Add fullscale_no_qp config

fullscale_no_qp/config.json

@@ -0,0 +1,180 @@
+{
+  "sam3_defaults": false,
+  "da3_model": "depth-anything/DA3NESTED-GIANT-LARGE",
+  "da3_resolution": 1008,
+  "use_cached_depth": true,
+  "da3_cache_name": "da3_nested_cache_1008",
+  "d_model": 256,
+  "n_heads": 8,
+  "num_decoder_layers": 6,
+  "num_queries": 10,
+  "multi_object": false,
+  "match_strategy": "hungarian",
+  "per_text_decode": false,
+  "grouped_text_attn": false,
+  "sam3_multi_object": false,
+  "num_objects": 1,
+  "max_objects": 0,
+  "train_seghead": false,
+  "dataset": "scannetpp",
+  "frames_per_sequence": 50,
+  "samples_per_sequence": 1,
+  "part_query_mode": "random",
+  "augment": false,
+  "data_root": null,
+  "split": "train_balanced_v5",
+  "views": 8,
+  "sampling_strategy": "stratified",
+  "da3_chunk_size": 8,
+  "max_scenes": 245,
+  "resolution": 1008,
+  "min_mask_coverage": 0.0,
+  "epochs": 100,
+  "stop_at_epoch": 0,
+  "val_every": 0,
+  "val_split": "val_cache_good",
+  "val_max_samples": null,
+  "save_best_val": true,
+  "no_save_best_val": false,
+  "min_ram_gb": 100.0,
+  "batch_size": 8,
+  "num_workers": 4,
+  "prefetch_factor": 2,
+  "no_persistent_workers": false,
+  "lr": 0.0002,
+  "lr_scheduler": "cosine",
+  "lr_warmup_epochs": 2,
+  "lr_min": 1e-06,
+  "lr_step_size": 10,
+  "lr_gamma": 0.5,
+  "grad_accum": 1,
+  "grad_clip": 1.0,
+  "run_name": "fullscale_no_qp_245s_100ep",
+  "checkpoint_dir": "/home/nvidia/mv_sam3/checkpoints",
+  "resume": "checkpoints/fullscale_no_qp_245s_100ep",
+  "load_weights": null,
+  "focal_weight": 0.0,
+  "focal_alpha": 0.75,
+  "focal_gamma": 2.0,
+  "dice_weight": 5.0,
+  "boundary_weight": 0.0,
+  "lovasz_weight": 0.0,
+  "use_point_sampling": false,
+  "num_sample_points": 4096,
+  "loss_at_native_res": false,
+  "attn_map_size": 64,
+  "seed": 42,
+  "use_presence_token": true,
+  "use_box_prompts": true,
+  "box_prompt_dropout": 0.2,
+  "use_point_prompts": true,
+  "point_prompt_dropout": 0.2,
+  "num_pos_points": 10,
+  "num_neg_points": 2,
+  "samples_per_scene": 1,
+  "min_category_samples": 1,
+  "exclude_categories": null,
+  "prompt_type": "all",
+  "use_world_pe": true,
+  "use_gasa": true,
+  "gasa_beta_init": 1.0,
+  "gasa_kernel_dim": 32,
+  "gasa_fixed_kernel": false,
+  "gasa_kernel_type": "learned",
+  "gasa_bidirectional": false,
+  "use_image_to_token": false,
+  "use_pos_refine": false,
+  "use_box_rpb": false,
+  "dim_feedforward": 2048,
+  "post_norm": true,
+  "ffn_fp32": true,
+  "use_query_pe": false,
+  "use_depth_crossattn": false,
+  "per_layer_text": true,
+  "no_initial_text": false,
+  "no_text_proj": false,
+  "clean_v": false,
+  "additive_pe": true,
+  "cross_view": false,
+  "use_da3_poses_for_gasa": true,
+  "use_gt_poses_for_gasa": false,
+  "no_gt_poses": false,
+  "sheaf_use_gt_poses": false,
+  "pe_type": "world",
+  "no_pointmap_normalize": false,
+  "use_iterative_pos": false,
+  "presence_weight": 2.0,
+  "presence_focal": true,
+  "presence_alpha": 0.5,
+  "presence_gamma": 0.0,
+  "use_centroid_head": true,
+  "centroid_weight": 0.5,
+  "mask_based_centroid": false,
+  "use_triangulation": false,
+  "eval_localization": false,
+  "use_iou_head": false,
+  "iou_head_weight": 1.0,
+  "pred_logits_source": "mask_mean",
+  "contrastive_weight": 0.0,
+  "contrastive_margin": 0.5,
+  "contrastive_source": "logits",
+  "no_object_weight": 1.0,
+  "align_weight": 1.0,
+  "align_alpha": 0.5,
+  "align_gamma": 2.0,
+  "align_tau": 2.0,
+  "per_layer_align": false,
+  "per_layer_align_weight": null,
+  "use_sheaf_loss": false,
+  "sheaf_weight": 0.1,
+  "sheaf_type": "constant",
+  "sheaf_d_edge": 32,
+  "sheaf_threshold": 0.1,
+  "sheaf_soft_correspondences": false,
+  "sheaf_sigma": 0.1,
+  "sheaf_detach_target": true,
+  "sheaf_max_frame_distance": 0,
+  "sheaf_symmetric_detach": true,
+  "sheaf_mutual_nn": true,
+  "output_localization": false,
+  "output_depth": false,
+  "output_pointcloud": false,
+  "semantic_union": false,
+  "mask_selection": "confidence",
+  "vis_every_epochs": 0,
+  "vis_every_batches": 0,
+  "use_lora": false,
+  "lora_rank": 8,
+  "lora_alpha": 16.0,
+  "lora_sam3": false,
+  "lora_da3": false,
+  "lora_mask_embed": false,
+  "query_proj_mlp": false,
+  "no_query_proj": true,
+  "train_mask_embed": true,
+  "use_mask_refiner": false,
+  "init_scoring_from_sam3": false,
+  "init_text_crossattn_from_sam3": false,
+  "init_decoder_from_sam3": false,
+  "torch_compile": false,
+  "compile_mode": "reduce-overhead",
+  "batch_views": false,
+  "profile": false,
+  "use_spatial_tokens": true,
+  "use_spatial_attn_bias": false,
+  "use_text_spatial_bias": true,
+  "use_spatial_points": false,
+  "use_object_aware_spatial": true,
+  "spatial_augment_prob": 0.5,
+  "spatial_multi_instance_only": true,
+  "spatial_relational_prob": 0.0,
+  "spatial_gt_aware": true,
+  "spatial_ranking_weight": 0.0,
+  "spatial_ranking_margin": 0.5,
+  "mask_smooth_kernel": 0,
+  "include_categories": null,
+  "class_balanced": true,
+  "class_balance_power": 0.7,
+  "world_size": 7,
+  "resumed_from": "checkpoints/fullscale_no_qp_245s_100ep"
+}