# Created by BVH & VG, Jul 2025. import os from hydra.core.config_store import ConfigStore from custom.experiment.template import template_any4d_2b ################################################################# debug = False ################################################################# job = dict( project='a4d2', group='cvpr', name='mixed_4v', prepend_datetime=True, ) wandb = dict( enabled=not debug, entity='tri', project='a4d2_cvpr', num_validation_logs=5, ) any4d_config = dict( # NOTE(bvh): pre-refactor flat fields are deprecated; semantics now in data_*_overrides. # override_zero_origin=True, # was commented out pre-refactor too (deliberate for mixed_*) # override_temporal_stride_val=1, data_val_overrides=dict(frame_stride=1), transforms='default_zo1', # transforms='default_fitorigin', dataloader='basile', vae='a4d_vae', video_entries=[ dict( rgb0=(0, 16, 'load', 'load'), rgb0_input_mask=(16, 17, 'load', None), rgb0_output_mask=(17, 18, 'zero', None), cams0=(18, 50, 'zero/load', 'zero/load'), cams0_input_mask=(50, 51, 'zero/load', None), cams0_output_mask=(51, 52, 'zero', None), ), dict( rgb1=(0, 16, 'copy:rgb0/load', 'copy:rgb0/load'), rgb1_input_mask=(16, 17, 'copy:rgb0_input_mask/load', None), rgb1_output_mask=(17, 18, 'zero', None), cams1=(18, 50, 'zero/load', 'zero/load'), cams1_input_mask=(50, 51, 'zero/load', None), cams1_output_mask=(51, 52, 'zero', None), ), dict( rgb2=(0, 16, 'copy:rgb1/load', 'copy:rgb1/load'), rgb2_input_mask=(16, 17, 'copy:rgb1_input_mask/load', None), rgb2_output_mask=(17, 18, 'zero', None), cams2=(18, 50, 'copy:cams1/load', 'copy:cams1/load'), cams2_input_mask=(50, 51, 'copy:cams1_input_mask/load', None), cams2_output_mask=(51, 52, 'zero', None), ), # TODO why does rgb3 say rgb2? looks like a bug / this specific file was never run? dict( rgb2=(0, 16, 'copy:rgb1/load', 'copy:rgb1/load'), rgb2_input_mask=(16, 17, 'copy:rgb1_input_mask/load', None), rgb2_output_mask=(17, 18, 'zero', None), cams2=(18, 50, 'copy:cams1/load', 'copy:cams1/load'), cams2_input_mask=(50, 51, 'copy:cams1_input_mask/load', None), cams2_output_mask=(51, 52, 'zero', None), ), ], # lowdim_adaln_entries=dict( # scale_factor=(0, 1, 0, 1, 'zero'), # scale_factor_input_mask=(0, 1, 1, 2, 'zero'), # ), num_views=4, video_concat_mode='view', video_proj_mode='per_view', loss_weights=dict( rgb0=1.0, rgb1=1.0, rgb2=1.0, rgb3=1.0, ), harmonize_streams=True, harmonize_frames=True, load_modals=['rgb','cams'], lbm_real_modals=['rgb','cams'], task_probs=dict( cross_modal=0.0, dyn_view_synth=1.0, forecast=0.0, pose_est=0.0, inv_dyn=0.0, policy=0.0, world_model=0.0, ), use_views='rand2', track_metrics=dict( rgb0=['psnr','ssim'], # rgb1=['psnr','ssim'], # rgb2=['psnr','ssim'], # rgb3=['psnr','ssim'], ), train_visuals_interval=0, train_visuals_detail=0, val_visuals_detail=1, train_directives=dict( num_pred_views=[1,2,3], ), val_directives=dict( tasks='dyn_view_synth', num_pred_views=2, remove_cond=True, ), ) ################################################################# S3_PRETRAINED_PREFIX = r's3://tri-ml-sandbox-16011-us-west-2-datasets/cosmos-predict-2/checkpoints' model = dict( # dit_path='', # dit_path=f'{S3_PRETRAINED_PREFIX}/nvidia/Cosmos-Predict2-2B-Video2World/model-480p-10fps.pt', # dit_path='s3://tri-ml-sandbox-16011-us-west-2-datasets/sagemaker/cosmos-predict2/a4d2/cvpr/10-17-01-54_official3/model/iter_000006000_003072000.pt', # dit_path='s3://tri-ml-sandbox-16011-us-west-2-datasets/sagemaker/cosmos-predict2/a4d2/cvpr/10-17-01-54_official3/model/iter_000008000_004096000.pt', # dit_path='s3://tri-ml-sandbox-16011-us-west-2-datasets/sagemaker/cosmos-predict2/a4d2/cvpr/10-17-01-54_official3/model/iter_000010000_005120000.pt', # dit_path='s3://tri-ml-sandbox-16011-us-west-2-datasets/sagemaker/cosmos-predict2/a4d2/cvpr/10-17-01-54_official3/model/iter_000014000_007168000.pt', # dit_path='s3://tri-ml-sandbox-16011-us-west-2-datasets/sagemaker/cosmos-predict2/a4d2/cvpr/10-17-01-54_official3/model/iter_000016000_008192000.pt', # dit_path='s3://tri-ml-sandbox-16011-us-west-2-datasets/sagemaker/cosmos-predict2/a4d2/cvpr/10-17-01-54_official3/model/iter_000020000_010240000.pt', # dit_path='s3://tri-ml-sandbox-16011-us-west-2-datasets/sagemaker/cosmos-predict2/a4d2/cvpr/10-23-03-23_officialA/model/iter_000002000_001024000.pt', # dit_path='s3://tri-ml-sandbox-16011-us-west-2-datasets/sagemaker/cosmos-predict2/a4d2/cvpr/10-24-19-03_officialC/model/iter_000001000_000512000.pt', # dit_path='s3://tri-ml-sandbox-16011-us-west-2-datasets/sagemaker/cosmos-predict2/a4d2/cvpr/10-24-20-37_officialC/model/iter_000002500_001280000.pt', # dit_path='s3://tri-ml-sandbox-16011-us-west-2-datasets/sagemaker/cosmos-predict2/a4d2/cvpr/10-25-17-13_officialD6/model/iter_000000200_000076800.pt', # dit_path='s3://tri-ml-sandbox-16011-us-west-2-datasets/sagemaker/cosmos-predict2/a4d2/cvpr/10-25-19-46_officialD8/model/iter_000000800_000409600.pt', # dit_path='s3://tri-ml-sandbox-16011-us-west-2-datasets/sagemaker/cosmos-predict2/a4d2/cvpr/10-25-17-13_officialD6/model/iter_000003000_001152000.pt', # dit_path='s3://tri-ml-sandbox-16011-us-west-2-datasets/sagemaker/cosmos-predict2/a4d2/cvpr/10-27-19-05_official16HR/model/iter_000000200_000102400.pt', # dit_path='s3://tri-ml-sandbox-16011-us-west-2-datasets/sagemaker/cosmos-predict2/a4d2/cvpr/10-28-02-33_officialHR6b/model/iter_000000700_000134400.pt', # dit_path='s3://tri-ml-sandbox-16011-us-west-2-datasets/sagemaker/cosmos-predict2/a4d2/cvpr/10-28-17-03_officialHR6c/model/iter_000000300_000057600.pt', # dit_path='s3://tri-ml-sandbox-16011-us-west-2-datasets/sagemaker/cosmos-predict2/a4d2/cvpr/10-29-01-04_officialHR16d/model/iter_000001500_000384000.pt', # dit_path='s3://tri-ml-sandbox-16011-us-west-2-datasets/sagemaker/cosmos-predict2/a4d2/cvpr/10-29-16-49_officialHR8e/model/iter_000000600_000153600.pt', # dit_path='s3://tri-ml-sandbox-16011-us-west-2-datasets/sagemaker/cosmos-predict2/a4d2/cvpr/10-29-01-04_officialHR16d/model/iter_000002250_000576000.pt', dit_path='s3://tri-ml-sandbox-16011-us-west-2-datasets/sagemaker/cosmos-predict2/a4d2/cvpr/10-28-17-03_officialHR6c/model/iter_000004700_000902400.pt', text_encoder_path='', # text_encoder_path=f'{S3_PRETRAINED_PREFIX}/google-t5/t5-11b', vae_path=f'{S3_PRETRAINED_PREFIX}/nvidia/Cosmos-Predict2-2B-Video2World/tokenizer/tokenizer.pth', fsdp_shard_size=8, ############ run_validation=not debug, validation_iter=1000, max_iter=25000, grad_accum_iter=1, context_parallel_size=1, device_monitor=0, manual_gc_iter=288, manual_gc_warm_up=-1, ############ state_t=11, tokenizer_chunk_duration=41, ) checkpoint = dict( save_iter=1000, s3_folder='s3://tri-ml-sandbox-16011-us-west-2-datasets/sagemaker/cosmos-predict2/', early_sanity_check=-1, ) optimizer = dict( lr=2.0e-5, ) scheduler = dict( warm_up_steps = [1000], # [1000], cycle_lengths = [model['max_iter']], f_start=[0.01], f_max=[1.0], f_min=[0.05], ) dataset_train = dict( config='custom/config/cvpr/mixed/train_4v_32gpu.yaml', # config='custom/config/cvpr/mixed/train_2v_32gpu.yaml', num_workers=4, batch_size=4, # batch_size=2, ) dataset_val = dict( config=dict( DDAD2v0 = 'custom/config/cvpr/mixed/val_4v/ddad2v0.yaml', DDAD2v1 = 'custom/config/cvpr/mixed/val_4v/ddad2v1.yaml', DDAD2v2 = 'custom/config/cvpr/mixed/val_4v/ddad2v2.yaml', DDAD2v3 = 'custom/config/cvpr/mixed/val_4v/ddad2v3.yaml', DDAD2v4 = 'custom/config/cvpr/mixed/val_4v/ddad2v4.yaml', DDAD2v5 = 'custom/config/cvpr/mixed/val_4v/ddad2v5.yaml', Waymo2v0 = 'custom/config/cvpr/mixed/val_4v/waymo2v0.yaml', Waymo2v1 = 'custom/config/cvpr/mixed/val_4v/waymo2v1.yaml', Waymo2v2 = 'custom/config/cvpr/mixed/val_4v/waymo2v2.yaml', Waymo2v3 = 'custom/config/cvpr/mixed/val_4v/waymo2v3.yaml', Waymo2v4 = 'custom/config/cvpr/mixed/val_4v/waymo2v4.yaml', Waymo2v5 = 'custom/config/cvpr/mixed/val_4v/waymo2v5.yaml', # ### DRIVING # ArgoVerse2sync = 'custom/config/cvpr/mixed/val_4v/argoverse2sync.yaml', # DDADfront = 'custom/config/cvpr/mixed/val_4v/ddad.yaml', # DL3DV1 = 'custom/config/cvpr/mixed/val_4v/dl3dv1.yaml', # # DDADside = 'custom/config/cvpr/mixed/val_4v/ddad_side.yaml', # # Waymo = 'custom/config/cvpr/mixed/val_4v/waymo.yaml', # # LyftL5 = 'custom/config/cvpr/mixed/val_4v/lyftl5.yaml', # # ### ROBOTICS # # DROIDCalibID = 'custom/config/cvpr/mixed/val_4v/droidcalib_id.yaml', # # DROIDCalibOOD = 'custom/config/cvpr/mixed/val_4v/droidcalib_ood.yaml', # # LBMv12 = 'custom/config/cvpr/mixed/val_4v/lbmv12.yaml', # # ### 3D # # # TartanAir = 'custom/config/cvpr/mixed/val_4v/tartanair.yaml', # # # DL3DV = 'custom/config/cvpr/mixed/val_4v/dl3dv.yaml', # # # MVImgNet = 'custom/config/cvpr/mixed/val_4v/mvimgnet.yaml', # # # RE10K = 'custom/config/cvpr/mixed/val_4v/re10k.yaml', # # # WildRGBD = 'custom/config/cvpr/mixed/val_4v/wildrgbd.yaml', # # # ScanNet = 'custom/config/cvpr/mixed/val_4v/scannet.yaml', # # ### 4D # # Kubric4D = 'custom/config/cvpr/mixed/val_4v/kubric4d.yaml', # # Kubric5D = 'custom/config/cvpr/mixed/val_4v/kubric5d.yaml', # # PD4D = 'custom/config/cvpr/mixed/val_4v/pd4d.yaml', # GCDKubric4Ddir = 'custom/config/cvpr/mixed/val_2v/gcd_kubric4d_dir.yaml', # GCDKubric4Dgrad = 'custom/config/cvpr/mixed/val_2v/gcd_kubric4d_grad.yaml', # GCDPD4Ddir = 'custom/config/cvpr/mixed/val_2v/gcd_pd4d_dir.yaml', # GCDPD4Dgrad = 'custom/config/cvpr/mixed/val_2v/gcd_pd4d_grad.yaml', # # EgoExo4D = 'custom/config/cvpr/mixed/val_4v/egoexo4d.yaml', # DyCheck = 'custom/config/cvpr/mixed/val_2v/dycheck.yaml', # # DyCheck5 = 'custom/config/cvpr/mixed/val_4v/dycheck5.yaml', ), # params=dict( # resize=1.5, # ), num_workers=1, batch_size=1, ) metrics = dict( gt_source='gt', modes=['r','rpv','apv'], # ,'ipv'], rgb=dict(), ) ################################################################# cs = ConfigStore.instance() this_config = template_any4d_2b( job, wandb, any4d_config, model, checkpoint, optimizer, scheduler, metrics, dataset_train, dataset_val) experiment_name = 'cvpr_' + os.path.splitext(os.path.basename(__file__))[0] cs.store( group='experiment', package='_global_', name=experiment_name, node=this_config, ) # dataset_val = dict( # config=dict( # ### DRIVING # ArgoVerse2sync_FL = 'custom/config/cvpr/mixed/val_4v/argoverse2sync_FL.yaml', # ArgoVerse2sync_FR = 'custom/config/cvpr/mixed/val_4v/argoverse2sync_FR.yaml', # ArgoVerse2sync_LF = 'custom/config/cvpr/mixed/val_4v/argoverse2sync_LF.yaml', # ArgoVerse2sync_RF = 'custom/config/cvpr/mixed/val_4v/argoverse2sync_RF.yaml', # DDAD_FL = 'custom/config/cvpr/mixed/val_4v/ddad_FL.yaml', # DDAD_FR = 'custom/config/cvpr/mixed/val_4v/ddad_FR.yaml', # DDAD_LF = 'custom/config/cvpr/mixed/val_4v/ddad_LF.yaml', # DDAD_RF = 'custom/config/cvpr/mixed/val_4v/ddad_RF.yaml', # Waymo_FL = 'custom/config/cvpr/mixed/val_4v/waymo_FL.yaml', # Waymo_FR = 'custom/config/cvpr/mixed/val_4v/waymo_FR.yaml', # Waymo_LF = 'custom/config/cvpr/mixed/val_4v/waymo_LF.yaml', # Waymo_RF = 'custom/config/cvpr/mixed/val_4v/waymo_RF.yaml', # LyftL5_FL = 'custom/config/cvpr/mixed/val_4v/lyftl5_FL.yaml', # LyftL5_FR = 'custom/config/cvpr/mixed/val_4v/lyftl5_FR.yaml', # LyftL5_LF = 'custom/config/cvpr/mixed/val_4v/lyftl5_LF.yaml', # LyftL5_RF = 'custom/config/cvpr/mixed/val_4v/lyftl5_RF.yaml', # ### ROBOTICS # DROID_ID_LR = 'custom/config/cvpr/mixed/val_4v/droidcalibv2_ID_LR.yaml', # DROID_ID_RL = 'custom/config/cvpr/mixed/val_4v/droidcalibv2_ID_RL.yaml', # DROID_OOD_LR = 'custom/config/cvpr/mixed/val_4v/droidcalibv2_OOD_LR.yaml', # DROID_OOD_RL = 'custom/config/cvpr/mixed/val_4v/droidcalibv2_OOD_RL.yaml', # LBMv12_LR = 'custom/config/cvpr/mixed/val_4v/lbmv12_LR.yaml', # LBMv12_RL = 'custom/config/cvpr/mixed/val_4v/lbmv12_RL.yaml', # ### 3D # GCDKubric4Dgrad = 'custom/config/cvpr/mixed/val_4v/gcd_kubric4d_grad.yaml', # GCDKubric4Ddir = 'custom/config/cvpr/mixed/val_4v/gcd_kubric4d_dir.yaml', # GCDPD4Dgrad = 'custom/config/cvpr/mixed/val_4v/gcd_pd4d_grad.yaml', # GCDPD4Ddir = 'custom/config/cvpr/mixed/val_4v/gcd_pd4d_dir.yaml', # # TartanAir = 'custom/config/cvpr/mixed/val_4v/tartanair.yaml', # # DL3DV = 'custom/config/cvpr/mixed/val_4v/dl3dv.yaml', # # MVImgNet = 'custom/config/cvpr/mixed/val_4v/mvimgnet.yaml', # # RE10K = 'custom/config/cvpr/mixed/val_4v/re10k.yaml', # # WildRGBD = 'custom/config/cvpr/mixed/val_4v/wildrgbd.yaml', # # ScanNet = 'custom/config/cvpr/mixed/val_4v/scannet.yaml', # ### 4D # # Kubric4D = 'custom/config/cvpr/mixed/val_4v/kubric4d.yaml', # # Kubric5D = 'custom/config/cvpr/mixed/val_4v/kubric5d.yaml', # # PD4D = 'custom/config/cvpr/mixed/val_4v/pd4d.yaml', # GCDKubric4Dgrad = 'custom/config/cvpr/mixed/val_4v/gcd_kubric4d_grad.yaml', # GCDKubric4Ddir = 'custom/config/cvpr/mixed/val_4v/gcd_kubric4d_dir.yaml', # GCDPD4Ddir = 'custom/config/cvpr/mixed/val_4v/gcd_pd4d_dir.yaml', # GCDPD4Dgrad = 'custom/config/cvpr/mixed/val_4v/gcd_pd4d_grad.yaml', # # EgoExo4D = 'custom/config/cvpr/mixed/val_4v/egoexo4d.yaml', # DyCheck = 'custom/config/cvpr/mixed/val_4v/dycheck.yaml', # # DyCheck5 = 'custom/config/cvpr/mixed/val_4v/dycheck5.yaml', # ), # params=dict( # resize=1.5, # ), # num_workers=1, # batch_size=1, # )