Fix #4

configs_pub/eccv20/MSRVTT_miech_trainval.json

@@ -0,0 +1,184 @@
+{
+ "n_gpu": 1,
+ "seed": 0,
+ "experts": {
+ "face_dim": 512,
+ "modalities": [
+ "face",
+ "ocr",
+ "rgb",
+ "s3d",
+ "scene",
+ "speech",
+ "vggish"
+ ]
+ },
+ "arch": {
+ "type": "CENet",
+ "args": {
+ "keep_missing_modalities": true,
+ "test_caption_mode": "indep",
+ "txt_inp": "bertftn",
+ "txt_agg": "bertftn",
+ "txt_wgh": "emb",
+ "vid_wgh": "none",
+ "vid_cont": "bert",
+ "vid_inp": "both",
+ "pos_enc": "tint",
+ "out_tok": "mxp",
+ "l2renorm": false,
+ "vid_bert_params": {
+ "vocab_size_or_config_json_file": 10,
+ "hidden_size": 512,
+ "num_hidden_layers": 4,
+ "num_attention_heads": 4,
+ "intermediate_size": 3072,
+ "hidden_act": "gelu",
+ "hidden_dropout_prob": 0.1,
+ "attention_probs_dropout_prob": 0.1,
+ "max_position_embeddings": 32,
+ "type_vocab_size": 19,
+ "initializer_range": 0.02,
+ "layer_norm_eps": 1e-12
+ },
+ "txt_pro": "gbn",
+ "txt_bert_params": {
+ "hidden_dropout_prob": 0.1,
+ "attention_probs_dropout_prob": 0.1
+ }
+ }
+ },
+ "train_sets": [
+ {
+ "type": "ExpertDataLoader",
+ "args": {
+ "mix": [
+ {
+ "dataset_name": "MSRVTT",
+ "cut_name": "miech",
+ "split_name": "trainval",
+ "data_dir": "data/MSRVTT/symlinked-feats",
+ "mix_weight": 1.0,
+ "query_shuffling": "shufk1",
+ "temporal_encoding_window": 1,
+ "max_text_words": 30,
+ "max_expert_tokens": 30
+ }
+ ],
+ "batch_size": 32,
+ "num_workers": 64,
+ "pin_memory": false
+ }
+ }
+ ],
+ "continuous_eval_sets": [
+ {
+ "type": "ExpertDataLoader",
+ "args": {
+ "mix": [
+ {
+ "dataset_name": "MSRVTT",
+ "cut_name": "miech",
+ "split_name": "trn",
+ "data_dir": "data/MSRVTT/symlinked-feats",
+ "captions_per_video": 1,
+ "query_shuffling": "indiv",
+ "temporal_encoding_window": 1,
+ "max_text_words": 30,
+ "max_expert_tokens": 30
+ }
+ ],
+ "batch_size": 32,
+ "num_workers": 64,
+ "pin_memory": false
+ }
+ },
+ {
+ "type": "ExpertDataLoader",
+ "args": {
+ "mix": [
+ {
+ "dataset_name": "MSRVTT",
+ "cut_name": "miech",
+ "split_name": "test",
+ "data_dir": "data/MSRVTT/symlinked-feats",
+ "captions_per_video": 1,
+ "query_shuffling": "indiv",
+ "temporal_encoding_window": 1,
+ "max_text_words": 30,
+ "max_expert_tokens": 30
+ }
+ ],
+ "batch_size": 32,
+ "num_workers": 64,
+ "pin_memory": false
+ }
+ }
+ ],
+ "final_eval_sets": [
+ {
+ "type": "ExpertDataLoader",
+ "args": {
+ "mix": [
+ {
+ "dataset_name": "MSRVTT",
+ "cut_name": "miech",
+ "split_name": "test",
+ "data_dir": "data/MSRVTT/symlinked-feats",
+ "captions_per_video": 1,
+ "query_shuffling": "indiv",
+ "temporal_encoding_window": 1,
+ "max_text_words": 30,
+ "max_expert_tokens": 30
+ }
+ ],
+ "batch_size": 32,
+ "num_workers": 0,
+ "pin_memory": false
+ }
+ }
+ ],
+ "optimizer": {
+ "type": "Adam",
+ "args": {
+ "lr": 5e-05,
+ "weight_decay": 0
+ }
+ },
+ "loss": {
+ "type": "MaxMarginRankingLoss",
+ "args": {
+ "margin": 0.05,
+ "fix_norm": true
+ }
+ },
+ "metrics": [
+ "t2v_metrics",
+ "v2t_metrics"
+ ],
+ "visualizer": {
+ "type": "Visualizer",
+ "args": {
+ "vis_vid_freq": 25.0,
+ "num_samples": 20
+ }
+ },
+ "lr_scheduler": {
+ "type": "StepLR",
+ "args": {
+ "step_size": 1,
+ "gamma": 0.95
+ }
+ },
+ "trainer": {
+ "epochs": 50,
+ "max_samples_per_epoch": 32000,
+ "save_period": 1,
+ "skip_first_n_saves": 0,
+ "include_optim_in_ckpts": true,
+ "verbosity": 2,
+ "tensorboard": true,
+ "monitor": "max epoch"
+ },
+ "cross_seed": 0
+}