config.json
| 1 | { |
| 2 | "architectures": [ |
| 3 | "LongT5ForConditionalGeneration" |
| 4 | ], |
| 5 | "d_ff": 2048, |
| 6 | "d_kv": 64, |
| 7 | "d_model": 768, |
| 8 | "decoder_start_token_id": 0, |
| 9 | "dense_act_fn": "gelu_new", |
| 10 | "dropout_rate": 0.1, |
| 11 | "encoder_attention_type": "transient-global", |
| 12 | "eos_token_id": 1, |
| 13 | "feed_forward_proj": "gated-gelu", |
| 14 | "global_block_size": 16, |
| 15 | "initializer_factor": 1.0, |
| 16 | "is_encoder_decoder": true, |
| 17 | "is_gated_act": true, |
| 18 | "layer_norm_epsilon": 1e-06, |
| 19 | "local_radius": 127, |
| 20 | "max_length": 1024, |
| 21 | "min_length": 8, |
| 22 | "model_type": "longt5", |
| 23 | "n_positions": 4096, |
| 24 | "num_decoder_layers": 12, |
| 25 | "num_heads": 12, |
| 26 | "num_layers": 12, |
| 27 | "output_past": true, |
| 28 | "pad_token_id": 0, |
| 29 | "relative_attention_max_distance": 128, |
| 30 | "relative_attention_num_buckets": 32, |
| 31 | "tie_word_embeddings": false, |
| 32 | "torch_dtype": "float32", |
| 33 | "transformers_version": "4.27.4", |
| 34 | "use_cache": true, |
| 35 | "vocab_size": 32128 |
| 36 | } |
| 37 | |