{ "_name_or_path": "ai4bharat/IndicBART", "activation_dropout": 0.1, "activation_function": "gelu", "architectures": [ "T5ForConditionalGeneration" ], "attention_dropout": 0.1, "bos_token_id": 64000, "classifier_dropout": 0.0, "d_ff": 2048, "d_kv": 64, "d_model": 1024, "decoder_attention_heads": 16, "decoder_ffn_dim": 4096, "decoder_layerdrop": 0.0, "decoder_layers": 6, "decoder_start_token_id": 0, "dense_act_fn": "relu", "dropout": 0.1, "dropout_rate": 0.1, "encoder_attention_heads": 16, "encoder_ffn_dim": 4096, "encoder_layerdrop": 0.0, "encoder_layers": 6, "eos_token_id": 64001, "feed_forward_proj": "relu", "gradient_checkpointing": false, "init_std": 0.02, "initializer_factor": 1.0, "is_encoder_decoder": true, "is_gated_act": false, "layer_norm_epsilon": 1e-06, "max_position_embeddings": 1024, "model_type": "t5", "num_decoder_layers": 6, "num_heads": 8, "num_layers": 6, "pad_token_id": 0, "relative_attention_max_distance": 128, "relative_attention_num_buckets": 32, "scale_embedding": false, "tokenizer_class": "AlbertTokenizer", "torch_dtype": "float32", "transformers_version": "4.42.4", "use_cache": true, "vocab_size": 64014 }