import os import toml import logging class ConfigLoader: """ Loader for configuration from `config.toml`. """ def __init__(self, config_path="config.toml"): if not os.path.exists(config_path): raise FileNotFoundError(f"Configuration file `{config_path}` not found!") self.config = toml.load(config_path) # --------------------------- # General parameters # --------------------------- general_cfg = self.config.get("general", {}) self.use_telegram = general_cfg.get("use_telegram", False) # --------------------------- # Common parameters # --------------------------- self.split = self.config.get("split", "train") # --------------------------- # Dataset paths # --------------------------- self.datasets = self.config.get("datasets", {}) # --------------------------- # Modalities and emotions # --------------------------- self.modalities = self.config.get("modalities", ["audio"]) self.emotion_columns = self.config.get( "emotion_columns", ["Neutral", "Anger", "Disgust", "Fear", "Happiness", "Sadness", "Surprise", "Other"], ) # --------------------------- # DataLoader # --------------------------- dataloader_cfg = self.config.get("dataloader", {}) self.num_workers = dataloader_cfg.get("num_workers", 0) self.shuffle = dataloader_cfg.get("shuffle", True) self.prepare_only = dataloader_cfg.get("prepare_only", False) self.average_features = dataloader_cfg.get("average_features", False) # --------------------------- # Training: general # --------------------------- train_general = self.config.get("train", {}).get("general", {}) self.random_seed = train_general.get("random_seed", 42) self.subset_size = train_general.get("subset_size", 0) self.merge_probability = train_general.get("merge_probability", 0) self.batch_size = train_general.get("batch_size", 8) self.num_epochs = train_general.get("num_epochs", 100) self.max_patience = train_general.get("max_patience", 10) self.save_best_model = train_general.get("save_best_model", False) self.save_prepared_data = train_general.get("save_prepared_data", True) self.save_feature_path = train_general.get("save_feature_path", "./features/") self.search_type = train_general.get("search_type", "none") self.smoothing_probability = train_general.get("smoothing_probability", 0) self.path_to_df_ls = train_general.get("path_to_df_ls", None) self.early_stop_on = train_general.get("early_stop_on", "dev") self.lambda_emotion = train_general.get("lambda_emotion", 1) self.lambda_personality = train_general.get("lambda_personality", 5) self.lambda_domain = train_general.get("lambda_domain", 0.1) self.checkpoint_dir = train_general.get("checkpoint_dir", "checkpoints") self.device = train_general.get("device", "cuda") self.selection_metric = train_general.get("selection_metric", "mean_combo") self.single_task = train_general.get("single_task", False) self.opt_set = train_general.get("opt_set", "dev") # --------------------------- # Training: model parameters # --------------------------- train_model = self.config.get("train", {}).get("model", {}) self.id_ablation_type_by_modality = train_model.get("id_ablation_type_by_modality", 0) self.id_ablation_type_by_component = train_model.get("id_ablation_type_by_component", 0) self.single_task_id = train_model.get("single_task_id", 0) self.model_name = train_model.get("model_name", "BiFormer") self.model_stage = train_model.get("model_stage", "emotion") self.path_to_saved_emotion_model = train_model.get("path_to_saved_emotion_model", None) self.path_to_saved_personality_model = train_model.get("path_to_saved_personality_model", None) self.per_activation = train_model.get("per_activation", "sigmoid") self.weight_emotion = train_model.get("weight_emotion", 1.0) self.weight_pers = train_model.get("weight_pers", 1.0) self.pers_loss_type = train_model.get("pers_loss_type", True) self.emotion_loss_type = train_model.get("emotion_loss_type", True) self.flag_emo_weight = train_model.get("flag_emo_weight", False) self.ssl_weight_emotion = train_model.get("ssl_weight_emotion", 1) self.ssl_weight_personality = train_model.get("ssl_weight_personality", 1) self.ssl_confidence_threshold_emo = train_model.get("ssl_confidence_threshold_emo", 0.6) self.ssl_confidence_threshold_pt = train_model.get("ssl_confidence_threshold_pt", 0.6) self.pers_loss_type = train_model.get("pers_loss_type", "mae") self.emotion_loss_type = train_model.get("emotion_loss_type", "CE") self.alpha_sup = train_model.get("alpha_sup", 1.0) self.w_lr_sup = train_model.get("w_lr_sup", 0.025) self.alpha_ssl = train_model.get("alpha_ssl", 0.5) self.w_lr_ssl = train_model.get("w_lr_ssl", 0.001) self.lambda_ssl = train_model.get("lambda_ssl", 0.2) self.w_floor = train_model.get("w_floor", 1e-3) self.hidden_dim = train_model.get("hidden_dim", 256) self.hidden_dim_gated = train_model.get("hidden_dim_gated", 256) self.num_transformer_heads = train_model.get("num_transformer_heads", 8) self.num_graph_heads = train_model.get("num_graph_heads", 8) self.tr_layer_number = train_model.get("tr_layer_number", 5) self.mamba_d_state = train_model.get("mamba_d_state", 16) self.mamba_ker_size = train_model.get("mamba_ker_size", 4) self.mamba_layer_number = train_model.get("mamba_layer_number", 3) self.positional_encoding = train_model.get("positional_encoding", True) self.dropout = train_model.get("dropout", 0.0) self.out_features = train_model.get("out_features", 128) self.mode = train_model.get("mode", "mean") self.fusion_dim = train_model.get("fusion_dim", 64) self.attention = train_model.get("attention", None) # Parameters for the best emotion/personality models self.hidden_dim_emo = train_model.get("hidden_dim_emo", 256) self.out_features_emo = train_model.get("out_features_emo", 256) self.name_best_emo_model = train_model.get("name_best_emo_model", "BiFormer") self.name_best_per_model = train_model.get("name_best_per_model", "BiFormer") self.path_to_saved_emotion_model = train_model.get("path_to_saved_emotion_model", None) self.path_to_saved_personality_model = train_model.get("path_to_saved_personality_model", None) self.num_transformer_heads_emo = train_model.get("num_transformer_heads_emo", 8) self.tr_layer_number_emo = train_model.get("tr_layer_number_emo", 1) self.positional_encoding_emo = train_model.get("positional_encoding_emo", True) self.mamba_d_state_emo = train_model.get("mamba_d_state_emo", 16) self.mamba_layer_number_emo = train_model.get("mamba_layer_number_emo", 3) self.hidden_dim_per = train_model.get("hidden_dim_per", 256) self.out_features_per = train_model.get("out_features_per", 256) self.num_transformer_heads_per = train_model.get("num_transformer_heads_per", 8) self.tr_layer_number_per = train_model.get("tr_layer_number_per", 1) self.positional_encoding_per = train_model.get("positional_encoding_per", True) self.mamba_d_state_per = train_model.get("mamba_d_state_per", 16) self.mamba_layer_number_per = train_model.get("mamba_layer_number_per", 3) self.best_per_activation = train_model.get("best_per_activation", "sigmoid") self.image_embedding_dim = train_model.get("image_embedding_dim", 2560) # --------------------------- # Training: optimizer # --------------------------- train_optimizer = self.config.get("train", {}).get("optimizer", {}) self.optimizer = train_optimizer.get("optimizer", "adam") self.lr = train_optimizer.get("lr", 1e-4) self.weight_decay = train_optimizer.get("weight_decay", 0.0) self.momentum = train_optimizer.get("momentum", 0.9) # --------------------------- # Training: scheduler # --------------------------- train_scheduler = self.config.get("train", {}).get("scheduler", {}) self.scheduler_type = train_scheduler.get("scheduler_type", "plateau") self.warmup_ratio = train_scheduler.get("warmup_ratio", 0.1) # --------------------------- # Embeddings # --------------------------- emb_cfg = self.config.get("embeddings", {}) self.audio_model_name = emb_cfg.get("audio_model", "amiriparian/ExHuBERT") self.text_model_name = emb_cfg.get("text_model", "jinaai/jina-embeddings-v3") self.audio_classifier_checkpoint = emb_cfg.get("audio_classifier_checkpoint", "best_audio_model.pt") self.text_classifier_checkpoint = emb_cfg.get("text_classifier_checkpoint", "best_text_model.pth") self.image_classifier_checkpoint = emb_cfg.get("image_classifier_checkpoint", "torchscript_model_0_66_37_wo_gl.pth") self.image_model_type = emb_cfg.get("image_model_type", "resnet50") self.cut_target_layer = emb_cfg.get("cut_target_layer", 2) self.roi_video = emb_cfg.get("roi_video", "face") self.counter_need_frames = emb_cfg.get("counter_need_frames", 20) self.image_size = emb_cfg.get("image_size", 224) self.audio_embedding_dim = emb_cfg.get("audio_embedding_dim", 1024) self.text_embedding_dim = emb_cfg.get("text_embedding_dim", 1024) self.emb_normalize = emb_cfg.get("emb_normalize", True) self.audio_pooling = emb_cfg.get("audio_pooling", None) self.text_pooling = emb_cfg.get("text_pooling", None) self.max_tokens = emb_cfg.get("max_tokens", 256) self.window_size = emb_cfg.get("window_size", 5) if __name__ == "__main__": self.log_config() def log_config(self): logging.info("=== CONFIGURATION ===") logging.info(f"Split: {self.split}") logging.info(f"Datasets: {list(self.datasets.keys())}") for name, ds in self.datasets.items(): logging.info(f"[Dataset: {name}]") logging.info(f" Base Dir: {ds.get('base_dir', 'N/A')}") logging.info(f" CSV Path: {ds.get('csv_path', '')}") logging.info(f" WAV Dir: {ds.get('wav_dir', 'N/A')}") logging.info(f" Video Dir: {ds.get('video_dir', '')}") logging.info(f" Audio Dir: {ds.get('audio_dir', '')}") # Training parameters logging.info("--- Training Config ---") logging.info(f"DataLoader: batch_size={self.batch_size}, num_workers={self.num_workers}, shuffle={self.shuffle}") logging.info(f"Model Name: {self.model_name}") logging.info(f"Random Seed: {self.random_seed}") logging.info(f"Hidden Dim: {self.hidden_dim}") logging.info(f"Gated Hidden Dim: {self.hidden_dim_gated}") logging.info(f"Transformer Heads: {self.num_transformer_heads}") logging.info(f"Graph Heads: {self.num_graph_heads}") logging.info(f"Stat Pooling Mode: {self.mode}") logging.info(f"Optimizer: {self.optimizer}") logging.info(f"Scheduler Type: {self.scheduler_type}") logging.info(f"Warmup Ratio: {self.warmup_ratio}") logging.info(f"Weight Decay: {self.weight_decay}") logging.info(f"Momentum (SGD): {self.momentum}") logging.info(f"Positional Encoding: {self.positional_encoding}") logging.info(f"Transformer Layers: {self.tr_layer_number}") logging.info(f"Mamba D State: {self.mamba_d_state}") logging.info(f"Mamba Kernel Size: {self.mamba_ker_size}") logging.info(f"Mamba Layers: {self.mamba_layer_number}") logging.info(f"Dropout: {self.dropout}") logging.info(f"Out Features: {self.out_features}") logging.info(f"Learning Rate: {self.lr}") logging.info(f"Epochs: {self.num_epochs}") logging.info(f"Merge Probability: {self.merge_probability}") logging.info(f"Smoothing Probability: {self.smoothing_probability}") logging.info(f"Max Patience: {self.max_patience}") logging.info(f"Save Prepared Data: {self.save_prepared_data}") logging.info(f"Features Save Path: {self.save_feature_path}") logging.info(f"Search Type: {self.search_type}") # Embeddings logging.info("--- Embeddings Config ---") logging.info(f"Audio Model: {self.audio_model_name}, Text Model: {self.text_model_name}") logging.info(f"Audio dim={self.audio_embedding_dim}, Text dim={self.text_embedding_dim}") logging.info(f"Audio pooling={self.audio_pooling}, Text pooling={self.text_pooling}") logging.info(f"Device={self.device}, Normalize={self.emb_normalize}") def show_config(self): self.log_config()