{ "architectures": [ "ScBassetForSequencePrediction" ], "batch_norm_eps": 0.001, "batch_norm_momentum": 0.1, "bos_token_id": 1, "bottleneck_size": 32, "dtype": "float32", "eos_token_id": 2, "head": { "act": null, "bias": true, "dropout": 0.0, "hidden_size": 32, "layer_norm_eps": 1e-12, "loss_weight": null, "num_labels": 2034, "output_name": null, "problem_type": "binary", "transform": null, "transform_act": "gelu", "type": null }, "hidden_act": "quick_gelu", "hidden_dropout": 0.2, "hidden_size": 32, "id2label": null, "label2id": null, "mask_token_id": 4, "model_type": "scbasset", "null_token_id": 5, "num_labels": 2034, "pad_token_id": 0, "pointwise_channels": 256, "sequence_length": 1344, "stem_channels": 288, "stem_kernel_size": 17, "stem_pool_size": 3, "tie_word_embeddings": true, "tower_channels": [ 288, 323, 363, 407, 456, 512 ], "tower_kernel_size": 5, "tower_pool_size": 2, "transformers_version": "5.7.0", "unk_token_id": 3, "vocab_size": 5 }