meterora_lib / mttl_config.json
kadarsh22's picture
Upload model
95a62d1 verified
{"base_model": "meta-llama/Meta-Llama-3-8B", "class_name": "mttl.models.expert_model.MultiExpertModelConfig", "default_expert_name": null, "expert_infos": [{"expert_name": "kadarsh22_meteora-llama3-8b-abstract_narrative_understanding-lora", "class_name": "mttl.models.library.expert_library.MetadataEntry", "expert_task_name": null, "parent_node": null, "expert_config": {"modify_modules": ".*", "class_name": "mttl.models.modifiers.lora.LoRAConfig", "modify_layers": "v_proj|o_proj|q_proj|up_proj|down_proj|k_proj|gate_proj", "tie_params": null, "lora_rank": 8, "lora_alpha": 16, "lora_dropout": 0.0, "lora_init_b_random": false}, "training_config": {}, "expert_model": "meta-llama/Meta-Llama-3-8B", "expert_deleted": false}, {"expert_name": "kadarsh22_meteora-llama3-8b-alpaca-lora", "class_name": "mttl.models.library.expert_library.MetadataEntry", "expert_task_name": null, "parent_node": null, "expert_config": {"modify_modules": ".*", "class_name": "mttl.models.modifiers.lora.LoRAConfig", "modify_layers": "v_proj|up_proj|o_proj|k_proj|gate_proj|down_proj|q_proj", "tie_params": null, "lora_rank": 8, "lora_alpha": 16, "lora_dropout": 0.0, "lora_init_b_random": false}, "training_config": {}, "expert_model": "meta-llama/Meta-Llama-3-8B", "expert_deleted": false}, {"expert_name": "kadarsh22_meteora-llama3-8b-cnn_dailymail-lora", "class_name": "mttl.models.library.expert_library.MetadataEntry", "expert_task_name": null, "parent_node": null, "expert_config": {"modify_modules": ".*", "class_name": "mttl.models.modifiers.lora.LoRAConfig", "modify_layers": "gate_proj|up_proj|o_proj|q_proj|v_proj|k_proj|down_proj", "tie_params": null, "lora_rank": 8, "lora_alpha": 16, "lora_dropout": 0.0, "lora_init_b_random": false}, "training_config": {}, "expert_model": "/data1/model/llama3/unsloth/Llama3-8b", "expert_deleted": false}, {"expert_name": "kadarsh22_meteora-llama3-8b-contextual_parametric_knowledge_conflicts-lora", "class_name": "mttl.models.library.expert_library.MetadataEntry", "expert_task_name": null, "parent_node": null, "expert_config": {"modify_modules": ".*", "class_name": "mttl.models.modifiers.lora.LoRAConfig", "modify_layers": "v_proj|down_proj|q_proj|k_proj|o_proj|gate_proj|up_proj", "tie_params": null, "lora_rank": 8, "lora_alpha": 16, "lora_dropout": 0.0, "lora_init_b_random": false}, "training_config": {}, "expert_model": "/data1/model/llama3/unsloth/Llama3-8b", "expert_deleted": false}, {"expert_name": "kadarsh22_meteora-llama3-8b-cs_algorithms-lora", "class_name": "mttl.models.library.expert_library.MetadataEntry", "expert_task_name": null, "parent_node": null, "expert_config": {"modify_modules": ".*", "class_name": "mttl.models.modifiers.lora.LoRAConfig", "modify_layers": "gate_proj|v_proj|o_proj|q_proj|up_proj|k_proj|down_proj", "tie_params": null, "lora_rank": 8, "lora_alpha": 16, "lora_dropout": 0.0, "lora_init_b_random": false}, "training_config": {}, "expert_model": "/data1/model/llama3/unsloth/Llama3-8b", "expert_deleted": false}, {"expert_name": "kadarsh22_meteora-llama3-8b-disfl_qa-lora", "class_name": "mttl.models.library.expert_library.MetadataEntry", "expert_task_name": null, "parent_node": null, "expert_config": {"modify_modules": ".*", "class_name": "mttl.models.modifiers.lora.LoRAConfig", "modify_layers": "up_proj|v_proj|k_proj|gate_proj|o_proj|down_proj|q_proj", "tie_params": null, "lora_rank": 8, "lora_alpha": 16, "lora_dropout": 0.0, "lora_init_b_random": false}, "training_config": {}, "expert_model": "/data1/model/llama3/unsloth/Llama3-8b", "expert_deleted": false}, {"expert_name": "kadarsh22_meteora-llama3-8b-elementary_math_qa-lora", "class_name": "mttl.models.library.expert_library.MetadataEntry", "expert_task_name": null, "parent_node": null, "expert_config": {"modify_modules": ".*", "class_name": "mttl.models.modifiers.lora.LoRAConfig", "modify_layers": "gate_proj|down_proj|k_proj|v_proj|o_proj|q_proj|up_proj", "tie_params": null, "lora_rank": 8, "lora_alpha": 16, "lora_dropout": 0.0, "lora_init_b_random": false}, "training_config": {}, "expert_model": "/data1/model/llama3/unsloth/Llama3-8b", "expert_deleted": false}, {"expert_name": "kadarsh22_meteora-llama3-8b-epistemic_reasoning-lora", "class_name": "mttl.models.library.expert_library.MetadataEntry", "expert_task_name": null, "parent_node": null, "expert_config": {"modify_modules": ".*", "class_name": "mttl.models.modifiers.lora.LoRAConfig", "modify_layers": "gate_proj|up_proj|v_proj|k_proj|o_proj|q_proj|down_proj", "tie_params": null, "lora_rank": 8, "lora_alpha": 16, "lora_dropout": 0.0, "lora_init_b_random": false}, "training_config": {}, "expert_model": "/data1/model/llama3/unsloth/Llama3-8b", "expert_deleted": false}, {"expert_name": "kadarsh22_meteora-llama3-8b-formal_fallacies_syllogisms_negation-lora", "class_name": "mttl.models.library.expert_library.MetadataEntry", "expert_task_name": null, "parent_node": null, "expert_config": {"modify_modules": ".*", "class_name": "mttl.models.modifiers.lora.LoRAConfig", "modify_layers": "v_proj|up_proj|down_proj|o_proj|q_proj|gate_proj|k_proj", "tie_params": null, "lora_rank": 8, "lora_alpha": 16, "lora_dropout": 0.0, "lora_init_b_random": false}, "training_config": {}, "expert_model": "/data1/model/llama3/unsloth/Llama3-8b", "expert_deleted": false}, {"expert_name": "kadarsh22_meteora-llama3-8b-goal_step_wikihow-lora", "class_name": "mttl.models.library.expert_library.MetadataEntry", "expert_task_name": null, "parent_node": null, "expert_config": {"modify_modules": ".*", "class_name": "mttl.models.modifiers.lora.LoRAConfig", "modify_layers": "o_proj|gate_proj|down_proj|q_proj|k_proj|up_proj|v_proj", "tie_params": null, "lora_rank": 8, "lora_alpha": 16, "lora_dropout": 0.0, "lora_init_b_random": false}, "training_config": {}, "expert_model": "/data1/model/llama3/unsloth/Llama3-8b", "expert_deleted": false}, {"expert_name": "kadarsh22_meteora-llama3-8b-gsm8k-lora", "class_name": "mttl.models.library.expert_library.MetadataEntry", "expert_task_name": null, "parent_node": null, "expert_config": {"modify_modules": ".*", "class_name": "mttl.models.modifiers.lora.LoRAConfig", "modify_layers": "q_proj|up_proj|v_proj|o_proj|down_proj|k_proj|gate_proj", "tie_params": null, "lora_rank": 8, "lora_alpha": 16, "lora_dropout": 0.0, "lora_init_b_random": false}, "training_config": {}, "expert_model": "/data1/model/llama3/unsloth/Llama3-8b", "expert_deleted": false}, {"expert_name": "kadarsh22_meteora-llama3-8b-language_identification-lora", "class_name": "mttl.models.library.expert_library.MetadataEntry", "expert_task_name": null, "parent_node": null, "expert_config": {"modify_modules": ".*", "class_name": "mttl.models.modifiers.lora.LoRAConfig", "modify_layers": "gate_proj|o_proj|up_proj|q_proj|down_proj|k_proj|v_proj", "tie_params": null, "lora_rank": 8, "lora_alpha": 16, "lora_dropout": 0.0, "lora_init_b_random": false}, "training_config": {}, "expert_model": "/data1/model/llama3/unsloth/Llama3-8b", "expert_deleted": false}, {"expert_name": "kadarsh22_meteora-llama3-8b-linguistics_puzzles-lora", "class_name": "mttl.models.library.expert_library.MetadataEntry", "expert_task_name": null, "parent_node": null, "expert_config": {"modify_modules": ".*", "class_name": "mttl.models.modifiers.lora.LoRAConfig", "modify_layers": "o_proj|down_proj|gate_proj|q_proj|v_proj|k_proj|up_proj", "tie_params": null, "lora_rank": 8, "lora_alpha": 16, "lora_dropout": 0.0, "lora_init_b_random": false}, "training_config": {}, "expert_model": "/data1/model/llama3/unsloth/Llama3-8b", "expert_deleted": false}, {"expert_name": "kadarsh22_meteora-llama3-8b-logical_deduction-lora", "class_name": "mttl.models.library.expert_library.MetadataEntry", "expert_task_name": null, "parent_node": null, "expert_config": {"modify_modules": ".*", "class_name": "mttl.models.modifiers.lora.LoRAConfig", "modify_layers": "q_proj|up_proj|o_proj|down_proj|v_proj|k_proj|gate_proj", "tie_params": null, "lora_rank": 8, "lora_alpha": 16, "lora_dropout": 0.0, "lora_init_b_random": false}, "training_config": {}, "expert_model": "/data1/model/llama3/unsloth/Llama3-8b", "expert_deleted": false}, {"expert_name": "kadarsh22_meteora-llama3-8b-news_commentary_de-lora", "class_name": "mttl.models.library.expert_library.MetadataEntry", "expert_task_name": null, "parent_node": null, "expert_config": {"modify_modules": ".*", "class_name": "mttl.models.modifiers.lora.LoRAConfig", "modify_layers": "gate_proj|up_proj|down_proj|o_proj|k_proj|q_proj|v_proj", "tie_params": null, "lora_rank": 8, "lora_alpha": 16, "lora_dropout": 0.0, "lora_init_b_random": false}, "training_config": {}, "expert_model": "/data1/model/llama3/unsloth/Llama3-8b", "expert_deleted": false}, {"expert_name": "kadarsh22_meteora-llama3-8b-news_commentary_es-lora", "class_name": "mttl.models.library.expert_library.MetadataEntry", "expert_task_name": null, "parent_node": null, "expert_config": {"modify_modules": ".*", "class_name": "mttl.models.modifiers.lora.LoRAConfig", "modify_layers": "up_proj|o_proj|q_proj|v_proj|k_proj|down_proj|gate_proj", "tie_params": null, "lora_rank": 8, "lora_alpha": 16, "lora_dropout": 0.0, "lora_init_b_random": false}, "training_config": {}, "expert_model": "/data1/model/llama3/unsloth/Llama3-8b", "expert_deleted": false}, {"expert_name": "kadarsh22_meteora-llama3-8b-news_commentary_it-lora", "class_name": "mttl.models.library.expert_library.MetadataEntry", "expert_task_name": null, "parent_node": null, "expert_config": {"modify_modules": ".*", "class_name": "mttl.models.modifiers.lora.LoRAConfig", "modify_layers": "k_proj|q_proj|gate_proj|o_proj|down_proj|v_proj|up_proj", "tie_params": null, "lora_rank": 8, "lora_alpha": 16, "lora_dropout": 0.0, "lora_init_b_random": false}, "training_config": {}, "expert_model": "/data1/model/llama3/unsloth/Llama3-8b", "expert_deleted": false}, {"expert_name": "kadarsh22_meteora-llama3-8b-object_counting-lora", "class_name": "mttl.models.library.expert_library.MetadataEntry", "expert_task_name": null, "parent_node": null, "expert_config": {"modify_modules": ".*", "class_name": "mttl.models.modifiers.lora.LoRAConfig", "modify_layers": "down_proj|up_proj|v_proj|o_proj|gate_proj|k_proj|q_proj", "tie_params": null, "lora_rank": 8, "lora_alpha": 16, "lora_dropout": 0.0, "lora_init_b_random": false}, "training_config": {}, "expert_model": "/data1/model/llama3/unsloth/Llama3-8b", "expert_deleted": false}, {"expert_name": "kadarsh22_meteora-llama3-8b-paragraph_segmentation-lora", "class_name": "mttl.models.library.expert_library.MetadataEntry", "expert_task_name": null, "parent_node": null, "expert_config": {"modify_modules": ".*", "class_name": "mttl.models.modifiers.lora.LoRAConfig", "modify_layers": "up_proj|o_proj|k_proj|q_proj|gate_proj|down_proj|v_proj", "tie_params": null, "lora_rank": 8, "lora_alpha": 16, "lora_dropout": 0.0, "lora_init_b_random": false}, "training_config": {}, "expert_model": "/data1/model/llama3/unsloth/Llama3-8b", "expert_deleted": false}, {"expert_name": "kadarsh22_meteora-llama3-8b-play_dialog_same_or_different-lora", "class_name": "mttl.models.library.expert_library.MetadataEntry", "expert_task_name": null, "parent_node": null, "expert_config": {"modify_modules": ".*", "class_name": "mttl.models.modifiers.lora.LoRAConfig", "modify_layers": "gate_proj|down_proj|o_proj|up_proj|q_proj|k_proj|v_proj", "tie_params": null, "lora_rank": 8, "lora_alpha": 16, "lora_dropout": 0.0, "lora_init_b_random": false}, "training_config": {}, "expert_model": "/data1/model/llama3/unsloth/Llama3-8b", "expert_deleted": false}, {"expert_name": "kadarsh22_meteora-llama3-8b-question_selection-lora", "class_name": "mttl.models.library.expert_library.MetadataEntry", "expert_task_name": null, "parent_node": null, "expert_config": {"modify_modules": ".*", "class_name": "mttl.models.modifiers.lora.LoRAConfig", "modify_layers": "gate_proj|v_proj|up_proj|k_proj|o_proj|q_proj|down_proj", "tie_params": null, "lora_rank": 8, "lora_alpha": 16, "lora_dropout": 0.0, "lora_init_b_random": false}, "training_config": {}, "expert_model": "/data1/model/llama3/unsloth/Llama3-8b", "expert_deleted": false}, {"expert_name": "kadarsh22_meteora-llama3-8b-reasoning_about_colored_objects-lora", "class_name": "mttl.models.library.expert_library.MetadataEntry", "expert_task_name": null, "parent_node": null, "expert_config": {"modify_modules": ".*", "class_name": "mttl.models.modifiers.lora.LoRAConfig", "modify_layers": "v_proj|k_proj|q_proj|down_proj|up_proj|o_proj|gate_proj", "tie_params": null, "lora_rank": 8, "lora_alpha": 16, "lora_dropout": 0.0, "lora_init_b_random": false}, "training_config": {}, "expert_model": "/data1/model/llama3/unsloth/Llama3-8b", "expert_deleted": false}, {"expert_name": "kadarsh22_meteora-llama3-8b-strategyqa-lora", "class_name": "mttl.models.library.expert_library.MetadataEntry", "expert_task_name": null, "parent_node": null, "expert_config": {"modify_modules": ".*", "class_name": "mttl.models.modifiers.lora.LoRAConfig", "modify_layers": "q_proj|gate_proj|down_proj|k_proj|up_proj|o_proj|v_proj", "tie_params": null, "lora_rank": 8, "lora_alpha": 16, "lora_dropout": 0.0, "lora_init_b_random": false}, "training_config": {}, "expert_model": "/data1/model/llama3/unsloth/Llama3-8b", "expert_deleted": false}, {"expert_name": "kadarsh22_meteora-llama3-8b-topical_chat-lora", "class_name": "mttl.models.library.expert_library.MetadataEntry", "expert_task_name": null, "parent_node": null, "expert_config": {"modify_modules": ".*", "class_name": "mttl.models.modifiers.lora.LoRAConfig", "modify_layers": "k_proj|up_proj|v_proj|q_proj|gate_proj|down_proj|o_proj", "tie_params": null, "lora_rank": 8, "lora_alpha": 16, "lora_dropout": 0.0, "lora_init_b_random": false}, "training_config": {}, "expert_model": "/data1/model/llama3/unsloth/Llama3-8b", "expert_deleted": false}, {"expert_name": "kadarsh22_meteora-llama3-8b-tracking_shuffled_objects-lora", "class_name": "mttl.models.library.expert_library.MetadataEntry", "expert_task_name": null, "parent_node": null, "expert_config": {"modify_modules": ".*", "class_name": "mttl.models.modifiers.lora.LoRAConfig", "modify_layers": "v_proj|gate_proj|q_proj|up_proj|down_proj|k_proj|o_proj", "tie_params": null, "lora_rank": 8, "lora_alpha": 16, "lora_dropout": 0.0, "lora_init_b_random": false}, "training_config": {}, "expert_model": "/data1/model/llama3/unsloth/Llama3-8b", "expert_deleted": false}, {"expert_name": "kadarsh22_meteora-llama3-8b-unit_conversion-lora", "class_name": "mttl.models.library.expert_library.MetadataEntry", "expert_task_name": null, "parent_node": null, "expert_config": {"modify_modules": ".*", "class_name": "mttl.models.modifiers.lora.LoRAConfig", "modify_layers": "down_proj|v_proj|up_proj|o_proj|gate_proj|q_proj|k_proj", "tie_params": null, "lora_rank": 8, "lora_alpha": 16, "lora_dropout": 0.0, "lora_init_b_random": false}, "training_config": {}, "expert_model": "/data1/model/llama3/unsloth/Llama3-8b", "expert_deleted": false}, {"expert_name": "kadarsh22_meteora-llama3-8b-vitaminc_fact_verification-lora", "class_name": "mttl.models.library.expert_library.MetadataEntry", "expert_task_name": null, "parent_node": null, "expert_config": {"modify_modules": ".*", "class_name": "mttl.models.modifiers.lora.LoRAConfig", "modify_layers": "o_proj|gate_proj|up_proj|k_proj|down_proj|q_proj|v_proj", "tie_params": null, "lora_rank": 8, "lora_alpha": 16, "lora_dropout": 0.0, "lora_init_b_random": false}, "training_config": {}, "expert_model": "/data1/model/llama3/unsloth/Llama3-8b", "expert_deleted": false}, {"expert_name": "kadarsh22_meteora-llama3-8b-winowhy-lora", "class_name": "mttl.models.library.expert_library.MetadataEntry", "expert_task_name": null, "parent_node": null, "expert_config": {"modify_modules": ".*", "class_name": "mttl.models.modifiers.lora.LoRAConfig", "modify_layers": "o_proj|gate_proj|up_proj|q_proj|v_proj|down_proj|k_proj", "tie_params": null, "lora_rank": 8, "lora_alpha": 16, "lora_dropout": 0.0, "lora_init_b_random": false}, "training_config": {}, "expert_model": "/data1/model/llama3/unsloth/Llama3-8b", "expert_deleted": false}], "selector_config": null}