49 lines
1.3 KiB
JSON
49 lines
1.3 KiB
JSON
|
{
|
||
|
|
"wandb": {
|
||
|
|
"name": "olvit",
|
||
|
|
"group": "dvd",
|
||
|
|
"tags": [],
|
||
|
|
"project": "olvit"
|
||
|
|
|
||
|
|
},
|
||
|
|
"model": {
|
||
|
|
"model_type": "discriminative",
|
||
|
|
"n_heads": 6,
|
||
|
|
"v_emb_dim": 36,
|
||
|
|
"dim_feedforward": 200,
|
||
|
|
"dropout_p": 0.1,
|
||
|
|
"fc_dim": 512,
|
||
|
|
"sample_rate_video": 20,
|
||
|
|
"n_transf_layers": 4,
|
||
|
|
"use_pretrained_lm": true,
|
||
|
|
"projection_as_in_aloe": true,
|
||
|
|
"pretrained_lm_name": "distilroberta-base",
|
||
|
|
"dataset": "dvd"
|
||
|
|
},
|
||
|
|
"extended_model": {
|
||
|
|
"hist_len_for_state_gen": 7,
|
||
|
|
"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": 128,
|
||
|
|
"seed": 12345,
|
||
|
|
"epochs": 1000
|
||
|
|
},
|
||
|
|
"checkpoint": {
|
||
|
|
"checkpoint_folder": "checkpoints/dvd",
|
||
|
|
"checkpoint_file_name": "olvit"
|
||
|
|
}
|
||
|
|
}
|