{ "actual_seq_length": 512, "architectures": [ "FNetForPreTraining" ], "bos_token_id": 1, "dtype": "float32", "eos_token_id": 2, "hidden_act": "gelu_new", "hidden_dropout_prob": 0.1, "hidden_size": 64, "initializer_range": 0.02, "intermediate_size": 128, "layer_norm_eps": 1e-12, "max_position_embeddings": 512, "model_type": "fnet", "num_hidden_layers": 2, "pad_token_id": 3, "tie_word_embeddings": true, "tpu_short_seq_length": 512, "transformers_version": "5.22.0.dev0", "type_vocab_size": 4, "use_fft": true, "use_latest": false, "use_tpu_fourier_optimizations": false, "vocab_size": 32000 }