{ "attn_implementation": "flash_attention_2", "vision_token_compressor": { "enabled": true, "algorithm": "qwen35_cdpruner_v1", "method": "cdpruner", "paper_url": "https://arxiv.org/abs/2506.10967", "official_repo": "Theia-4869/CDPruner", "official_repo_revision": "9541616c40fcd5625de1cdb8ea6c33c129eb7864", "adaptation_scope": "qwen35_single_image_post_native_merger_cdpruner_learnable_nearest_retained_merge_pre_llm", "original_method_reproduction_claim_allowed": false, "route_schema_version": "vision_opd_cdpruner_learnable_merge_zero_residual_curriculum_route_v1", "transport_key": "dart_merge_routes", "placement": "post_native_merger_pre_llm", "budget_policy": "per_image_min_N_max_32_ceil_active_optimizer_step_retention_bps_N_v1", "retention_ratio": 0.05, "minimum_tokens_per_image": 32, "distance": null, "distance_chunk_tokens": null, "routing_autograd": "detached_no_grad_fp32", "knn_k": null, "center_score": null, "assignment": "greedy_conditional_dpp_stable_original_index_tie", "reduction": "differentiable_fp32_zero_init_signed_residual", "position_policy": "selected_original_mrope", "query_conditioned": true, "attention_conditioned": false, "bbox_conditioned": false, "persistent_n_by_n_matrix_forbidden": true, "kernel_jitter": 1e-06, "relevance_epsilon": 1e-06, "residual_epsilon": 1e-12, "selector_dtype": "float32", "merge": "learnable_nearest_retained_zero_init_residual", "query_policy": "user_question_options_only_excluding_special_and_image_tokens", "relevance_fraction": null, "candidate_multiplier": null, "query_weight": null, "attention_weight": null, "bbox_weight": null, "deterministic": true, "data_manifest_path": null, "data_manifest_schema": null, "conditional_objective": "instruction_relevance_weighted_visual_diversity_greedy_dpp_matrix_free_fp32", "assignment_chunk_tokens": 2048, "query_route_schema_version": "vision_opd_route_query_v2", "summary": { "enabled": false }, "curriculum": { "enabled": true, "schema_version": "vision_opd_v8_visual_token_curriculum_v2", "driver": "completed_optimizer_steps_before_current_update", "schedule_type": "warmup_hold_raised_cosine_frozen_bps", "total_optimizer_steps": 188, "start_retention_bps": 2500, "final_retention_bps": 500, "warmup_hold_optimizer_steps": 14, "final_plateau_start_completed_steps": 100, "rounding_mode": "round_half_up", "minimum_tokens_per_image": 32 }, "learnable_merge": { "enabled": true, "aggregation": "zero_init_signed_residual_v1", "initialization": "pretrained_mlp_gate_rows_zero_output_v1", "input_hidden_size": 2560, "hidden_size": 64, "temperature": 1.0, "assignment": "cosine_nearest_retained_original_index_tie", "activation_checkpointing": true, "chunk_tokens": 1024 } } }