61 lines
No EOL
1.7 KiB
JSON
61 lines
No EOL
1.7 KiB
JSON
{
|
|
"wandb": {
|
|
"name": "olvit",
|
|
"group": "simmc2",
|
|
"tags": [],
|
|
"project": "olvit"
|
|
|
|
},
|
|
"model": {
|
|
"model_type": "generative",
|
|
"dataset": "simmc2",
|
|
"feature_type": "object_text_features",
|
|
"object_feature_generator_dim": 50,
|
|
"n_object_feature_generator_layers": 2,
|
|
"n_heads": 6,
|
|
"v_emb_dim": 516,
|
|
"emb_dim": 216,
|
|
"dim_feedforward": 200,
|
|
"dropout_p": 0.1,
|
|
"fc_dim": 512,
|
|
"sample_rate_video": 1,
|
|
"n_encoder_layers": 4,
|
|
"n_decoder_layers": 4,
|
|
"use_pretrained_lm": true,
|
|
"vocab_size": 50265,
|
|
"projection_as_in_aloe": false,
|
|
"pretrained_lm_name": "distilroberta-base"
|
|
},
|
|
"extended_model": {
|
|
"hist_len_for_state_gen": 3,
|
|
"number_of_relevant_emb": 2,
|
|
"num_layers_v_state": 2,
|
|
"num_layers_d_state": 2,
|
|
"combiner_option": "OptionA",
|
|
"state_tracker_type": "Transformer",
|
|
"use_v_state": true,
|
|
"use_d_state": true,
|
|
"n_heads_combiner_transformer": 8,
|
|
"n_heads_state_tracker": 6,
|
|
"dim_feedforward_v_transformer": 140,
|
|
"dim_feedforward_d_transformer": 60
|
|
},
|
|
"training": {
|
|
"lr": 1e-4,
|
|
"warmup_steps": 4000,
|
|
"total_steps": 200000,
|
|
"batch_size": 8,
|
|
"seed": 12345,
|
|
"epochs": 1000
|
|
},
|
|
"datamodule": {
|
|
"fea_dir": "data/simmc/visual_features_resnet50_simmc2.1.pt",
|
|
"data_dir": "data/simmc/dialogs"
|
|
},
|
|
"checkpoint": {
|
|
"checkpoint_folder": "checkpoints/simmc/",
|
|
"checkpoint_file_name": "olvit",
|
|
"output_path": "output/simmc/",
|
|
"checkpoint_path": "TO_BE_DETERMINED"
|
|
}
|
|
} |