Config file for Synthetic Datasets

[wentinghome]

Hi,
I'm wondering if there is a config file for training the synthetic dataset? Thank you.

[youngfish42]

Maybe you can refer to the config files for real-wolrd dataset.

[youngfish42]

For example, for task of "density"

{
    "data": {
        "task":"density"
        },
    
    "tb": {
        "tb_logging":true,
        "dir":"tensorboard",
        "name": "S_density_optuna"
    },

    "optuna":{
        "opt_n_trials" : 50,
        "opt_n_cores" : 1,
        "monitor_metric" : "val_micro_f1",
        "opt_direction" : "maximize" ,
        "sampler" : "random",
        "pruning" : false
    },

    "hyperparams_fix" : {
        "max_epochs": 200,
        "use_neighborhood": false,
        "use_structure": true,
        "use_position": false,
        "seed": 3,
        "node_embed_size": 32,
        "structure_patch_type": "triangular_random_walk",
        "lstm_aggregator": "last",
        "n_processes": 16,
        "resample_anchor_patches": false,
        "freeze_node_embeds":false,
        "use_mpn_projection":true,
        "print_train_times":false,
        "compute_similarities":false,
        "sample_walk_len":50,
        "n_triangular_walks": 5,
        "random_walk_len":10, 
        "rw_beta" : 0.65,
        "ff_attn": false,
        "max_sim_epochs": 5,
        "embedding_type": "graphsaint"      

     
    },

    "hyperparams_optuna": {
        "batch_size" :{
            "type" : "suggest_categorical",
            "args" : [[ 64, 128] ]
        },
        "learning_rate" :{
            "type" : "suggest_float",
            "args" : [1e-4, 1e-3],
            "kwargs" : {"log":true}
        },
        "grad_clip" : {
            "type" : "suggest_float",
            "args" : [0.0, 0.5]
        },
        "n_layers": {
            "type" : "suggest_int",
            "args" : [1,5]
        },
        "neigh_sample_border_size" : {
            "type" : "suggest_int",
            "args" : [1,2]
        },
        "n_anchor_patches_pos_out": {
            "type" : "suggest_int",
            "args" : [50,200]
        }, 
        "n_anchor_patches_pos_in": {
            "type" : "suggest_int",
            "args" : [25,75]
        },
        "n_anchor_patches_N_in": {
            "type" : "suggest_int",
            "args" : [10,25]
        }, 
        "n_anchor_patches_N_out": {
            "type" : "suggest_int",
            "args" : [25,75]
        }, 
        "n_anchor_patches_structure": {
            "type" : "suggest_int",
            "args" : [15,45]
        }, 
 
        "linear_hidden_dim_1" : {
            "type" : "suggest_categorical",
            "args" : [[64]]
        },
        "linear_hidden_dim_2" : {
            "type" : "suggest_categorical",
            "args" : [[32, 64]]
        },
        "lstm_dropout" : {
            "type" : "suggest_float",
            "args" : [0.0, 0.4]
        },
        "lstm_n_layers" : {
            "type" : "suggest_int",
            "args" : [1,2]
        },
        "lin_dropout" : {
            "type" : "suggest_float",
            "args" : [0.0, 0.4]
        },
        "cc_aggregator":{
            "type" : "suggest_categorical",
            "args" : [[ "sum","max"] ]
        },
        "trainable_cc":{
            "type" : "suggest_categorical",
            "args" : [[ true,false] ]
        },
        "auto_lr_find":{
            "type" : "suggest_categorical",
            "args" : [[ true,false] ]
        }
    }
}