sandwich_data', pad_image_to_square=True, template_map_fn=dict( template='xtuner.utils.PROMPT_TEMPLATE.vicuna', type='xtuner.dataset.map_fns.template_map_fn_factory'), tokenizer=dict( padding_side='right', pretrained_model_name_or_path= '/mnt/program/wlx_data/experiments/language-1022/rwkv-1b6/iter_745524.hf', trust_remote_code=True, type='transformers.AutoTokenizer.from_pretrained'), type='xtuner.dataset.AuroraDataset'), dict( data_path= '/mnt/program/wlx_data/datasets/lmms-lab/LLaVA-Video-178K/llava_video_frames/0_30_s_perceptiontest/converted_jsons/converted_0_30_s_perceptiontest_mc_qa_processed.json', dataset_map_fn='xtuner.dataset.map_fns.aurora_map_fn', image_folder= '/mnt/program/wlx_data/datasets/lmms-lab/LLaVA-Video-178K/llava_video_frames/0_30_s_perceptiontest/', image_processor=dict( crop_size=384, pretrained_model_name_or_path= 'laion/CLIP-ViT-bigG-14-laion2B-39B-b160k', size=384, trust_remote_code=True, type='transformers.CLIPImageProcessor.from_pretrained'), max_length=4096, offline_processed_text_folder= '/mnt/program/wlx_data/datasets/lmms-lab/LLaVA-Video-178K/llava_video_frames/0_30_s_perceptiontest/tokenized-by-rwkv-v6/mc/tokenized_sandwich_data', pad_image_to_square=True, template_map_fn=dict( template='xtuner.utils.PROMPT_TEMPLATE.vicuna', type='xtuner.dataset.map_fns.template_map_fn_factory'), tokenizer=dict( padding_side='right', pretrained_model_name_or_path= '/mnt/program/wlx_data/experiments/language-1022/rwkv-1b6/iter_745524.hf', trust_remote_code=True, type='transformers.AutoTokenizer.from_pretrained'), type='xtuner.dataset.AuroraDataset'), dict( data_path= '/mnt/program/wlx_data/datasets/lmms-lab/LLaVA-Video-178K/llava_video_frames/0_30_s_youtube_v0_1/converted_jsons/filtered_converted_0_30_s_youtube_mc_v0_1_qa_processed.json', dataset_map_fn='xtuner.dataset.map_fns.aurora_map_fn', image_folder= '/mnt/program/wlx_data/datasets/lmms-lab/LLaVA-Video-178K/llava_video_frames/0_30_s_youtube_v0_1/', image_processor=dict( crop_size=384, pretrained_model_name_or_path= 'laion/CLIP-ViT-bigG-14-laion2B-39B-b160k', size=384, trust_remote_code=True, type='transformers.CLIPImageProcessor.from_pretrained'), max_length=4096, offline_processed_text_folder= '/mnt/program/wlx_data/datasets/lmms-lab/LLaVA-Video-178K/llava_video_frames/0_30_s_youtube_v0_1/tokenized-by-rwkv-v6/mc/tokenized_sandwich_data', pad_image_to_square=True, template_map_fn=dict( template='xtuner.utils.PROMPT_TEMPLATE.vicuna', type='xtuner.dataset.map_fns.template_map_fn_factory'), tokenizer=dict( padding_side='right', pretrained_model_name_or_path= '/mnt/program/wlx_data/experiments/language-1022/rwkv-1b6/iter_745524.hf', trust_remote_code=True, type='transformers.AutoTokenizer.from_pretrained'), type='xtuner.dataset.AuroraDataset'), dict( data_path= '/mnt/program/wlx_data/datasets/lmms-lab/LLaVA-Video-178K/llava_video_frames/1_2_m_academic_v0_1/converted_jsons/converted_1_2_m_academic_mc_v0_1_qa_processed.json', dataset_map_fn='xtuner.dataset.map_fns.aurora_map_fn', image_folder= '/mnt/program/wlx_data/datasets/lmms-lab/LLaVA-Video-178K/llava_video_frames/1_2_m_academic_v0_1/', image_processor=dict( crop_size=384, pretrained_model_name_or_path= 'laion/CLIP-ViT-bigG-14-laion2B-39B-b160k', size=384, trust_remote_code=True, type='transformers.CLIPImageProcessor.from_pretrained'), max_length=4096, offline_processed_text_folder= '/mnt/program/wlx_data/datasets/lmms-lab/LLaVA-Video-178K/llava_video_frames/1_2_m_academic_v0_1/tokenized-by-rwkv-v6/mc/tokenized_sandwich_data', pad_image_to_square=True, template_map_fn=dict( template='xtuner.utils.PROMPT_TEMPLATE.vicuna', type='xtuner.dataset.map_fns.template_map_fn_factory'), tokenizer=dict( padding_side='right', pretrained_model_name_or_path= '/mnt/program/wlx_data/experiments/language-1022/rwkv-1b6/iter_745524.hf', trust_remote_code=True, type='transformers.AutoTokenizer.from_pretrained'), type='xtuner.dataset.AuroraDataset'), dict( data_path= '/mnt/program/wlx_data/datasets/lmms-lab/LLaVA-Video-178K/llava_video_frames/1_2_m_nextqa/converted_jsons/converted_1_2_m_nextqa_mc_qa_processed.json', dataset_map_fn='xtuner.dataset.map_fns.aurora_map_fn', image_folder= '/mnt/program/wlx_data/datasets/lmms-lab/LLaVA-Video-178K/llava_video_frames/1_2_m_nextqa/', image_processor=dict( crop_size=384, pretrained_model_name_or_path= 'laion/CLIP-ViT-bigG-14-laion2B-39B-b160k', size=384, trust_remote_code=True, type='transformers.CLIPImageProcessor.from_pretrained'), max_length=4096, offline_processed_text_folder= '/mnt/program/wlx_data/datasets/lmms-lab/LLaVA-Video-178K/llava_video_frames/1_2_m_nextqa/tokenized-by-rwkv-v6/mc/tokenized_sandwich_data', pad_image_to_square=True, template_map_fn=dict( template='xtuner.utils.PROMPT_TEMPLATE.vicuna', type='xtuner.dataset.map_fns.template_map_fn_factory'), tokenizer=dict( padding_side='right', pretrained_model_name_or_path= '/mnt/program/wlx_data/experiments/language-1022/rwkv-1b6/iter_745524.hf', trust_remote_code=True, type='transformers.AutoTokenizer.from_pretrained'), type='xtuner.dataset.AuroraDataset'), dict( data_path= '/mnt/program/wlx_data/datasets/lmms-lab/LLaVA-Video-178K/llava_video_frames/1_2_m_youtube_v0_1/converted_jsons/converted_1_2_m_youtube_mc_v0_1_qa_processed.json', dataset_map_fn='xtuner.dataset.map_fns.aurora_map_fn', image_folder= '/mnt/program/wlx_data/datasets/lmms-lab/LLaVA-Video-178K/llava_video_frames/1_2_m_youtube_v0_1/', image_processor=dict( crop_size=384, pretrained_model_name_or_path= 'laion/CLIP-ViT-bigG-14-laion2B-39B-b160k', size=384, trust_remote_code=True, type='transformers.CLIPImageProcessor.from_pretrained'), max_length=4096, offline_processed_text_folder= '/mnt/program/wlx_data/datasets/lmms-lab/LLaVA-Video-178K/llava_video_frames/1_2_m_youtube_v0_1/tokenized-by-rwkv-v6/mc/tokenized_sandwich_data', pad_image_to_square=True, template_map_fn=dict( template='xtuner.utils.PROMPT_TEMPLATE.vicuna', type='xtuner.dataset.map_fns.template_map_fn_factory'), tokenizer=dict( padding_side='right', pretrained_model_name_or_path= '/mnt/program/wlx_data/experiments/language-1022/rwkv-1b6/iter_745524.hf', trust_remote_code=True, type='transformers.AutoTokenizer.from_pretrained'), type='xtuner.dataset.AuroraDataset'), dict( data_path= '/mnt/program/wlx_data/datasets/lmms-lab/LLaVA-Video-178K/llava_video_frames/2_3_m_academic_v0_1/converted_jsons/converted_2_3_m_academic_mc_v0_1_qa_processed.json', dataset_map_fn='xtuner.dataset.map_fns.aurora_map_fn', image_folder= '/mnt/program/wlx_data/datasets/lmms-lab/LLaVA-Video-178K/llava_video_frames/2_3_m_academic_v0_1/', image_processor=dict( crop_size=384, pretrained_model_name_or_path= 'laion/CLIP-ViT-bigG-14-laion2B-39B-b160k', size=384, trust_remote_code=True, type='transformers.CLIPImageProcessor.from_pretrained'), max_length=4096, offline_processed_text_folder= '/mnt/program/wlx_data/datasets/lmms-lab/LLaVA-Video-178K/llava_video_frames/2_3_m_academic_v0_1/tokenized-by-rwkv-v6/mc/tokenized_sandwich_data', pad_image_to_square=True, template_map_fn=dict( template='xtuner.utils.PROMPT_TEMPLATE.vicuna', type='xtuner.dataset.map_fns.template_map_fn_factory'), tokenizer=dict( padding_side='right', pretrained_model_name_or_path= '/mnt/program/wlx_data/experiments/language-1022/rwkv-1b6/iter_745524.hf', trust_remote_code=True, type='transformers.AutoTokenizer.from_pretrained'), type='xtuner.dataset.AuroraDataset'), dict( data_path= '/mnt/program/wlx_data/datasets/lmms-lab/LLaVA-Video-178K/llava_video_frames/2_3_m_nextqa/converted_jsons/converted_2_3_m_nextqa_mc_qa_processed.json', dataset_map_fn='xtuner.dataset.map_fns.aurora_map_fn', image_folder= '/mnt/program/wlx_data/datasets/lmms-lab/LLaVA-Video-178K/llava_video_frames/2_3_m_nextqa/', image_processor=dict( crop_size=384, pretrained_model_name_or_path= 'laion/CLIP-ViT-bigG-14-laion2B-39B-b160k', size=384, trust_remote_code=True, type='transformers.CLIPImageProcessor.from_pretrained'), max_length=4096, offline_processed_text_folder= '/mnt/program/wlx_data/datasets/lmms-lab/LLaVA-Video-178K/llava_video_frames/2_3_m_nextqa/tokenized-by-rwkv-v6/mc/tokenized_sandwich_data', pad_image_to_square=True, template_map_fn=dict( template='xtuner.utils.PROMPT_TEMPLATE.vicuna', type='xtuner.dataset.map_fns.template_map_fn_factory'), tokenizer=dict( padding_side='right', pretrained_model_name_or_path= '/mnt/program/wlx_data/experiments/language-1022/rwkv-1b6/iter_745524.hf', trust_remote_code=True, type='transformers.AutoTokenizer.from_pretrained'), type='xtuner.dataset.AuroraDataset'), dict( data_path= '/mnt/program/wlx_data/datasets/lmms-lab/LLaVA-Video-178K/llava_video_frames/2_3_m_youtube_v0_1/converted_jsons/converted_2_3_m_youtube_mc_v0_1_qa_processed.json', dataset_map_fn='xtuner.dataset.map_fns.aurora_map_fn', image_folder= '/mnt/program/wlx_data/datasets/lmms-lab/LLaVA-Video-178K/llava_video_frames/2_3_m_youtube_v0_1/', image_processor=dict( crop_size=384, pretrained_model_name_or_path= 'laion/CLIP-ViT-bigG-14-laion2B-39B-b160k', size=384, trust_remote_code=True, type='transformers.CLIPImageProcessor.from_pretrained'), max_length=4096, offline_processed_text_folder= '/mnt/program/wlx_data/datasets/lmms-lab/LLaVA-Video-178K/llava_video_frames/2_3_m_youtube_v0_1/tokenized-by-rwkv-v6/mc/tokenized_sandwich_data', pad_image_to_square=True, template_map_fn=dict( template='xtuner.utils.PROMPT_TEMPLATE.vicuna', type='xtuner.dataset.map_fns.template_map_fn_factory'), tokenizer=dict( padding_side='right', pretrained_model_name_or_path= '/mnt/program/wlx_data/experiments/language-1022/rwkv-1b6/iter_745524.hf', trust_remote_code=True, type='transformers.AutoTokenizer.from_pretrained'), type='xtuner.dataset.AuroraDataset'), dict( data_path= '/mnt/program/wlx_data/datasets/lmms-lab/LLaVA-Video-178K/llava_video_frames/30_60_s_academic_v0_1/converted_jsons/converted_30_60_s_academic_mc_v0_1_qa_processed.json', dataset_map_fn='xtuner.dataset.map_fns.aurora_map_fn', image_folder= '/mnt/program/wlx_data/datasets/lmms-lab/LLaVA-Video-178K/llava_video_frames/30_60_s_academic_v0_1/', image_processor=dict( crop_size=384, pretrained_model_name_or_path= 'laion/CLIP-ViT-bigG-14-laion2B-39B-b160k', size=384, trust_remote_code=True, type='transformers.CLIPImageProcessor.from_pretrained'), max_length=4096, offline_processed_text_folder= '/mnt/program/wlx_data/datasets/lmms-lab/LLaVA-Video-178K/llava_video_frames/30_60_s_academic_v0_1/tokenized-by-rwkv-v6/mc/tokenized_sandwich_data', pad_image_to_square=True, template_map_fn=dict( template='xtuner.utils.PROMPT_TEMPLATE.vicuna', type='xtuner.dataset.map_fns.template_map_fn_factory'), tokenizer=dict( padding_side='right', pretrained_model_name_or_path= '/mnt/program/wlx_data/experiments/language-1022/rwkv-1b6/iter_745524.hf', trust_remote_code=True, type='transformers.AutoTokenizer.from_pretrained'), type='xtuner.dataset.AuroraDataset'), dict( data_path= '/mnt/program/wlx_data/datasets/lmms-lab/LLaVA-Video-178K/llava_video_frames/30_60_s_nextqa/converted_jsons/converted_30_60_s_nextqa_mc_qa_processed.json', dataset_map_fn='xtuner.dataset.map_fns.aurora_map_fn', image_folder= '/mnt/program/wlx_data/datasets/lmms-lab/LLaVA-Video-178K/llava_video_frames/30_60_s_nextqa/', image_processor=dict( crop_size=384, pretrained_model_name_or_path= 'laion/CLIP-ViT-bigG-14-laion2B-39B-b160k', size=384, trust_remote_code=True, type='transformers.CLIPImageProcessor.from_pretrained'), max_length=4096, offline_processed_text_folder= '/mnt/program/wlx_data/datasets/lmms-lab/LLaVA-Video-178K/llava_video_frames/30_60_s_nextqa/tokenized-by-rwkv-v6/mc/tokenized_sandwich_data', pad_image_to_square=True, template_map_fn=dict( template='xtuner.utils.PROMPT_TEMPLATE.vicuna', type='xtuner.dataset.map_fns.template_map_fn_factory'), tokenizer=dict( padding_side='right', pretrained_model_name_or_path= '/mnt/program/wlx_data/experiments/language-1022/rwkv-1b6/iter_745524.hf', trust_remote_code=True, type='transformers.AutoTokenizer.from_pretrained'), type='xtuner.dataset.AuroraDataset'), dict( data_path= '/mnt/program/wlx_data/datasets/lmms-lab/LLaVA-Video-178K/llava_video_frames/30_60_s_perceptiontest/converted_jsons/converted_30_60_s_perceptiontest_mc_qa_processed.json', dataset_map_fn='xtuner.dataset.map_fns.aurora_map_fn', image_folder= '/mnt/program/wlx_data/datasets/lmms-lab/LLaVA-Video-178K/llava_video_frames/30_60_s_perceptiontest/', image_processor=dict( crop_size=384, pretrained_model_name_or_path= 'laion/CLIP-ViT-bigG-14-laion2B-39B-b160k', size=384, trust_remote_code=True, type='transformers.CLIPImageProcessor.from_pretrained'), max_length=4096, offline_processed_text_folder= '/mnt/program/wlx_data/datasets/lmms-lab/LLaVA-Video-178K/llava_video_frames/30_60_s_perceptiontest/tokenized-by-rwkv-v6/mc/tokenized_sandwich_data', pad_image_to_square=True, template_map_fn=dict( template='xtuner.utils.PROMPT_TEMPLATE.vicuna', type='xtuner.dataset.map_fns.template_map_fn_factory'), tokenizer=dict( padding_side='right', pretrained_model_name_or_path= '/mnt/program/wlx_data/experiments/language-1022/rwkv-1b6/iter_745524.hf', trust_remote_code=True, type='transformers.AutoTokenizer.from_pretrained'), type='xtuner.dataset.AuroraDataset'), dict( data_path= '/mnt/program/wlx_data/datasets/lmms-lab/LLaVA-Video-178K/llava_video_frames/30_60_s_youtube_v0_1/converted_jsons/converted_30_60_s_youtube_mc_v0_1_qa_processed.json', dataset_map_fn='xtuner.dataset.map_fns.aurora_map_fn', image_folder= '/mnt/program/wlx_data/datasets/lmms-lab/LLaVA-Video-178K/llava_video_frames/30_60_s_youtube_v0_1/', image_processor=dict( crop_size=384, pretrained_model_name_or_path= 'laion/CLIP-ViT-bigG-14-laion2B-39B-b160k', size=384, trust_remote_code=True, type='transformers.CLIPImageProcessor.from_pretrained'), max_length=4096, offline_processed_text_folder= '/mnt/program/wlx_data/datasets/lmms-lab/LLaVA-Video-178K/llava_video_frames/30_60_s_youtube_v0_1/tokenized-by-rwkv-v6/mc/tokenized_sandwich_data', pad_image_to_square=True, template_map_fn=dict( template='xtuner.utils.PROMPT_TEMPLATE.vicuna', type='xtuner.dataset.map_fns.template_map_fn_factory'), tokenizer=dict( padding_side='right', pretrained_model_name_or_path= '/mnt/program/wlx_data/experiments/language-1022/rwkv-1b6/iter_745524.hf', trust_remote_code=True, type='transformers.AutoTokenizer.from_pretrained'), type='xtuner.dataset.AuroraDataset'), dict( data_path= '/mnt/program/wlx_data/datasets/lmms-lab/LLaVA-Video-178K/llava_video_frames/0_30_s_academic_v0_1/converted_jsons/filtered_converted_0_30_s_academic_oe_v0_1_qa_processed.json', dataset_map_fn='xtuner.dataset.map_fns.aurora_map_fn', image_folder= '/mnt/program/wlx_data/datasets/lmms-lab/LLaVA-Video-178K/llava_video_frames/0_30_s_academic_v0_1/', image_processor=dict( crop_size=384, pretrained_model_name_or_path= 'laion/CLIP-ViT-bigG-14-laion2B-39B-b160k', size=384, trust_remote_code=True, type='transformers.CLIPImageProcessor.from_pretrained'), max_length=4096, offline_processed_text_folder= '/mnt/program/wlx_data/datasets/lmms-lab/LLaVA-Video-178K/llava_video_frames/0_30_s_academic_v0_1/tokenized-by-rwkv-v6/oe/tokenized_sandwich_data', pad_image_to_square=True, template_map_fn=dict( template='xtuner.utils.PROMPT_TEMPLATE.vicuna', type='xtuner.dataset.map_fns.template_map_fn_factory'), tokenizer=dict( padding_side='right', pretrained_model_name_or_path= '/mnt/program/wlx_data/experiments/language-1022/rwkv-1b6/iter_745524.hf', trust_remote_code=True, type='transformers.AutoTokenizer.from_pretrained'), type='xtuner.dataset.AuroraDataset'), dict( data_path= '/mnt/program/wlx_data/datasets/lmms-lab/LLaVA-Video-178K/llava_video_frames/0_30_s_activitynetqa/converted_jsons/converted_0_30_s_activitynetqa_oe_qa_processed.json', dataset_map_fn='xtuner.dataset.map_fns.aurora_map_fn', image_folder= '/mnt/program/wlx_data/datasets/lmms-lab/LLaVA-Video-178K/llava_video_frames/0_30_s_activitynetqa/', image_processor=dict( crop_size=384, pretrained_model_name_or_path= 'laion/CLIP-ViT-bigG-14-laion2B-39B-b160k', size=384, trust_remote_code=True, type='transformers.CLIPImageProcessor.from_pretrained'), max_length=4096, offline_processed_text_folder= '/mnt/program/wlx_data/datasets/lmms-lab/LLaVA-Video-178K/llava_video_frames/0_30_s_activitynetqa/tokenized-by-rwkv-v6/oe/tokenized_sandwich_data', pad_image_to_square=True, template_map_fn=dict( template='xtuner.utils.PROMPT_TEMPLATE.vicuna', type='xtuner.dataset.map_fns.template_map_fn_factory'), tokenizer=dict( padding_side='right', pretrained_model_name_or_path= '/mnt/program/wlx_data/experiments/language-1022/rwkv-1b6/iter_745524.hf', trust_remote_code=True, type='transformers.AutoTokenizer.from_pretrained'), type='xtuner.dataset.AuroraDataset'), dict( data_path= '/mnt/program/wlx_data/datasets/lmms-lab/LLaVA-Video-178K/llava_video_frames/0_30_s_nextqa/converted_jsons/converted_0_30_s_nextqa_oe_qa_processed.json', dataset_map_fn='xtuner.dataset.map_fns.aurora_map_fn', image_folder= '/mnt/program/wlx_data/datasets/lmms-lab/LLaVA-Video-178K/llava_video_frames/0_30_s_nextqa/', image_processor=dict( crop_size=384, pretrained_model_name_or_path= 'laion/CLIP-ViT-bigG-14-laion2B-39B-b160k', size=384, trust_remote_code=True, type='transformers.CLIPImageProcessor.from_pretrained'), max_length=4096, offline_processed_text_folder= '/mnt/program/wlx_data/datasets/lmms-lab/LLaVA-Video-178K/llava_video_frames/0_30_s_nextqa/tokenized-by-rwkv-v6/oe/tokenized_sandwich_data', pad_image_to_square=True, template_map_fn=dict( template='xtuner.utils.PROMPT_TEMPLATE.vicuna', type='xtuner.dataset.map_fns.template_map_fn_factory'), tokenizer=dict( padding_side='right', pretrained_model_name_or_path= '/mnt/program/wlx_data/experiments/language-1022/rwkv-1b6/iter_745524.hf', trust_remote_code=True, type='transformers.AutoTokenizer.from_pretrained'), type='xtuner.dataset.AuroraDataset'), dict( data_path= '/mnt/program/wlx_data/datasets/lmms-lab/LLaVA-Video-178K/llava_video_frames/0_30_s_youtube_v0_1/converted_jsons/filtered_converted_0_30_s_youtube_oe_v0_1_qa_processed.json', dataset_map_fn='xtuner.dataset.map_fns.aurora_map_fn', image_folder= '/mnt/program/wlx_data/datasets/lmms-lab/LLaVA-Video-178K/llava_video_frames/0_30_s_youtube_v0_1/', image_processor=dict( crop_size=384, pretrained_model_name_or_path= 'laion/CLIP-ViT-bigG-14-laion2B-39B-b160k', size=384, trust_remote_code=True, type='transformers.CLIPImageProcessor.from_pretrained'), max_length=4096, offline_processed_text_folder= '/mnt/program/wlx_data/datasets/lmms-lab/LLaVA-Video-178K/llava_video_frames/0_30_s_youtube_v0_1/tokenized-by-rwkv-v6/oe/tokenized_sandwich_data', pad_image_to_square=True, template_map_fn=dict( template='xtuner.utils.PROMPT_TEMPLATE.vicuna', type='xtuner.dataset.map_fns.template_map_fn_factory'), tokenizer=dict( padding_side='right', pretrained_model_name_or_path= '/mnt/program/wlx_data/experiments/language-1022/rwkv-1b6/iter_745524.hf', trust_remote_code=True, type='transformers.AutoTokenizer.from_pretrained'), type='xtuner.dataset.AuroraDataset'), dict( data_path= '/mnt/program/wlx_data/datasets/lmms-lab/LLaVA-Video-178K/llava_video_frames/1_2_m_academic_v0_1/converted_jsons/converted_1_2_m_academic_oe_v0_1_qa_processed.json', dataset_map_fn='xtuner.dataset.map_fns.aurora_map_fn', image_folder= '/mnt/program/wlx_data/datasets/lmms-lab/LLaVA-Video-178K/llava_video_frames/1_2_m_academic_v0_1/', image_processor=dict( crop_size=384, pretrained_model_name_or_path= 'laion/CLIP-ViT-bigG-14-laion2B-39B-b160k', size=384, trust_remote_code=True, type='transformers.CLIPImageProcessor.from_pretrained'), max_length=4096, offline_processed_text_folder= '/mnt/program/wlx_data/datasets/lmms-lab/LLaVA-Video-178K/llava_video_frames/1_2_m_academic_v0_1/tokenized-by-rwkv-v6/oe/tokenized_sandwich_data', pad_image_to_square=True, template_map_fn=dict( template='xtuner.utils.PROMPT_TEMPLATE.vicuna', type='xtuner.dataset.map_fns.template_map_fn_factory'), tokenizer=dict( padding_side='right', pretrained_model_name_or_path= '/mnt/program/wlx_data/experiments/language-1022/rwkv-1b6/iter_745524.hf', trust_remote_code=True, type='transformers.AutoTokenizer.from_pretrained'), type='xtuner.dataset.AuroraDataset'), dict( data_path= '/mnt/program/wlx_data/datasets/lmms-lab/LLaVA-Video-178K/llava_video_frames/1_2_m_activitynetqa/converted_jsons/converted_1_2_m_activitynetqa_oe_qa_processed.json', dataset_map_fn='xtuner.dataset.map_fns.aurora_map_fn', image_folder= '/mnt/program/wlx_data/datasets/lmms-lab/LLaVA-Video-178K/llava_video_frames/1_2_m_activitynetqa/', image_processor=dict( crop_size=384, pretrained_model_name_or_path= 'laion/CLIP-ViT-bigG-14-laion2B-39B-b160k', size=384, trust_remote_code=True, type='transformers.CLIPImageProcessor.from_pretrained'), max_length=4096, offline_processed_text_folder= '/mnt/program/wlx_data/datasets/lmms-lab/LLaVA-Video-178K/llava_video_frames/1_2_m_activitynetqa/tokenized-by-rwkv-v6/oe/tokenized_sandwich_data', pad_image_to_square=True, template_map_fn=dict( template='xtuner.utils.PROMPT_TEMPLATE.vicuna', type='xtuner.dataset.map_fns.template_map_fn_factory'), tokenizer=dict( padding_side='right', pretrained_model_name_or_path= '/mnt/program/wlx_data/experiments/language-1022/rwkv-1b6/iter_745524.hf', trust_remote_code=True, type='transformers.AutoTokenizer.from_pretrained'), type='xtuner.dataset.AuroraDataset'), dict( data_path= '/mnt/program/wlx_data/datasets/lmms-lab/LLaVA-Video-178K/llava_video_frames/1_2_m_nextqa/converted_jsons/converted_1_2_m_nextqa_oe_qa_processed.json', dataset_map_fn='xtuner.dataset.map_fns.aurora_map_fn', image_folder= '/mnt/program/wlx_data/datasets/lmms-lab/LLaVA-Video-178K/llava_video_frames/1_2_m_nextqa/', image_processor=dict( crop_size=384, pretrained_model_name_or_path= 'laion/CLIP-ViT-bigG-14-laion2B-39B-b160k', size=384, trust_remote_code=True, type='transformers.CLIPImageProcessor.from_pretrained'), max_length=4096, offline_processed_text_folder= '/mnt/program/wlx_data/datasets/lmms-lab/LLaVA-Video-178K/llava_video_frames/1_2_m_nextqa/tokenized-by-rwkv-v6/oe/tokenized_sandwich_data', pad_image_to_square=True, template_map_fn=dict( template='xtuner.utils.PROMPT_TEMPLATE.vicuna', type='xtuner.dataset.map_fns.template_map_fn_factory'), tokenizer=dict( padding_side='right', pretrained_model_name_or_path= '/mnt/program/wlx_data/experiments/language-1022/rwkv-1b6/iter_745524.hf', trust_remote_code=True, type='transformers.AutoTokenizer.from_pretrained'), type='xtuner.dataset.AuroraDataset'), dict( data_path= '/mnt/program/wlx_data/datasets/lmms-lab/LLaVA-Video-178K/llava_video_frames/1_2_m_youtube_v0_1/converted_jsons/converted_1_2_m_youtube_oe_v0_1_qa_processed.json', dataset_map_fn='xtuner.dataset.map_fns.aurora_map_fn', image_folder= '/mnt/program/wlx_data/datasets/lmms-lab/LLaVA-Video-178K/llava_video_frames/1_2_m_youtube_v0_1/', image_processor=dict( crop_size=384, pretrained_model_name_or_path= 'laion/CLIP-ViT-bigG-14-laion2B-39B-b160k', size=384, trust_remote_code=True, type='transformers.CLIPImageProcessor.from_pretrained'), max_length=4096, offline_processed_text_folder= '/mnt/program/wlx_data/datasets/lmms-lab/LLaVA-Video-178K/llava_video_frames/1_2_m_youtube_v0_1/tokenized-by-rwkv-v6/oe/tokenized_sandwich_data', pad_image_to_square=True, template_map_fn=dict( template='xtuner.utils.PROMPT_TEMPLATE.vicuna', type='xtuner.dataset.map_fns.template_map_fn_factory'), tokenizer=dict( padding_side='right', pretrained_model_name_or_path= '/mnt/program/wlx_data/experiments/language-1022/rwkv-1b6/iter_745524.hf', trust_remote_code=True, type='transformers.AutoTokenizer.from_pretrained'), type='xtuner.dataset.AuroraDataset'), dict( data_path= '/mnt/program/wlx_data/datasets/lmms-lab/LLaVA-Video-178K/llava_video_frames/2_3_m_academic_v0_1/converted_jsons/converted_2_3_m_academic_oe_v0_1_qa_processed.json', dataset_map_fn='xtuner.dataset.map_fns.aurora_map_fn', image_folder= '/mnt/program/wlx_data/datasets/lmms-lab/LLaVA-Video-178K/llava_video_frames/2_3_m_academic_v0_1/', image_processor=dict( crop_size=384, pretrained_model_name_or_path= 'laion/CLIP-ViT-bigG-14-laion2B-39B-b160k', size=384, trust_remote_code=True, type='transformers.CLIPImageProcessor.from_pretrained'), max_length=4096, offline_processed_text_folder= '/mnt/program/wlx_data/datasets/lmms-lab/LLaVA-Video-178K/llava_video_frames/2_3_m_academic_v0_1/tokenized-by-rwkv-v6/oe/tokenized_sandwich_data', pad_image_to_square=True, template_map_fn=dict( template='xtuner.utils.PROMPT_TEMPLATE.vicuna', type='xtuner.dataset.map_fns.template_map_fn_factory'), tokenizer=dict( padding_side='right', pretrained_model_name_or_path= '/mnt/program/wlx_data/experiments/language-1022/rwkv-1b6/iter_745524.hf', trust_remote_code=True, type='transformers.AutoTokenizer.from_pretrained'), type='xtuner.dataset.AuroraDataset'), dict( data_path= '/mnt/program/wlx_data/datasets/lmms-lab/LLaVA-Video-178K/llava_video_frames/2_3_m_activitynetqa/converted_jsons/converted_2_3_m_activitynetqa_oe_qa_processed.json', dataset_map_fn='xtuner.dataset.map_fns.aurora_map_fn', image_folder= '/mnt/program/wlx_data/datasets/lmms-lab/LLaVA-Video-178K/llava_video_frames/2_3_m_activitynetqa/', image_processor=dict( crop_size=384, pretrained_model_name_or_path= 'laion/CLIP-ViT-bigG-14-laion2B-39B-b160k', size=384, trust_remote_code=True, type='transformers.CLIPImageProcessor.from_pretrained'), max_length=4096, offline_processed_text_folder= '/mnt/program/wlx_data/datasets/lmms-lab/LLaVA-Video-178K/llava_video_frames/2_3_m_activitynetqa/tokenized-by-rwkv-v6/oe/tokenized_sandwich_data', pad_image_to_square=True, template_map_fn=dict( template='xtuner.utils.PROMPT_TEMPLATE.vicuna', type='xtuner.dataset.map_fns.template_map_fn_factory'), tokenizer=dict( padding_side='right', pretrained_model_name_or_path= '/mnt/program/wlx_data/experiments/language-1022/rwkv-1b6/iter_745524.hf', trust_remote_code=True, type='transformers.AutoTokenizer.from_pretrained'), type='xtuner.dataset.AuroraDataset'), dict( data_path= '/mnt/program/wlx_data/datasets/lmms-lab/LLaVA-Video-178K/llava_video_frames/2_3_m_nextqa/converted_jsons/converted_2_3_m_nextqa_oe_qa_processed.json', dataset_map_fn='xtuner.dataset.map_fns.aurora_map_fn', image_folder= '/mnt/program/wlx_data/datasets/lmms-lab/LLaVA-Video-178K/llava_video_frames/2_3_m_nextqa/', image_processor=dict( crop_size=384, pretrained_model_name_or_path= 'laion/CLIP-ViT-bigG-14-laion2B-39B-b160k', size=384, trust_remote_code=True, type='transformers.CLIPImageProcessor.from_pretrained'), max_length=4096, offline_processed_text_folder= '/mnt/program/wlx_data/datasets/lmms-lab/LLaVA-Video-178K/llava_video_frames/2_3_m_nextqa/tokenized-by-rwkv-v6/oe/tokenized_sandwich_data', pad_image_to_square=True, template_map_fn=dict( template='xtuner.utils.PROMPT_TEMPLATE.vicuna', type='xtuner.dataset.map_fns.template_map_fn_factory'), tokenizer=dict( padding_side='right', pretrained_model_name_or_path= '/mnt/program/wlx_data/experiments/language-1022/rwkv-1b6/iter_745524.hf', trust_remote_code=True, type='transformers.AutoTokenizer.from_pretrained'), type='xtuner.dataset.AuroraDataset'), dict( data_path= '/mnt/program/wlx_data/datasets/lmms-lab/LLaVA-Video-178K/llava_video_frames/2_3_m_youtube_v0_1/converted_jsons/converted_2_3_m_youtube_oe_v0_1_qa_processed.json', dataset_map_fn='xtuner.dataset.map_fns.aurora_map_fn', image_folder= '/mnt/program/wlx_data/datasets/lmms-lab/LLaVA-Video-178K/llava_video_frames/2_3_m_youtube_v0_1/', image_processor=dict( crop_size=384, pretrained_model_name_or_path= 'laion/CLIP-ViT-bigG-14-laion2B-39B-b160k', size=384, trust_remote_code=True, type='transformers.CLIPImageProcessor.from_pretrained'), max_length=4096, offline_processed_text_folder= '/mnt/program/wlx_data/datasets/lmms-lab/LLaVA-Video-178K/llava_video_frames/2_3_m_youtube_v0_1/tokenized-by-rwkv-v6/oe/tokenized_sandwich_data', pad_image_to_square=True, template_map_fn=dict( template='xtuner.utils.PROMPT_TEMPLATE.vicuna', type='xtuner.dataset.map_fns.template_map_fn_factory'), tokenizer=dict( padding_side='right', pretrained_model_name_or_path= '/mnt/program/wlx_data/experiments/language-1022/rwkv-1b6/iter_745524.hf', trust_remote_code=True, type='transformers.AutoTokenizer.from_pretrained'), type='xtuner.dataset.AuroraDataset'), dict( data_path= '/mnt/program/wlx_data/datasets/lmms-lab/LLaVA-Video-178K/llava_video_frames/30_60_s_academic_v0_1/converted_jsons/converted_30_60_s_academic_oe_v0_1_qa_processed.json', dataset_map_fn='xtuner.dataset.map_fns.aurora_map_fn', image_folder= '/mnt/program/wlx_data/datasets/lmms-lab/LLaVA-Video-178K/llava_video_frames/30_60_s_academic_v0_1/', image_processor=dict( crop_size=384, pretrained_model_name_or_path= 'laion/CLIP-ViT-bigG-14-laion2B-39B-b160k', size=384, trust_remote_code=True, type='transformers.CLIPImageProcessor.from_pretrained'), max_length=4096, offline_processed_text_folder= '/mnt/program/wlx_data/datasets/lmms-lab/LLaVA-Video-178K/llava_video_frames/30_60_s_academic_v0_1/tokenized-by-rwkv-v6/oe/tokenized_sandwich_data', pad_image_to_square=True, template_map_fn=dict( template='xtuner.utils.PROMPT_TEMPLATE.vicuna', type='xtuner.dataset.map_fns.template_map_fn_factory'), tokenizer=dict( padding_side='right', pretrained_model_name_or_path= '/mnt/program/wlx_data/experiments/language-1022/rwkv-1b6/iter_745524.hf', trust_remote_code=True, type='transformers.AutoTokenizer.from_pretrained'), type='xtuner.dataset.AuroraDataset'), dict( data_path= '/mnt/program/wlx_data/datasets/lmms-lab/LLaVA-Video-178K/llava_video_frames/30_60_s_activitynetqa/converted_jsons/converted_30_60_s_activitynetqa_oe_qa_processed.json', dataset_map_fn='xtuner.dataset.map_fns.aurora_map_fn', image_folder= '/mnt/program/wlx_data/datasets/lmms-lab/LLaVA-Video-178K/llava_video_frames/30_60_s_activitynetqa/', image_processor=dict( crop_size=384, pretrained_model_name_or_path= 'laion/CLIP-ViT-bigG-14-laion2B-39B-b160k', size=384, trust_remote_code=True, type='transformers.CLIPImageProcessor.from_pretrained'), max_length=4096, offline_processed_text_folder= '/mnt/program/wlx_data/datasets/lmms-lab/LLaVA-Video-178K/llava_video_frames/30_60_s_activitynetqa/tokenized-by-rwkv-v6/oe/tokenized_sandwich_data', pad_image_to_square=True, template_map_fn=dict( template='xtuner.utils.PROMPT_TEMPLATE.vicuna', type='xtuner.dataset.map_fns.template_map_fn_factory'), tokenizer=dict( padding_side='right', pretrained_model_name_or_path= '/mnt/program/wlx_data/experiments/language-1022/rwkv-1b6/iter_745524.hf', trust_remote_code=True, type='transformers.AutoTokenizer.from_pretrained'), type='xtuner.dataset.AuroraDataset'), dict( data_path= '/mnt/program/wlx_data/datasets/lmms-lab/LLaVA-Video-178K/llava_video_frames/30_60_s_nextqa/converted_jsons/converted_30_60_s_nextqa_oe_qa_processed.json', dataset_map_fn='xtuner.dataset.map_fns.aurora_map_fn', image_folder= '/mnt/program/wlx_data/datasets/lmms-lab/LLaVA-Video-178K/llava_video_frames/30_60_s_nextqa/', image_processor=dict( crop_size=384, pretrained_model_name_or_path= 'laion/CLIP-ViT-bigG-14-laion2B-39B-b160k', size=384, trust_remote_code=True, type='transformers.CLIPImageProcessor.from_pretrained'), max_length=4096, offline_processed_text_folder= '/mnt/program/wlx_data/datasets/lmms-lab/LLaVA-Video-178K/llava_video_frames/30_60_s_nextqa/tokenized-by-rwkv-v6/oe/tokenized_sandwich_data', pad_image_to_square=True, template_map_fn=dict( template='xtuner.utils.PROMPT_TEMPLATE.vicuna', type='xtuner.dataset.map_fns.template_map_fn_factory'), tokenizer=dict( padding_side='right', pretrained_model_name_or_path= '/mnt/program/wlx_data/experiments/language-1022/rwkv-1b6/iter_745524.hf', trust_remote_code=True, type='transformers.AutoTokenizer.from_pretrained'), type='xtuner.dataset.AuroraDataset'), dict( data_path= '/mnt/program/wlx_data/datasets/lmms-lab/LLaVA-Video-178K/llava_video_frames/30_60_s_youtube_v0_1/converted_jsons/filtered_converted_30_60_s_youtube_oe_v0_1_qa_processed.json', dataset_map_fn='xtuner.dataset.map_fns.aurora_map_fn', image_folder= '/mnt/program/wlx_data/datasets/lmms-lab/LLaVA-Video-178K/llava_video_frames/30_60_s_youtube_v0_1/', image_processor=dict( crop_size=384, pretrained_model_name_or_path= 'laion/CLIP-ViT-bigG-14-laion2B-39B-b160k', size=384, trust_remote_code=True, type='transformers.CLIPImageProcessor.from_pretrained'), max_length=4096, offline_processed_text_folder= '/mnt/program/wlx_data/datasets/lmms-lab/LLaVA-Video-178K/llava_video_frames/30_60_s_youtube_v0_1/tokenized-by-rwkv-v6/oe/tokenized_sandwich_data', pad_image_to_square=True, template_map_fn=dict( template='xtuner.utils.PROMPT_TEMPLATE.vicuna', type='xtuner.dataset.map_fns.template_map_fn_factory'), tokenizer=dict( padding_side='right', pretrained_model_name_or_path= '/mnt/program/wlx_data/experiments/language-1022/rwkv-1b6/iter_745524.hf', trust_remote_code=True, type='transformers.AutoTokenizer.from_pretrained'), type='xtuner.dataset.AuroraDataset'), dict( data_path= '/mnt/program/wlx_data/datasets/lmms-lab/LLaVA-Video-178K/llava_video_frames/0_30_s_academic_v0_1/converted_jsons/filtered_converted_0_30_s_academic_v0_1_cap_processed.json', dataset_map_fn='xtuner.dataset.map_fns.aurora_map_fn', image_folder= '/mnt/program/wlx_data/datasets/lmms-lab/LLaVA-Video-178K/llava_video_frames/0_30_s_academic_v0_1/', image_processor=dict( crop_size=384, pretrained_model_name_or_path= 'laion/CLIP-ViT-bigG-14-laion2B-39B-b160k', size=384, trust_remote_code=True, type='transformers.CLIPImageProcessor.from_pretrained'), max_length=4096, offline_processed_text_folder= '/mnt/program/wlx_data/datasets/lmms-lab/LLaVA-Video-178K/llava_video_frames/0_30_s_academic_v0_1/tokenized-by-rwkv-v6/cap/tokenized_sandwich_data', pad_image_to_square=True, template_map_fn=dict( template='xtuner.utils.PROMPT_TEMPLATE.vicuna', type='xtuner.dataset.map_fns.template_map_fn_factory'), tokenizer=dict( padding_side='right', pretrained_model_name_or_path= '/mnt/program/wlx_data/experiments/language-1022/rwkv-1b6/iter_745524.hf', trust_remote_code=True, type='transformers.AutoTokenizer.from_pretrained'), type='xtuner.dataset.AuroraDataset'), dict( data_path= '/mnt/program/wlx_data/datasets/lmms-lab/LLaVA-Video-178K/llava_video_frames/0_30_s_youtube_v0_1/converted_jsons/filtered_converted_0_30_s_youtube_v0_1_cap_processed.json', dataset_map_fn='xtuner.dataset.map_fns.aurora_map_fn', image_folder= '/mnt/program/wlx_data/datasets/lmms-lab/LLaVA-Video-178K/llava_video_frames/0_30_s_youtube_v0_1/', image_processor=dict( crop_size=384, pretrained_model_name_or_path= 'laion/CLIP-ViT-bigG-14-laion2B-39B-b160k', size=384, trust_remote_code=True, type='transformers.CLIPImageProcessor.from_pretrained'), max_length=4096, offline_processed_text_folder= '/mnt/program/wlx_data/datasets/lmms-lab/LLaVA-Video-178K/llava_video_frames/0_30_s_youtube_v0_1/tokenized-by-rwkv-v6/cap/tokenized_sandwich_data', pad_image_to_square=True, template_map_fn=dict( template='xtuner.utils.PROMPT_TEMPLATE.vicuna', type='xtuner.dataset.map_fns.template_map_fn_factory'), tokenizer=dict( padding_side='right', pretrained_model_name_or_path= '/mnt/program/wlx_data/experiments/language-1022/rwkv-1b6/iter_745524.hf', trust_remote_code=True, type='transformers.AutoTokenizer.from_pretrained'), type='xtuner.dataset.AuroraDataset'), dict( data_path= '/mnt/program/wlx_data/datasets/lmms-lab/LLaVA-Video-178K/llava_video_frames/1_2_m_academic_v0_1/converted_jsons/converted_1_2_m_academic_v0_1_cap_processed.json', dataset_map_fn='xtuner.dataset.map_fns.aurora_map_fn', image_folder= '/mnt/program/wlx_data/datasets/lmms-lab/LLaVA-Video-178K/llava_video_frames/1_2_m_academic_v0_1/', image_processor=dict( crop_size=384, pretrained_model_name_or_path= 'laion/CLIP-ViT-bigG-14-laion2B-39B-b160k', size=384, trust_remote_code=True, type='transformers.CLIPImageProcessor.from_pretrained'), max_length=4096, offline_processed_text_folder= '/mnt/program/wlx_data/datasets/lmms-lab/LLaVA-Video-178K/llava_video_frames/1_2_m_academic_v0_1/tokenized-by-rwkv-v6/cap/tokenized_sandwich_data', pad_image_to_square=True, template_map_fn=dict( template='xtuner.utils.PROMPT_TEMPLATE.vicuna', type='xtuner.dataset.map_fns.template_map_fn_factory'), tokenizer=dict( padding_side='right', pretrained_model_name_or_path= '/mnt/program/wlx_data/experiments/language-1022/rwkv-1b6/iter_745524.hf', trust_remote_code=True, type='transformers.AutoTokenizer.from_pretrained'), type='xtuner.dataset.AuroraDataset'), dict( data_path= '/mnt/program/wlx_data/datasets/lmms-lab/LLaVA-Video-178K/llava_video_frames/1_2_m_youtube_v0_1/converted_jsons/converted_1_2_m_youtube_v0_1_cap_processed.json', dataset_map_fn='xtuner.dataset.map_fns.aurora_map_fn', image_folder= '/mnt/program/wlx_data/datasets/lmms-lab/LLaVA-Video-178K/llava_video_frames/1_2_m_youtube_v0_1/', image_processor=dict( crop_size=384, pretrained_model_name_or_path= 'laion/CLIP-ViT-bigG-14-laion2B-39B-b160k', size=384, trust_remote_code=True, type='transformers.CLIPImageProcessor.from_pretrained'), max_length=4096, offline_processed_text_folder= '/mnt/program/wlx_data/datasets/lmms-lab/LLaVA-Video-178K/llava_video_frames/1_2_m_youtube_v0_1/tokenized-by-rwkv-v6/cap/tokenized_sandwich_data', pad_image_to_square=True, template_map_fn=dict( template='xtuner.utils.PROMPT_TEMPLATE.vicuna', type='xtuner.dataset.map_fns.template_map_fn_factory'), tokenizer=dict( padding_side='right', pretrained_model_name_or_path= '/mnt/program/wlx_data/experiments/language-1022/rwkv-1b6/iter_745524.hf', trust_remote_code=True, type='transformers.AutoTokenizer.from_pretrained'), type='xtuner.dataset.AuroraDataset'), dict( data_path= '/mnt/program/wlx_data/datasets/lmms-lab/LLaVA-Video-178K/llava_video_frames/2_3_m_academic_v0_1/converted_jsons/converted_2_3_m_academic_v0_1_cap_processed.json', dataset_map_fn='xtuner.dataset.map_fns.aurora_map_fn', image_folder= '/mnt/program/wlx_data/datasets/lmms-lab/LLaVA-Video-178K/llava_video_frames/2_3_m_academic_v0_1/', image_processor=dict( crop_size=384, pretrained_model_name_or_path= 'laion/CLIP-ViT-bigG-14-laion2B-39B-b160k', size=384, trust_remote_code=True, type='transformers.CLIPImageProcessor.from_pretrained'), max_length=4096, offline_processed_text_folder= '/mnt/program/wlx_data/datasets/lmms-lab/LLaVA-Video-178K/llava_video_frames/2_3_m_academic_v0_1/tokenized-by-rwkv-v6/cap/tokenized_sandwich_data', pad_image_to_square=True, template_map_fn=dict( template='xtuner.utils.PROMPT_TEMPLATE.vicuna', type='xtuner.dataset.map_fns.template_map_fn_factory'), tokenizer=dict( padding_side='right', pretrained_model_name_or_path= '/mnt/program/wlx_data/experiments/language-1022/rwkv-1b6/iter_745524.hf', trust_remote_code=True, type='transformers.AutoTokenizer.from_pretrained'), type='xtuner.dataset.AuroraDataset'), dict( data_path= '/mnt/program/wlx_data/datasets/lmms-lab/LLaVA-Video-178K/llava_video_frames/2_3_m_youtube_v0_1/converted_jsons/converted_2_3_m_youtube_v0_1_cap_processed.json', dataset_map_fn='xtuner.dataset.map_fns.aurora_map_fn', image_folder= '/mnt/program/wlx_data/datasets/lmms-lab/LLaVA-Video-178K/llava_video_frames/2_3_m_youtube_v0_1/', image_processor=dict( crop_size=384, pretrained_model_name_or_path= 'laion/CLIP-ViT-bigG-14-laion2B-39B-b160k', size=384, trust_remote_code=True, type='transformers.CLIPImageProcessor.from_pretrained'), max_length=4096, offline_processed_text_folder= '/mnt/program/wlx_data/datasets/lmms-lab/LLaVA-Video-178K/llava_video_frames/2_3_m_youtube_v0_1/tokenized-by-rwkv-v6/cap/tokenized_sandwich_data', pad_image_to_square=True, template_map_fn=dict( template='xtuner.utils.PROMPT_TEMPLATE.vicuna', type='xtuner.dataset.map_fns.template_map_fn_factory'), tokenizer=dict( padding_side='right', pretrained_model_name_or_path= '/mnt/program/wlx_data/experiments/language-1022/rwkv-1b6/iter_745524.hf', trust_remote_code=True, type='transformers.AutoTokenizer.from_pretrained'), type='xtuner.dataset.AuroraDataset'), dict( data_path= '/mnt/program/wlx_data/datasets/lmms-lab/LLaVA-Video-178K/llava_video_frames/30_60_s_academic_v0_1/converted_jsons/converted_30_60_s_academic_v0_1_cap_processed.json', dataset_map_fn='xtuner.dataset.map_fns.aurora_map_fn', image_folder= '/mnt/program/wlx_data/datasets/lmms-lab/LLaVA-Video-178K/llava_video_frames/30_60_s_academic_v0_1/', image_processor=dict( crop_size=384, pretrained_model_name_or_path= 'laion/CLIP-ViT-bigG-14-laion2B-39B-b160k', size=384, trust_remote_code=True, type='transformers.CLIPImageProcessor.from_pretrained'), max_length=4096, offline_processed_text_folder= '/mnt/program/wlx_data/datasets/lmms-lab/LLaVA-Video-178K/llava_video_frames/30_60_s_academic_v0_1/tokenized-by-rwkv-v6/cap/tokenized_sandwich_data', pad_image_to_square=True, template_map_fn=dict( template='xtuner.utils.PROMPT_TEMPLATE.vicuna', type='xtuner.dataset.map_fns.template_map_fn_factory'), tokenizer=dict( padding_side='right', pretrained_model_name_or_path= '/mnt/program/wlx_data/experiments/language-1022/rwkv-1b6/iter_745524.hf', trust_remote_code=True, type='transformers.AutoTokenizer.from_pretrained'), type='xtuner.dataset.AuroraDataset'), dict( data_path= '/mnt/program/wlx_data/datasets/lmms-lab/LLaVA-Video-178K/llava_video_frames/30_60_s_youtube_v0_1/converted_jsons/filtered_converted_30_60_s_youtube_v0_1_cap_processed.json', dataset_map_fn='xtuner.dataset.map_fns.aurora_map_fn', image_folder= '/mnt/program/wlx_data/datasets/lmms-lab/LLaVA-Video-178K/llava_video_frames/30_60_s_youtube_v0_1/', image_processor=dict( crop_size=384, pretrained_model_name_or_path= 'laion/CLIP-ViT-bigG-14-laion2B-39B-b160k', size=384, trust_remote_code=True, type='transformers.CLIPImageProcessor.from_pretrained'), max_length=4096, offline_processed_text_folder= '/mnt/program/wlx_data/datasets/lmms-lab/LLaVA-Video-178K/llava_video_frames/30_60_s_youtube_v0_1/tokenized-by-rwkv-v6/cap/tokenized_sandwich_data', pad_image_to_square=True, template_map_fn=dict( template='xtuner.utils.PROMPT_TEMPLATE.vicuna', type='xtuner.dataset.map_fns.template_map_fn_factory'), tokenizer=dict( padding_side='right', pretrained_model_name_or_path= '/mnt/program/wlx_data/experiments/language-1022/rwkv-1b6/iter_745524.hf', trust_remote_code=True, type='transformers.AutoTokenizer.from_pretrained'), type='xtuner.dataset.AuroraDataset'), ], type='xtuner.dataset.ConcatDataset'), num_workers=8, sampler=dict(shuffle=True, type='mmengine.dataset.DefaultSampler')) train_dataset = dict( datasets=[ dict( data_path= '/mnt/program/wlx_data/datasets/lmms-lab/LLaVA-Video-178K/llava_video_frames/0_30_s_academic_v0_1/converted_jsons/filtered_converted_0_30_s_academic_mc_v0_1_qa_processed.json', dataset_map_fn='xtuner.dataset.map_fns.aurora_map_fn', image_folder= '/mnt/program/wlx_data/datasets/lmms-lab/LLaVA-Video-178K/llava_video_frames/0_30_s_academic_v0_1/', image_processor=dict( crop_size=384, pretrained_model_name_or_path= 'laion/CLIP-ViT-bigG-14-laion2B-39B-b160k', size=384, trust_remote_code=True, type='transformers.CLIPImageProcessor.from_pretrained'), max_length=4096, offline_processed_text_folder= '/mnt/program/wlx_data/datasets/lmms-lab/LLaVA-Video-178K/llava_video_frames/0_30_s_academic_v0_1/tokenized-by-rwkv-v6/mc/tokenized_sandwich_data', pad_image_to_square=True, template_map_fn=dict( template='xtuner.utils.PROMPT_TEMPLATE.vicuna', type='xtuner.dataset.map_fns.template_map_fn_factory'), tokenizer=dict( padding_side='right', pretrained_model_name_or_path= '/mnt/program/wlx_data/experiments/language-1022/rwkv-1b6/iter_745524.hf', trust_remote_code=True, type='transformers.AutoTokenizer.from_pretrained'), type='xtuner.dataset.AuroraDataset'), dict( data_path= '/mnt/program/wlx_data/datasets/lmms-lab/LLaVA-Video-178K/llava_video_frames/0_30_s_nextqa/converted_jsons/converted_0_30_s_nextqa_mc_qa_processed.json', dataset_map_fn='xtuner.dataset.map_fns.aurora_map_fn', image_folder= '/mnt/program/wlx_data/datasets/lmms-lab/LLaVA-Video-178K/llava_video_frames/0_30_s_nextqa/', image_processor=dict( crop_size=384, pretrained_model_name_or_path= 'laion/CLIP-ViT-bigG-14-laion2B-39B-b160k', size=384, trust_remote_code=True, type='transformers.CLIPImageProcessor.from_pretrained'), max_length=4096, offline_processed_text_folder= '/mnt/program/wlx_data/datasets/lmms-lab/LLaVA-Video-178K/llava_video_frames/0_30_s_nextqa/tokenized-by-rwkv-v6/mc/tokenized_sandwich_data', pad_image_to_square=True, template_map_fn=dict( template='xtuner.utils.PROMPT_TEMPLATE.vicuna', type='xtuner.dataset.map_fns.template_map_fn_factory'), tokenizer=dict( padding_side='right', pretrained_model_name_or_path= '/mnt/program/wlx_data/experiments/language-1022/rwkv-1b6/iter_745524.hf', trust_remote_code=True, type='transformers.AutoTokenizer.from_pretrained'), type='xtuner.dataset.AuroraDataset'), dict( data_path= '/mnt/program/wlx_data/datasets/lmms-lab/LLaVA-Video-178K/llava_video_frames/0_30_s_perceptiontest/converted_jsons/converted_0_30_s_perceptiontest_mc_qa_processed.json', dataset_map_fn='xtuner.dataset.map_fns.aurora_map_fn', image_folder= '/mnt/program/wlx_data/datasets/lmms-lab/LLaVA-Video-178K/llava_video_frames/0_30_s_perceptiontest/', image_processor=dict( crop_size=384, pretrained_model_name_or_path= 'laion/CLIP-ViT-bigG-14-laion2B-39B-b160k', size=384, trust_remote_code=True, type='transformers.CLIPImageProcessor.from_pretrained'), max_length=4096, offline_processed_text_folder= '/mnt/program/wlx_data/datasets/lmms-lab/LLaVA-Video-178K/llava_video_frames/0_30_s_perceptiontest/tokenized-by-rwkv-v6/mc/tokenized_sandwich_data', pad_image_to_square=True, template_map_fn=dict( template='xtuner.utils.PROMPT_TEMPLATE.vicuna', type='xtuner.dataset.map_fns.template_map_fn_factory'), tokenizer=dict( padding_side='right', pretrained_model_name_or_path= '/mnt/program/wlx_data/experiments/language-1022/rwkv-1b6/iter_745524.hf', trust_remote_code=True, type='transformers.AutoTokenizer.from_pretrained'), type='xtuner.dataset.AuroraDataset'), dict( data_path= '/mnt/program/wlx_data/datasets/lmms-lab/LLaVA-Video-178K/llava_video_frames/0_30_s_youtube_v0_1/converted_jsons/filtered_converted_0_30_s_youtube_mc_v0_1_qa_processed.json', dataset_map_fn='xtuner.dataset.map_fns.aurora_map_fn', image_folder= '/mnt/program/wlx_data/datasets/lmms-lab/LLaVA-Video-178K/llava_video_frames/0_30_s_youtube_v0_1/', image_processor=dict( crop_size=384, pretrained_model_name_or_path= 'laion/CLIP-ViT-bigG-14-laion2B-39B-b160k', size=384, trust_remote_code=True, type='transformers.CLIPImageProcessor.from_pretrained'), max_length=4096, offline_processed_text_folder= '/mnt/program/wlx_data/datasets/lmms-lab/LLaVA-Video-178K/llava_video_frames/0_30_s_youtube_v0_1/tokenized-by-rwkv-v6/mc/tokenized_sandwich_data', pad_image_to_square=True, template_map_fn=dict( template='xtuner.utils.PROMPT_TEMPLATE.vicuna', type='xtuner.dataset.map_fns.template_map_fn_factory'), tokenizer=dict( padding_side='right', pretrained_model_name_or_path= '/mnt/program/wlx_data/experiments/language-1022/rwkv-1b6/iter_745524.hf', trust_remote_code=True, type='transformers.AutoTokenizer.from_pretrained'), type='xtuner.dataset.AuroraDataset'), dict( data_path= '/mnt/program/wlx_data/datasets/lmms-lab/LLaVA-Video-178K/llava_video_frames/1_2_m_academic_v0_1/converted_jsons/converted_1_2_m_academic_mc_v0_1_qa_processed.json', dataset_map_fn='xtuner.dataset.map_fns.aurora_map_fn', image_folder= '/mnt/program/wlx_data/datasets/lmms-lab/LLaVA-Video-178K/llava_video_frames/1_2_m_academic_v0_1/', image_processor=dict( crop_size=384, pretrained_model_name_or_path= 'laion/CLIP-ViT-bigG-14-laion2B-39B-b160k', size=384, trust_remote_code=True, type='transformers.CLIPImageProcessor.from_pretrained'), max_length=4096, offline_processed_text_folder= '/mnt/program/wlx_data/datasets/lmms-lab/LLaVA-Video-178K/llava_video_frames/1_2_m_academic_v0_1/tokenized-by-rwkv-v6/mc/tokenized_sandwich_data', pad_image_to_square=True, template_map_fn=dict( template='xtuner.utils.PROMPT_TEMPLATE.vicuna', type='xtuner.dataset.map_fns.template_map_fn_factory'), tokenizer=dict( padding_side='right', pretrained_model_name_or_path= '/mnt/program/wlx_data/experiments/language-1022/rwkv-1b6/iter_745524.hf', trust_remote_code=True, type='transformers.AutoTokenizer.from_pretrained'), type='xtuner.dataset.AuroraDataset'), dict( data_path= '/mnt/program/wlx_data/datasets/lmms-lab/LLaVA-Video-178K/llava_video_frames/1_2_m_nextqa/converted_jsons/converted_1_2_m_nextqa_mc_qa_processed.json', dataset_map_fn='xtuner.dataset.map_fns.aurora_map_fn', image_folder= '/mnt/program/wlx_data/datasets/lmms-lab/LLaVA-Video-178K/llava_video_frames/1_2_m_nextqa/', image_processor=dict( crop_size=384, pretrained_model_name_or_path= 'laion/CLIP-ViT-bigG-14-laion2B-39B-b160k', size=384, trust_remote_code=True, type='transformers.CLIPImageProcessor.from_pretrained'), max_length=4096, offline_processed_text_folder= '/mnt/program/wlx_data/datasets/lmms-lab/LLaVA-Video-178K/llava_video_frames/1_2_m_nextqa/tokenized-by-rwkv-v6/mc/tokenized_sandwich_data', pad_image_to_square=True, template_map_fn=dict( template='xtuner.utils.PROMPT_TEMPLATE.vicuna', type='xtuner.dataset.map_fns.template_map_fn_factory'), tokenizer=dict( padding_side='right', pretrained_model_name_or_path= '/mnt/program/wlx_data/experiments/language-1022/rwkv-1b6/iter_745524.hf', trust_remote_code=True, type='transformers.AutoTokenizer.from_pretrained'), type='xtuner.dataset.AuroraDataset'), dict( data_path= '/mnt/program/wlx_data/datasets/lmms-lab/LLaVA-Video-178K/llava_video_frames/1_2_m_youtube_v0_1/converted_jsons/converted_1_2_m_youtube_mc_v0_1_qa_processed.json', dataset_map_fn='xtuner.dataset.map_fns.aurora_map_fn', image_folder= '/mnt/program/wlx_data/datasets/lmms-lab/LLaVA-Video-178K/llava_video_frames/1_2_m_youtube_v0_1/', image_processor=dict( crop_size=384, pretrained_model_name_or_path= 'laion/CLIP-ViT-bigG-14-laion2B-39B-b160k', size=384, trust_remote_code=True, type='transformers.CLIPImageProcessor.from_pretrained'), max_length=4096, offline_processed_text_folder= '/mnt/program/wlx_data/datasets/lmms-lab/LLaVA-Video-178K/llava_video_frames/1_2_m_youtube_v0_1/tokenized-by-rwkv-v6/mc/tokenized_sandwich_data', pad_image_to_square=True, template_map_fn=dict( template='xtuner.utils.PROMPT_TEMPLATE.vicuna', type='xtuner.dataset.map_fns.template_map_fn_factory'), tokenizer=dict( padding_side='right', pretrained_model_name_or_path= '/mnt/program/wlx_data/experiments/language-1022/rwkv-1b6/iter_745524.hf', trust_remote_code=True, type='transformers.AutoTokenizer.from_pretrained'), type='xtuner.dataset.AuroraDataset'), dict( data_path= '/mnt/program/wlx_data/datasets/lmms-lab/LLaVA-Video-178K/llava_video_frames/2_3_m_academic_v0_1/converted_jsons/converted_2_3_m_academic_mc_v0_1_qa_processed.json', dataset_map_fn='xtuner.dataset.map_fns.aurora_map_fn', image_folder= '/mnt/program/wlx_data/datasets/lmms-lab/LLaVA-Video-178K/llava_video_frames/2_3_m_academic_v0_1/', image_processor=dict( crop_size=384, pretrained_model_name_or_path= 'laion/CLIP-ViT-bigG-14-laion2B-39B-b160k', size=384, trust_remote_code=True, type='transformers.CLIPImageProcessor.from_pretrained'), max_length=4096, offline_processed_text_folder= '/mnt/program/wlx_data/datasets/lmms-lab/LLaVA-Video-178K/llava_video_frames/2_3_m_academic_v0_1/tokenized-by-rwkv-v6/mc/tokenized_sandwich_data', pad_image_to_square=True, template_map_fn=dict( template='xtuner.utils.PROMPT_TEMPLATE.vicuna', type='xtuner.dataset.map_fns.template_map_fn_factory'), tokenizer=dict( padding_side='right', pretrained_model_name_or_path= '/mnt/program/wlx_data/experiments/language-1022/rwkv-1b6/iter_745524.hf', trust_remote_code=True, type='transformers.AutoTokenizer.from_pretrained'), type='xtuner.dataset.AuroraDataset'), dict( data_path= '/mnt/program/wlx_data/datasets/lmms-lab/LLaVA-Video-178K/llava_video_frames/2_3_m_nextqa/converted_jsons/converted_2_3_m_nextqa_mc_qa_processed.json', dataset_map_fn='xtuner.dataset.map_fns.aurora_map_fn', image_folder= '/mnt/program/wlx_data/datasets/lmms-lab/LLaVA-Video-178K/llava_video_frames/2_3_m_nextqa/', image_processor=dict( crop_size=384, pretrained_model_name_or_path= 'laion/CLIP-ViT-bigG-14-laion2B-39B-b160k', size=384, trust_remote_code=True, type='transformers.CLIPImageProcessor.from_pretrained'), max_length=4096, offline_processed_text_folder= '/mnt/program/wlx_data/datasets/lmms-lab/LLaVA-Video-178K/llava_video_frames/2_3_m_nextqa/tokenized-by-rwkv-v6/mc/tokenized_sandwich_data', pad_image_to_square=True, template_map_fn=dict( template='xtuner.utils.PROMPT_TEMPLATE.vicuna', type='xtuner.dataset.map_fns.template_map_fn_factory'), tokenizer=dict( padding_side='right', pretrained_model_name_or_path= '/mnt/program/wlx_data/experiments/language-1022/rwkv-1b6/iter_745524.hf', trust_remote_code=True, type='transformers.AutoTokenizer.from_pretrained'), type='xtuner.dataset.AuroraDataset'), dict( data_path= '/mnt/program/wlx_data/datasets/lmms-lab/LLaVA-Video-178K/llava_video_frames/2_3_m_youtube_v0_1/converted_jsons/converted_2_3_m_youtube_mc_v0_1_qa_processed.json', dataset_map_fn='xtuner.dataset.map_fns.aurora_map_fn', image_folder= '/mnt/program/wlx_data/datasets/lmms-lab/LLaVA-Video-178K/llava_video_frames/2_3_m_youtube_v0_1/', image_processor=dict( crop_size=384, pretrained_model_name_or_path= 'laion/CLIP-ViT-bigG-14-laion2B-39B-b160k', size=384, trust_remote_code=True, type='transformers.CLIPImageProcessor.from_pretrained'), max_length=4096, offline_processed_text_folder= '/mnt/program/wlx_data/datasets/lmms-lab/LLaVA-Video-178K/llava_video_frames/2_3_m_youtube_v0_1/tokenized-by-rwkv-v6/mc/tokenized_sandwich_data', pad_image_to_square=True, template_map_fn=dict( template='xtuner.utils.PROMPT_TEMPLATE.vicuna', type='xtuner.dataset.map_fns.template_map_fn_factory'), tokenizer=dict( padding_side='right', pretrained_model_name_or_path= '/mnt/program/wlx_data/experiments/language-1022/rwkv-1b6/iter_745524.hf', trust_remote_code=True, type='transformers.AutoTokenizer.from_pretrained'), type='xtuner.dataset.AuroraDataset'), dict( data_path= '/mnt/program/wlx_data/datasets/lmms-lab/LLaVA-Video-178K/llava_video_frames/30_60_s_academic_v0_1/converted_jsons/converted_30_60_s_academic_mc_v0_1_qa_processed.json', dataset_map_fn='xtuner.dataset.map_fns.aurora_map_fn', image_folder= '/mnt/program/wlx_data/datasets/lmms-lab/LLaVA-Video-178K/llava_video_frames/30_60_s_academic_v0_1/', image_processor=dict( crop_size=384, pretrained_model_name_or_path= 'laion/CLIP-ViT-bigG-14-laion2B-39B-b160k', size=384, trust_remote_code=True, type='transformers.CLIPImageProcessor.from_pretrained'), max_length=4096, offline_processed_text_folder= '/mnt/program/wlx_data/datasets/lmms-lab/LLaVA-Video-178K/llava_video_frames/30_60_s_academic_v0_1/tokenized-by-rwkv-v6/mc/tokenized_sandwich_data', pad_image_to_square=True, template_map_fn=dict( template='xtuner.utils.PROMPT_TEMPLATE.vicuna', type='xtuner.dataset.map_fns.template_map_fn_factory'), tokenizer=dict( padding_side='right', pretrained_model_name_or_path= '/mnt/program/wlx_data/experiments/language-1022/rwkv-1b6/iter_745524.hf', trust_remote_code=True, type='transformers.AutoTokenizer.from_pretrained'), type='xtuner.dataset.AuroraDataset'), dict( data_path= '/mnt/program/wlx_data/datasets/lmms-lab/LLaVA-Video-178K/llava_video_frames/30_60_s_nextqa/converted_jsons/converted_30_60_s_nextqa_mc_qa_processed.json', dataset_map_fn='xtuner.dataset.map_fns.aurora_map_fn', image_folder= '/mnt/program/wlx_data/datasets/lmms-lab/LLaVA-Video-178K/llava_video_frames/30_60_s_nextqa/', image_processor=dict( crop_size=384, pretrained_model_name_or_path= 'laion/CLIP-ViT-bigG-14-laion2B-39B-b160k', size=384, trust_remote_code=True, type='transformers.CLIPImageProcessor.from_pretrained'), max_length=4096, offline_processed_text_folder= '/mnt/program/wlx_data/datasets/lmms-lab/LLaVA-Video-178K/llava_video_frames/30_60_s_nextqa/tokenized-by-rwkv-v6/mc/tokenized_sandwich_data', pad_image_to_square=True, template_map_fn=dict( template='xtuner.utils.PROMPT_TEMPLATE.vicuna', type='xtuner.dataset.map_fns.template_map_fn_factory'), tokenizer=dict( padding_side='right', pretrained_model_name_or_path= '/mnt/program/wlx_data/experiments/language-1022/rwkv-1b6/iter_745524.hf', trust_remote_code=True, type='transformers.AutoTokenizer.from_pretrained'), type='xtuner.dataset.AuroraDataset'), dict( data_path= '/mnt/program/wlx_data/datasets/lmms-lab/LLaVA-Video-178K/llava_video_frames/30_60_s_perceptiontest/converted_jsons/converted_30_60_s_perceptiontest_mc_qa_processed.json', dataset_map_fn='xtuner.dataset.map_fns.aurora_map_fn', image_folder= '/mnt/program/wlx_data/datasets/lmms-lab/LLaVA-Video-178K/llava_video_frames/30_60_s_perceptiontest/', image_processor=dict( crop_size=384, pretrained_model_name_or_path= 'laion/CLIP-ViT-bigG-14-laion2B-39B-b160k', size=384, trust_remote_code=True, type='transformers.CLIPImageProcessor.from_pretrained'), max_length=4096, offline_processed_text_folder= '/mnt/program/wlx_data/datasets/lmms-lab/LLaVA-Video-178K/llava_video_frames/30_60_s_perceptiontest/tokenized-by-rwkv-v6/mc/tokenized_sandwich_data', pad_image_to_square=True, template_map_fn=dict( template='xtuner.utils.PROMPT_TEMPLATE.vicuna', type='xtuner.dataset.map_fns.template_map_fn_factory'), tokenizer=dict( padding_side='right', pretrained_model_name_or_path= '/mnt/program/wlx_data/experiments/language-1022/rwkv-1b6/iter_745524.hf', trust_remote_code=True, type='transformers.AutoTokenizer.from_pretrained'), type='xtuner.dataset.AuroraDataset'), dict( data_path= '/mnt/program/wlx_data/datasets/lmms-lab/LLaVA-Video-178K/llava_video_frames/30_60_s_youtube_v0_1/converted_jsons/converted_30_60_s_youtube_mc_v0_1_qa_processed.json', dataset_map_fn='xtuner.dataset.map_fns.aurora_map_fn', image_folder= '/mnt/program/wlx_data/datasets/lmms-lab/LLaVA-Video-178K/llava_video_frames/30_60_s_youtube_v0_1/', image_processor=dict( crop_size=384, pretrained_model_name_or_path= 'laion/CLIP-ViT-bigG-14-laion2B-39B-b160k', size=384, trust_remote_code=True, type='transformers.CLIPImageProcessor.from_pretrained'), max_length=4096, offline_processed_text_folder= '/mnt/program/wlx_data/datasets/lmms-lab/LLaVA-Video-178K/llava_video_frames/30_60_s_youtube_v0_1/tokenized-by-rwkv-v6/mc/tokenized_sandwich_data', pad_image_to_square=True, template_map_fn=dict( template='xtuner.utils.PROMPT_TEMPLATE.vicuna', type='xtuner.dataset.map_fns.template_map_fn_factory'), tokenizer=dict( padding_side='right', pretrained_model_name_or_path= '/mnt/program/wlx_data/experiments/language-1022/rwkv-1b6/iter_745524.hf', trust_remote_code=True, type='transformers.AutoTokenizer.from_pretrained'), type='xtuner.dataset.AuroraDataset'), dict( data_path= '/mnt/program/wlx_data/datasets/lmms-lab/LLaVA-Video-178K/llava_video_frames/0_30_s_academic_v0_1/converted_jsons/filtered_converted_0_30_s_academic_oe_v0_1_qa_processed.json', dataset_map_fn='xtuner.dataset.map_fns.aurora_map_fn', image_folder= '/mnt/program/wlx_data/datasets/lmms-lab/LLaVA-Video-178K/llava_video_frames/0_30_s_academic_v0_1/', image_processor=dict( crop_size=384, pretrained_model_name_or_path= 'laion/CLIP-ViT-bigG-14-laion2B-39B-b160k', size=384, trust_remote_code=True, type='transformers.CLIPImageProcessor.from_pretrained'), max_length=4096, offline_processed_text_folder= '/mnt/program/wlx_data/datasets/lmms-lab/LLaVA-Video-178K/llava_video_frames/0_30_s_academic_v0_1/tokenized-by-rwkv-v6/oe/tokenized_sandwich_data', pad_image_to_square=True, template_map_fn=dict( template='xtuner.utils.PROMPT_TEMPLATE.vicuna', type='xtuner.dataset.map_fns.template_map_fn_factory'), tokenizer=dict( padding_side='right', pretrained_model_name_or_path= '/mnt/program/wlx_data/experiments/language-1022/rwkv-1b6/iter_745524.hf', trust_remote_code=True, type='transformers.AutoTokenizer.from_pretrained'), type='xtuner.dataset.AuroraDataset'), dict( data_path= '/mnt/program/wlx_data/datasets/lmms-lab/LLaVA-Video-178K/llava_video_frames/0_30_s_activitynetqa/converted_jsons/converted_0_30_s_activitynetqa_oe_qa_processed.json', dataset_map_fn='xtuner.dataset.map_fns.aurora_map_fn', image_folder= '/mnt/program/wlx_data/datasets/lmms-lab/LLaVA-Video-178K/llava_video_frames/0_30_s_activitynetqa/', image_processor=dict( crop_size=384, pretrained_model_name_or_path= 'laion/CLIP-ViT-bigG-14-laion2B-39B-b160k', size=384, trust_remote_code=True, type='transformers.CLIPImageProcessor.from_pretrained'), max_length=4096, offline_processed_text_folder= '/mnt/program/wlx_data/datasets/lmms-lab/LLaVA-Video-178K/llava_video_frames/0_30_s_activitynetqa/tokenized-by-rwkv-v6/oe/tokenized_sandwich_data', pad_image_to_square=True, template_map_fn=dict( template='xtuner.utils.PROMPT_TEMPLATE.vicuna', type='xtuner.dataset.map_fns.template_map_fn_factory'), tokenizer=dict( padding_side='right', pretrained_model_name_or_path= '/mnt/program/wlx_data/experiments/language-1022/rwkv-1b6/iter_745524.hf', trust_remote_code=True, type='transformers.AutoTokenizer.from_pretrained'), type='xtuner.dataset.AuroraDataset'), dict( data_path= '/mnt/program/wlx_data/datasets/lmms-lab/LLaVA-Video-178K/llava_video_frames/0_30_s_nextqa/converted_jsons/converted_0_30_s_nextqa_oe_qa_processed.json', dataset_map_fn='xtuner.dataset.map_fns.aurora_map_fn', image_folder= '/mnt/program/wlx_data/datasets/lmms-lab/LLaVA-Video-178K/llava_video_frames/0_30_s_nextqa/', image_processor=dict( crop_size=384, pretrained_model_name_or_path= 'laion/CLIP-ViT-bigG-14-laion2B-39B-b160k', size=384, trust_remote_code=True, type='transformers.CLIPImageProcessor.from_pretrained'), max_length=4096, offline_processed_text_folder= '/mnt/program/wlx_data/datasets/lmms-lab/LLaVA-Video-178K/llava_video_frames/0_30_s_nextqa/tokenized-by-rwkv-v6/oe/tokenized_sandwich_data', pad_image_to_square=True, template_map_fn=dict( template='xtuner.utils.PROMPT_TEMPLATE.vicuna', type='xtuner.dataset.map_fns.template_map_fn_factory'), tokenizer=dict( padding_side='right', pretrained_model_name_or_path= '/mnt/program/wlx_data/experiments/language-1022/rwkv-1b6/iter_745524.hf', trust_remote_code=True, type='transformers.AutoTokenizer.from_pretrained'), type='xtuner.dataset.AuroraDataset'), dict( data_path= '/mnt/program/wlx_data/datasets/lmms-lab/LLaVA-Video-178K/llava_video_frames/0_30_s_youtube_v0_1/converted_jsons/filtered_converted_0_30_s_youtube_oe_v0_1_qa_processed.json', dataset_map_fn='xtuner.dataset.map_fns.aurora_map_fn', image_folder= '/mnt/program/wlx_data/datasets/lmms-lab/LLaVA-Video-178K/llava_video_frames/0_30_s_youtube_v0_1/', image_processor=dict( crop_size=384, pretrained_model_name_or_path= 'laion/CLIP-ViT-bigG-14-laion2B-39B-b160k', size=384, trust_remote_code=True, type='transformers.CLIPImageProcessor.from_pretrained'), max_length=4096, offline_processed_text_folder= '/mnt/program/wlx_data/datasets/lmms-lab/LLaVA-Video-178K/llava_video_frames/0_30_s_youtube_v0_1/tokenized-by-rwkv-v6/oe/tokenized_sandwich_data', pad_image_to_square=True, template_map_fn=dict( template='xtuner.utils.PROMPT_TEMPLATE.vicuna', type='xtuner.dataset.map_fns.template_map_fn_factory'), tokenizer=dict( padding_side='right', pretrained_model_name_or_path= '/mnt/program/wlx_data/experiments/language-1022/rwkv-1b6/iter_745524.hf', trust_remote_code=True, type='transformers.AutoTokenizer.from_pretrained'), type='xtuner.dataset.AuroraDataset'), dict( data_path= '/mnt/program/wlx_data/datasets/lmms-lab/LLaVA-Video-178K/llava_video_frames/1_2_m_academic_v0_1/converted_jsons/converted_1_2_m_academic_oe_v0_1_qa_processed.json', dataset_map_fn='xtuner.dataset.map_fns.aurora_map_fn', image_folder= '/mnt/program/wlx_data/datasets/lmms-lab/LLaVA-Video-178K/llava_video_frames/1_2_m_academic_v0_1/', image_processor=dict( crop_size=384, pretrained_model_name_or_path= 'laion/CLIP-ViT-bigG-14-laion2B-39B-b160k', size=384, trust_remote_code=True, type='transformers.CLIPImageProcessor.from_pretrained'), max_length=4096, offline_processed_text_folder= '/mnt/program/wlx_data/datasets/lmms-lab/LLaVA-Video-178K/llava_video_frames/1_2_m_academic_v0_1/tokenized-by-rwkv-v6/oe/tokenized_sandwich_data', pad_image_to_square=True, template_map_fn=dict( template='xtuner.utils.PROMPT_TEMPLATE.vicuna', type='xtuner.dataset.map_fns.template_map_fn_factory'), tokenizer=dict( padding_side='right', pretrained_model_name_or_path= '/mnt/program/wlx_data/experiments/language-1022/rwkv-1b6/iter_745524.hf', trust_remote_code=True, type='transformers.AutoTokenizer.from_pretrained'), type='xtuner.dataset.AuroraDataset'), dict( data_path= '/mnt/program/wlx_data/datasets/lmms-lab/LLaVA-Video-178K/llava_video_frames/1_2_m_activitynetqa/converted_jsons/converted_1_2_m_activitynetqa_oe_qa_processed.json', dataset_map_fn='xtuner.dataset.map_fns.aurora_map_fn', image_folder= '/mnt/program/wlx_data/datasets/lmms-lab/LLaVA-Video-178K/llava_video_frames/1_2_m_activitynetqa/', image_processor=dict( crop_size=384, pretrained_model_name_or_path= 'laion/CLIP-ViT-bigG-14-laion2B-39B-b160k', size=384, trust_remote_code=True, type='transformers.CLIPImageProcessor.from_pretrained'), max_length=4096, offline_processed_text_folder= '/mnt/program/wlx_data/datasets/lmms-lab/LLaVA-Video-178K/llava_video_frames/1_2_m_activitynetqa/tokenized-by-rwkv-v6/oe/tokenized_sandwich_data', pad_image_to_square=True, template_map_fn=dict( template='xtuner.utils.PROMPT_TEMPLATE.vicuna', type='xtuner.dataset.map_fns.template_map_fn_factory'), tokenizer=dict( padding_side='right', pretrained_model_name_or_path= '/mnt/program/wlx_data/experiments/language-1022/rwkv-1b6/iter_745524.hf', trust_remote_code=True, type='transformers.AutoTokenizer.from_pretrained'), type='xtuner.dataset.AuroraDataset'), dict( data_path= '/mnt/program/wlx_data/datasets/lmms-lab/LLaVA-Video-178K/llava_video_frames/1_2_m_nextqa/converted_jsons/converted_1_2_m_nextqa_oe_qa_processed.json', dataset_map_fn='xtuner.dataset.map_fns.aurora_map_fn', image_folder= '/mnt/program/wlx_data/datasets/lmms-lab/LLaVA-Video-178K/llava_video_frames/1_2_m_nextqa/', image_processor=dict( crop_size=384, pretrained_model_name_or_path= 'laion/CLIP-ViT-bigG-14-laion2B-39B-b160k', size=384, trust_remote_code=True, type='transformers.CLIPImageProcessor.from_pretrained'), max_length=4096, offline_processed_text_folder= '/mnt/program/wlx_data/datasets/lmms-lab/LLaVA-Video-178K/llava_video_frames/1_2_m_nextqa/tokenized-by-rwkv-v6/oe/tokenized_sandwich_data', pad_image_to_square=True, template_map_fn=dict( template='xtuner.utils.PROMPT_TEMPLATE.vicuna', type='xtuner.dataset.map_fns.template_map_fn_factory'), tokenizer=dict( padding_side='right', pretrained_model_name_or_path= '/mnt/program/wlx_data/experiments/language-1022/rwkv-1b6/iter_745524.hf', trust_remote_code=True, type='transformers.AutoTokenizer.from_pretrained'), type='xtuner.dataset.AuroraDataset'), dict( data_path= '/mnt/program/wlx_data/datasets/lmms-lab/LLaVA-Video-178K/llava_video_frames/1_2_m_youtube_v0_1/converted_jsons/converted_1_2_m_youtube_oe_v0_1_qa_processed.json', dataset_map_fn='xtuner.dataset.map_fns.aurora_map_fn', image_folder= '/mnt/program/wlx_data/datasets/lmms-lab/LLaVA-Video-178K/llava_video_frames/1_2_m_youtube_v0_1/', image_processor=dict( crop_size=384, pretrained_model_name_or_path= 'laion/CLIP-ViT-bigG-14-laion2B-39B-b160k', size=384, trust_remote_code=True, type='transformers.CLIPImageProcessor.from_pretrained'), max_length=4096, offline_processed_text_folder= '/mnt/program/wlx_data/datasets/lmms-lab/LLaVA-Video-178K/llava_video_frames/1_2_m_youtube_v0_1/tokenized-by-rwkv-v6/oe/tokenized_sandwich_data', pad_image_to_square=True, template_map_fn=dict( template='xtuner.utils.PROMPT_TEMPLATE.vicuna', type='xtuner.dataset.map_fns.template_map_fn_factory'), tokenizer=dict( padding_side='right', pretrained_model_name_or_path= '/mnt/program/wlx_data/experiments/language-1022/rwkv-1b6/iter_745524.hf', trust_remote_code=True, type='transformers.AutoTokenizer.from_pretrained'), type='xtuner.dataset.AuroraDataset'), dict( data_path= '/mnt/program/wlx_data/datasets/lmms-lab/LLaVA-Video-178K/llava_video_frames/2_3_m_academic_v0_1/converted_jsons/converted_2_3_m_academic_oe_v0_1_qa_processed.json', dataset_map_fn='xtuner.dataset.map_fns.aurora_map_fn', image_folder= '/mnt/program/wlx_data/datasets/lmms-lab/LLaVA-Video-178K/llava_video_frames/2_3_m_academic_v0_1/', image_processor=dict( crop_size=384, pretrained_model_name_or_path= 'laion/CLIP-ViT-bigG-14-laion2B-39B-b160k', size=384, trust_remote_code=True, type='transformers.CLIPImageProcessor.from_pretrained'), max_length=4096, offline_processed_text_folder= '/mnt/program/wlx_data/datasets/lmms-lab/LLaVA-Video-178K/llava_video_frames/2_3_m_academic_v0_1/tokenized-by-rwkv-v6/oe/tokenized_sandwich_data', pad_image_to_square=True, template_map_fn=dict( template='xtuner.utils.PROMPT_TEMPLATE.vicuna', type='xtuner.dataset.map_fns.template_map_fn_factory'), tokenizer=dict( padding_side='right', pretrained_model_name_or_path= '/mnt/program/wlx_data/experiments/language-1022/rwkv-1b6/iter_745524.hf', trust_remote_code=True, type='transformers.AutoTokenizer.from_pretrained'), type='xtuner.dataset.AuroraDataset'), dict( data_path= '/mnt/program/wlx_data/datasets/lmms-lab/LLaVA-Video-178K/llava_video_frames/2_3_m_activitynetqa/converted_jsons/converted_2_3_m_activitynetqa_oe_qa_processed.json', dataset_map_fn='xtuner.dataset.map_fns.aurora_map_fn', image_folder= '/mnt/program/wlx_data/datasets/lmms-lab/LLaVA-Video-178K/llava_video_frames/2_3_m_activitynetqa/', image_processor=dict( crop_size=384, pretrained_model_name_or_path= 'laion/CLIP-ViT-bigG-14-laion2B-39B-b160k', size=384, trust_remote_code=True, type='transformers.CLIPImageProcessor.from_pretrained'), max_length=4096, offline_processed_text_folder= '/mnt/program/wlx_data/datasets/lmms-lab/LLaVA-Video-178K/llava_video_frames/2_3_m_activitynetqa/tokenized-by-rwkv-v6/oe/tokenized_sandwich_data', pad_image_to_square=True, template_map_fn=dict( template='xtuner.utils.PROMPT_TEMPLATE.vicuna', type='xtuner.dataset.map_fns.template_map_fn_factory'), tokenizer=dict( padding_side='right', pretrained_model_name_or_path= '/mnt/program/wlx_data/experiments/language-1022/rwkv-1b6/iter_745524.hf', trust_remote_code=True, type='transformers.AutoTokenizer.from_pretrained'), type='xtuner.dataset.AuroraDataset'), dict( data_path= '/mnt/program/wlx_data/datasets/lmms-lab/LLaVA-Video-178K/llava_video_frames/2_3_m_nextqa/converted_jsons/converted_2_3_m_nextqa_oe_qa_processed.json', dataset_map_fn='xtuner.dataset.map_fns.aurora_map_fn', image_folder= '/mnt/program/wlx_data/datasets/lmms-lab/LLaVA-Video-178K/llava_video_frames/2_3_m_nextqa/', image_processor=dict( crop_size=384, pretrained_model_name_or_path= 'laion/CLIP-ViT-bigG-14-laion2B-39B-b160k', size=384, trust_remote_code=True, type='transformers.CLIPImageProcessor.from_pretrained'), max_length=4096, offline_processed_text_folder= '/mnt/program/wlx_data/datasets/lmms-lab/LLaVA-Video-178K/llava_video_frames/2_3_m_nextqa/tokenized-by-rwkv-v6/oe/tokenized_sandwich_data', pad_image_to_square=True, template_map_fn=dict( template='xtuner.utils.PROMPT_TEMPLATE.vicuna', type='xtuner.dataset.map_fns.template_map_fn_factory'), tokenizer=dict( padding_side='right', pretrained_model_name_or_path= '/mnt/program/wlx_data/experiments/language-1022/rwkv-1b6/iter_745524.hf', trust_remote_code=True, type='transformers.AutoTokenizer.from_pretrained'), type='xtuner.dataset.AuroraDataset'), dict( data_path= '/mnt/program/wlx_data/datasets/lmms-lab/LLaVA-Video-178K/llava_video_frames/2_3_m_youtube_v0_1/converted_jsons/converted_2_3_m_youtube_oe_v0_1_qa_processed.json', dataset_map_fn='xtuner.dataset.map_fns.aurora_map_fn', image_folder= '/mnt/program/wlx_data/datasets/lmms-lab/LLaVA-Video-178K/llava_video_frames/2_3_m_youtube_v0_1/', image_processor=dict( crop_size=384, pretrained_model_name_or_path= 'laion/CLIP-ViT-bigG-14-laion2B-39B-b160k', size=384, trust_remote_code=True, type='transformers.CLIPImageProcessor.from_pretrained'), max_length=4096, offline_processed_text_folder= '/mnt/program/wlx_data/datasets/lmms-lab/LLaVA-Video-178K/llava_video_frames/2_3_m_youtube_v0_1/tokenized-by-rwkv-v6/oe/tokenized_sandwich_data', pad_image_to_square=True, template_map_fn=dict( template='xtuner.utils.PROMPT_TEMPLATE.vicuna', type='xtuner.dataset.map_fns.template_map_fn_factory'), tokenizer=dict( padding_side='right', pretrained_model_name_or_path= '/mnt/program/wlx_data/experiments/language-1022/rwkv-1b6/iter_745524.hf', trust_remote_code=True, type='transformers.AutoTokenizer.from_pretrained'), type='xtuner.dataset.AuroraDataset'), dict( data_path= '/mnt/program/wlx_data/datasets/lmms-lab/LLaVA-Video-178K/llava_video_frames/30_60_s_academic_v0_1/converted_jsons/converted_30_60_s_academic_oe_v0_1_qa_processed.json', dataset_map_fn='xtuner.dataset.map_fns.aurora_map_fn', image_folder= '/mnt/program/wlx_data/datasets/lmms-lab/LLaVA-Video-178K/llava_video_frames/30_60_s_academic_v0_1/', image_processor=dict( crop_size=384, pretrained_model_name_or_path= 'laion/CLIP-ViT-bigG-14-laion2B-39B-b160k', size=384, trust_remote_code=True, type='transformers.CLIPImageProcessor.from_pretrained'), max_length=4096, offline_processed_text_folder= '/mnt/program/wlx_data/datasets/lmms-lab/LLaVA-Video-178K/llava_video_frames/30_60_s_academic_v0_1/tokenized-by-rwkv-v6/oe/tokenized_sandwich_data', pad_image_to_square=True, template_map_fn=dict( template='xtuner.utils.PROMPT_TEMPLATE.vicuna', type='xtuner.dataset.map_fns.template_map_fn_factory'), tokenizer=dict( padding_side='right', pretrained_model_name_or_path= '/mnt/program/wlx_data/experiments/language-1022/rwkv-1b6/iter_745524.hf', trust_remote_code=True, type='transformers.AutoTokenizer.from_pretrained'), type='xtuner.dataset.AuroraDataset'), dict( data_path= '/mnt/program/wlx_data/datasets/lmms-lab/LLaVA-Video-178K/llava_video_frames/30_60_s_activitynetqa/converted_jsons/converted_30_60_s_activitynetqa_oe_qa_processed.json', dataset_map_fn='xtuner.dataset.map_fns.aurora_map_fn', image_folder= '/mnt/program/wlx_data/datasets/lmms-lab/LLaVA-Video-178K/llava_video_frames/30_60_s_activitynetqa/', image_processor=dict( crop_size=384, pretrained_model_name_or_path= 'laion/CLIP-ViT-bigG-14-laion2B-39B-b160k', size=384, trust_remote_code=True, type='transformers.CLIPImageProcessor.from_pretrained'), max_length=4096, offline_processed_text_folder= '/mnt/program/wlx_data/datasets/lmms-lab/LLaVA-Video-178K/llava_video_frames/30_60_s_activitynetqa/tokenized-by-rwkv-v6/oe/tokenized_sandwich_data', pad_image_to_square=True, template_map_fn=dict( template='xtuner.utils.PROMPT_TEMPLATE.vicuna', type='xtuner.dataset.map_fns.template_map_fn_factory'), tokenizer=dict( padding_side='right', pretrained_model_name_or_path= '/mnt/program/wlx_data/experiments/language-1022/rwkv-1b6/iter_745524.hf', trust_remote_code=True, type='transformers.AutoTokenizer.from_pretrained'), type='xtuner.dataset.AuroraDataset'), dict( data_path= '/mnt/program/wlx_data/datasets/lmms-lab/LLaVA-Video-178K/llava_video_frames/30_60_s_nextqa/converted_jsons/converted_30_60_s_nextqa_oe_qa_processed.json', dataset_map_fn='xtuner.dataset.map_fns.aurora_map_fn', image_folder= '/mnt/program/wlx_data/datasets/lmms-lab/LLaVA-Video-178K/llava_video_frames/30_60_s_nextqa/', image_processor=dict( crop_size=384, pretrained_model_name_or_path= 'laion/CLIP-ViT-bigG-14-laion2B-39B-b160k', size=384, trust_remote_code=True, type='transformers.CLIPImageProcessor.from_pretrained'), max_length=4096, offline_processed_text_folder= '/mnt/program/wlx_data/datasets/lmms-lab/LLaVA-Video-178K/llava_video_frames/30_60_s_nextqa/tokenized-by-rwkv-v6/oe/tokenized_sandwich_data', pad_image_to_square=True, template_map_fn=dict( template='xtuner.utils.PROMPT_TEMPLATE.vicuna', type='xtuner.dataset.map_fns.template_map_fn_factory'), tokenizer=dict( padding_side='right', pretrained_model_name_or_path= '/mnt/program/wlx_data/experiments/language-1022/rwkv-1b6/iter_745524.hf', trust_remote_code=True, type='transformers.AutoTokenizer.from_pretrained'), type='xtuner.dataset.AuroraDataset'), dict( data_path= '/mnt/program/wlx_data/datasets/lmms-lab/LLaVA-Video-178K/llava_video_frames/30_60_s_youtube_v0_1/converted_jsons/filtered_converted_30_60_s_youtube_oe_v0_1_qa_processed.json', dataset_map_fn='xtuner.dataset.map_fns.aurora_map_fn', image_folder= '/mnt/program/wlx_data/datasets/lmms-lab/LLaVA-Video-178K/llava_video_frames/30_60_s_youtube_v0_1/', image_processor=dict( crop_size=384, pretrained_model_name_or_path= 'laion/CLIP-ViT-bigG-14-laion2B-39B-b160k', size=384, trust_remote_code=True, type='transformers.CLIPImageProcessor.from_pretrained'), max_length=4096, offline_processed_text_folder= '/mnt/program/wlx_data/datasets/lmms-lab/LLaVA-Video-178K/llava_video_frames/30_60_s_youtube_v0_1/tokenized-by-rwkv-v6/oe/tokenized_sandwich_data', pad_image_to_square=True, template_map_fn=dict( template='xtuner.utils.PROMPT_TEMPLATE.vicuna', type='xtuner.dataset.map_fns.template_map_fn_factory'), tokenizer=dict( padding_side='right', pretrained_model_name_or_path= '/mnt/program/wlx_data/experiments/language-1022/rwkv-1b6/iter_745524.hf', trust_remote_code=True, type='transformers.AutoTokenizer.from_pretrained'), type='xtuner.dataset.AuroraDataset'), dict( data_path= '/mnt/program/wlx_data/datasets/lmms-lab/LLaVA-Video-178K/llava_video_frames/0_30_s_academic_v0_1/converted_jsons/filtered_converted_0_30_s_academic_v0_1_cap_processed.json', dataset_map_fn='xtuner.dataset.map_fns.aurora_map_fn', image_folder= '/mnt/program/wlx_data/datasets/lmms-lab/LLaVA-Video-178K/llava_video_frames/0_30_s_academic_v0_1/', image_processor=dict( crop_size=384, pretrained_model_name_or_path= 'laion/CLIP-ViT-bigG-14-laion2B-39B-b160k', size=384, trust_remote_code=True, type='transformers.CLIPImageProcessor.from_pretrained'), max_length=4096, offline_processed_text_folder= '/mnt/program/wlx_data/datasets/lmms-lab/LLaVA-Video-178K/llava_video_frames/0_30_s_academic_v0_1/tokenized-by-rwkv-v6/cap/tokenized_sandwich_data', pad_image_to_square=True, template_map_fn=dict( template='xtuner.utils.PROMPT_TEMPLATE.vicuna', type='xtuner.dataset.map_fns.template_map_fn_factory'), tokenizer=dict( padding_side='right', pretrained_model_name_or_path= '/mnt/program/wlx_data/experiments/language-1022/rwkv-1b6/iter_745524.hf', trust_remote_code=True, type='transformers.AutoTokenizer.from_pretrained'), type='xtuner.dataset.AuroraDataset'), dict( data_path= '/mnt/program/wlx_data/datasets/lmms-lab/LLaVA-Video-178K/llava_video_frames/0_30_s_youtube_v0_1/converted_jsons/filtered_converted_0_30_s_youtube_v0_1_cap_processed.json', dataset_map_fn='xtuner.dataset.map_fns.aurora_map_fn', image_folder= '/mnt/program/wlx_data/datasets/lmms-lab/LLaVA-Video-178K/llava_video_frames/0_30_s_youtube_v0_1/', image_processor=dict( crop_size=384, pretrained_model_name_or_path= 'laion/CLIP-ViT-bigG-14-laion2B-39B-b160k', size=384, trust_remote_code=True, type='transformers.CLIPImageProcessor.from_pretrained'), max_length=4096, offline_processed_text_folder= '/mnt/program/wlx_data/datasets/lmms-lab/LLaVA-Video-178K/llava_video_frames/0_30_s_youtube_v0_1/tokenized-by-rwkv-v6/cap/tokenized_sandwich_data', pad_image_to_square=True, template_map_fn=dict( template='xtuner.utils.PROMPT_TEMPLATE.vicuna', type='xtuner.dataset.map_fns.template_map_fn_factory'), tokenizer=dict( padding_side='right', pretrained_model_name_or_path= '/mnt/program/wlx_data/experiments/language-1022/rwkv-1b6/iter_745524.hf', trust_remote_code=True, type='transformers.AutoTokenizer.from_pretrained'), type='xtuner.dataset.AuroraDataset'), dict( data_path= '/mnt/program/wlx_data/datasets/lmms-lab/LLaVA-Video-178K/llava_video_frames/1_2_m_academic_v0_1/converted_jsons/converted_1_2_m_academic_v0_1_cap_processed.json', dataset_map_fn='xtuner.dataset.map_fns.aurora_map_fn', image_folder= '/mnt/program/wlx_data/datasets/lmms-lab/LLaVA-Video-178K/llava_video_frames/1_2_m_academic_v0_1/', image_processor=dict( crop_size=384, pretrained_model_name_or_path= 'laion/CLIP-ViT-bigG-14-laion2B-39B-b160k', size=384, trust_remote_code=True, type='transformers.CLIPImageProcessor.from_pretrained'), max_length=4096, offline_processed_text_folder= '/mnt/program/wlx_data/datasets/lmms-lab/LLaVA-Video-178K/llava_video_frames/1_2_m_academic_v0_1/tokenized-by-rwkv-v6/cap/tokenized_sandwich_data', pad_image_to_square=True, template_map_fn=dict( template='xtuner.utils.PROMPT_TEMPLATE.vicuna', type='xtuner.dataset.map_fns.template_map_fn_factory'), tokenizer=dict( padding_side='right', pretrained_model_name_or_path= '/mnt/program/wlx_data/experiments/language-1022/rwkv-1b6/iter_745524.hf', trust_remote_code=True, type='transformers.AutoTokenizer.from_pretrained'), type='xtuner.dataset.AuroraDataset'), dict( data_path= '/mnt/program/wlx_data/datasets/lmms-lab/LLaVA-Video-178K/llava_video_frames/1_2_m_youtube_v0_1/converted_jsons/converted_1_2_m_youtube_v0_1_cap_processed.json', dataset_map_fn='xtuner.dataset.map_fns.aurora_map_fn', image_folder= '/mnt/program/wlx_data/datasets/lmms-lab/LLaVA-Video-178K/llava_video_frames/1_2_m_youtube_v0_1/', image_processor=dict( crop_size=384, pretrained_model_name_or_path= 'laion/CLIP-ViT-bigG-14-laion2B-39B-b160k', size=384, trust_remote_code=True, type='transformers.CLIPImageProcessor.from_pretrained'), max_length=4096, offline_processed_text_folder= '/mnt/program/wlx_data/datasets/lmms-lab/LLaVA-Video-178K/llava_video_frames/1_2_m_youtube_v0_1/tokenized-by-rwkv-v6/cap/tokenized_sandwich_data', pad_image_to_square=True, template_map_fn=dict( template='xtuner.utils.PROMPT_TEMPLATE.vicuna', type='xtuner.dataset.map_fns.template_map_fn_factory'), tokenizer=dict( padding_side='right', pretrained_model_name_or_path= '/mnt/program/wlx_data/experiments/language-1022/rwkv-1b6/iter_745524.hf', trust_remote_code=True, type='transformers.AutoTokenizer.from_pretrained'), type='xtuner.dataset.AuroraDataset'), dict( data_path= '/mnt/program/wlx_data/datasets/lmms-lab/LLaVA-Video-178K/llava_video_frames/2_3_m_academic_v0_1/converted_jsons/converted_2_3_m_academic_v0_1_cap_processed.json', dataset_map_fn='xtuner.dataset.map_fns.aurora_map_fn', image_folder= '/mnt/program/wlx_data/datasets/lmms-lab/LLaVA-Video-178K/llava_video_frames/2_3_m_academic_v0_1/', image_processor=dict( crop_size=384, pretrained_model_name_or_path= 'laion/CLIP-ViT-bigG-14-laion2B-39B-b160k', size=384, trust_remote_code=True, type='transformers.CLIPImageProcessor.from_pretrained'), max_length=4096, offline_processed_text_folder= '/mnt/program/wlx_data/datasets/lmms-lab/LLaVA-Video-178K/llava_video_frames/2_3_m_academic_v0_1/tokenized-by-rwkv-v6/cap/tokenized_sandwich_data', pad_image_to_square=True, template_map_fn=dict( template='xtuner.utils.PROMPT_TEMPLATE.vicuna', type='xtuner.dataset.map_fns.template_map_fn_factory'), tokenizer=dict( padding_side='right', pretrained_model_name_or_path= '/mnt/program/wlx_data/experiments/language-1022/rwkv-1b6/iter_745524.hf', trust_remote_code=True, type='transformers.AutoTokenizer.from_pretrained'), type='xtuner.dataset.AuroraDataset'), dict( data_path= '/mnt/program/wlx_data/datasets/lmms-lab/LLaVA-Video-178K/llava_video_frames/2_3_m_youtube_v0_1/converted_jsons/converted_2_3_m_youtube_v0_1_cap_processed.json', dataset_map_fn='xtuner.dataset.map_fns.aurora_map_fn', image_folder= '/mnt/program/wlx_data/datasets/lmms-lab/LLaVA-Video-178K/llava_video_frames/2_3_m_youtube_v0_1/', image_processor=dict( crop_size=384, pretrained_model_name_or_path= 'laion/CLIP-ViT-bigG-14-laion2B-39B-b160k', size=384, trust_remote_code=True, type='transformers.CLIPImageProcessor.from_pretrained'), max_length=4096, offline_processed_text_folder= '/mnt/program/wlx_data/datasets/lmms-lab/LLaVA-Video-178K/llava_video_frames/2_3_m_youtube_v0_1/tokenized-by-rwkv-v6/cap/tokenized_sandwich_data', pad_image_to_square=True, template_map_fn=dict( template='xtuner.utils.PROMPT_TEMPLATE.vicuna', type='xtuner.dataset.map_fns.template_map_fn_factory'), tokenizer=dict( padding_side='right', pretrained_model_name_or_path= '/mnt/program/wlx_data/experiments/language-1022/rwkv-1b6/iter_745524.hf', trust_remote_code=True, type='transformers.AutoTokenizer.from_pretrained'), type='xtuner.dataset.AuroraDataset'), dict( data_path= '/mnt/program/wlx_data/datasets/lmms-lab/LLaVA-Video-178K/llava_video_frames/30_60_s_academic_v0_1/converted_jsons/converted_30_60_s_academic_v0_1_cap_processed.json', dataset_map_fn='xtuner.dataset.map_fns.aurora_map_fn', image_folder= '/mnt/program/wlx_data/datasets/lmms-lab/LLaVA-Video-178K/llava_video_frames/30_60_s_academic_v0_1/', image_processor=dict( crop_size=384, pretrained_model_name_or_path= 'laion/CLIP-ViT-bigG-14-laion2B-39B-b160k', size=384, trust_remote_code=True, type='transformers.CLIPImageProcessor.from_pretrained'), max_length=4096, offline_processed_text_folder= '/mnt/program/wlx_data/datasets/lmms-lab/LLaVA-Video-178K/llava_video_frames/30_60_s_academic_v0_1/tokenized-by-rwkv-v6/cap/tokenized_sandwich_data', pad_image_to_square=True, template_map_fn=dict( template='xtuner.utils.PROMPT_TEMPLATE.vicuna', type='xtuner.dataset.map_fns.template_map_fn_factory'), tokenizer=dict( padding_side='right', pretrained_model_name_or_path= '/mnt/program/wlx_data/experiments/language-1022/rwkv-1b6/iter_745524.hf', trust_remote_code=True, type='transformers.AutoTokenizer.from_pretrained'), type='xtuner.dataset.AuroraDataset'), dict( data_path= '/mnt/program/wlx_data/datasets/lmms-lab/LLaVA-Video-178K/llava_video_frames/30_60_s_youtube_v0_1/converted_jsons/filtered_converted_30_60_s_youtube_v0_1_cap_processed.json', dataset_map_fn='xtuner.dataset.map_fns.aurora_map_fn', image_folder= '/mnt/program/wlx_data/datasets/lmms-lab/LLaVA-Video-178K/llava_video_frames/30_60_s_youtube_v0_1/', image_processor=dict( crop_size=384, pretrained_model_name_or_path= 'laion/CLIP-ViT-bigG-14-laion2B-39B-b160k', size=384, trust_remote_code=True, type='transformers.CLIPImageProcessor.from_pretrained'), max_length=4096, offline_processed_text_folder= '/mnt/program/wlx_data/datasets/lmms-lab/LLaVA-Video-178K/llava_video_frames/30_60_s_youtube_v0_1/tokenized-by-rwkv-v6/cap/tokenized_sandwich_data', pad_image_to_square=True, template_map_fn=dict( template='xtuner.utils.PROMPT_TEMPLATE.vicuna', type='xtuner.dataset.map_fns.template_map_fn_factory'), tokenizer=dict( padding_side='right', pretrained_model_name_or_path= '/mnt/program/wlx_data/experiments/language-1022/rwkv-1b6/iter_745524.hf', trust_remote_code=True, type='transformers.AutoTokenizer.from_pretrained'), type='xtuner.dataset.AuroraDataset'), ], type='xtuner.dataset.ConcatDataset') visual_encoder_name_or_path = '/mnt/program/wlx_data/experiments/language-1022/rwkv-1b6/iter_745524.hf/visual_encoder/' visual_token_merge_ratio = 0.1 visualizer = None warmup_ratio = 0.03 weight_decay = 0 work_dir = '/mnt/program/wlx_data/experiments/language-llava-video-178k-no-sort-unfreeze-llm-1112'