{"LanguageModel": {"input": {"vocab_size": 49152, "pad_vocab_size_multiple": 8, "tie_word_embeddings": true, "lm_head_bias": false}}, "MixerModel": {"input": {"d_model": 2048, "n_layer": 24, "lm_head_prenorm": "rms", "rms_norm_eps": 1e-05}}, "Block1": {"n_layers": 24, "BlockType": "modules.llama_block", "block_input": {"resid_dropout": 0.0, "rms_norm_eps": 1e-05}, "mlp": {"intermediate_size": 8192}, "CoreType": "modules.mixers.discrete_mamba2", "core_input": {"d_state": 96, "n_v_heads": 32, "n_qk_heads": 32, "d_conv": 4, "conv_bias": true, "expand": 1, "chunk_size": 128, "activation": "identity", "bias": false}}}