| { |
| "activation_dropout": 0.0, |
| "activation_function": "gelu", |
| "add_cross_attention": true, |
| "architectures": [ |
| "STLDecoderModel" |
| ], |
| "attention_dropout": 0.0, |
| "auto_map": { |
| "AutoConfig": "configuration_stldec.STLDecoderConfig", |
| "AutoModel": "modeling_stldec.STLDecoderModel", |
| "AutoModelForCausalLM": "modeling_stldec.STLDecoderModel", |
| "AutoTokenizer": [ |
| "tokenizer_stldec.STLTokenizer", |
| null |
| ] |
| }, |
| "bos_token_id": 2, |
| "dropout": 0.1, |
| "dtype": "float32", |
| "embedding_dim_target": 1024, |
| "encoder_hidden_size": 1024, |
| "eos_token_id": 3, |
| "hidden_size": 1024, |
| "init_std": 0.02, |
| "intermediate_size": 4096, |
| "is_decoder": true, |
| "max_position_embeddings": 512, |
| "model_type": "stl_decoder", |
| "num_attention_heads": 16, |
| "num_hidden_layers": 12, |
| "pad_token_id": 1, |
| "scale_embedding": false, |
| "transformers_version": "4.57.3", |
| "use_cache": true, |
| "vocab_size": 35 |
| } |
|
|
