{ "architectures": [ "AmplifyForPreTraining" ], "attention_bias": false, "attention_dropout": 0.0, "bos_token_id": 1, "dtype": "float32", "eos_token_id": 2, "feedforward_bias": false, "head": null, "hidden_act": "swiglu", "hidden_dropout": 0.0, "hidden_size": 640, "id2label": null, "initializer_range": 0.02, "intermediate_size": 2560, "label2id": null, "layer_norm_after_embedding": false, "layer_norm_before_last_layer": true, "layer_norm_eps": 1e-05, "lm_head": { "act": null, "bias": true, "dropout": 0.0, "hidden_size": null, "layer_norm_eps": 1e-12, "loss_weight": null, "num_labels": 37, "output_name": null, "transform": null, "transform_act": null }, "mask_token_id": 4, "max_position_embeddings": 2048, "model_type": "amplify", "null_token_id": 5, "num_attention_heads": 10, "num_hidden_layers": 24, "num_labels": 1, "pad_token_id": 0, "rms_norm": true, "tie_word_embeddings": false, "transformers_version": "5.7.0", "unk_token_id": 3, "vocab_size": 37 }